Skip to content

O
OpenXG-RAN
Commit fa3a637b authored by Raymond Knopp's avatar Raymond Knopp
Browse files
Options

cleanup of generator, added directory/file creation for output

parent 550ab500
Show whitespace changes
Inline Side-by-side
Showing with 434 additions and 420 deletions
openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/cnProc_gen.c
View file @ fa3a637b
...@@ -11,7 +11,19 @@ ...@@ -11,7 +11,19 @@
void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
{ {
printf("void nrLDPC_cnProc_BG1_Z%d\n",Z); const char *ratestr[3]={"13","23","89"};
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/nrLDPC_cnProc_BG1_Z%d_%s.c",Z,ratestr[R]);
FILE *fd=fopen(fname,"w");
if (fd == NULL) {printf("Cannot create %s\n");abort();}
fprintf(fd,"void nrLDPC_cnProc_BG1_Z%d_%s(t_nrLDPC_procBuf* p_procBuf) {\n",Z,ratestr[R]);
const uint8_t* lut_numCnInCnGroups; const uint8_t* lut_numCnInCnGroups;
const uint32_t* lut_startAddrCnGroups = lut_startAddrCnGroups_BG1; const uint32_t* lut_startAddrCnGroups = lut_startAddrCnGroups_BG1;
...@@ -19,7 +31,7 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -19,7 +31,7 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
if (R==0) lut_numCnInCnGroups = lut_numCnInCnGroups_BG1_R13; if (R==0) lut_numCnInCnGroups = lut_numCnInCnGroups_BG1_R13;
else if (R==1) lut_numCnInCnGroups = lut_numCnInCnGroups_BG1_R23; else if (R==1) lut_numCnInCnGroups = lut_numCnInCnGroups_BG1_R23;
else if (R==2) lut_numCnInCnGroups = lut_numCnInCnGroups_BG1_R89; else if (R==2) lut_numCnInCnGroups = lut_numCnInCnGroups_BG1_R89;
else { printf("aborting, illegal R %d\n",R); abort();} else { printf("aborting, illegal R %d\n",R); fclose(fd);abort();}
int8_t* cnProcBuf = p_procBuf->cnProcBuf; int8_t* cnProcBuf = p_procBuf->cnProcBuf;
int8_t* cnProcBufRes = p_procBuf->cnProcBufRes; int8_t* cnProcBufRes = p_procBuf->cnProcBufRes;
...@@ -44,7 +56,7 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -44,7 +56,7 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
// LUT with offsets for bits that need to be processed // LUT with offsets for bits that need to be processed
// 1. bit proc requires LLRs of 2. and 3. bit, 2.bits of 1. and 3. etc. // 1. bit proc requires LLRs of 2. and 3. bit, 2.bits of 1. and 3. etc.
// Offsets are in units of bitOffsetInGroup (1*384/32) // Offsets are in units of bitOffsetInGroup (1*384/32)
// const uint8_t lut_idxCnProcG3[3][2] = {{12,24}, {0,24}, {0,12}}; // const uint8_t lut_idxCnProcG3[3][2] = {{12,24}, {0,24}, {0,12}};
// ===================================================================== // =====================================================================
// Process group with 3 BNs // Process group with 3 BNs
...@@ -54,9 +66,9 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -54,9 +66,9 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
// Offsets are in units of bitOffsetInGroup (1*384/32) // Offsets are in units of bitOffsetInGroup (1*384/32)
const uint8_t lut_idxCnProcG3[3][2] = {{12,24}, {0,24}, {0,12}}; const uint8_t lut_idxCnProcG3[3][2] = {{12,24}, {0,24}, {0,12}};
printf(" __m256i ymm0, min, sgn;\n"); fprintf(fd," __m256i ymm0, min, sgn;\n");
printf(" const __m256i* p_ones = (__m256i*) ones256_epi8;\n"); fprintf(fd," const __m256i* p_ones = (__m256i*) ones256_epi8;\n");
printf(" const __m256i* p_maxLLR = (__m256i*) maxLLR256_epi8;\n"); fprintf(fd," const __m256i* p_maxLLR = (__m256i*) maxLLR256_epi8;\n");
if (lut_numCnInCnGroups[0] > 0) if (lut_numCnInCnGroups[0] > 0)
{ {
...@@ -83,29 +95,29 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -83,29 +95,29 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[0]+lut_idxCnProcG3[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[0]+lut_idxCnProcG3[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// 32 CNs of second BN // 32 CNs of second BN
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][1] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][1] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[0]+lut_idxCnProcG3[j][1] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[0]+lut_idxCnProcG3[j][1] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcB ufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[0]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcB ufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[0]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -141,32 +153,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -141,32 +153,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[1]+lut_idxCnProcG4[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[1]+lut_idxCnProcG4[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<3; k++) for (k=1; k<3; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[1]+lut_idxCnProcG4[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[1]+lut_idxCnProcG4[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[1]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[1]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -205,32 +217,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -205,32 +217,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[2]+lut_idxCnProcG5[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[2]+lut_idxCnProcG5[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<4; k++) for (k=1; k<4; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[2]+lut_idxCnProcG5[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[2]+lut_idxCnProcG5[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[2]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[2]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -269,32 +281,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -269,32 +281,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[3]+lut_idxCnProcG6[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[3]+lut_idxCnProcG6[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<5; k++) for (k=1; k<5; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[3]+lut_idxCnProcG6[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[3]+lut_idxCnProcG6[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[3]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[3]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -336,32 +348,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -336,32 +348,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[4]+lut_idxCnProcG7[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[4]+lut_idxCnProcG7[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<6; k++) for (k=1; k<6; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[4]+lut_idxCnProcG7[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[4]+lut_idxCnProcG7[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[4]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[4]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -404,37 +416,37 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -404,37 +416,37 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[5]+lut_idxCnProcG8[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[5]+lut_idxCnProcG8[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<7; k++) for (k=1; k<7; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[5]+lut_idxCnProcG8[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[5]+lut_idxCnProcG8[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[5]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[5]+(j*bitOffsetInGroup)+i);
} }
} }
} }
// ===================================================================== // =====================================================================
// Process group with 9 BNs // Process group with 9 BNs
// Offset is 2*384/32 = 24 // Offset is 2*384/32 = 24
...@@ -473,32 +485,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -473,32 +485,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[6]+lut_idxCnProcG9[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[6]+lut_idxCnProcG9[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<8; k++) for (k=1; k<8; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[6]+lut_idxCnProcG9[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[6]+lut_idxCnProcG9[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[6]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[6]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -542,32 +554,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -542,32 +554,32 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[7]+lut_idxCnProcG10[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[7]+lut_idxCnProcG10[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<9; k++) for (k=1; k<9; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[7]+lut_idxCnProcG10[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[7]+lut_idxCnProcG10[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[7]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[7]+(j*bitOffsetInGroup)+i);
} }
} }
} }
...@@ -613,34 +625,36 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R) ...@@ -613,34 +625,36 @@ void nrLDPC_cnProc_BG1_generator(t_nrLDPC_procBuf* p_procBuf, uint16_t Z,int R)
for (i=0; i<M; i++,iprime++) for (i=0; i<M; i++,iprime++)
{ {
// Abs and sign of 32 CNs (first BN) // Abs and sign of 32 CNs (first BN)
// ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i]; // ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[8]+lut_idxCnProcG19[j][0] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[8]+lut_idxCnProcG19[j][0] + i);
// sgn = _mm256_sign_epi8(*p_ones, ymm0); // sgn = _mm256_sign_epi8(*p_ones, ymm0);
printf(" sgn = _mm256_sign_epi8(*p_ones, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(*p_ones, ymm0);\n");
// min = _mm256_abs_epi8(ymm0); // min = _mm256_abs_epi8(ymm0);
printf(" min = _mm256_abs_epi8(ymm0);\n"); fprintf(fd," min = _mm256_abs_epi8(ymm0);\n");
// Loop over BNs // Loop over BNs
for (k=1; k<18; k++) for (k=1; k<18; k++)
{ {
printf(" ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[8]+lut_idxCnProcG19[j][k] + i); fprintf(fd," ymm0 = ((__m256i*)&cnProcBuf)[%d];\n",lut_startAddrCnGroups[8]+lut_idxCnProcG19[j][k] + i);
// min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0)); // min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));
printf(" min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n"); fprintf(fd," min = _mm256_min_epu8(min, _mm256_abs_epi8(ymm0));\n");
// sgn = _mm256_sign_epi8(sgn, ymm0); // sgn = _mm256_sign_epi8(sgn, ymm0);
printf(" sgn = _mm256_sign_epi8(sgn, ymm0);\n"); fprintf(fd," sgn = _mm256_sign_epi8(sgn, ymm0);\n");
} }
// Store result // Store result
// min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127 // min = _mm256_min_epu8(min, *p_maxLLR); // 128 in epi8 is -127
printf(" min = _mm256_min_epu8(min, *p_maxLLR);\n"); fprintf(fd," min = _mm256_min_epu8(min, *p_maxLLR);\n");
// *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn); // *p_cnProcBufResBit = _mm256_sign_epi8(min, sgn);
// p_cnProcBufResBit++; // p_cnProcBufResBit++;
printf(" ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[8]+(j*bitOffsetInGroup)+i); fprintf(fd," ((__m256i*)cnProcBufRestBit)[%d] = _mm256_sign_epi8(min, sgn);\n",lut_numCnInCnGroups[8]+(j*bitOffsetInGroup)+i);
} }
} }
} }
fprintf(fd,"}\n");
fclose(fd);
}//end of the function nrLDPC_cnProc_BG1 }//end of the function nrLDPC_cnProc_BG1
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment