diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.c
index feb0fa7089338aa0ebfa855ac9318ed7d9b68604..fd015580495e5c1832ff666653ad45e189d5b88a 100644
--- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.c
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.c
@@ -71,7 +71,7 @@ fprintf(fd, "// Process group with 1 CNs \n");
// Process group with 2 CNs
-
+/*
if (lut_numBnInBnGroups[0] > 0)
{
// If elements in group move to next address
@@ -123,7 +123,7 @@ fprintf(fd, "// Process group with 1 CNs \n");
}
// =====================================================================
// Process group with 2 CNs
-
+*/
fprintf(fd, "// Process group with 2 CNs \n");
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.o b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.o
new file mode 100644
index 0000000000000000000000000000000000000000..8caf4b85a31f5704239e0d6aa2071be5529f93a2
Binary files /dev/null and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG1_avx2.o differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG2_avx2.o b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG2_avx2.o
new file mode 100644
index 0000000000000000000000000000000000000000..2f6a08ea358ebdad5d329ffe676e4f502c3abf58
Binary files /dev/null and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProcPc_gen_BG2_avx2.o differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_BG1_avx2.o b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_BG1_avx2.o
new file mode 100644
index 0000000000000000000000000000000000000000..d40fd0a3af91fdbdb82afdac3ac2f5500eec1dc6
Binary files /dev/null and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_BG1_avx2.o differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_BG2_avx2.o b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_BG2_avx2.o
new file mode 100644
index 0000000000000000000000000000000000000000..248c7244cf105f023b1896cd4308aee4e0e8ce83
Binary files /dev/null and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_BG2_avx2.o differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_avx2 b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_avx2
index d2fd3730144602aa16b4728e7acbcaff53d6e823..8e7d6324bd6cb8fa9fbb81d57e9d9c56413530da 100755
Binary files a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_avx2 and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/bnProc_gen_avx2 differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/main.o b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/main.o
new file mode 100644
index 0000000000000000000000000000000000000000..c9ea8704ae2d57a0c36127605f696ab9b1febb99
Binary files /dev/null and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/main.o differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/sauvegarde.tar.gz b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/sauvegarde.tar.gz
new file mode 100644
index 0000000000000000000000000000000000000000..438d0b4b2111a33261ab02399920965505229458
Binary files /dev/null and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc/sauvegarde.tar.gz differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProcPc_gen_BG1_avx512.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProcPc_gen_BG1_avx512.c
new file mode 100644
index 0000000000000000000000000000000000000000..09bd99521639960ad989eb46a832cf33ee5c96af
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProcPc_gen_BG1_avx512.c
@@ -0,0 +1,1768 @@
+
+
+
+
+#include <stdint.h>
+#include <immintrin.h>
+#include "../../nrLDPCdecoder_defs.h"
+#include "../../nrLDPC_types.h"
+
+
+void nrLDPC_bnProcPc_BG1_generator_AVX512(int R)
+{
+ const char *ratestr[3]={"13","23","89"};
+
+ if (R<0 || R>2) {printf("Illegal R %d\n",R); abort();}
+
+
+
+
+ char fname[50];
+ sprintf(fname,"../ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R%s_AVX512.h",ratestr[R]);
+ FILE *fd=fopen(fname,"w");
+ if (fd == NULL) {printf("Cannot create \n");abort();}
+
+
+
+ fprintf(fd,"static inline void nrLDPC_bnProcPc_BG1_R%s_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {\n",ratestr[R]);
+ const uint8_t* lut_numBnInBnGroups;
+ const uint32_t* lut_startAddrBnGroups;
+ const uint16_t* lut_startAddrBnGroupsLlr;
+ if (R==0) {
+
+
+ lut_numBnInBnGroups = lut_numBnInBnGroups_BG1_R13;
+ lut_startAddrBnGroups = lut_startAddrBnGroups_BG1_R13;
+ lut_startAddrBnGroupsLlr = lut_startAddrBnGroupsLlr_BG1_R13;
+
+ }
+ else if (R==1){
+
+ lut_numBnInBnGroups = lut_numBnInBnGroups_BG1_R23;
+ lut_startAddrBnGroups = lut_startAddrBnGroups_BG1_R23;
+ lut_startAddrBnGroupsLlr = lut_startAddrBnGroupsLlr_BG1_R23;
+ }
+ else if (R==2) {
+
+ lut_numBnInBnGroups = lut_numBnInBnGroups_BG1_R89;
+ lut_startAddrBnGroups = lut_startAddrBnGroups_BG1_R89;
+ lut_startAddrBnGroupsLlr = lut_startAddrBnGroupsLlr_BG1_R89;
+ }
+ else { printf("aborting, illegal R %d\n",R); fclose(fd);abort();}
+ // Number of BNs in Groups
+// uint32_t M;
+ //uint32_t M32rem;
+ //uint32_t i,j;
+ uint32_t k;
+ // Offset to each bit within a group in terms of 32 Byte
+ uint32_t cnOffsetInGroup;
+ uint8_t idxBnGroup = 0;
+
+ fprintf(fd," __m512i zmm0, zmm1, zmmRes0, zmmRes1; \n");
+
+
+ fprintf(fd," __m256i* p_bnProcBuf; \n");
+ fprintf(fd," __m256i* p_llrProcBuf;\n");
+ fprintf(fd," __m512i* p_llrRes; \n");
+ fprintf(fd," uint32_t M ;\n");
+
+
+fprintf(fd, "// Process group with 1 CNs \n");
+
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[0] > 0)
+ {
+ // If elements in group move to next address
+ // idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[0] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[0]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<1; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+
+ fprintf(fd,"}\n");
+ }
+ // =====================================================================
+ // Process group with 2 CNs
+
+
+fprintf(fd, "// Process group with 2 CNs \n");
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[1] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[1] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[1]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<2; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 3 CNs
+
+
+fprintf(fd, "// Process group with 3 CNs \n");
+
+ // Process group with 3 CNs
+
+ if (lut_numBnInBnGroups[2] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[2] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[2]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<3; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 4 CNs
+
+fprintf(fd, "// Process group with 4 CNs \n");
+
+ // Process group with 4 CNs
+
+ if (lut_numBnInBnGroups[3] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[3] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[3]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<4; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 5 CNs
+
+fprintf(fd, "// Process group with 5 CNs \n");
+
+ // Process group with 5 CNs
+
+ if (lut_numBnInBnGroups[4] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[4] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[4]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<5; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 6 CNs
+
+fprintf(fd, "// Process group with 6 CNs \n");
+
+ // Process group with 6 CNs
+
+ if (lut_numBnInBnGroups[5] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[5] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[5]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<6; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 7 CNs
+
+fprintf(fd, "// Process group with 7 CNs \n");
+
+ // Process group with 7 CNs
+
+ if (lut_numBnInBnGroups[6] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[6] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[6]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<7; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ //fprintf(fd," (__m512i*) &llrRes[%d + i] = _mm512_permutex_epi64(zmm0, 0xD8);\n",lut_startAddrBnGroupsLlr[idxBnGroup]>>6 );
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 8 CNs
+
+fprintf(fd, "// Process group with 8 CNs \n");
+
+ // Process group with 8 CNs
+
+ if (lut_numBnInBnGroups[7] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[7] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[7]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<8; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ //fprintf(fd," (__m512i*) &llrRes[%d + i] = _mm512_permutex_epi64(zmm0, 0xD8);\n",lut_startAddrBnGroupsLlr[idxBnGroup]>>6 );
+
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 9 CNs
+
+fprintf(fd, "// Process group with 9 CNs \n");
+
+ // Process group with 9 CNs
+
+ if (lut_numBnInBnGroups[8] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[8] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[8]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<9; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ //fprintf(fd," (__m512i*) &llrRes[%d + i] = _mm512_permutex_epi64(zmm0, 0xD8);\n",lut_startAddrBnGroupsLlr[idxBnGroup]>>6 );
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 10 CNs
+
+fprintf(fd, "// Process group with 10 CNs \n");
+
+ // Process group with 10 CNs
+
+ if (lut_numBnInBnGroups[9] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[9] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[9]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<10; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+
+fprintf(fd, "// Process group with 11 CNs \n");
+
+
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[10] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[10] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[10]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<11; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+ // =====================================================================
+ // Process group with 2 CNs
+
+
+fprintf(fd, "// Process group with 12 CNs \n");
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[11] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[11] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[11]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<12; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 13 CNs
+
+
+fprintf(fd, "// Process group with 13 CNs \n");
+
+ // Process group with 3 CNs
+
+ if (lut_numBnInBnGroups[12] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[12] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[12]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<13; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 4 CNs
+
+fprintf(fd, "// Process group with 14 CNs \n");
+
+ // Process group with 4 CNs
+
+ if (lut_numBnInBnGroups[13] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[13] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[13]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<14; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 5 CNs
+
+fprintf(fd, "// Process group with 15 CNs \n");
+
+ // Process group with 5 CNs
+
+ if (lut_numBnInBnGroups[14] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[14] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[14]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<15; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 6 CNs
+
+fprintf(fd, "// Process group with 16 CNs \n");
+
+ // Process group with 6 CNs
+
+ if (lut_numBnInBnGroups[15] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[15] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[15]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<16; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 17 CNs
+
+fprintf(fd, "// Process group with 17 CNs \n");
+
+ // Process group with 17 CNs
+
+ if (lut_numBnInBnGroups[16] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[16] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[16]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<17; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 18 CNs
+
+fprintf(fd, "// Process group with 18 CNs \n");
+
+ // Process group with 8 CNs
+
+ if (lut_numBnInBnGroups[17] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[17] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[17]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<18; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 9 CNs
+
+fprintf(fd, "// Process group with 19 CNs \n");
+
+ // Process group with 9 CNs
+
+ if (lut_numBnInBnGroups[18] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[18] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[18]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<19; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 20 CNs
+
+fprintf(fd, "// Process group with 20 CNs \n");
+
+ // Process group with 20 CNs
+
+ if (lut_numBnInBnGroups[19] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[19] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[19]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<20; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+
+
+ // =====================================================================
+
+fprintf(fd, "// Process group with 21 CNs \n");
+
+
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[20] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[20] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[20]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<21; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+ // =====================================================================
+ // Process group with 2 CNs
+
+
+fprintf(fd, "// Process group with 22 CNs \n");
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[21] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[21] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[21]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<22; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 13 CNs
+
+
+fprintf(fd, "// Process group with <23 CNs \n");
+
+ // Process group with 3 CNs
+
+ if (lut_numBnInBnGroups[22] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[22] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[22]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<23; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 4 CNs
+
+fprintf(fd, "// Process group with 24 CNs \n");
+
+ // Process group with 4 CNs
+
+ if (lut_numBnInBnGroups[23] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[23] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[23]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<24; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 5 CNs
+
+fprintf(fd, "// Process group with 25 CNs \n");
+
+ // Process group with 5 CNs
+
+ if (lut_numBnInBnGroups[24] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[24] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[24]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<25; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 6 CNs
+
+fprintf(fd, "// Process group with 26 CNs \n");
+
+ // Process group with 6 CNs
+
+ if (lut_numBnInBnGroups[25] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[25] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[25]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<26; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 17 CNs
+
+fprintf(fd, "// Process group with 27 CNs \n");
+
+ // Process group with 17 CNs
+
+ if (lut_numBnInBnGroups[26] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[26] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[26]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<27; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 18 CNs
+
+fprintf(fd, "// Process group with 28 CNs \n");
+
+ // Process group with 8 CNs
+
+ if (lut_numBnInBnGroups[27] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[27] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[27]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<28; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 9 CNs
+
+fprintf(fd, "// Process group with 29 CNs \n");
+
+ // Process group with 9 CNs
+
+ if (lut_numBnInBnGroups[28] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[28] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[28]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<29; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 20 CNs
+
+fprintf(fd, "// Process group with 30 CNs \n");
+
+ // Process group with 20 CNs
+
+ if (lut_numBnInBnGroups[29] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[29] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[29]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<30; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);\n");
+ // zmm0 = [zmmRes1[255:256] zmmRes0[255:256] zmmRes1[127:0] zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256] zmmRes1[127:0] zmmRes0[255:256] zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ fprintf(fd,"}\n");
+ fclose(fd);
+}//end of the function nrLDPC_bnProcPc_BG1
+
+
+
+
+
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProcPc_gen_BG2_avx512.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProcPc_gen_BG2_avx512.c
new file mode 100644
index 0000000000000000000000000000000000000000..a025ae8efb3b1b653720ae741c66e2d8f1ce38a6
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProcPc_gen_BG2_avx512.c
@@ -0,0 +1,1770 @@
+
+
+
+#include <stdint.h>
+#include <immintrin.h>
+#include "../../nrLDPCdecoder_defs.h"
+#include "../../nrLDPC_types.h"
+
+
+void nrLDPC_bnProcPc_BG2_generator_AVX512(int R)
+{
+ const char *ratestr[3]={"15","13","23"};
+
+ if (R<0 || R>2) {printf("Illegal R %d\n",R); abort();}
+
+
+ // system("mkdir -p ../ldpc_gen_files");
+
+ char fname[50];
+ sprintf(fname,"../ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R%s_AVX512.h",ratestr[R]);
+ FILE *fd=fopen(fname,"w");
+ if (fd == NULL) {printf("Cannot create \n");abort();}
+
+// fprintf(fd,"#include <stdint.h>\n");
+ //fprintf(fd,"#include <immintrin.h>\n");
+
+ fprintf(fd,"static inline void nrLDPC_bnProcPc_BG2_R%s_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {\n",ratestr[R]);
+ const uint8_t* lut_numBnInBnGroups;
+ const uint32_t* lut_startAddrBnGroups;
+ const uint16_t* lut_startAddrBnGroupsLlr;
+ if (R==0) {
+
+
+ lut_numBnInBnGroups = lut_numBnInBnGroups_BG2_R15;
+ lut_startAddrBnGroups = lut_startAddrBnGroups_BG2_R15;
+ lut_startAddrBnGroupsLlr = lut_startAddrBnGroupsLlr_BG2_R15;
+
+ }
+ else if (R==1){
+
+ lut_numBnInBnGroups = lut_numBnInBnGroups_BG2_R13;
+ lut_startAddrBnGroups = lut_startAddrBnGroups_BG2_R13;
+ lut_startAddrBnGroupsLlr = lut_startAddrBnGroupsLlr_BG2_R13;
+ }
+ else if (R==2) {
+
+ lut_numBnInBnGroups = lut_numBnInBnGroups_BG2_R23;
+ lut_startAddrBnGroups = lut_startAddrBnGroups_BG2_R23;
+ lut_startAddrBnGroupsLlr = lut_startAddrBnGroupsLlr_BG2_R23;
+ }
+ else { printf("aborting, illegal R %d\n",R); fclose(fd);abort();}
+
+
+ uint32_t k;
