removed buffer read overflow (harmless but flagged by ASAN)

Merge remote-tracking branch 'origin/ldpc_enc_avx512' into rk_integration1224
aarch64 testing and corrections
2026-07-16 14:10:28 +00:00 · 2025-01-31 21:37:52 +01:00 · 2025-01-30 20:41:37 +01:00 · 2025-01-30 09:27:26 +00:00 · 2025-01-30 06:21:40 +01:00 · 2025-01-29 22:32:03 +01:00
23 changed files with 3034 additions and 772 deletions
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -146,9 +146,25 @@ elseif(CMAKE_SYSTEM_PROCESSOR STREQUAL "x86_64")
    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -DSIMDE_X86_SSSE3_NATIVE")
  endif()
 elseif (CMAKE_SYSTEM_PROCESSOR STREQUAL "armv7l")
-  set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mfloat-abi=hard -mfpu=neon -lgcc -lrt")
+    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mfloat-abi=hard -mfpu=neon -lgcc -lrt")
 elseif (CMAKE_SYSTEM_PROCESSOR STREQUAL "aarch64")
-  set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -lgcc -lrt")
+  execute_process(COMMAND
+	          "lscpu"
+                  OUTPUT_VARIABLE
+                  CMAKE_NEOVERSE_INFO
+                  OUTPUT_STRIP_TRAILING_WHITESPACE
+                 )  
+  if (CMAKE_NEOVERSE_INFO MATCHES "Neoverse-V1")	
+    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mcpu=neoverse-v1 -lgcc -lrt")
+  elseif (CMAKE_NEOVERSE_INFO MATCHES "Neoverse-V2")
+    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mcpu=neoverse-v2 -ftree-vectorize -lgcc -lrt")
+  elseif (CMAKE_NEOVERSE_INFO MATCHES "Neoverse-N1")
+    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mcpu=neoverse-n1  -lgcc -lrt")
+  elseif (CMAKE_NEOVERSE_INFO MATCHES "Neoverse-N2")
+    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mcpu=neoverse-n2 -ftree-vectorize -lgcc -lrt")
+  else ()
+    set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -gdwarf-2 -mcpu=native -lgcc -lrt")
+  endif()
 else()
  message(FATAL_ERROR "compile for CPU architecture ${CPUARCH}, CMAKE_SYSTEM_PROCESSOR ${CMAKE_SYSTEM_PROCESSOR}")
 endif()
--- a/openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder.c
+++ b/openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nrLDPC_coding_segment_encoder.c
@@ -66,98 +66,442 @@ static void ldpc8blocks_coding_segment(void *p)

  // nrLDPC_encoder output is in "d"
  // let's make this interface happy!
-  uint8_t tmp[8][68 * 384] __attribute__((aligned(32)));
-  uint8_t *d[impp->n_segments];
-  for (int rr = impp->macro_num * 8, i = 0; rr < impp->n_segments && rr < (impp->macro_num + 1) * 8; rr++, i++)
-    d[rr] = tmp[i];
-
+  uint8_t d[68 * 384]__attribute__((aligned(64)));
+  uint8_t *dp[2];
+  dp[0]=&d[0];
  uint8_t *c[nrLDPC_TB_encoding_parameters->C];
+  unsigned int macro_segment, macro_segment_end;
+
+  
+  macro_segment = 8*impp->macro_num;
+  macro_segment_end = (impp->n_segments > 8*(impp->macro_num+1)) ? 8*(impp->macro_num+1) : impp->n_segments;
  for (int r = 0; r < nrLDPC_TB_encoding_parameters->C; r++)
-    c[r] = nrLDPC_TB_encoding_parameters->segments[r].c;
-  start_meas(&nrLDPC_TB_encoding_parameters->segments[impp->macro_num * 8].ts_ldpc_encode);
-  ldpc_interface_segment.LDPCencoder(c, d, impp);
-  stop_meas(&nrLDPC_TB_encoding_parameters->segments[impp->macro_num * 8].ts_ldpc_encode);
+    c[r]=nrLDPC_TB_encoding_parameters->segments[r].c;
+  start_meas(&nrLDPC_TB_encoding_parameters->segments[impp->macro_num*8].ts_ldpc_encode);
+  ldpc_interface_segment.LDPCencoder(c, dp, impp);
+  stop_meas(&nrLDPC_TB_encoding_parameters->segments[impp->macro_num*8].ts_ldpc_encode);
  // Compute where to place in output buffer that is concatenation of all segments

-  uint32_t r_offset = 0;
-  for (int i = 0; i < impp->macro_num * 8; i++)
-    r_offset += nrLDPC_TB_encoding_parameters->segments[i].E;
-  for (int rr = impp->macro_num * 8; rr < impp->n_segments && rr < (impp->macro_num + 1) * 8; rr++) {
-    if (impp->F > 0) {
-      // writing into positions d[r][k-2Zc] as in clause 5.3.2 step 2) in 38.212
-      memset(&d[rr][impp->K - impp->F - 2 * impp->Zc], NR_NULL, impp->F);
-    }
+
+  if (impp->F>0) {
+      // writing into positions d[k-2Zc] as in clause 5.3.2 step 2) in 38.212
+      memset(&d[impp->K - impp->F - 2 * impp->Zc], NR_NULL, impp->F);
+  }

 #ifdef DEBUG_LDPC_ENCODING
-    LOG_D(PHY, "rvidx in encoding = %d\n", rel15->rvIndex[0]);
+  LOG_D(PHY,"rvidx in encoding = %d\n", nrLDPC_TB_encoding_parameters->rv_index);
 #endif
-    uint32_t E = nrLDPC_TB_encoding_parameters->segments[rr].E;
-    LOG_D(NR_PHY,
-          "Rate Matching, Code segment %d/%d (coded bits (G) %u, E %d, Filler bits %d, Filler offset %d mod_order %d, nb_rb "
+  const uint32_t E = nrLDPC_TB_encoding_parameters->segments[macro_segment].E;
+  uint32_t E2=E,E2_first_segment=macro_segment_end-macro_segment;
+  bool Eshift=false;
+  for (int s=macro_segment;s<macro_segment_end;s++)
+      if (nrLDPC_TB_encoding_parameters->segments[s].E != E) {
+	 E2=nrLDPC_TB_encoding_parameters->segments[s].E;
+         Eshift=true;
+	 E2_first_segment = s-macro_segment;
+         break;
+      }	 
+    
+
+  LOG_D(NR_PHY,
+        "Rate Matching, Code segment %d...%d/%d (coded bits (G) %u, E %d, E2 %d Filler bits %d, Filler offset %d mod_order %d, nb_rb "
          "%d,nrOfLayer %d)...\n",
-          rr,
-          impp->n_segments,
-          G,
-          E,
-          impp->F,
-          impp->K - impp->F - 2 * impp->Zc,
-          mod_order,
-          nb_rb,
-          nrLDPC_TB_encoding_parameters->nb_layers);
+        macro_segment,
+        macro_segment_end,
+        impp->n_segments,
+        G,
+        E,E2,
+        impp->F,
+        impp->K - impp->F - 2 * impp->Zc,
+        mod_order,
+        nb_rb,
+        nrLDPC_TB_encoding_parameters->nb_layers);
+/*
+  printf("Rate Matching, Code segment %d...%d/%d (coded bits (G) %u, E %d, E2 %d Filler bits %d, Filler offset %d mod_order %d, nb_rb "
+          "%d,nrOfLayer %d)...\n",
+        macro_segment,
+        macro_segment_end,
+        impp->n_segments,
+        G,
+        E,E2,
+        impp->F,
+        impp->K - impp->F - 2 * impp->Zc,
+        mod_order,
+        nb_rb,
+        nrLDPC_TB_encoding_parameters->nb_layers);
+*/
+  uint32_t Tbslbrm = nrLDPC_TB_encoding_parameters->tbslbrm;

-    uint32_t Tbslbrm = nrLDPC_TB_encoding_parameters->tbslbrm;
-
-    uint8_t e[E];
-    bzero(e, E);
-    start_meas(&nrLDPC_TB_encoding_parameters->segments[rr].ts_rate_match);
+  uint8_t e[E]__attribute__((aligned(64)));
+  uint8_t f[E+64]__attribute__((aligned(64)));
+  uint8_t e2[E2]__attribute__((aligned(64)));
+  uint8_t f2[E2+64]__attribute__((aligned(64)));
+  bzero (e, E);
+  if (Eshift) bzero (e2,E2);
+  start_meas(&nrLDPC_TB_encoding_parameters->segments[macro_segment].ts_rate_match);
+  nr_rate_matching_ldpc(Tbslbrm,
+                        impp->BG,
+                        impp->Zc,
+                        d,
+                        e,
+                        impp->n_segments,
+                        impp->F,
+                        impp->K - impp->F - 2 * impp->Zc,
+                        nrLDPC_TB_encoding_parameters->rv_index,
+                        E);
+  if (Eshift)
    nr_rate_matching_ldpc(Tbslbrm,
                          impp->BG,
                          impp->Zc,
-                          d[rr],
-                          e,
+                          d,
+                          e2,
                          impp->n_segments,
                          impp->F,
                          impp->K - impp->F - 2 * impp->Zc,
                          nrLDPC_TB_encoding_parameters->rv_index,
-                          E);
-    stop_meas(&nrLDPC_TB_encoding_parameters->segments[rr].ts_rate_match);
-    if (impp->K - impp->F - 2 * impp->Zc > E) {
-      LOG_E(PHY, "dlsch coding A %d K %d G %d (nb_rb %d, mod_order %d)\n", A, impp->K, G, nb_rb, (int)mod_order);
+                          E2);

-      LOG_E(NR_PHY,
-            "Rate Matching, Code segment %d/%d (coded bits (G) %u, E %d, K %d, Filler bits %d, Filler offset %d mod_order %d, "
-            "nb_rb %d)...\n",
-            rr,
-            impp->n_segments,
-            G,
-            E,
-            impp->K,
-            impp->F,
-            impp->K - impp->F - 2 * impp->Zc,
-            mod_order,
-            nb_rb);
-    }
-#ifdef DEBUG_LDPC_ENCODING
-
-    for (int i = 0; i < 16; i++)
-      printf("output ratematching e[%d]= %d r_offset %u\n", i, e[i], r_offset);
+  stop_meas(&nrLDPC_TB_encoding_parameters->segments[macro_segment].ts_rate_match);
+  if (impp->K - impp->F - 2 * impp->Zc > E) {
+    LOG_E(PHY,
+          "dlsch coding A %d  Kr %d G %d (nb_rb %d, mod_order %d)\n",
+          A,
+          impp->K,
+          G,
+          nb_rb,
+          (int)mod_order);