+ // Offset to each bit within a group in terms of 32 Byte
+ uint32_t cnOffsetInGroup;
+ uint8_t idxBnGroup = 0;
+
+ fprintf(fd," __m512i zmm0,zmm1,zmmRes0,zmmRes1; \n");
+
+
+ fprintf(fd," __m256i* p_bnProcBuf; \n");
+ fprintf(fd," __m256i* p_llrProcBuf;\n");
+ fprintf(fd," __m512i* p_llrRes; \n");
+ fprintf(fd," uint32_t M ;\n");
+
+
+fprintf(fd, "// Process group with 1 CNs \n");
+
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[0] > 0)
+ {
+ // If elements in group move to next address
+ // idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[0] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[0]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+
+ // Loop over CNs
+ /*for (k=1; k<1; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+*/
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[j+1]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+
+ fprintf(fd,"}\n");
+ }
+ // =====================================================================
+ // Process group with 2 CNs
+
+
+fprintf(fd, "// Process group with 2 CNs \n");
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[1] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[1] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[1]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<2; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 3 CNs
+
+
+fprintf(fd, "// Process group with 3 CNs \n");
+
+ // Process group with 3 CNs
+
+ if (lut_numBnInBnGroups[2] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[2] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[2]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<3; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 4 CNs
+
+fprintf(fd, "// Process group with 4 CNs \n");
+
+ // Process group with 4 CNs
+
+ if (lut_numBnInBnGroups[3] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[3] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[3]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<4; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 5 CNs
+
+fprintf(fd, "// Process group with 5 CNs \n");
+
+ // Process group with 5 CNs
+
+ if (lut_numBnInBnGroups[4] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[4] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[4]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<5; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 6 CNs
+
+fprintf(fd, "// Process group with 6 CNs \n");
+
+ // Process group with 6 CNs
+
+ if (lut_numBnInBnGroups[5] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[5] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[5]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<6; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 7 CNs
+
+fprintf(fd, "// Process group with 7 CNs \n");
+
+ // Process group with 7 CNs
+
+ if (lut_numBnInBnGroups[6] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[6] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[6]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<7; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ //fprintf(fd," (__m512i*) &llrRes[%d + i] = _mm512_permutex_epi64(zmm0, 0xD8);\n",lut_startAddrBnGroupsLlr[idxBnGroup]>>5 );
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 8 CNs
+
+fprintf(fd, "// Process group with 8 CNs \n");
+
+ // Process group with 8 CNs
+
+ if (lut_numBnInBnGroups[7] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[7] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[7]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<8; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ //fprintf(fd," (__m512i*) &llrRes[%d + i] = _mm512_permutex_epi64(zmm0, 0xD8);\n",lut_startAddrBnGroupsLlr[idxBnGroup]>>5 );
+
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 9 CNs
+
+fprintf(fd, "// Process group with 9 CNs \n");
+
+ // Process group with 9 CNs
+
+ if (lut_numBnInBnGroups[8] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[8] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[8]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<9; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ //fprintf(fd," (__m512i*) &llrRes[%d + i] = _mm512_permutex_epi64(zmm0, 0xD8);\n",lut_startAddrBnGroupsLlr[idxBnGroup]>>5 );
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 10 CNs
+
+fprintf(fd, "// Process group with 10 CNs \n");
+
+ // Process group with 10 CNs
+
+ if (lut_numBnInBnGroups[9] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[9] );
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[9]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<10; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+
+fprintf(fd, "// Process group with 11 CNs \n");
+
+
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[10] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[10] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[10]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<11; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+ // =====================================================================
+ // Process group with 2 CNs
+
+
+fprintf(fd, "// Process group with 12 CNs \n");
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[11] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[11] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[11]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<12; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 13 CNs
+
+
+fprintf(fd, "// Process group with 13 CNs \n");
+
+ // Process group with 3 CNs
+
+ if (lut_numBnInBnGroups[12] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[12] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[12]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<13; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 4 CNs
+
+fprintf(fd, "// Process group with 14 CNs \n");
+
+ // Process group with 4 CNs
+
+ if (lut_numBnInBnGroups[13] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[13] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[13]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<14; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 5 CNs
+
+fprintf(fd, "// Process group with 15 CNs \n");
+
+ // Process group with 5 CNs
+
+ if (lut_numBnInBnGroups[14] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[14] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[14]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<15; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 6 CNs
+
+fprintf(fd, "// Process group with 16 CNs \n");
+
+ // Process group with 6 CNs
+
+ if (lut_numBnInBnGroups[15] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[15] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[15]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<16; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 17 CNs
+
+fprintf(fd, "// Process group with 17 CNs \n");
+
+ // Process group with 17 CNs
+
+ if (lut_numBnInBnGroups[16] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[16] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[16]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<17; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 18 CNs
+
+fprintf(fd, "// Process group with 18 CNs \n");
+
+ // Process group with 8 CNs
+
+ if (lut_numBnInBnGroups[17] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[17] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[17]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<18; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 9 CNs
+
+fprintf(fd, "// Process group with 19 CNs \n");
+
+ // Process group with 9 CNs
+
+ if (lut_numBnInBnGroups[18] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[18] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[18]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<19; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 20 CNs
+
+fprintf(fd, "// Process group with 20 CNs \n");
+
+ // Process group with 20 CNs
+
+ if (lut_numBnInBnGroups[19] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[19] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[19]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<20; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+
+
+ // =====================================================================
+
+fprintf(fd, "// Process group with 21 CNs \n");
+
+
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[20] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[20] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[20]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<21; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+ // =====================================================================
+ // Process group with 2 CNs
+
+
+fprintf(fd, "// Process group with 22 CNs \n");
+
+ // Process group with 2 CNs
+
+ if (lut_numBnInBnGroups[21] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[21] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[21]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<22; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 13 CNs
+
+
+fprintf(fd, "// Process group with <23 CNs \n");
+
+ // Process group with 3 CNs
+
+ if (lut_numBnInBnGroups[22] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[22] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[22]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<23; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 4 CNs
+
+fprintf(fd, "// Process group with 24 CNs \n");