+    LOG_E(NR_PHY,
+          "Rate Matching, Code segments %d...%d/%d (coded bits (G) %u, E %d, Kr %d, Filler bits %d, Filler offset %d mod_order %d, "
+          "nb_rb %d)...\n",
+          macro_segment,
+	  macro_segment_end,
+          impp->n_segments,
+          G,
+          E,
+          impp->K,
+          impp->F,
+          impp->K - impp->F - 2 * impp->Zc,
+          mod_order,
+          nb_rb);
+  }
+  start_meas(&nrLDPC_TB_encoding_parameters->segments[macro_segment].ts_interleave);
+  nr_interleaving_ldpc(E,
+                       mod_order,
+                       e,
+                       f);
+  if (Eshift)
+    nr_interleaving_ldpc(E2,
+                         mod_order,
+                         e2,
+                         f2);
+  stop_meas(&nrLDPC_TB_encoding_parameters->segments[macro_segment].ts_interleave);
+  if(impp->toutput != NULL) start_meas(impp->toutput);
+/*
+  for (int i=0;i<16;i++)
+     for (int s=0;s<macro_segment_end-macro_segment;s++) 
+        printf("i %d: segment %d : f[%d] %d\n",i,macro_segment+s,i,(f[i]>>s)&1);
+*/
+  // information part and puncture columns
+ 
+  uint32_t Eoffset=0;
+  for (int s=0; s<macro_segment; s++)
+    Eoffset += (nrLDPC_TB_encoding_parameters->segments[s].E); 
+#ifdef __AVX512F__
+  uint64_t *output_p = (uint64_t*)impp->output;
+  int i=0;
+  __m512i bitperm[8];
+  uint32_t Eoffset2[8];
+  uint32_t Eoffset2_bit[8];
+  bitperm[0] = _mm512_set1_epi64(0x3830282018100800);
+  __m512i inc   = _mm512_set1_epi8(0x1);
+  Eoffset2[0] = Eoffset>>6;
+  Eoffset2_bit[0] = Eoffset&63;
+  //printf("E2_first_segment %d,Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2_first_segment,Eoffset,Eoffset2[0],Eoffset2_bit[0]);
+  for (int n=1;n<E2_first_segment;n++) {
+    bitperm[n]   = _mm512_add_epi8(bitperm[n-1],inc);
+    Eoffset += E;
+    Eoffset2[n]  = Eoffset>>6;
+    Eoffset2_bit[n] = Eoffset&63;
+    //printf("E %d : n %d, Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E,n,Eoffset,Eoffset2[n],Eoffset2_bit[n]);
+  } 
+  if (Eshift) {
+    bitperm[E2_first_segment]   = _mm512_add_epi8(bitperm[E2_first_segment-1],inc);
+    Eoffset += E;
+    Eoffset2[E2_first_segment] = Eoffset>>6;
+    Eoffset2_bit[E2_first_segment] = Eoffset &63;
+    //printf("E2 %d : Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2,Eoffset,Eoffset2[E2_first_segment],Eoffset2_bit[E2_first_segment]);
+  }
+  for (int n=E2_first_segment+1;n<macro_segment_end-macro_segment;n++) {
+    bitperm[n]   = _mm512_add_epi8(bitperm[n-1],inc);
+    Eoffset+=E2;
+    Eoffset2[n] = Eoffset>>6;
+    Eoffset2_bit[n] = Eoffset&63;
+    //printf("E2 %d (macro_segment_end %d, macro_segment %d) : n %d, Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2,macro_segment_end,macro_segment,n,Eoffset,Eoffset2[n],Eoffset2_bit[n]);
+  }
+  int i2=0;
+  __m64 tmp;
+  for (i=0;i<E2;i+=64,i2++) {
+     if (i<E) {
+      for (int j=0; j < E2_first_segment; j++) {
+#ifdef DEBUG_BIT_INTERLEAVE
+        printf("segment %d : qword %d, first bit %d last bit %d\n",j, 
+            Eoffset2[j]+(i>>6),
+            (Eoffset2_bit[j] + i)&63,
+            (i<=(E-64)) ? 63 : (Eoffset2_bit[j] + E-1)&63);
+        if  (Eoffset2_bit[j] > 0) 
+          printf("segment %d : qword %d, first bit %d last bit %d\n",j,
+                 1+Eoffset2[j]+(i>>6),
+                 0,
+                 ((Eoffset2_bit[j] + i)&63) - 1);
+#endif 
+        // Note: Here and below for AVX2, we are using the 64-bit SIMD instruction
+        // instead of C >>/<< because when the Eoffset2_bit is 64 or 0, the <<
+        // and >> operations are undefined and in fact don't give "0" which is
+        // what we want here. The SIMD version do give 0 when the shift is 64
+        tmp = (__m64)_mm512_bitshuffle_epi64_mask(((__m512i *)f)[i2],bitperm[j]);
+        *(__m64*)(output_p + Eoffset2[j])   = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]),_mm_slli_si64(tmp,Eoffset2_bit[j]));
+        *(__m64*)(output_p + Eoffset2[j]+1) = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]+1),_mm_srli_si64(tmp,(64-Eoffset2_bit[j])));
+#ifdef DEBUG_BIT_INTERLEAVE
+       if (j<=1) { 
+           printf("pos : %d, i2 %d, j %d\n",(int)(output_p + Eoffset2[j]-(uint64_t*)impp->output),i2,j);
+           printf("i : %x.%x.%x.%x.%x.%x.%x.%x\n",
+           ((uint8_t*)(output_p + Eoffset2[j]))[0],
+           ((uint8_t*)(output_p + Eoffset2[j]))[1],
+           ((uint8_t*)(output_p + Eoffset2[j]))[2],
+           ((uint8_t*)(output_p + Eoffset2[j]))[3],
+           ((uint8_t*)(output_p + Eoffset2[j]))[4],
+           ((uint8_t*)(output_p + Eoffset2[j]))[5],
+           ((uint8_t*)(output_p + Eoffset2[j]))[6],
+           ((uint8_t*)(output_p + Eoffset2[j]))[7]);
+           printf("i+1 : %x.%x.%x.%x.%x.%x.%x.%x\n",
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[0],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[1],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[2],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[3],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[4],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[5],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[6],
+           ((uint8_t*)(output_p + Eoffset2[j]+1))[7]);
+           printf("%x.%x.%x.%x.%x.%x.%x.%x\n",
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),0)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),1)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),2)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),3)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),4)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),5)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),6)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),7)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),8)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),9)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),10)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),11)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),12)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),13)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),14)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),15)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),16)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),17)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),18)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),19)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),20)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),21)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),22)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),23)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),24)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),25)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),26)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),27)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),28)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),29)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),30)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],0),31)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),0)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),1)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),2)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),3)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),4)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),5)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),6)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),7)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),8)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),9)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),10)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),11)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),12)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),13)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),14)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),15)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),16)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),17)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),18)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),19)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),20)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),21)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),22)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),23)<<(7-j))&128),
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),24)<<(0-j))&1)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),25)<<(1-j))&2)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),26)<<(2-j))&4)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),27)<<(3-j))&8)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),28)<<(4-j))&16)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),29)<<(5-j))&32)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),30)<<(6-j))&64)+
+               ((_mm256_extract_epi8(_mm512_extracti32x8_epi32(((__m512i *)f)[i2],1),31)<<(7-j))&128));
+       }
 #endif
-    start_meas(&nrLDPC_TB_encoding_parameters->segments[rr].ts_interleave);
-    nr_interleaving_ldpc(E, mod_order, e, impp->output + r_offset);
-    stop_meas(&nrLDPC_TB_encoding_parameters->segments[rr].ts_interleave);
-#ifdef DEBUG_LDPC_ENCODING
-
-    for (int i = 0; i < 16; i++)
-      printf("output interleaving f[%d]= %d r_offset %u\n", i, impp->output[i + r_offset], r_offset);
-
-    if (r == impp->n_segments - 1)
-      write_output("enc_output.m", "enc", impp->output, G, 1, 4);
-
+      } //for (int j=0;
+     } // if (i < E)
+     for (int j=E2_first_segment; j < macro_segment_end-macro_segment; j++) {
+#ifdef DEBUG_BIT_INTERLEAVE
+       printf("segment %d : qword %d, first bit %d last bit %d\n",j,
+           Eoffset2[j]+(i>>6),
+           (Eoffset2_bit[j] + i)&63,
+           (i<=(E-64)) ? 63 : (Eoffset2_bit[j] + E-1)&63);
+       if (Eoffset2_bit[j] > 0) 
+         printf("segment %d : qword %d, first bit %d last bit %d\n",j,
+                1+Eoffset2[j]+(i>>6),
+                0,
+                ((Eoffset2_bit[j] + i)&63) - 1);
 #endif
-    r_offset += E;
+       tmp = (__m64)_mm512_bitshuffle_epi64_mask(((__m512i *)f2)[i2],bitperm[j]);
+       *(__m64*)(output_p + Eoffset2[j])   = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]),
+                                                          _mm_slli_si64(tmp,Eoffset2_bit[j]));
+       *(__m64*)(output_p + Eoffset2[j]+1) = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]+1),_mm_srli_si64(tmp,(64-Eoffset2_bit[j])));
+     }
+     output_p++;
  }

+#elif defined(__aarch64__)
+  uint16_t *output_p = (uint16_t*)impp->output;
+  int i=0;
+  uint32_t Eoffset2[8];
+  uint32_t Eoffset2_bit[8];
+  Eoffset2[0] = Eoffset>>4;
+  Eoffset2_bit[0] = Eoffset&15;
+  //printf("E2_first_segment %d,Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2_first_segment,Eoffset,Eoffset2[0],Eoffset2_bit[0]);
+  for (int n=1;n<E2_first_segment;n++) {
+    Eoffset += E;
+    Eoffset2[n]  = Eoffset>>4;
+    Eoffset2_bit[n] = Eoffset&15;
+    //printf("E %d : n %d, Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E,n,Eoffset,Eoffset2[n],Eoffset2_bit[n]);
+  } 
+  if (Eshift) {
+    Eoffset += E;
+    Eoffset2[E2_first_segment] = Eoffset>>4;
+    Eoffset2_bit[E2_first_segment] = Eoffset &15;
+    //printf("E2 %d : Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2,Eoffset,Eoffset2[E2_first_segment],Eoffset2_bit[E2_first_segment]);
+  }
+  for (int n=E2_first_segment+1;n<macro_segment_end-macro_segment;n++) {
+    Eoffset+=E2;
+    Eoffset2[n] = Eoffset>>4;
+    Eoffset2_bit[n] = Eoffset&15;
+    //printf("E2 %d (macro_segment_end %d, macro_segment %d) : n %d, Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2,macro_segment_end,macro_segment,n,Eoffset,Eoffset2[n],Eoffset2_bit[n]);
+  }
+  int i2=0;
+  const int8_t __attribute__ ((aligned (16))) ucShift[8][16] = {
+    {0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7},     // segment 0
+    {-1,0,1,2,3,4,5,6,-1,0,1,2,3,4,5,6},   // segment 1
+    {-2,-1,0,1,2,3,4,5,-2,-1,0,1,2,3,4,5}, // segment 2
+    {-3,-2,-1,0,1,2,3,4,-3,-2,-1,0,1,2,3,4}, // segment 3
+    {-4,-3,-2,-1,0,1,2,3,-4,-3,-2,-1,0,1,2,3}, // segment 4
+    {-5,-4,-3,-2,-1,0,1,2,-5,-4,-3,-2,-1,0,1,2}, // segment 5
+    {-6,-5,-4,-3,-2,-1,0,1,-6,-5,-4,-3,-2,-1,0,1}, // segment 6
+    {-7,-6,-5,-4,-3,-2,-1,0,-7,-6,-5,-4,-3,-2,-1,0}}; // segment 7
+  const uint8_t __attribute__ ((aligned (16))) masks[16] = 
+      {0x1,0x2,0x4,0x8,0x10,0x20,0x40,0x80,0x1,0x2,0x4,0x8,0x10,0x20,0x40,0x80};
+  int8x16_t vshift[8];
+  for (int n=0;n<8;n++) vshift[n] = vld1q_s8(ucShift[n]);
+  uint8x16_t vmask  = vld1q_u8(masks);
+
+  int32_t tmp;
+  uint8x16_t cshift;
+  for (i=0;i<E2;i+=16,i2++) {
+     if (i<E) {	
+       for (int j=0; j < E2_first_segment; j++) {
+         cshift = vandq_u8(vshlq_u8(((uint8x16_t*)f)[i2],vshift[j]),vmask);
+         tmp = (int)vaddv_u8(vget_low_u8(cshift));
+         tmp += (int)(vaddv_u8(vget_high_u8(cshift))<<8);
+         *(output_p + Eoffset2[j])   |= (uint16_t)(tmp<<Eoffset2_bit[j]);
+         *(output_p + Eoffset2[j]+1) |= (uint16_t)(tmp>>(16-Eoffset2_bit[j]));
+       }
+     }
+     for (int j=E2_first_segment; j < macro_segment_end-macro_segment; j++) {
+       cshift = vandq_u8(vshlq_u8(((uint8x16_t*)f2)[i2],vshift[j]),vmask);
+       tmp = (int)vaddv_u8(vget_low_u8(cshift));
+       tmp += (int)(vaddv_u8(vget_high_u8(cshift))<<8);
+       *(output_p + Eoffset2[j])   |= (uint16_t)(tmp<<Eoffset2_bit[j]);
+       *(output_p + Eoffset2[j]+1) |= (uint16_t)(tmp>>(16-Eoffset2_bit[j]));
+     }
+     output_p++;
+  }
+       
+#else
+  uint32_t *output_p = (uint32_t*)impp->output;
+  int i=0;
+  uint32_t Eoffset2[8];
+  uint32_t Eoffset2_bit[8];
+  Eoffset2[0] = Eoffset>>5;
+  Eoffset2_bit[0] = Eoffset&31;
+  //printf("E2_first_segment %d,Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2_first_segment,Eoffset,Eoffset2[0],Eoffset2_bit[0]);
+  for (int n=1;n<E2_first_segment;n++) {
+    Eoffset += E;
+    Eoffset2[n]  = Eoffset>>5;
+    Eoffset2_bit[n] = Eoffset&31;
+    //printf("E %d : n %d, Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E,n,Eoffset,Eoffset2[n],Eoffset2_bit[n]);
+  } 
+  if (Eshift) {
+    Eoffset += E;
+    Eoffset2[E2_first_segment] = Eoffset>>5;
+    Eoffset2_bit[E2_first_segment] = Eoffset &31;
+    //printf("E2 %d : Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2,Eoffset,Eoffset2[E2_first_segment],Eoffset2_bit[E2_first_segment]);
+  }
+  for (int n=E2_first_segment+1;n<macro_segment_end-macro_segment;n++) {
+    Eoffset+=E2;
+    Eoffset2[n] = Eoffset>>5;
+    Eoffset2_bit[n] = Eoffset&31;
+    //printf("E2 %d (macro_segment_end %d, macro_segment %d) : n %d, Eoffset %d, Eoffset2 %d, Eoffset2_bit %d\n",E2,macro_segment_end,macro_segment,n,Eoffset,Eoffset2[n],Eoffset2_bit[n]);
+  }
+  int i2=0;
+  int tmp;
+  __m64 tmp64,tmp64b,tmp64c;
+  
+  for (i=0;i<E2;i+=32,i2++) {
+     if (i<E) {
+       for (int j=0; j < E2_first_segment; j++) {
+         // Note: Here and below, we are using the 64-bit SIMD instruction
+         // instead of C >>/<< because when the Eoffset2_bit is 64 or 0, the <<
+         // and >> operations are undefined and in fact don't give "0" which is
+         // what we want here. The SIMD version do give 0 when the shift is 64
+         tmp = _mm256_movemask_epi8(_mm256_slli_epi16(((__m256i *)f)[i2],7-j));
+         tmp64=_mm_set1_pi32(tmp);
+         tmp64b  = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]),_mm_slli_pi32(tmp64,Eoffset2_bit[j]));
+         tmp64c = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]),_mm_srli_pi32(tmp64,(32-Eoffset2_bit[j])));
+         *(output_p + Eoffset2[j])   = _m_to_int(tmp64b);
+         *(output_p + Eoffset2[j]+1) = _m_to_int(_mm_srli_si64(tmp64c,32));
+       }
+    } 
+    for (int j=E2_first_segment; j < macro_segment_end-macro_segment; j++) {
+       tmp = _mm256_movemask_epi8(_mm256_slli_epi16(((__m256i *)f2)[i2],7-j));
+       tmp64=_mm_set1_pi32(tmp);
+       tmp64b  = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]),_mm_slli_pi32(tmp64,Eoffset2_bit[j]));
+       tmp64c = _mm_or_si64(*(__m64*)(output_p + Eoffset2[j]),_mm_srli_pi32(tmp64,(32-Eoffset2_bit[j])));
+       *(output_p + Eoffset2[j])   = _m_to_int(tmp64b);
+       *(output_p + Eoffset2[j]+1) = _m_to_int(_mm_srli_si64(tmp64c,32));
+    }
+    output_p++;
+  }
+
+  
+#endif
+
+  if(impp->toutput != NULL) stop_meas(impp->toutput);
+/*
+  printf("E %d (mod16 %d) F %d\n",E,E&15,impp->F); 
+  for (int i=0;i<16;i++)
+     for (int s=0;s<macro_segment_end-macro_segment;s++) 
+        printf("i %d: segment %d : output[%d] %d\n",i,macro_segment+s,i,output_offset[i+s*E]);
+*/
  // Task running in // completed
  completed_task_ans(impp->ans);
 }
@@ -180,7 +524,7 @@ static int nrLDPC_prepare_TB_encoding(nrLDPC_slot_encoding_parameters_t *nrLDPC_
  NR_DL_gNB_HARQ_t harq;
  impp.harq = &harq;
  impp.BG = nrLDPC_TB_encoding_parameters->BG;
-  impp.output = nrLDPC_TB_encoding_parameters->segments->output;
+  impp.output = nrLDPC_TB_encoding_parameters->segments->output; 
  impp.K = nrLDPC_TB_encoding_parameters->K;
  impp.F = nrLDPC_TB_encoding_parameters->F;

@@ -205,6 +549,7 @@ static int nrLDPC_prepare_TB_encoding(nrLDPC_slot_encoding_parameters_t *nrLDPC_
 int nrLDPC_coding_encoder(nrLDPC_slot_encoding_parameters_t *nrLDPC_slot_encoding_parameters)
 {
  int nbTasks = 0;
+
  for (int dlsch_id = 0; dlsch_id < nrLDPC_slot_encoding_parameters->nb_TBs; dlsch_id++) {
    nrLDPC_TB_encoding_parameters_t *nrLDPC_TB_encoding_parameters = &nrLDPC_slot_encoding_parameters->TBs[dlsch_id];
    size_t n_seg = (nrLDPC_TB_encoding_parameters->C / 8 + ((nrLDPC_TB_encoding_parameters->C & 7) == 0 ? 0 : 1));
--- a/openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nr_rate_matching.c
+++ b/openair1/PHY/CODING/nrLDPC_coding/nrLDPC_coding_segment/nr_rate_matching.c
@@ -36,43 +36,138 @@ void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f)
  memset(f, 0, E * sizeof(uint8_t));
  uint8_t *e0, *e1, *e2, *e3, *e4, *e5, *e6, *e7;
  uint8_t *fp;
-#if 0 // def __WASAVX2__
-  simde__m256i tmp0,tmp1,tmp2,tmp0b,tmp1b,tmp3,tmp4,tmp5;
-  simde__m256i *e0_256,*e1_256,*e2_256,*e3_256,*e4_256,*e5_256,*e6_256,*e7_256;
+  int j,k;
+#if 1 //def __WASAVX2__
+#ifdef __AVX512VBMI__
+  __m512i tmp0,tmp1,tmp2,tmp3,tmp4,tmp5;
+  __m512i *e0_512,*e1_512,*e2_512,*e3_512,*e4_512,*e5_512,*e6_512,*e7_512;
+  __m512i *f_512=(__m512i *)f;
+#else
+  simde__m128i tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6,tmp7;
+  simde__m128i *e0_128,*e1_128,*e2_128,*e3_128,*e4_128,*e5_128,*e6_128,*e7_128;

-  simde__m256i *f_256=(simde__m256i *)f;
-
-  uint8_t *fp2;
+  simde__m128i *f_128=(simde__m128i *)f;
+#endif
  switch(Qm) {
  case 2:
    e0=e;
    e1=e0+EQm;
-    e0_256=(simde__m256i *)e0;
-    e1_256=(simde__m256i *)e1;
-    for (int k=0,j=0;j<EQm>>5;j++,k+=2) {
-      f_256[k]   = simde_mm256_unpacklo_epi8(e0_256[j],e1_256[j]);
-      f_256[k+1] = simde_mm256_unpackhi_epi8(e0_256[j],e1_256[j]); 
+#ifdef __AVX512VBMI__
+    e0_512=(__m512i *)e0;
+    e1_512=(__m512i *)e1;
+    __m512i p8a = _mm512_set_epi8(95,31,94,30,93,29,92,28,91,27,90,26,89,25,88,24,87,23,86,22,85,21,84,20,83,19,82,18,81,17,80,16,79,15,78,14,77,13,76,12,75,11,74,10,73,9,72,8,71,7,70,6,69,5,68,4,67,3,66,2,65,1,64,0);
+    __m512i p8b = _mm512_set_epi8(127,63,126,62,125,61,124,60,123,59,122,58,121,57,120,56,119,55,118,54,117,53,116,52,115,51,114,50,113,49,112,48,111,47,110,46,109,45,108,44,107,43,106,42,105,41,104,40,103,39,102,38,101,37,100,36,99,35,98,34,97,33,96,32);
+
+    __m512i e0j,e1j;
+    for (k=0,j=0;j<EQm>>6;j++,k+=2) {
+      e0j    = _mm512_loadu_si512(e0_512+j);	    
+      e1j    = _mm512_loadu_si512(e1_512+j);	    
+      _mm512_storeu_si512(f_512+k,_mm512_permutex2var_epi8(e0j,p8a,e1j)); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
+      _mm512_storeu_si512(f_512+k+1,_mm512_permutex2var_epi8(e0j,p8b,e1j)); 
    }
+    if ((j<<6) != EQm) {
+      int k2=k<<6;
+      fp = &f[k2];
+      for (int j2=(j<<6) ; j2< EQm ; j2++) {
+        *fp++ = e0[j2];
+        *fp++ = e1[j2];
+      }
+    }
+#else
+    e0_128=(simde__m128i *)e0;
+    e1_128=(simde__m128i *)e1;
+    simde__m128i e0j,e1j;
+    for (k=0,j=0;j<EQm>>4;j++,k+=2) {
+      e0j=simde_mm_loadu_si128(e0_128+j);
+      e1j=simde_mm_loadu_si128(e1_128+j);
+      simde_mm_storeu_si128(f_128+k,simde_mm_unpacklo_epi8(e0j,e1j));
+      simde_mm_storeu_si128(f_128+k+1,simde_mm_unpackhi_epi8(e0j,e1j)); 
+    }
+    if ((j<<4) != EQm) {
+      int k2=k<<4;
+      fp = &f[k2];
+      for (int j2=(j<<4) ; j2< EQm ; j2++) {
+        *fp++ = e0[j2];
+        *fp++ = e1[j2];
+      }
+    }
+#endif
    break;
  case 4:
    e0=e;
    e1=e0+EQm;
    e2=e1+EQm;
    e3=e2+EQm;
-    e0_256=(simde__m256i *)e0;
-    e1_256=(simde__m256i *)e1;
-    e2_256=(simde__m256i *)e2;
-    e3_256=(simde__m256i *)e3;
-    for (int k=0,j=0;j<EQm>>5;j++,k+=4) {
-      tmp0   = simde_mm256_unpacklo_epi8(e0_256[j],e1_256[j]); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
-      tmp1   = simde_mm256_unpacklo_epi8(e2_256[j],e3_256[j]); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+15) e3(i+15)
-      f_256[k]   = simde_mm256_unpacklo_epi8(tmp0,tmp1);   // e0(i) e1(i) e2(i) e3(i) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
-      f_256[k+1] = simde_mm256_unpackhi_epi8(tmp0,tmp1);   // e0(i+8) e1(i+8) e2(i+8) e3(i+8) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15)
-      tmp0   = simde_mm256_unpackhi_epi8(e0_256[j],e1_256[j]); // e0(i+16) e1(i+16) e0(i+17) e1(i+17) .... e0(i+31) e1(i+31)
-      tmp1   = simde_mm256_unpackhi_epi8(e2_256[j],e3_256[j]); // e2(i+16) e3(i+16) e2(i+17) e3(i+17) .... e2(i+31) e3(i+31)
-      f_256[k+2] = simde_mm256_unpacklo_epi8(tmp0,tmp1);
-      f_256[k+3] = simde_mm256_unpackhi_epi8(tmp0,tmp1); 
+#ifdef __AVX512VBMI__
+    e0_512=(__m512i *)e0;
+    e1_512=(__m512i *)e1;
+    e2_512=(__m512i *)e2;
+    e3_512=(__m512i *)e3;
+    p8a = _mm512_set_epi8(95,31,94,30,93,29,92,28,91,27,90,26,89,25,88,24,87,23,86,22,85,21,84,20,83,19,82,18,81,17,80,16,79,15,78,14,77,13,76,12,75,11,74,10,73,9,72,8,71,7,70,6,69,5,68,4,67,3,66,2,65,1,64,0);
+    p8b = _mm512_set_epi8(127,63,126,62,125,61,124,60,123,59,122,58,121,57,120,56,119,55,118,54,117,53,116,52,115,51,114,50,113,49,112,48,111,47,110,46,109,45,108,44,107,43,106,42,105,41,104,40,103,39,102,38,101,37,100,36,99,35,98,34,97,33,96,32);
+
+    __m512i p16a = _mm512_set_epi16(47,15,46,14,45,13,44,12,43,11,42,10,41,9,40,8,39,7,38,6,37,5,36,4,35,3,34,2,33,1,32,0);
+    __m512i p16b = _mm512_set_epi16(63,31,62,30,61,29,60,28,59,27,58,26,57,25,56,24,55,23,54,22,53,21,52,20,51,19,50,18,49,17,48,16);
+
+    __m512i p32a = _mm512_set_epi32(23,7,22,6,21,5,20,4,19,3,18,2,17,1,16,0);
+    __m512i p32b = _mm512_set_epi32(31,15,30,14,29,13,28,12,27,11,26,10,25,9,24,8);
+    __m512i e2j,e3j;
+    for (k=0,j=0;j<EQm>>6;j++,k+=4) {
+      e0j    = _mm512_loadu_si512(e0_512+j);	    
+      e1j    = _mm512_loadu_si512(e1_512+j);	    
+      e2j    = _mm512_loadu_si512(e2_512+j);	    
+      e3j    = _mm512_loadu_si512(e3_512+j);	    
+      tmp0   = _mm512_permutex2var_epi8(e0j,p8a,e1j); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
+      tmp1   = _mm512_permutex2var_epi8(e2j,p8a,e3j); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+15) e3(i+15)
+      _mm512_storeu_si512(f_512+k,_mm512_permutex2var_epi16(tmp0,p16a,tmp1));  // e0(i) e1(i) e2(i) e3(i) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
+      _mm512_storeu_si512(f_512+k+1,_mm512_permutex2var_epi16(tmp0,p16b,tmp1)); 
+
+      tmp0   = _mm512_permutex2var_epi8(e0j,p8b,e1j); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
+      tmp1   = _mm512_permutex2var_epi8(e2j,p8b,e3j); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+15) e3(i+15)
+      _mm512_storeu_si512(f_512+k+2,_mm512_permutex2var_epi16(tmp0,p16a,tmp1));  // e0(i) e1(i) e2(i) e3(i) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
+      _mm512_storeu_si512(f_512+k+3,_mm512_permutex2var_epi16(tmp0,p16b,tmp1)); 
    }
+    if ((j<<6) != EQm) {
+      int k2=k<<6;
+      fp = &f[k2];
+      for (int j2=(j<<6) ; j2< EQm ; j2++) {
+        *fp++ = e0[j2];
+        *fp++ = e1[j2];
+        *fp++ = e2[j2];
+        *fp++ = e3[j2];
+      }
+    }
+#else
+    e0_128=(simde__m128i *)e0;
+    e1_128=(simde__m128i *)e1;
+    e2_128=(simde__m128i *)e2;
+    e3_128=(simde__m128i *)e3;
+    simde__m128i e2j,e3j;
+    for (k=0,j=0;j<EQm>>4;j++,k+=4) {
+      e0j    = simde_mm_loadu_si128(e0_128+j);	    
+      e1j    = simde_mm_loadu_si128(e1_128+j);	    
+      e2j    = simde_mm_loadu_si128(e2_128+j);	    
+      e3j    = simde_mm_loadu_si128(e3_128+j);	    
+      tmp0   = simde_mm_unpacklo_epi8(e0j,e1j); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+7) e1(i+7)
+      tmp1   = simde_mm_unpacklo_epi8(e2j,e3j); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+7) e3(i+7)
+      simde_mm_storeu_si128(f_128+k,simde_mm_unpacklo_epi16(tmp0,tmp1));   // e0(i) e1(i) e2(i) e3(i) ... e0(i+3) e1(i+3) e2(i+3) e3(i+3)
+      simde_mm_storeu_si128(f_128+k+1,simde_mm_unpackhi_epi16(tmp0,tmp1));   // e0(i+4) e1(i+4) e2(i+4) e3(i+4) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
+      tmp0   = simde_mm_unpackhi_epi8(e0j,e1j); // e0(i+8) e1(i+8) e0(i+9) e1(i+9) .... e0(i+15) e1(i+15)
+      tmp1   = simde_mm_unpackhi_epi8(e2j,e3j); // e2(i+8) e3(i+9) e2(i+10) e3(i+10) .... e2(i+31) e3(i+31)
+      simde_mm_storeu_si128(f_128+k+2,simde_mm_unpacklo_epi16(tmp0,tmp1));
+      simde_mm_storeu_si128(f_128+k+3,simde_mm_unpackhi_epi16(tmp0,tmp1)); 
+    }
+    if ((j<<4) != EQm) {
+      int k2=k<<4;
+      fp = &f[k2];
+      for (int j2=(j<<4) ; j2< EQm ; j2++) {
+        *fp++ = e0[j2];
+        *fp++ = e1[j2];
+        *fp++ = e2[j2];
+        *fp++ = e3[j2];
+      }
+    }
+#endif
    break;
  case 6:
    e0=e;
@@ -81,95 +176,77 @@ void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f)
    e3=e2+EQm;
    e4=e3+EQm;
    e5=e4+EQm;
-    e0_256=(simde__m256i *)e0;
-    e1_256=(simde__m256i *)e1;
-    e2_256=(simde__m256i *)e2;
-    e3_256=(simde__m256i *)e3;
-    e4_256=(simde__m256i *)e4;
-    e5_256=(simde__m256i *)e5;
+#ifdef __AVX512VBMI__
+    e0_512=(__m512i *)e0;
+    e1_512=(__m512i *)e1;
+    e2_512=(__m512i *)e2;
+    e3_512=(__m512i *)e3;
+    e4_512=(__m512i *)e4;
+    e5_512=(__m512i *)e5;
+    __m512i e4j,e5j;
+    p8a = _mm512_set_epi8(95,31,94,30,93,29,92,28,91,27,90,26,89,25,88,24,87,23,86,22,85,21,84,20,83,19,82,18,81,17,80,16,79,15,78,14,77,13,76,12,75,11,74,10,73,9,72,8,71,7,70,6,69,5,68,4,67,3,66,2,65,1,64,0);
+    p8b = _mm512_set_epi8(127,63,126,62,125,61,124,60,123,59,122,58,121,57,120,56,119,55,118,54,117,53,116,52,115,51,114,50,113,49,112,48,111,47,110,46,109,45,108,44,107,43,106,42,105,41,104,40,103,39,102,38,101,37,100,36,99,35,98,34,97,33,96,32);