+
+ // Process group with 4 CNs
+
+ if (lut_numBnInBnGroups[23] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[23] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[23]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<24; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 5 CNs
+
+fprintf(fd, "// Process group with 25 CNs \n");
+
+ // Process group with 5 CNs
+
+ if (lut_numBnInBnGroups[24] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[24] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[24]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<25; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+
+ // =====================================================================
+ // Process group with 6 CNs
+
+fprintf(fd, "// Process group with 26 CNs \n");
+
+ // Process group with 6 CNs
+
+ if (lut_numBnInBnGroups[25] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[25] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[25]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<26; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 17 CNs
+
+fprintf(fd, "// Process group with 27 CNs \n");
+
+ // Process group with 17 CNs
+
+ if (lut_numBnInBnGroups[26] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[26] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[26]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<27; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 18 CNs
+
+fprintf(fd, "// Process group with 28 CNs \n");
+
+ // Process group with 8 CNs
+
+ if (lut_numBnInBnGroups[27] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[27] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[27]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<28; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ // =====================================================================
+ // Process group with 9 CNs
+
+fprintf(fd, "// Process group with 29 CNs \n");
+
+ // Process group with 9 CNs
+
+ if (lut_numBnInBnGroups[28] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[28] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[28]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<29; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+
+ // =====================================================================
+ // Process group with 20 CNs
+
+fprintf(fd, "// Process group with 30 CNs \n");
+
+ // Process group with 20 CNs
+
+ if (lut_numBnInBnGroups[29] > 0)
+ {
+ // If elements in group move to next address
+ idxBnGroup++;
+
+ // Number of groups of 32 BNs for parallel processing
+ fprintf(fd," M = (%d*Z + 63)>>6;\n",lut_numBnInBnGroups[29] );;
+
+ // Set the offset to each CN within a group in terms of 16 Byte
+ cnOffsetInGroup = (lut_numBnInBnGroups[29]*NR_LDPC_ZMAX)>>5;
+
+ // Set pointers to start of group 2
+ fprintf(fd," p_bnProcBuf = (__m256i*) &bnProcBuf [%d];\n",lut_startAddrBnGroups[idxBnGroup]);
+ fprintf(fd," p_llrProcBuf = (__m256i*) &llrProcBuf [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ fprintf(fd," p_llrRes = (__m512i*) &llrRes [%d];\n",lut_startAddrBnGroupsLlr[idxBnGroup]);
+ // Loop over BNs
+ fprintf(fd," for (int i=0,j=0;i<M;i++,j+=2) {\n");
+ // First 16 LLRs of first CN
+ fprintf(fd," zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);\n");
+ fprintf(fd," zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);\n");
+
+ // Loop over CNs
+ for (k=1; k<30; k++)
+ {
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j]);\n", k*cnOffsetInGroup);
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[%d + j +1]);\n", k*cnOffsetInGroup);
+
+ fprintf(fd, " zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1); \n");
+ }
+
+ // Add LLR from receiver input
+ fprintf(fd," zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);\n");
+ fprintf(fd," zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);\n");
+
+ fprintf(fd," zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);\n");
+ fprintf(fd," zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);\n");
+
+ // Pack results back to epi8
+ fprintf(fd," zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);\n");
+ //zmm0 = [zmmRes1[255:256]zmmRes0[255:256]zmmRes1[127:0]zmmRes0[127:0]]
+ // p_llrRes = [zmmRes1[255:256]zmmRes1[127:0]zmmRes0[255:256]zmmRes0[127:0]]
+ fprintf(fd," p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);\n");
+
+ fprintf(fd,"}\n");
+ }
+
+ fprintf(fd,"}\n");
+ fclose(fd);
+}//end of the function nrLDPC_bnProcPc_BG2
+
+
+
+
+
+
+
+
+
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProc_gen_avx512 b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProc_gen_avx512
index f686a271f18d31892f83a65819d96e3439325fd9..dabfabe18fdbb098b9ea9a0996079e0fde890445 100755
Binary files a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProc_gen_avx512 and b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/bnProc_gen_avx512 differ
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/main.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/main.c
index 3b2bc3ede498433e920ff2431f816ea7aa1a0d53..2bc870f17b10d31ab6e6ad4ac0bb761f5c0ef012 100644
--- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/main.c
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_bnProc_avx512/main.c
@@ -4,8 +4,8 @@
#define NB_R 3
void nrLDPC_bnProc_BG1_generator_AVX512(int);
void nrLDPC_bnProc_BG2_generator_AVX512(int);
-//void nrLDPC_bnProcPc_BG1_generator_AVX2(int);
-//void nrLDPC_bnProcPc_BG2_generator_AVX2(int);
+void nrLDPC_bnProcPc_BG1_generator_AVX512(int);
+void nrLDPC_bnProcPc_BG2_generator_AVX512(int);
int main()
{
@@ -15,8 +15,8 @@ int main()
nrLDPC_bnProc_BG1_generator_AVX512(R[i]);
nrLDPC_bnProc_BG2_generator_AVX512(R[i]);
-// nrLDPC_bnProcPc_BG1_generator_AVX2(R[i]);
-// nrLDPC_bnProcPc_BG2_generator_AVX2(R[i]);
+ nrLDPC_bnProcPc_BG1_generator_AVX512(R[i]);
+ nrLDPC_bnProcPc_BG2_generator_AVX512(R[i]);
}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R13_AVX2.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R13_AVX2.c
new file mode 100644
index 0000000000000000000000000000000000000000..3f1c421cebdee768cc056933786826f33ddf9fe6
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R13_AVX2.c
@@ -0,0 +1,440 @@
+#include <stdint.h>
+#include <immintrin.h>
+void nrLDPC_bnProc_BG2_R13_AVX2(int8_t* bnProcBuf,int8_t* bnProcBufRes, int8_t* llrRes, uint16_t Z ) {
+ __m256i* p_bnProcBuf;
+ __m256i* p_bnProcBufRes;
+ __m256i* p_llrRes;
+ __m256i* p_res;
+ uint32_t M, i;
+// Process group with 2 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [6912];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [6912];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [6912];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [6912];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+// Process group with 3 CNs
+// Process group with 4 CNs
+ M = (2*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [7680];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [7680];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [7296];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [7296];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [7296];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [7296];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+// Process group with 5 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [10752];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [10752];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [8064];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [8064];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [8064];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [8064];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [8064];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+// Process group with 6 CNs
+ M = (5*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [12672];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [12672];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [8448];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [8448];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [8448];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[180];
+ p_llrRes = (__m256i*) &llrRes [8448];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[180 + i]);
+}
+ p_res = &p_bnProcBufRes[240];
+ p_llrRes = (__m256i*) &llrRes [8448];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[240 + i]);
+}
+ p_res = &p_bnProcBufRes[300];
+ p_llrRes = (__m256i*) &llrRes [8448];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[300 + i]);
+}
+// Process group with 7 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [24192];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [24192];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [10368];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+// Process group with 8 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [26880];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [26880];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [10752];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+// Process group with 9 CNs
+// Process group with 10 CNs
+// Process group with 11 CNs
+// Process group with 12 CNs