-    for (int j=0,k=0;j<EQm>>5;j++,k+=192) {
-      fp  = f+k;
-      fp2 = fp+96;
+    p16a = _mm512_set_epi16(47,15,46,14,45,13,44,12,43,11,42,10,41,9,40,8,39,7,38,6,37,5,36,4,35,3,34,2,33,1,32,0);
+    p16b = _mm512_set_epi16(63,31,62,30,61,29,60,28,59,27,58,26,57,25,56,24,55,23,54,22,53,21,52,20,51,19,50,18,49,17,48,16);

-      tmp0   = simde_mm256_unpacklo_epi8(e0_256[j],e1_256[j]); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
-      tmp1   = simde_mm256_unpacklo_epi8(e2_256[j],e3_256[j]); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+15) e3(i+15)
-      tmp0b  = simde_mm256_unpacklo_epi16(tmp0,tmp1); // e0(i) e1(i) e2(i) e3(i) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
-      tmp1b  = simde_mm256_unpackhi_epi16(tmp0,tmp1); // e0(i+8) e1(i+8) e2(i+8) e3(i+8) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15)
-      tmp0   = simde_mm256_unpacklo_epi8(e4_256[j],e5_256[j]); // e4(i) e5(i) e4(i+1) e5(i+1) .... e4(i+15) e5(i+15)
-      *((uint32_t*)fp)      = simde_mm256_extract_epi32(tmp0b,0);
-      *((uint16_t*)(fp+4))  = simde_mm256_extract_epi16(tmp0,0);
-      *((uint32_t*)(fp+6))  = simde_mm256_extract_epi32(tmp0b,1);
-      *((uint16_t*)(fp+10)) = simde_mm256_extract_epi16(tmp0,1);
-      *((uint32_t*)(fp+12)) = simde_mm256_extract_epi32(tmp0b,2);
-      *((uint16_t*)(fp+16)) = simde_mm256_extract_epi16(tmp0,2);
-      *((uint32_t*)(fp+18)) = simde_mm256_extract_epi32(tmp0b,3);
-      *((uint16_t*)(fp+22)) = simde_mm256_extract_epi16(tmp0,3);
-      *((uint32_t*)(fp+24)) = simde_mm256_extract_epi32(tmp0b,4);
-      *((uint16_t*)(fp+26)) = simde_mm256_extract_epi16(tmp0,4);
-      *((uint32_t*)(fp+30)) = simde_mm256_extract_epi32(tmp0b,5);
-      *((uint16_t*)(fp+34)) = simde_mm256_extract_epi16(tmp0,5);
-      *((uint32_t*)(fp+36)) = simde_mm256_extract_epi32(tmp0,6);
-      *((uint16_t*)(fp+40)) = simde_mm256_extract_epi16(tmp0,6);
-      *((uint32_t*)(fp+42)) = simde_mm256_extract_epi32(tmp0b,7);
-      *((uint16_t*)(fp+46)) = simde_mm256_extract_epi16(tmp0,7);
+    __m512i p16c  = _mm512_set_epi16(21,20,41,19,18,40,17,16,39,15,14,38,13,12,37,11,10,36,9,8,35,7,6,34,5,4,33,3,2,32,1,0);
+    __m512i p16d  = _mm512_set_epi16(10,52,9,8,51,7,6,50,5,4,49,3,2,48,1,0,47,31,30,46,29,28,45,27,26,44,25,24,43,23,22,42);
+    __m512i p16d2 = _mm512_set_epi16(32+10,30,32+9,32+8,27,32+7,32+6,24,32+5,32+4,21,32+3,32+2,18,32+1,32+0,15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0);
+    __m512i p16e  = _mm512_set_epi16(63,31,30,62,29,28,61,27,26,60,25,24,59,23,22,58,21,20,57,19,18,56,17,16,55,15,14,54,13,12,53,11);

-      *((uint32_t*)(fp+48)) = simde_mm256_extract_epi32(tmp1b,0);
-      *((uint16_t*)(fp+52)) = simde_mm256_extract_epi16(tmp0,8);
-      *((uint32_t*)(fp+56)) = simde_mm256_extract_epi32(tmp1b,1);
-      *((uint16_t*)(fp+60)) = simde_mm256_extract_epi16(tmp0,9);
-      *((uint32_t*)(fp+62)) = simde_mm256_extract_epi32(tmp1b,2);
-      *((uint16_t*)(fp+66)) = simde_mm256_extract_epi16(tmp0,10);
-      *((uint32_t*)(fp+68)) = simde_mm256_extract_epi32(tmp1b,3);
-      *((uint16_t*)(fp+72)) = simde_mm256_extract_epi16(tmp0,11);
-      *((uint32_t*)(fp+74)) = simde_mm256_extract_epi32(tmp1b,4);
-      *((uint16_t*)(fp+76)) = simde_mm256_extract_epi16(tmp0,12);
-      *((uint32_t*)(fp+80)) = simde_mm256_extract_epi32(tmp1b,5);
-      *((uint16_t*)(fp+82)) = simde_mm256_extract_epi16(tmp0,13);
-      *((uint32_t*)(fp+86)) = simde_mm256_extract_epi32(tmp1b,6);
-      *((uint16_t*)(fp+90)) = simde_mm256_extract_epi16(tmp0,14);
-      *((uint32_t*)(fp+92)) = simde_mm256_extract_epi32(tmp1b,7);
-      *((uint16_t*)(fp+94)) = simde_mm256_extract_epi16(tmp0,15);
+    for (j=0,k=0;j<EQm>>6;j++,k+=6) {
+      e0j    = _mm512_loadu_si512(e0_512+j);	    
+      e1j    = _mm512_loadu_si512(e1_512+j);	    
+      e2j    = _mm512_loadu_si512(e2_512+j);	    
+      e3j    = _mm512_loadu_si512(e3_512+j);	    
+      e4j    = _mm512_loadu_si512(e4_512+j);	    
+      e5j    = _mm512_loadu_si512(e5_512+j);	    
+      tmp0   = _mm512_permutex2var_epi8(e0j,p8a,e1j); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+31) e1(i+31)
+      tmp1   = _mm512_permutex2var_epi8(e2j,p8a,e3j); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+31) e3(i+31)
+      tmp2   = _mm512_permutex2var_epi8(e4j,p8a,e5j); // e4(i) e5(i) e4(i+1) e5(i+1) .... e4(i+31) e5(i+31)
+      tmp3   = _mm512_permutex2var_epi16(tmp0,p16a,tmp1);  // e0(i) e1(i) e2(i) e3(i) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15)
+      tmp4   = _mm512_permutex2var_epi16(tmp0,p16b,tmp1);  // e0(i+16) e1(i+16) e2(i+16) e3(i+16) ... e0(i+31) e1(i+31) e2(i+31) e3(i+31)
+      _mm512_storeu_si512(f_512+k,_mm512_permutex2var_epi16(tmp3,p16c,tmp2));  // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) ... e0(i+9) e1(i+9) e2(i+9) e3(i+9) e4(i+9) e5(i+9) e0(i+10) e1(i+10)
+      tmp5   = _mm512_permutex2var_epi16(tmp3,p16d,tmp2);  // e2(i+10) e3(i+10) e4(i+10) e5(i+10) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15) e4(i+15) e5(i+15) x x x x e4(i+16) e5(i+16) x x x x .... e4(i+20) e5(i+20) x x
+      _mm512_storeu_si512(f_512+k+1,_mm512_permutex2var_epi16(tmp5,p16d2,tmp4)); // e2(i+10) e3(i+10) e4(i+10) e5(i+10) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15) e4(i+15) e5(i+15) e0(i+16) e1(i+16) e2(i+16) e3(i+16) e4(i+16) e5(i+16)  e0(i+20) e1(i+20) e2(i+20) e3(i+20) e4(i+20) e5(i+20) e0(i+21) e1(i+21)  
+      _mm512_storeu_si512(f_512+k+2,_mm512_permutex2var_epi16(tmp4,p16e,tmp2)); // e2(i+21) e3(i+21) e4(i+21) e5(i+21) .... e0(i+31) e1(i+31) e2(i+31) e3(i+31) e4(i+31) e5(i+31) 