+// Process group with 13 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [29952];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [29952];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [11136];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+// Process group with 14 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [34944];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [34944];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[156];
+ p_llrRes = (__m256i*) &llrRes [11520];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[156 + i]);
+}
+// Process group with 15 CNs
+// Process group with 16 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [40320];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [40320];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[156];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[156 + i]);
+}
+ p_res = &p_bnProcBufRes[168];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[168 + i]);
+}
+ p_res = &p_bnProcBufRes[180];
+ p_llrRes = (__m256i*) &llrRes [11904];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[180 + i]);
+}
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R15_AVX2.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R15_AVX2.c
new file mode 100644
index 0000000000000000000000000000000000000000..f649102df7a8cde518242a39292f8180778425e4
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R15_AVX2.c
@@ -0,0 +1,799 @@
+#include <stdint.h>
+#include <immintrin.h>
+void nrLDPC_bnProc_BG2_R15_AVX2(int8_t* bnProcBuf,int8_t* bnProcBufRes, int8_t* llrRes, uint16_t Z ) {
+ __m256i* p_bnProcBuf;
+ __m256i* p_bnProcBufRes;
+ __m256i* p_llrRes;
+ __m256i* p_res;
+ uint32_t M, i;
+// Process group with 2 CNs
+// Process group with 3 CNs
+// Process group with 4 CNs
+// Process group with 5 CNs
+ M = (2*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [14592];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [14592];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [14592];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [14592];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [14592];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [14592];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [14592];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+// Process group with 6 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [18432];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [18432];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [15360];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [15360];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [15360];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [15360];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [15360];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [15360];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+// Process group with 7 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [20736];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [20736];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [15744];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+// Process group with 8 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [23424];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [23424];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [16128];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+// Process group with 9 CNs
+ M = (2*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [26496];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [26496];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[168];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[168 + i]);
+}
+ p_res = &p_bnProcBufRes[192];
+ p_llrRes = (__m256i*) &llrRes [16512];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[192 + i]);
+}
+// Process group with 10 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [33408];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [33408];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [17280];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+// Process group with 11 CNs
+// Process group with 12 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [37248];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [37248];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [17664];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+// Process group with 13 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [41856];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [41856];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [18048];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+// Process group with 14 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [46848];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [46848];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[156];
+ p_llrRes = (__m256i*) &llrRes [18432];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[156 + i]);
+}
+// Process group with 15 CNs
+// Process group with 16 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [52224];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [52224];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[156];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[156 + i]);
+}
+ p_res = &p_bnProcBufRes[168];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[168 + i]);
+}
+ p_res = &p_bnProcBufRes[180];
+ p_llrRes = (__m256i*) &llrRes [18816];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[180 + i]);
+}
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [58368];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [58368];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[156];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[156 + i]);
+}
+ p_res = &p_bnProcBufRes[168];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[168 + i]);
+}
+ p_res = &p_bnProcBufRes[180];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[180 + i]);
+}
+ p_res = &p_bnProcBufRes[192];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[192 + i]);
+}
+ p_res = &p_bnProcBufRes[204];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[204 + i]);
+}
+ p_res = &p_bnProcBufRes[216];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[216 + i]);
+}
+ p_res = &p_bnProcBufRes[228];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[228 + i]);
+}
+ p_res = &p_bnProcBufRes[240];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[240 + i]);
+}
+ p_res = &p_bnProcBufRes[252];
+ p_llrRes = (__m256i*) &llrRes [19200];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[252 + i]);
+}
+// Process group with <23 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [66816];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [66816];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[84];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[84 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+ p_res = &p_bnProcBufRes[132];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[132 + i]);
+}
+ p_res = &p_bnProcBufRes[144];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[144 + i]);
+}
+ p_res = &p_bnProcBufRes[156];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[156 + i]);
+}
+ p_res = &p_bnProcBufRes[168];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[168 + i]);
+}
+ p_res = &p_bnProcBufRes[180];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[180 + i]);
+}
+ p_res = &p_bnProcBufRes[192];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[192 + i]);
+}
+ p_res = &p_bnProcBufRes[204];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[204 + i]);
+}
+ p_res = &p_bnProcBufRes[216];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[216 + i]);
+}
+ p_res = &p_bnProcBufRes[228];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[228 + i]);
+}
+ p_res = &p_bnProcBufRes[240];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[240 + i]);
+}
+ p_res = &p_bnProcBufRes[252];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[252 + i]);
+}
+ p_res = &p_bnProcBufRes[264];
+ p_llrRes = (__m256i*) &llrRes [19584];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[264 + i]);
+}
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R23_AVX2.c b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R23_AVX2.c
new file mode 100644
index 0000000000000000000000000000000000000000..59a1613099c8da476887c855153e5a4b2ebdd575
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProc/nrLDPC_bnProc_BG2_R23_AVX2.c
@@ -0,0 +1,153 @@
+#include <stdint.h>
+#include <immintrin.h>
+void nrLDPC_bnProc_BG2_R23_AVX2(int8_t* bnProcBuf,int8_t* bnProcBufRes, int8_t* llrRes, uint16_t Z ) {
+ __m256i* p_bnProcBuf;
+ __m256i* p_bnProcBufRes;
+ __m256i* p_llrRes;
+ __m256i* p_res;
+ uint32_t M, i;
+// Process group with 2 CNs
+ M = (3*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [1152];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [1152];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [1152];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [1152];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+// Process group with 3 CNs
+ M = (5*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [3456];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [3456];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [2304];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [2304];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+ p_res = &p_bnProcBufRes[120];
+ p_llrRes = (__m256i*) &llrRes [2304];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[120 + i]);
+}
+// Process group with 4 CNs