-      tmp0   = simde_mm256_unpackhi_epi8(e0_256[j],e1_256[j]); // e0(i+16) e1(i+16) e0(i+17) e1(i+17) .... e0(i+31) e1(i+31)
-      tmp1   = simde_mm256_unpackhi_epi8(e2_256[j],e3_256[j]); // e2(i+16) e3(i+16) e2(i+17) e3(i+17) .... e2(i+31) e3(i+31)
-      tmp0b  = simde_mm256_unpacklo_epi16(tmp0,tmp1); // e0(i+16) e1(i+16) e2(i+16) e3(i+16) ... e0(i+23) e1(i+23) e2(i+23) e3(i+23)
-      tmp1b  = simde_mm256_unpackhi_epi16(tmp0,tmp1); // e0(i+24) e1(i+24) e2(i+24) e3(i+24) ... e0(i+31) e1(i+31) e2(i+31) e3(i+31)
-      tmp0   = simde_mm256_unpackhi_epi8(e4_256[j],e5_256[j]); // e4(i+16) e5(i+16) e4(i+17) e5(i+17) .... e4(i+31) e5(i+31)
-      *((uint32_t*)fp2)      = simde_mm256_extract_epi32(tmp0b,0);
-      *((uint16_t*)(fp2+4))  = simde_mm256_extract_epi16(tmp0,0);
-      *((uint32_t*)(fp2+6))  = simde_mm256_extract_epi32(tmp0b,1);
-      *((uint16_t*)(fp2+10)) = simde_mm256_extract_epi16(tmp0,1);
-      *((uint32_t*)(fp2+12)) = simde_mm256_extract_epi32(tmp0b,2);
-      *((uint16_t*)(fp2+16)) = simde_mm256_extract_epi16(tmp0,2);
-      *((uint32_t*)(fp2+18)) = simde_mm256_extract_epi32(tmp0b,3);
-      *((uint16_t*)(fp2+22)) = simde_mm256_extract_epi16(tmp0,3);
-      *((uint32_t*)(fp2+24)) = simde_mm256_extract_epi32(tmp0b,4);
-      *((uint16_t*)(fp2+26)) = simde_mm256_extract_epi16(tmp0,4);
-      *((uint32_t*)(fp2+30)) = simde_mm256_extract_epi32(tmp0b,5);
-      *((uint16_t*)(fp2+34)) = simde_mm256_extract_epi16(tmp0,5);
-      *((uint32_t*)(fp2+36)) = simde_mm256_extract_epi32(tmp0,6);
-      *((uint16_t*)(fp2+40)) = simde_mm256_extract_epi16(tmp0,6);
-      *((uint32_t*)(fp2+42)) = simde_mm256_extract_epi32(tmp0b,7);
-      *((uint16_t*)(fp2+46)) = simde_mm256_extract_epi16(tmp0,7);
+      tmp0   = _mm512_permutex2var_epi8(e0j,p8b,e1j); // e0(i+32) e1(i+32) e0(i+32) e1(i+32) .... e0(i+63) e1(i+63)
+      tmp1   = _mm512_permutex2var_epi8(e2j,p8b,e3j); // e2(i+32) e3(i+32) e2(i+32) e3(i+32) .... e2(i+63) e3(i+63)
+      tmp2   = _mm512_permutex2var_epi8(e4j,p8b,e5j); // e4(i+32) e5(i+32) e4(i+32) e5(i+32) .... e4(i+63) e5(i+63)
+      tmp3   = _mm512_permutex2var_epi16(tmp0,p16a,tmp1);  // e0(i) e1(i) e2(i) e3(i) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15)
+      tmp4   = _mm512_permutex2var_epi16(tmp0,p16b,tmp1);  // e0(i+16) e1(i+16) e2(i+16) e3(i+16) ... e0(i+31) e1(i+31) e2(i+31) e3(i+31)

-      *((uint32_t*)(fp2+48)) = simde_mm256_extract_epi32(tmp1b,0);
-      *((uint16_t*)(fp2+52)) = simde_mm256_extract_epi16(tmp0,8);
-      *((uint32_t*)(fp2+56)) = simde_mm256_extract_epi32(tmp1b,1);
-      *((uint16_t*)(fp2+60)) = simde_mm256_extract_epi16(tmp0,9);
-      *((uint32_t*)(fp2+62)) = simde_mm256_extract_epi32(tmp1b,2);
-      *((uint16_t*)(fp2+66)) = simde_mm256_extract_epi16(tmp0,10);
-      *((uint32_t*)(fp2+68)) = simde_mm256_extract_epi32(tmp1b,3);
-      *((uint16_t*)(fp2+72)) = simde_mm256_extract_epi16(tmp0,11);
-      *((uint32_t*)(fp2+74)) = simde_mm256_extract_epi32(tmp1b,4);
-      *((uint16_t*)(fp2+76)) = simde_mm256_extract_epi16(tmp0,12);
-      *((uint32_t*)(fp2+80)) = simde_mm256_extract_epi32(tmp1b,5);
-      *((uint16_t*)(fp2+82)) = simde_mm256_extract_epi16(tmp0,13);
-      *((uint32_t*)(fp2+86)) = simde_mm256_extract_epi32(tmp1b,6);
-      *((uint16_t*)(fp2+90)) = simde_mm256_extract_epi16(tmp0,14);
-      *((uint32_t*)(fp2+92)) = simde_mm256_extract_epi32(tmp1b,7);
-      *((uint16_t*)(fp2+94)) = simde_mm256_extract_epi16(tmp0,15);
+      _mm512_storeu_si512(f_512+k+3,_mm512_permutex2var_epi16(tmp3,p16c,tmp2));  // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) ... e0(i+9) e1(i+9) e2(i+9) e3(i+9) e4(i+9) e5(i+9) e0(i+10) e1(i+10)
+      tmp5   = _mm512_permutex2var_epi16(tmp3,p16d,tmp2);  // e2(i+10) e3(i+10) e4(i+10) e5(i+10) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15) e4(i+15) e5(i+15) x x x x e4(i+16) e5(i+16) x x x x .... e4(i+20) e5(i+20) x x
+      _mm512_storeu_si512(f_512+k+4,_mm512_permutex2var_epi16(tmp5,p16d2,tmp4)); // e2(i+10) e3(i+10) e4(i+10) e5(i+10) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15) e4(i+15) e5(i+15) e0(i+16) e1(i+16) e2(i+16) e3(i+16) e4(i+16) e5(i+16)  e0(i+20) e1(i+20) e2(i+20) e3(i+20) e4(i+20) e5(i+20) e0(i+21) e1(i+21)  
+      _mm512_storeu_si512(f_512+k+5,_mm512_permutex2var_epi16(tmp4,p16e,tmp2)); // e2(i+21) e3(i+21) e4(i+21) e5(i+21) .... e0(i+31) e1(i+31) e2(i+31) e3(i+31) e4(i+31) e5(i+31) 
    }
+    if ((j<<6) != EQm) {
+      int k2=k<<6;
+      fp = &f[k2];
+      for (int j2=(j<<6) ; j2< EQm ; j2++) {
+        *fp++ = e0[j2];
+        *fp++ = e1[j2];
+        *fp++ = e2[j2];
+        *fp++ = e3[j2];
+        *fp++ = e4[j2];
+        *fp++ = e5[j2];
+      }
+    }
+#else
+    fp = f;
+    for (int j = 0; j< EQm; j++){
+      *fp++ = e0[j];
+      *fp++ = e1[j];
+      *fp++ = e2[j];
+      *fp++ = e3[j];
+      *fp++ = e4[j];
+      *fp++ = e5[j];
+    }
+
+#endif
    break;
  case 8:
    e0=e;
@@ -181,43 +258,137 @@ void nr_interleaving_ldpc(uint32_t E, uint8_t Qm, uint8_t *e, uint8_t *f)
    e6=e5+EQm;
    e7=e6+EQm;

-    e0_256=(simde__m256i *)e0;
-    e1_256=(simde__m256i *)e1;
-    e2_256=(simde__m256i *)e2;
-    e3_256=(simde__m256i *)e3;
-    e4_256=(simde__m256i *)e4;
-    e5_256=(simde__m256i *)e5;
-    e6_256=(simde__m256i *)e6;
-    e7_256=(simde__m256i *)e7;
-    for (int k=0,j=0;j<EQm>>5;j++,k+=8) {
-      tmp0   = simde_mm256_unpacklo_epi8(e0_256[j],e1_256[j]); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
-      tmp1   = simde_mm256_unpacklo_epi8(e2_256[j],e3_256[j]); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+15) e3(i+15)
-      tmp2   = simde_mm256_unpacklo_epi8(e4_256[j],e5_256[j]); // e4(i) e5(i) e4(i+1) e5(i+1) .... e4(i+15) e5(i+15)
-      tmp3   = simde_mm256_unpacklo_epi8(e6_256[j],e7_256[j]); // e6(i) e7(i) e6(i+1) e7(i+1) .... e6(i+15) e7(i+15)
-      tmp4   = simde_mm256_unpacklo_epi16(tmp0,tmp1);  // e0(i) e1(i) e2(i) e3(i) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
-      tmp5   = simde_mm256_unpacklo_epi16(tmp2,tmp3);  // e4(i) e5(i) e6(i) e7(i) ... e4(i+7) e5(i+7) e6(i+7) e7(i+7)
-      f_256[k]   = simde_mm256_unpacklo_epi16(tmp4,tmp5);  // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) e6(i) e7(i)... e0(i+3) e1(i+3) e2(i+3) e3(i+3) e4(i+3) e5(i+3) e6(i+3) e7(i+3))
-      f_256[k+1] = simde_mm256_unpackhi_epi16(tmp4,tmp5);  // e0(i+4) e1(i+4) e2(i+4) e3(i+4) e4(i+4) e5(i+4) e6(i+4) e7(i+4)... e0(i+7) e1(i+7) e2(i+7) e3(i+7) e4(i+7) e5(i+7) e6(i+7) e7(i+7))
+#ifdef __AVX512VBMI__
+    e0_512=(__m512i *)e0;
+    e1_512=(__m512i *)e1;
+    e2_512=(__m512i *)e2;
+    e3_512=(__m512i *)e3;
+    e4_512=(__m512i *)e4;
+    e5_512=(__m512i *)e5;
+    e6_512=(__m512i *)e6;
+    e7_512=(__m512i *)e7;
+    __m512i e6j,e7j;
+    p8a = _mm512_set_epi8(95,31,94,30,93,29,92,28,91,27,90,26,89,25,88,24,87,23,86,22,85,21,84,20,83,19,82,18,81,17,80,16,79,15,78,14,77,13,76,12,75,11,74,10,73,9,72,8,71,7,70,6,69,5,68,4,67,3,66,2,65,1,64,0);
+    p8b = _mm512_set_epi8(127,63,126,62,125,61,124,60,123,59,122,58,121,57,120,56,119,55,118,54,117,53,116,52,115,51,114,50,113,49,112,48,111,47,110,46,109,45,108,44,107,43,106,42,105,41,104,40,103,39,102,38,101,37,100,36,99,35,98,34,97,33,96,32);

-      tmp4   = simde_mm256_unpackhi_epi16(tmp0,tmp1);  // e0(i+8) e1(i+8) e2(i+8) e3(i+8) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15)
-      tmp5   = simde_mm256_unpackhi_epi16(tmp2,tmp3);  // e4(i+8) e5(i+8) e6(i+8) e7(i+8) ... e4(i+15) e5(i+15) e6(i+15) e7(i+15)
-      f_256[k+2]   = simde_mm256_unpacklo_epi16(tmp4,tmp5);  // e0(i+8) e1(i+8) e2(i+8) e3(i+8) e4(i+8) e5(i+8) e6(i+8) e7(i+8)... e0(i+11) e1(i+11) e2(i+11) e3(i+11) e4(i+11) e5(i+11) e6(i+11) e7(i+11))
-      f_256[k+3] = simde_mm256_unpackhi_epi16(tmp4,tmp5);  // e0(i+12) e1(i+12) e2(i+12) e3(i+12) e4(i+12) e5(i+12) e6(i+12) e7(i+12)... e0(i+15) e1(i+15) e2(i+15) e3(i+15) e4(i+15) e5(i+15) e6(i+15) e7(i+15))
+    p16a = _mm512_set_epi16(47,15,46,14,45,13,44,12,43,11,42,10,41,9,40,8,39,7,38,6,37,5,36,4,35,3,34,2,33,1,32,0);
+    p16b = _mm512_set_epi16(63,31,62,30,61,29,60,28,59,27,58,26,57,25,56,24,55,23,54,22,53,21,52,20,51,19,50,18,49,17,48,16);