+ M = (3*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [9216];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [9216];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [4224];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [4224];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [4224];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[108];
+ p_llrRes = (__m256i*) &llrRes [4224];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[108 + i]);
+}
+// Process group with 5 CNs
+ M = (2*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [13824];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [13824];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [5376];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [5376];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [5376];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[72];
+ p_llrRes = (__m256i*) &llrRes [5376];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[72 + i]);
+}
+ p_res = &p_bnProcBufRes[96];
+ p_llrRes = (__m256i*) &llrRes [5376];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[96 + i]);
+}
+// Process group with 6 CNs
+ M = (1*Z + 31)>>5;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [17664];
+ p_bnProcBufRes = (__m256i*) &bnProcBufRes [17664];
+ p_res = &p_bnProcBufRes[0];
+ p_llrRes = (__m256i*) &llrRes [6144];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[0 + i]);
+}
+ p_res = &p_bnProcBufRes[12];
+ p_llrRes = (__m256i*) &llrRes [6144];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[12 + i]);
+}
+ p_res = &p_bnProcBufRes[24];
+ p_llrRes = (__m256i*) &llrRes [6144];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[24 + i]);
+}
+ p_res = &p_bnProcBufRes[36];
+ p_llrRes = (__m256i*) &llrRes [6144];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[36 + i]);
+}
+ p_res = &p_bnProcBufRes[48];
+ p_llrRes = (__m256i*) &llrRes [6144];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[48 + i]);
+}
+ p_res = &p_bnProcBufRes[60];
+ p_llrRes = (__m256i*) &llrRes [6144];
+ for (i=0;i<M;i++) {
+ p_res[i] = _mm256_subs_epi8(p_llrRes[i], p_bnProcBuf[60 + i]);
+}
+// Process group with 7 CNs
+// Process group with 8 CNs
+// Process group with 9 CNs
+// Process group with 10 CNs
+// Process group with 11 CNs
+// Process group with 12 CNs
+// Process group with 13 CNs
+// Process group with 14 CNs
+// Process group with 15 CNs
+// Process group with 16 CNs
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R13_AVX2.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R13_AVX2.h
index a4508d2d7a3db278a3700ccba3ae4c0c0fad8168..b10a842171b778f7d4323b6e1ad35ac6deb01b95 100644
--- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R13_AVX2.h
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R13_AVX2.h
@@ -5,20 +5,6 @@ static inline void nrLDPC_bnProcPc_BG1_R13_AVX2(int8_t* bnProcBuf,int8_t* llrRes
__m256i* p_llrRes;
uint32_t M ;
// Process group with 1 CNs
- M = (42*Z + 31)>>5;
- p_bnProcBuf = (__m128i*) &bnProcBuf [0];
- p_llrProcBuf = (__m128i*) &llrProcBuf [0];
- p_llrRes = (__m256i*) &llrRes [0];
- for (int i=0,j=0;i<M;i++,j+=2) {
- ymm0 = _mm256_cvtepi8_epi16(p_bnProcBuf [j]);
- ymm1 = _mm256_cvtepi8_epi16(p_bnProcBuf[j + 1]);
- ymm0 = _mm256_cvtepi8_epi16(p_llrProcBuf[j]);
- ymmRes0 = _mm256_adds_epi16(ymmRes0, ymm0);
- ymm1 = _mm256_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
- ymmRes1 = _mm256_adds_epi16(ymmRes1, ymm1);
- ymm0 = _mm256_packs_epi16(ymmRes0, ymmRes1);
- p_llrRes[i] = _mm256_permute4x64_epi64(ymm0, 0xD8);
-}
// Process group with 2 CNs
// Process group with 3 CNs
// Process group with 4 CNs
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R23_AVX2.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R23_AVX2.h
index 47fd449fe44011f19cfdd68225d5c79fb62e37d5..8647e8cde6a758fa7e38dbbda53184d7559cfc38 100644
--- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R23_AVX2.h
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R23_AVX2.h
@@ -5,20 +5,6 @@ static inline void nrLDPC_bnProcPc_BG1_R23_AVX2(int8_t* bnProcBuf,int8_t* llrRes
__m256i* p_llrRes;
uint32_t M ;
// Process group with 1 CNs
- M = (9*Z + 31)>>5;
- p_bnProcBuf = (__m128i*) &bnProcBuf [0];
- p_llrProcBuf = (__m128i*) &llrProcBuf [0];
- p_llrRes = (__m256i*) &llrRes [0];
- for (int i=0,j=0;i<M;i++,j+=2) {
- ymm0 = _mm256_cvtepi8_epi16(p_bnProcBuf [j]);
- ymm1 = _mm256_cvtepi8_epi16(p_bnProcBuf[j + 1]);
- ymm0 = _mm256_cvtepi8_epi16(p_llrProcBuf[j]);
- ymmRes0 = _mm256_adds_epi16(ymmRes0, ymm0);
- ymm1 = _mm256_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
- ymmRes1 = _mm256_adds_epi16(ymmRes1, ymm1);
- ymm0 = _mm256_packs_epi16(ymmRes0, ymmRes1);
- p_llrRes[i] = _mm256_permute4x64_epi64(ymm0, 0xD8);
-}
// Process group with 2 CNs
M = (1*Z + 31)>>5;
p_bnProcBuf = (__m128i*) &bnProcBuf [3456];
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R89_AVX2.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R89_AVX2.h
index 8c7fa43f5d51237654926b1f51cc3b66b907533c..aee271db84d036d34c317d326ede442d84b51056 100644
--- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R89_AVX2.h
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bnProcPc/nrLDPC_bnProcPc_BG1_R89_AVX2.h
@@ -5,20 +5,6 @@ static inline void nrLDPC_bnProcPc_BG1_R89_AVX2(int8_t* bnProcBuf,int8_t* llrRes
__m256i* p_llrRes;
uint32_t M ;
// Process group with 1 CNs
- M = (1*Z + 31)>>5;
- p_bnProcBuf = (__m128i*) &bnProcBuf [0];
- p_llrProcBuf = (__m128i*) &llrProcBuf [0];
- p_llrRes = (__m256i*) &llrRes [0];
- for (int i=0,j=0;i<M;i++,j+=2) {
- ymm0 = _mm256_cvtepi8_epi16(p_bnProcBuf [j]);
- ymm1 = _mm256_cvtepi8_epi16(p_bnProcBuf[j + 1]);
- ymm0 = _mm256_cvtepi8_epi16(p_llrProcBuf[j]);
- ymmRes0 = _mm256_adds_epi16(ymmRes0, ymm0);
- ymm1 = _mm256_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
- ymmRes1 = _mm256_adds_epi16(ymmRes1, ymm1);
- ymm0 = _mm256_packs_epi16(ymmRes0, ymmRes1);
- p_llrRes[i] = _mm256_permute4x64_epi64(ymm0, 0xD8);
-}
// Process group with 2 CNs
M = (3*Z + 31)>>5;
p_bnProcBuf = (__m128i*) &bnProcBuf [384];
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R13_AVX512.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R13_AVX512.h
new file mode 100644
index 0000000000000000000000000000000000000000..1db81b7a061e6c15bcbe6d842f7765a708aa16b3
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R13_AVX512.h
@@ -0,0 +1,743 @@
+static inline void nrLDPC_bnProcPc_BG1_R13_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {
+ __m512i zmm0, zmm1, zmmRes0, zmmRes1;
+ __m256i* p_bnProcBuf;
+ __m256i* p_llrProcBuf;
+ __m512i* p_llrRes;
+ uint32_t M ;
+// Process group with 1 CNs
+ M = (42*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [0];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [0];
+ p_llrRes = (__m512i*) &llrRes [0];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 2 CNs
+// Process group with 3 CNs
+// Process group with 4 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [16128];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [16128];
+ p_llrRes = (__m512i*) &llrRes [16128];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 5 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [17664];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [16512];
+ p_llrRes = (__m512i*) &llrRes [16512];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 6 CNs
+ M = (2*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [19584];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [16896];
+ p_llrRes = (__m512i*) &llrRes [16896];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 7 CNs
+ M = (4*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [24192];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [17664];
+ p_llrRes = (__m512i*) &llrRes [17664];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 8 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [34944];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [19200];
+ p_llrRes = (__m512i*) &llrRes [19200];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 9 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [44160];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [20352];
+ p_llrRes = (__m512i*) &llrRes [20352];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 10 CNs
+ M = (4*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [47616];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [20736];
+ p_llrRes = (__m512i*) &llrRes [20736];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[384 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[384 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[432 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[432 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 11 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [62976];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [22272];
+ p_llrRes = (__m512i*) &llrRes [22272];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[324 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[324 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[360 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[360 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 12 CNs
+ M = (4*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [75648];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [23424];