-      tmp0   = simde_mm256_unpackhi_epi8(e0_256[j],e1_256[j]); // e0(i+16) e1(i+16) e0(i+17) e1(i+17) .... e0(i+31) e1(i+31)
-      tmp1   = simde_mm256_unpackhi_epi8(e2_256[j],e3_256[j]); // e2(i+16) e3(i+16) e2(i+17) e3(i+17) .... e2(i+31) e3(i+31)
-      tmp2   = simde_mm256_unpackhi_epi8(e4_256[j],e5_256[j]); // e4(i+16) e5(i+16) e4(i+17) e5(i+17) .... e4(i+31) e5(i+31)
-      tmp3   = simde_mm256_unpackhi_epi8(e6_256[j],e7_256[j]); // e6(i+16) e7(i+16) e6(i+17) e7(i+17) .... e6(i+31) e7(i+31)
-      tmp4   = simde_mm256_unpacklo_epi16(tmp0,tmp1);  // e0(i+!6) e1(i+16) e2(i+16) e3(i+16) ... e0(i+23) e1(i+23) e2(i+23) e3(i+23)
-      tmp5   = simde_mm256_unpacklo_epi16(tmp2,tmp3);  // e4(i+16) e5(i+16) e6(i+16) e7(i+16) ... e4(i+23) e5(i+23) e6(i+23) e7(i+23)
-      f_256[k+4] = simde_mm256_unpacklo_epi16(tmp4,tmp5);  // e0(i+16) e1(i+16) e2(i+16) e3(i+16) e4(i+16) e5(i+16) e6(i+16) e7(i+16)... e0(i+19) e1(i+19) e2(i+19) e3(i+19) e4(i+19) e5(i+19) e6(i+19) e7(i+19))
-      f_256[k+5] = simde_mm256_unpackhi_epi16(tmp4,tmp5);  // e0(i+20) e1(i+20) e2(i+20) e3(i+20) e4(i+20) e5(i+20) e6(i+20) e7(i+20)... e0(i+23) e1(i+23) e2(i+23) e3(i+23) e4(i+23) e5(i+23) e6(i+23) e7(i+23))
+    p32a = _mm512_set_epi32(23,7,22,6,21,5,20,4,19,3,18,2,17,1,16,0);
+    p32b = _mm512_set_epi32(31,15,30,14,29,13,28,12,27,11,26,10,25,9,24,8);
+    for (k=0,j=0;j<EQm>>6;j++,k+=8) {
+      e0j    = _mm512_loadu_si512(e0_512+j);	    
+      e1j    = _mm512_loadu_si512(e1_512+j);	    
+      e2j    = _mm512_loadu_si512(e2_512+j);	    
+      e3j    = _mm512_loadu_si512(e3_512+j);	    
+      e4j    = _mm512_loadu_si512(e4_512+j);	    
+      e5j    = _mm512_loadu_si512(e5_512+j);	    
+      e6j    = _mm512_loadu_si512(e6_512+j);	    
+      e7j    = _mm512_loadu_si512(e7_512+j);	    
+      tmp0   = _mm512_permutex2var_epi8(e0j,p8a,e1j); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+15) e1(i+15)
+      tmp1   = _mm512_permutex2var_epi8(e2j,p8a,e3j); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+15) e3(i+15)
+      tmp2   = _mm512_permutex2var_epi8(e4j,p8a,e5j); // e4(i) e5(i) e4(i+1) e5(i+1) .... e4(i+15) e5(i+15)
+      tmp3   = _mm512_permutex2var_epi8(e6j,p8a,e7j); // e6(i) e7(i) e6(i+1) e7(i+1) .... e6(i+15) e7(i+15)
+      tmp4   = _mm512_permutex2var_epi16(tmp0,p16a,tmp1);  // e0(i) e1(i) e2(i) e3(i) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)
+      tmp5   = _mm512_permutex2var_epi16(tmp2,p16a,tmp3);  // e4(i) e5(i) e6(i) e7(i) ... e4(i+7) e5(i+7) e6(i+7) e7(i+7)
+      _mm512_storeu_si512(f_512+k,_mm512_permutex2var_epi32(tmp4,p32a,tmp5));  // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) e6(i) e7(i)... e0(i+3) e1(i+3) e2(i+3) e3(i+3) e4(i+3) e5(i+3) e6(i+3) e7(i+3))
+      _mm512_storeu_si512(f_512+k+1,_mm512_permutex2var_epi32(tmp4,p32b,tmp5));  // e0(i+4) e1(i+4) e2(i+4) e3(i+4) e4(i+4) e5(i+4) e6(i+4) e7(i+4)... e0(i+7) e1(i+7) e2(i+7) e3(i+7) e4(i+7) e5(i+7) e6(i+7) e7(i+7))

-      tmp4   = simde_mm256_unpackhi_epi16(tmp0,tmp1);  // e0(i+24) e1(i+24) e2(i+24) e3(i+24) ... e0(i+31) e1(i+31) e2(i+31) e3(i+31)
-      tmp5   = simde_mm256_unpackhi_epi16(tmp2,tmp3);  // e4(i+24) e5(i+24) e6(i+24) e7(i+24) ... e4(i+31) e5(i+31) e6(i+31) e7(i+31)
-      f_256[k+6] = simde_mm256_unpacklo_epi16(tmp4,tmp5);  // e0(i+24) e1(i+24) e2(i+24) e3(i+24) e4(i+24) e5(i+24) e6(i+24) e7(i+24)... e0(i+27) e1(i+27) e2(i+27) e3(i+27) e4(i+27) e5(i+27) e6(i+27) e7(i+27))
-      f_256[k+7] = simde_mm256_unpackhi_epi16(tmp4,tmp5);  // e0(i+28) e1(i+28) e2(i+28) e3(i+28) e4(i+28) e5(i+28) e6(i+28) e7(i+28)... e0(i+31) e1(i+31) e2(i+31) e3(i+31) e4(i+31) e5(i+31) e6(i+31) e7(i+31))
+      tmp4   = _mm512_permutex2var_epi16(tmp0,p16b,tmp1);  // e0(i+8) e1(i+8) e2(i+8) e3(i+8) ... e0(i+15) e1(i+15) e2(i+15) e3(i+15)
+      tmp5   = _mm512_permutex2var_epi16(tmp2,p16b,tmp3);  // e4(i+8) e5(i+8) e6(i+8) e7(i+8) ... e4(i+15) e5(i+15) e6(i+15) e7(i+15)
+      _mm512_storeu_si512(f_512+k+2,_mm512_permutex2var_epi32(tmp4,p32a,tmp5));  // e0(i+8) e1(i+8) e2(i+8) e3(i+8) e4(i+8) e5(i+8) e6(i+8) e7(i+8)... e0(i+11) e1(i+11) e2(i+11) e3(i+11) e4(i+11) e5(i+11) e6(i+11) e7(i+11))
+      _mm512_storeu_si512(f_512+k+3,_mm512_permutex2var_epi32(tmp4,p32b,tmp5));  // e0(i+12) e1(i+12) e2(i+12) e3(i+12) e4(i+12) e5(i+12) e6(i+12) e7(i+12)... e0(i+15) e1(i+15) e2(i+15) e3(i+15) e4(i+15) e5(i+15) e6(i+15) e7(i+15))
+
+      tmp0   = _mm512_permutex2var_epi8(e0j,p8b,e1j); // e0(i+16) e1(i+16) e0(i+17) e1(i+17) .... e0(i+31) e1(i+31)
+      tmp1   = _mm512_permutex2var_epi8(e2j,p8b,e3j); // e2(i+16) e3(i+16) e2(i+17) e3(i+17) .... e2(i+31) e3(i+31)
+      tmp2   = _mm512_permutex2var_epi8(e4j,p8b,e5j); // e4(i+16) e5(i+16) e4(i+17) e5(i+17) .... e4(i+31) e5(i+31)
+      tmp3   = _mm512_permutex2var_epi8(e6j,p8b,e7j); // e6(i+16) e7(i+16) e6(i+17) e7(i+17) .... e6(i+31) e7(i+31)
+      tmp4   = _mm512_permutex2var_epi16(tmp0,p16a,tmp1);  // e0(i+!6) e1(i+16) e2(i+16) e3(i+16) ... e0(i+23) e1(i+23) e2(i+23) e3(i+23)
+      tmp5   = _mm512_permutex2var_epi16(tmp2,p16a,tmp3);  // e4(i+16) e5(i+16) e6(i+16) e7(i+16) ... e4(i+23) e5(i+23) e6(i+23) e7(i+23)
+      _mm512_storeu_si512(f_512+k+4,_mm512_permutex2var_epi32(tmp4,p32a,tmp5));  // e0(i+16) e1(i+16) e2(i+16) e3(i+16) e4(i+16) e5(i+16) e6(i+16) e7(i+16)... e0(i+19) e1(i+19) e2(i+19) e3(i+19) e4(i+19) e5(i+19) e6(i+19) e7(i+19))
+      _mm512_storeu_si512(f_512+k+5,_mm512_permutex2var_epi32(tmp4,p32b,tmp5));  // e0(i+20) e1(i+20) e2(i+20) e3(i+20) e4(i+20) e5(i+20) e6(i+20) e7(i+20)... e0(i+23) e1(i+23) e2(i+23) e3(i+23) e4(i+23) e5(i+23) e6(i+23) e7(i+23))
+
+      tmp4   = _mm512_permutex2var_epi16(tmp0,p16b,tmp1);  // e0(i+24) e1(i+24) e2(i+24) e3(i+24) ... e0(i+31) e1(i+31) e2(i+31) e3(i+31)
+      tmp5   = _mm512_permutex2var_epi16(tmp2,p16b,tmp3);  // e4(i+24) e5(i+24) e6(i+24) e7(i+24) ... e4(i+31) e5(i+31) e6(i+31) e7(i+31)
+      _mm512_storeu_si512(f_512+k+6,_mm512_permutex2var_epi32(tmp4,p32a,tmp5));  // e0(i+24) e1(i+24) e2(i+24) e3(i+24) e4(i+24) e5(i+24) e6(i+24) e7(i+24)... e0(i+27) e1(i+27) e2(i+27) e3(i+27) e4(i+27) e5(i+27) e6(i+27) e7(i+27))
+      _mm512_storeu_si512(f_512+k+7,_mm512_permutex2var_epi32(tmp4,p32b,tmp5));  // e0(i+28) e1(i+28) e2(i+28) e3(i+28) e4(i+28) e5(i+28) e6(i+28) e7(i+28)... e0(i+31) e1(i+31) e2(i+31) e3(i+31) e4(i+31) e5(i+31) e6(i+31) e7(i+31))
    }
+    if ((j<<6) != EQm) {
+      int k2=k<<6;
+      fp=&f[k2];
+      for (int j2=(j<<6) ; j2< EQm ; j2++) {
+         *fp++ = e0[j2];
+         *fp++ = e1[j2];
+         *fp++ = e2[j2];
+         *fp++ = e3[j2];
+         *fp++ = e4[j2];
+         *fp++ = e5[j2];
+         *fp++ = e6[j2];
+         *fp++ = e7[j2];
+      }
+    }
+#else
+    e0_128=(simde__m128i *)e0;
+    e1_128=(simde__m128i *)e1;
+    e2_128=(simde__m128i *)e2;
+    e3_128=(simde__m128i *)e3;
+    e4_128=(simde__m128i *)e4;
+    e5_128=(simde__m128i *)e5;
+    e6_128=(simde__m128i *)e6;
+    e7_128=(simde__m128i *)e7;
+    simde__m128i e4j,e5j,e6j,e7j;
+    
+    for (k=0,j=0;j<EQm>>4;j++,k+=8) {
+      e0j    = simde_mm_loadu_si128(e0_128+j);	    
+      e1j    = simde_mm_loadu_si128(e1_128+j);	    
+      e2j    = simde_mm_loadu_si128(e2_128+j);	    
+      e3j    = simde_mm_loadu_si128(e3_128+j);	    
+      e4j    = simde_mm_loadu_si128(e4_128+j);	    
+      e5j    = simde_mm_loadu_si128(e5_128+j);	    
+      e6j    = simde_mm_loadu_si128(e6_128+j);	    
+      e7j    = simde_mm_loadu_si128(e7_128+j);	    
+      tmp0   = simde_mm_unpacklo_epi8(e0j,e1j); // e0(i) e1(i) e0(i+1) e1(i+1) .... e0(i+7) e1(i+7)
+      tmp1   = simde_mm_unpacklo_epi8(e2j,e3j); // e2(i) e3(i) e2(i+1) e3(i+1) .... e2(i+7) e3(i+7)
+      tmp2   = simde_mm_unpacklo_epi16(tmp0,tmp1);// e0(i) e1(i) e2(i) e3(i) e0(i+1) e1(i+1) e2(i+1) e3(i+1) ... e0(i+3) e1(i+3) e2(i+3) e3(i+3)	
+      tmp3   = simde_mm_unpackhi_epi16(tmp0,tmp1);// e0(i+4) e1(i+4) e2(i+4) e3(i+4) e0(i+5) e1(i+5) e2(i+5) e3(i+5) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)	
+      tmp4   = simde_mm_unpacklo_epi8(e4j,e5j); // e4(i) e5(i) e4(i+1) e5(i+1) .... e4(i+7) e5(i+7)
+      tmp5   = simde_mm_unpacklo_epi8(e6j,e7j); // e6(i) e7(i) e6(i+1) e7(i+1) .... e6(i+7) e7(i+7)
+      tmp6   = simde_mm_unpacklo_epi16(tmp4,tmp5);// e4(i) e5(i) e6(i) e7(i) e4(i+1) e5(i+1) e6(i+1) e7(i+1) ... e4(i+3) e5(i+3) e6(i+) e7(i+3)	
+      tmp7   = simde_mm_unpackhi_epi16(tmp4,tmp5);// e4(i+4) e5(i+4) e6(i+4) e7(i+4) e4(i+5) e5(i+5) e6(i+5) e7(i+5) ... e4(i+7) e5(i+7) e6(i+7) e7(i+7)	
+      simde_mm_storeu_si128(f_128+k,simde_mm_unpacklo_epi32(tmp2,tmp6));   // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) e6(i) e7(i) e0(i+1) ... e7(i+1)
+      simde_mm_storeu_si128(f_128+k+1,simde_mm_unpackhi_epi32(tmp2,tmp6));   // e0(i+2) e1(i+2) e2(i+2) e3(i+2) e4(i+2) e5(i+2) e6(i+2) e7(i+2) e0(i+3) e1(i+3) ... e7(i+3)
+      simde_mm_storeu_si128(f_128+k+2,simde_mm_unpacklo_epi32(tmp3,tmp7));   // e0(i+4) e1(i+4) e2(i+4) e3(i+4) e4(i+4) e5(i+4) e6(i+4) e7+4(i) e0(i+5) ... e7(i+5)
+      simde_mm_storeu_si128(f_128+k+3,simde_mm_unpackhi_epi32(tmp3,tmp7));   // e0(i+6) e1(i+6) e2(i+6) e3(i+6) e4(i+6) e5(i+6) e6(i+6) e7(i+6) e0(i+7) e0(i+7) ... e7(i+7)
+
+
+      tmp0   = simde_mm_unpackhi_epi8(e0j,e1j); // e0(i+8) e1(i+8) e0(i+9) e1(i+9) .... e0(i+15) e1(i+15)
+      tmp1   = simde_mm_unpackhi_epi8(e2j,e3j); // e2(i+8) e3(i+8) e2(i+9) e3(i+9) .... e2(i+15) e3(i+15)
+      tmp2   = simde_mm_unpacklo_epi16(tmp0,tmp1);// e0(i) e1(i) e2(i) e3(i) e0(i+1) e1(i+1) e2(i+1) e3(i+1) ... e0(i+3) e1(i+3) e2(i+3) e3(i+3)	
+      tmp3   = simde_mm_unpackhi_epi16(tmp0,tmp1);// e0(i+4) e1(i+4) e2(i+4) e3(i+4) e0(i+5) e1(i+5) e2(i+5) e3(i+5) ... e0(i+7) e1(i+7) e2(i+7) e3(i+7)	
+      tmp4   = simde_mm_unpackhi_epi8(e4j,e5j); // e4(i+8) e5(i+8) e4(i+9) e5(i+9) .... e4(i+15) e515i+7)
+      tmp5   = simde_mm_unpackhi_epi8(e6j,e7j); // e6(i+8) e7(i+8) e6(i+9) e7(i+9) .... e6(i+15) e7(i+15)
+      tmp6   = simde_mm_unpacklo_epi16(tmp4,tmp5);// e4(i) e5(i) e6(i) e7(i) e4(i+1) e5(i+1) e6(i+1) e7(i+1) ... e4(i+3) e5(i+3) e6(i+) e7(i+3)	
+      tmp7   = simde_mm_unpackhi_epi16(tmp4,tmp5);// e4(i+4) e5(i+4) e6(i+4) e7(i+4) e4(i+5) e5(i+5) e6(i+5) e7(i+5) ... e4(i+7) e5(i+7) e6(i+7) e7(i+7)	
+      simde_mm_storeu_si128(f_128+k+4,simde_mm_unpacklo_epi32(tmp2,tmp6));   // e0(i) e1(i) e2(i) e3(i) e4(i) e5(i) e6(i) e7(i) e0(i+1) ... e7(i+1)
+      simde_mm_storeu_si128(f_128+k+5,simde_mm_unpackhi_epi32(tmp2,tmp6));   // e0(i+2) e1(i+2) e2(i+2) e3(i+2) e4(i+2) e5(i+2) e6(i+2) e7(i+2) e0(i+3) e1(i+3) ... e7(i+3)
+      simde_mm_storeu_si128(f_128+k+6,simde_mm_unpacklo_epi32(tmp3,tmp7));   // e0(i+4) e1(i+4) e2(i+4) e3(i+4) e4(i+4) e5(i+4) e6(i+4) e7+4(i) e0(i+5) ... e7(i+5)
+      simde_mm_storeu_si128(f_128+k+7,simde_mm_unpackhi_epi32(tmp3,tmp7));   // e0(i+6) e1(i+6) e2(i+6) e3(i+6) e4(i+6) e5(i+6) e6(i+6) e7(i+6) e0(i+7) e0(i+7) ... e7(i+7)
+    }
+    if ((j<<4) != EQm) {
+      int k2=k<<4;
+      fp=&f[k2];
+      for (int j2=(j<<4) ; j2< EQm ; j2++) {
+         *fp++ = e0[j2];
+         *fp++ = e1[j2];
+         *fp++ = e2[j2];
+         *fp++ = e3[j2];
+         *fp++ = e4[j2];
+         *fp++ = e5[j2];
+         *fp++ = e6[j2];
+         *fp++ = e7[j2];
+      }
+    }
+#endif
    break;
  default: AssertFatal(1==0,"Should be here!\n");
  }
--- a/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_cnProc/cnProc_gen_BG1_128.c
+++ b/openair1/PHY/CODING/nrLDPC_decoder/nrLDPC_tools/generator_cnProc/cnProc_gen_BG1_128.c
@@ -126,7 +126,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[0]>>4)+lut_idxCnProcG3[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -138,7 +138,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	  
 	  //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	  // Store result
 	  //                min = simde_mm_min_epu8(min, maxLLR); // 128 in epi8 is -127
 #ifndef DROP_MAXLLR
@@ -146,7 +146,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[0]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[0]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
        }
    }
@@ -185,7 +185,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[1]>>4)+lut_idxCnProcG4[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -199,7 +199,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -209,7 +209,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif	  
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	      //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[1]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[1]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
        }
    }
@@ -252,7 +252,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[2]>>4)+lut_idxCnProcG5[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -266,7 +266,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -276,7 +276,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif	  
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[2]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[2]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"           }\n");
        }
    }
@@ -319,7 +319,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[3]>>4)+lut_idxCnProcG6[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -333,7 +333,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -343,7 +343,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif	  
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[3]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[3]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
 	}
    }
@@ -389,7 +389,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[4]>>4)+lut_idxCnProcG7[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -403,7 +403,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -413,7 +413,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif	  
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[4]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[4]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
 	}
    }
@@ -460,7 +460,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[5]>>4)+lut_idxCnProcG8[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -474,7 +474,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -484,7 +484,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[5]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[5]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"              }\n");
        }
    }
@@ -531,7 +531,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[6]>>4)+lut_idxCnProcG9[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -545,7 +545,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -555,7 +555,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[6]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[6]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
 	}
    }
@@ -603,7 +603,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[7]>>4)+lut_idxCnProcG10[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -617,7 +617,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -627,7 +627,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif	  
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[7]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[7]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
        }
    }
@@ -678,7 +678,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	  //                ymm0 = p_cnProcBuf[lut_idxCnProcG3[j][0] + i];
 	  fprintf(fd,"                ymm0 = ((simde__m128i*)cnProcBuf)[%d+i];\n",(lut_startAddrCnGroups[8]>>4)+lut_idxCnProcG19[j][0]*2);
 	  //                sgn  = simde_mm_sign_epi8(ones, ymm0);
-	  fprintf(fd,"                sgn  = simde_mm_sign_epi8(ones, ymm0);\n");
+	  fprintf(fd,"                sgn  = oai_mm_sign_epi8(ones, ymm0);\n");
 	  //                min  = simde_mm_abs_epi8(ymm0);
 	  fprintf(fd,"                min  = simde_mm_abs_epi8(ymm0);\n");
 	  
@@ -692,7 +692,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 	      fprintf(fd,"                min  = simde_mm_min_epu8(min, simde_mm_abs_epi8(ymm0));\n");
 	      
 	      //                sgn  = simde_mm_sign_epi8(sgn, ymm0);
-  	      fprintf(fd,"                sgn  = simde_mm_sign_epi8(sgn, ymm0);\n");
+  	      fprintf(fd,"                sgn  = oai_mm_sign_epi8(sgn, ymm0);\n");
 	    }
 	  