+ p_llrRes = (__m512i*) &llrRes [23424];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[384 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[384 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[432 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[432 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[480 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[480 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[528 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[528 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 13 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [94080];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [24960];
+ p_llrRes = (__m512i*) &llrRes [24960];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 14 CNs
+// Process group with 15 CNs
+// Process group with 16 CNs
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [99072];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [25344];
+ p_llrRes = (__m512i*) &llrRes [25344];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[264 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[264 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[276 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[276 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[300 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[300 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[312 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[312 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[324 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[324 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 29 CNs
+// Process group with 30 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [109824];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [25728];
+ p_llrRes = (__m512i*) &llrRes [25728];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[264 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[264 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[276 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[276 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[300 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[300 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[312 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[312 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[324 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[324 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[348 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[348 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R23_AVX512.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R23_AVX512.h
new file mode 100644
index 0000000000000000000000000000000000000000..902da267d9f611fe0a5a830a8047fe34d401a243
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R23_AVX512.h
@@ -0,0 +1,293 @@
+static inline void nrLDPC_bnProcPc_BG1_R23_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {
+ __m512i zmm0, zmm1, zmmRes0, zmmRes1;
+ __m256i* p_bnProcBuf;
+ __m256i* p_llrProcBuf;
+ __m512i* p_llrRes;
+ uint32_t M ;
+// Process group with 1 CNs
+ M = (9*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [0];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [0];
+ p_llrRes = (__m512i*) &llrRes [0];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 2 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [3456];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [3456];
+ p_llrRes = (__m512i*) &llrRes [3456];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 3 CNs
+ M = (5*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [4224];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [3840];
+ p_llrRes = (__m512i*) &llrRes [3840];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 4 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [9984];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [5760];
+ p_llrRes = (__m512i*) &llrRes [5760];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 5 CNs
+ M = (7*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [14592];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [6912];
+ p_llrRes = (__m512i*) &llrRes [6912];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[336 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 6 CNs
+ M = (8*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [28032];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [9600];
+ p_llrRes = (__m512i*) &llrRes [9600];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[288 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[384 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[384 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[480 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[480 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 7 CNs
+// Process group with 8 CNs
+// Process group with 9 CNs
+// Process group with 10 CNs
+// Process group with 11 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [46464];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [12672];
+ p_llrRes = (__m512i*) &llrRes [12672];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 12 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [50688];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [13056];
+ p_llrRes = (__m512i*) &llrRes [13056];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 13 CNs
+// Process group with 14 CNs
+// Process group with 15 CNs
+// Process group with 16 CNs
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R89_AVX512.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R89_AVX512.h
new file mode 100644
index 0000000000000000000000000000000000000000..4731c417df7a969b3f533550589b3314e0d04400
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG1_R89_AVX512.h
@@ -0,0 +1,147 @@
+static inline void nrLDPC_bnProcPc_BG1_R89_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {
+ __m512i zmm0, zmm1, zmmRes0, zmmRes1;
+ __m256i* p_bnProcBuf;
+ __m256i* p_llrProcBuf;
+ __m512i* p_llrRes;
+ uint32_t M ;
+// Process group with 1 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [0];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [0];
+ p_llrRes = (__m512i*) &llrRes [0];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 2 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [384];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [384];
+ p_llrRes = (__m512i*) &llrRes [384];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 3 CNs
+ M = (21*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [2688];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [1536];
+ p_llrRes = (__m512i*) &llrRes [1536];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[504 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[504 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 4 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [26880];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [9600];
+ p_llrRes = (__m512i*) &llrRes [9600];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 5 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [28416];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [9984];
+ p_llrRes = (__m512i*) &llrRes [9984];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0, zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1, zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0, zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 6 CNs
+// Process group with 7 CNs
+// Process group with 8 CNs
+// Process group with 9 CNs
+// Process group with 10 CNs
+// Process group with 11 CNs
+// Process group with 12 CNs
+// Process group with 13 CNs
+// Process group with 14 CNs
+// Process group with 15 CNs
+// Process group with 16 CNs
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R13_AVX512.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R13_AVX512.h
new file mode 100644
index 0000000000000000000000000000000000000000..67fc3eea3d3edb3a7aabbe62e6e42f06340c045c
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R13_AVX512.h
@@ -0,0 +1,441 @@
+static inline void nrLDPC_bnProcPc_BG2_R13_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {
+ __m512i zmm0,zmm1,zmmRes0,zmmRes1;
+ __m256i* p_bnProcBuf;
+ __m256i* p_llrProcBuf;
+ __m512i* p_llrRes;
+ uint32_t M ;
+// Process group with 1 CNs
+ M = (18*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [0];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [0];
+ p_llrRes = (__m512i*) &llrRes [0];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[j+1]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 2 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [6912];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [6912];
+ p_llrRes = (__m512i*) &llrRes [6912];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 3 CNs
+// Process group with 4 CNs
+ M = (2*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [7680];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [7296];
+ p_llrRes = (__m512i*) &llrRes [7296];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 5 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [10752];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [8064];
+ p_llrRes = (__m512i*) &llrRes [8064];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 6 CNs
+ M = (5*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [12672];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [8448];
+ p_llrRes = (__m512i*) &llrRes [8448];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[300 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[300 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 7 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [24192];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [10368];
+ p_llrRes = (__m512i*) &llrRes [10368];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 8 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [26880];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [10752];
+ p_llrRes = (__m512i*) &llrRes [10752];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 9 CNs
+// Process group with 10 CNs
+// Process group with 11 CNs
+// Process group with 12 CNs
+// Process group with 13 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [29952];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [11136];
+ p_llrRes = (__m512i*) &llrRes [11136];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 14 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [34944];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [11520];
+ p_llrRes = (__m512i*) &llrRes [11520];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 15 CNs
+// Process group with 16 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [40320];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [11904];
+ p_llrRes = (__m512i*) &llrRes [11904];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R15_AVX512.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R15_AVX512.h
new file mode 100644
index 0000000000000000000000000000000000000000..b6fce03ca4c4d4fca379c89b0f3eec94ca9d9658
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R15_AVX512.h
@@ -0,0 +1,751 @@
+static inline void nrLDPC_bnProcPc_BG2_R15_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {
+ __m512i zmm0,zmm1,zmmRes0,zmmRes1;
+ __m256i* p_bnProcBuf;
+ __m256i* p_llrProcBuf;
+ __m512i* p_llrRes;
+ uint32_t M ;
+// Process group with 1 CNs
+ M = (38*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [0];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [0];
+ p_llrRes = (__m512i*) &llrRes [0];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[j+1]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 2 CNs
+// Process group with 3 CNs
+// Process group with 4 CNs
+// Process group with 5 CNs
+ M = (2*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [14592];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [14592];
+ p_llrRes = (__m512i*) &llrRes [14592];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 6 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [18432];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [15360];
+ p_llrRes = (__m512i*) &llrRes [15360];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 7 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [20736];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [15744];
+ p_llrRes = (__m512i*) &llrRes [15744];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 8 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [23424];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [16128];
+ p_llrRes = (__m512i*) &llrRes [16128];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 9 CNs
+ M = (2*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [26496];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [16512];
+ p_llrRes = (__m512i*) &llrRes [16512];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 10 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [33408];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [17280];
+ p_llrRes = (__m512i*) &llrRes [17280];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 11 CNs
+// Process group with 12 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [37248];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [17664];
+ p_llrRes = (__m512i*) &llrRes [17664];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 13 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [41856];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [18048];
+ p_llrRes = (__m512i*) &llrRes [18048];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 14 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [46848];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [18432];
+ p_llrRes = (__m512i*) &llrRes [18432];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 15 CNs
+// Process group with 16 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [52224];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [18816];
+ p_llrRes = (__m512i*) &llrRes [18816];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [58368];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [19200];
+ p_llrRes = (__m512i*) &llrRes [19200];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with <23 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [66816];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [19584];
+ p_llrRes = (__m512i*) &llrRes [19584];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[84 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[132 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[144 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[156 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[168 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[180 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[192 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[204 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[216 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[228 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[240 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[252 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[264 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[264 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}
diff --git a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R23_AVX512.h b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R23_AVX512.h
new file mode 100644
index 0000000000000000000000000000000000000000..6330988dd3a1f306137541710367bd604541651b
--- /dev/null
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/ldpc_gen_files/bn_avx512/nrLDPC_bnProcPc_BG2_R23_AVX512.h
@@ -0,0 +1,181 @@
+static inline void nrLDPC_bnProcPc_BG2_R23_AVX512(int8_t* bnProcBuf,int8_t* llrRes , int8_t* llrProcBuf, uint16_t Z ) {
+ __m512i zmm0,zmm1,zmmRes0,zmmRes1;
+ __m256i* p_bnProcBuf;
+ __m256i* p_llrProcBuf;
+ __m512i* p_llrRes;
+ uint32_t M ;
+// Process group with 1 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [0];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [0];
+ p_llrRes = (__m512i*) &llrRes [0];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[j+1]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 2 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [1152];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [1152];
+ p_llrRes = (__m512i*) &llrRes [1152];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[j + 1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 3 CNs
+ M = (5*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [3456];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [2304];
+ p_llrRes = (__m512i*) &llrRes [2304];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[120 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 4 CNs
+ M = (3*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [9216];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [4224];
+ p_llrRes = (__m512i*) &llrRes [4224];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[108 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 5 CNs
+ M = (2*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [13824];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [5376];
+ p_llrRes = (__m512i*) &llrRes [5376];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[72 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[96 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 6 CNs
+ M = (1*Z + 63)>>6;
+ p_bnProcBuf = (__m256i*) &bnProcBuf [17664];
+ p_llrProcBuf = (__m256i*) &llrProcBuf [6144];
+ p_llrRes = (__m512i*) &llrRes [6144];
+ for (int i=0,j=0;i<M;i++,j+=2) {
+ zmmRes0 = _mm512_cvtepi8_epi16(p_bnProcBuf [j]);
+ zmmRes1 = _mm512_cvtepi8_epi16(p_bnProcBuf [j +1]);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[12 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[24 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[36 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[48 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_bnProcBuf[60 + j +1]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_cvtepi8_epi16(p_llrProcBuf[j]);
+ zmmRes0 = _mm512_adds_epi16(zmmRes0,zmm0);
+ zmm1 = _mm512_cvtepi8_epi16(p_llrProcBuf[j +1 ]);
+ zmmRes1 = _mm512_adds_epi16(zmmRes1,zmm1);
+ zmm0 = _mm512_packs_epi16(zmmRes0,zmmRes1);
+ p_llrRes[i] = _mm512_permutex_epi64(zmm0, 0xD8);
+}
+// Process group with 7 CNs
+// Process group with 8 CNs
+// Process group with 9 CNs
+// Process group with 10 CNs
+// Process group with 11 CNs
+// Process group with 12 CNs
+// Process group with 13 CNs
+// Process group with 14 CNs
+// Process group with 15 CNs
+// Process group with 16 CNs
+// Process group with 17 CNs
+// Process group with 18 CNs
+// Process group with 19 CNs
+// Process group with 20 CNs
+// Process group with 21 CNs
+// Process group with 22 CNs
+// Process group with <23 CNs
+// Process group with 24 CNs
+// Process group with 25 CNs
+// Process group with 26 CNs
+// Process group with 27 CNs
+// Process group with 28 CNs
+// Process group with 29 CNs
+// Process group with 30 CNs
+}