 	  // Store result
@@ -702,7 +702,7 @@ void nrLDPC_cnProc_BG1_generator_128(const char* dir, int R)
 #endif	  
 	  //                *p_cnProcBufResBit = simde_mm_sign_epi8(min, sgn);
 	  //                p_cnProcBufResBit++;
-	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = simde_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[8]>>4)+(j*bitOffsetInGroup));
+	  fprintf(fd,"                ((simde__m128i*)cnProcBufRes)[%d+i] = oai_mm_sign_epi8(min, sgn);\n",(lut_startAddrCnGroups[8]>>4)+(j*bitOffsetInGroup));
 	  fprintf(fd,"            }\n");
        }
    }
--- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc384_simd512_byte.c
+++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc384_simd512_byte.c
@@ -0,0 +1,151 @@
+#include "PHY/sse_intrin.h"
+// generated code for Zc=384, byte encoding
+static inline void ldpc384_byte(uint8_t *c,uint8_t *d) {
+  __m512i *csimd=(__m512i *)c,*dsimd=(__m512i *)d;
+
+  __m512i *c2,*d2;
+
+  int i2;
+  for (i2=0; i2<6; i2++) {
+     c2=&csimd[i2];
+     d2=&dsimd[i2];
+
+//row: 0
+     d2[0]=_mm512_xor_si512(c2[13468],_mm512_xor_si512(c2[3169],_mm512_xor_si512(c2[3435],_mm512_xor_si512(c2[5284],_mm512_xor_si512(c2[5028],_mm512_xor_si512(c2[15327],_mm512_xor_si512(c2[6085],_mm512_xor_si512(c2[13224],_mm512_xor_si512(c2[3193],_mm512_xor_si512(c2[2141],_mm512_xor_si512(c2[12977],_mm512_xor_si512(c2[2413],_mm512_xor_si512(c2[36],_mm512_xor_si512(c2[8500],_mm512_xor_si512(c2[3221],_mm512_xor_si512(c2[5068],_mm512_xor_si512(c2[14054],_mm512_xor_si512(c2[4286],_mm512_xor_si512(c2[64],_mm512_xor_si512(c2[6411],_mm512_xor_si512(c2[8786],_mm512_xor_si512(c2[1923],_mm512_xor_si512(c2[2993],_mm512_xor_si512(c2[2992],_mm512_xor_si512(c2[6686],_mm512_xor_si512(c2[3005],_mm512_xor_si512(c2[8546],_mm512_xor_si512(c2[14880],_mm512_xor_si512(c2[16216],_mm512_xor_si512(c2[13310],_mm512_xor_si512(c2[16740],_mm512_xor_si512(c2[8572],_mm512_xor_si512(c2[386],_mm512_xor_si512(c2[1178],_mm512_xor_si512(c2[12013],_mm512_xor_si512(c2[10432],_mm512_xor_si512(c2[13072],_mm512_xor_si512(c2[4632],_mm512_xor_si512(c2[5957],_mm512_xor_si512(c2[10443],_mm512_xor_si512(c2[9929],_mm512_xor_si512(c2[2271],_mm512_xor_si512(c2[5705],_mm512_xor_si512(c2[6771],_mm512_xor_si512(c2[6505],_mm512_xor_si512(c2[5451],_mm512_xor_si512(c2[6255],_mm512_xor_si512(c2[9421],_mm512_xor_si512(c2[14172],_mm512_xor_si512(c2[11281],_mm512_xor_si512(c2[9173],_mm512_xor_si512(c2[12868],_mm512_xor_si512(c2[13405],_mm512_xor_si512(c2[998],_mm512_xor_si512(c2[11824],_mm512_xor_si512(c2[13419],_mm512_xor_si512(c2[12891],_mm512_xor_si512(c2[4180],_mm512_xor_si512(c2[13958],_mm512_xor_si512(c2[3137],_mm512_xor_si512(c2[3663],_mm512_xor_si512(c2[2885],_mm512_xor_si512(c2[3672],_mm512_xor_si512(c2[10536],_mm512_xor_si512(c2[7121],_mm512_xor_si512(c2[12925],c2[10025]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 1
+     d2[6]=_mm512_xor_si512(c2[13468],_mm512_xor_si512(c2[13732],_mm512_xor_si512(c2[3433],_mm512_xor_si512(c2[3699],_mm512_xor_si512(c2[5548],_mm512_xor_si512(c2[5028],_mm512_xor_si512(c2[5292],_mm512_xor_si512(c2[15591],_mm512_xor_si512(c2[6349],_mm512_xor_si512(c2[13224],_mm512_xor_si512(c2[13488],_mm512_xor_si512(c2[3457],_mm512_xor_si512(c2[2405],_mm512_xor_si512(c2[12977],_mm512_xor_si512(c2[13241],_mm512_xor_si512(c2[2677],_mm512_xor_si512(c2[300],_mm512_xor_si512(c2[8764],_mm512_xor_si512(c2[3485],_mm512_xor_si512(c2[5332],_mm512_xor_si512(c2[14054],_mm512_xor_si512(c2[14318],_mm512_xor_si512(c2[4550],_mm512_xor_si512(c2[328],_mm512_xor_si512(c2[6411],_mm512_xor_si512(c2[6675],_mm512_xor_si512(c2[9050],_mm512_xor_si512(c2[2187],_mm512_xor_si512(c2[3257],_mm512_xor_si512(c2[3256],_mm512_xor_si512(c2[6950],_mm512_xor_si512(c2[3269],_mm512_xor_si512(c2[8810],_mm512_xor_si512(c2[15144],_mm512_xor_si512(c2[16216],_mm512_xor_si512(c2[16480],_mm512_xor_si512(c2[13574],_mm512_xor_si512(c2[109],_mm512_xor_si512(c2[8572],_mm512_xor_si512(c2[8836],_mm512_xor_si512(c2[650],_mm512_xor_si512(c2[1442],_mm512_xor_si512(c2[12013],_mm512_xor_si512(c2[12277],_mm512_xor_si512(c2[10696],_mm512_xor_si512(c2[13336],_mm512_xor_si512(c2[4632],_mm512_xor_si512(c2[4896],_mm512_xor_si512(c2[6221],_mm512_xor_si512(c2[10707],_mm512_xor_si512(c2[9929],_mm512_xor_si512(c2[10193],_mm512_xor_si512(c2[2535],_mm512_xor_si512(c2[5969],_mm512_xor_si512(c2[7035],_mm512_xor_si512(c2[6769],_mm512_xor_si512(c2[5715],_mm512_xor_si512(c2[6255],_mm512_xor_si512(c2[6519],_mm512_xor_si512(c2[9685],_mm512_xor_si512(c2[14436],_mm512_xor_si512(c2[11281],_mm512_xor_si512(c2[11545],_mm512_xor_si512(c2[9437],_mm512_xor_si512(c2[13132],_mm512_xor_si512(c2[13669],_mm512_xor_si512(c2[1262],_mm512_xor_si512(c2[12088],_mm512_xor_si512(c2[13419],_mm512_xor_si512(c2[13683],_mm512_xor_si512(c2[13155],_mm512_xor_si512(c2[4444],_mm512_xor_si512(c2[13958],_mm512_xor_si512(c2[14222],_mm512_xor_si512(c2[3401],_mm512_xor_si512(c2[3927],_mm512_xor_si512(c2[2885],_mm512_xor_si512(c2[3149],_mm512_xor_si512(c2[3936],_mm512_xor_si512(c2[10800],_mm512_xor_si512(c2[7121],_mm512_xor_si512(c2[7385],_mm512_xor_si512(c2[13189],c2[10289])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 2
+     d2[12]=_mm512_xor_si512(c2[13732],_mm512_xor_si512(c2[3433],_mm512_xor_si512(c2[3435],_mm512_xor_si512(c2[3699],_mm512_xor_si512(c2[5284],_mm512_xor_si512(c2[5548],_mm512_xor_si512(c2[5292],_mm512_xor_si512(c2[15327],_mm512_xor_si512(c2[15591],_mm512_xor_si512(c2[6085],_mm512_xor_si512(c2[6349],_mm512_xor_si512(c2[13488],_mm512_xor_si512(c2[3457],_mm512_xor_si512(c2[2141],_mm512_xor_si512(c2[2405],_mm512_xor_si512(c2[13241],_mm512_xor_si512(c2[2677],_mm512_xor_si512(c2[36],_mm512_xor_si512(c2[300],_mm512_xor_si512(c2[8764],_mm512_xor_si512(c2[3221],_mm512_xor_si512(c2[3485],_mm512_xor_si512(c2[5068],_mm512_xor_si512(c2[5332],_mm512_xor_si512(c2[14318],_mm512_xor_si512(c2[4550],_mm512_xor_si512(c2[64],_mm512_xor_si512(c2[328],_mm512_xor_si512(c2[6675],_mm512_xor_si512(c2[8786],_mm512_xor_si512(c2[9050],_mm512_xor_si512(c2[1923],_mm512_xor_si512(c2[2187],_mm512_xor_si512(c2[3257],_mm512_xor_si512(c2[2992],_mm512_xor_si512(c2[3256],_mm512_xor_si512(c2[6686],_mm512_xor_si512(c2[6950],_mm512_xor_si512(c2[3269],_mm512_xor_si512(c2[8546],_mm512_xor_si512(c2[8810],_mm512_xor_si512(c2[14880],_mm512_xor_si512(c2[15144],_mm512_xor_si512(c2[16480],_mm512_xor_si512(c2[13574],_mm512_xor_si512(c2[16740],_mm512_xor_si512(c2[109],_mm512_xor_si512(c2[8836],_mm512_xor_si512(c2[386],_mm512_xor_si512(c2[650],_mm512_xor_si512(c2[1178],_mm512_xor_si512(c2[1442],_mm512_xor_si512(c2[12277],_mm512_xor_si512(c2[10696],_mm512_xor_si512(c2[13072],_mm512_xor_si512(c2[13336],_mm512_xor_si512(c2[4896],_mm512_xor_si512(c2[6221],_mm512_xor_si512(c2[10443],_mm512_xor_si512(c2[10707],_mm512_xor_si512(c2[10193],_mm512_xor_si512(c2[2271],_mm512_xor_si512(c2[2535],_mm512_xor_si512(c2[5705],_mm512_xor_si512(c2[5969],_mm512_xor_si512(c2[7035],_mm512_xor_si512(c2[6505],_mm512_xor_si512(c2[6769],_mm512_xor_si512(c2[5451],_mm512_xor_si512(c2[5715],_mm512_xor_si512(c2[6519],_mm512_xor_si512(c2[9685],_mm512_xor_si512(c2[14172],_mm512_xor_si512(c2[14436],_mm512_xor_si512(c2[11545],_mm512_xor_si512(c2[9437],_mm512_xor_si512(c2[12868],_mm512_xor_si512(c2[13132],_mm512_xor_si512(c2[13669],_mm512_xor_si512(c2[998],_mm512_xor_si512(c2[1262],_mm512_xor_si512(c2[11824],_mm512_xor_si512(c2[12088],_mm512_xor_si512(c2[13683],_mm512_xor_si512(c2[12891],_mm512_xor_si512(c2[13155],_mm512_xor_si512(c2[4180],_mm512_xor_si512(c2[4444],_mm512_xor_si512(c2[14222],_mm512_xor_si512(c2[3401],_mm512_xor_si512(c2[3663],_mm512_xor_si512(c2[3927],_mm512_xor_si512(c2[3149],_mm512_xor_si512(c2[3672],_mm512_xor_si512(c2[3936],_mm512_xor_si512(c2[10536],_mm512_xor_si512(c2[10800],_mm512_xor_si512(c2[7385],_mm512_xor_si512(c2[13189],_mm512_xor_si512(c2[10025],c2[10289]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 3
+     d2[18]=_mm512_xor_si512(c2[13732],_mm512_xor_si512(c2[3433],_mm512_xor_si512(c2[3699],_mm512_xor_si512(c2[5284],_mm512_xor_si512(c2[5548],_mm512_xor_si512(c2[5292],_mm512_xor_si512(c2[15591],_mm512_xor_si512(c2[6085],_mm512_xor_si512(c2[6349],_mm512_xor_si512(c2[13488],_mm512_xor_si512(c2[3457],_mm512_xor_si512(c2[2405],_mm512_xor_si512(c2[13241],_mm512_xor_si512(c2[2677],_mm512_xor_si512(c2[36],_mm512_xor_si512(c2[300],_mm512_xor_si512(c2[8764],_mm512_xor_si512(c2[3485],_mm512_xor_si512(c2[5068],_mm512_xor_si512(c2[5332],_mm512_xor_si512(c2[14318],_mm512_xor_si512(c2[4550],_mm512_xor_si512(c2[328],_mm512_xor_si512(c2[6675],_mm512_xor_si512(c2[9050],_mm512_xor_si512(c2[1923],_mm512_xor_si512(c2[2187],_mm512_xor_si512(c2[3257],_mm512_xor_si512(c2[3256],_mm512_xor_si512(c2[6686],_mm512_xor_si512(c2[6950],_mm512_xor_si512(c2[3269],_mm512_xor_si512(c2[8810],_mm512_xor_si512(c2[14880],_mm512_xor_si512(c2[15144],_mm512_xor_si512(c2[16480],_mm512_xor_si512(c2[13574],_mm512_xor_si512(c2[109],_mm512_xor_si512(c2[8836],_mm512_xor_si512(c2[650],_mm512_xor_si512(c2[1178],_mm512_xor_si512(c2[1442],_mm512_xor_si512(c2[12277],_mm512_xor_si512(c2[10696],_mm512_xor_si512(c2[13072],_mm512_xor_si512(c2[13336],_mm512_xor_si512(c2[4896],_mm512_xor_si512(c2[6221],_mm512_xor_si512(c2[10443],_mm512_xor_si512(c2[10707],_mm512_xor_si512(c2[10193],_mm512_xor_si512(c2[2535],_mm512_xor_si512(c2[5705],_mm512_xor_si512(c2[5969],_mm512_xor_si512(c2[7035],_mm512_xor_si512(c2[6769],_mm512_xor_si512(c2[5451],_mm512_xor_si512(c2[5715],_mm512_xor_si512(c2[6519],_mm512_xor_si512(c2[9685],_mm512_xor_si512(c2[14436],_mm512_xor_si512(c2[11545],_mm512_xor_si512(c2[9437],_mm512_xor_si512(c2[12868],_mm512_xor_si512(c2[13132],_mm512_xor_si512(c2[13669],_mm512_xor_si512(c2[1262],_mm512_xor_si512(c2[11824],_mm512_xor_si512(c2[12088],_mm512_xor_si512(c2[13683],_mm512_xor_si512(c2[13155],_mm512_xor_si512(c2[4180],_mm512_xor_si512(c2[4444],_mm512_xor_si512(c2[14222],_mm512_xor_si512(c2[3401],_mm512_xor_si512(c2[3927],_mm512_xor_si512(c2[3149],_mm512_xor_si512(c2[3936],_mm512_xor_si512(c2[10536],_mm512_xor_si512(c2[10800],_mm512_xor_si512(c2[7385],_mm512_xor_si512(c2[13189],_mm512_xor_si512(c2[10025],c2[10289])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 4
+     d2[24]=_mm512_xor_si512(c2[3173],c2[14006]);
+
+//row: 5
+     d2[30]=_mm512_xor_si512(c2[4225],_mm512_xor_si512(c2[10827],_mm512_xor_si512(c2[11093],_mm512_xor_si512(c2[12936],_mm512_xor_si512(c2[795],_mm512_xor_si512(c2[12686],_mm512_xor_si512(c2[6084],_mm512_xor_si512(c2[13743],_mm512_xor_si512(c2[3708],_mm512_xor_si512(c2[3987],_mm512_xor_si512(c2[10851],_mm512_xor_si512(c2[9793],_mm512_xor_si512(c2[3734],_mm512_xor_si512(c2[10071],_mm512_xor_si512(c2[7694],_mm512_xor_si512(c2[13501],_mm512_xor_si512(c2[16152],_mm512_xor_si512(c2[10873],_mm512_xor_si512(c2[12720],_mm512_xor_si512(c2[4817],_mm512_xor_si512(c2[11944],_mm512_xor_si512(c2[7716],_mm512_xor_si512(c2[14069],_mm512_xor_si512(c2[16444],_mm512_xor_si512(c2[9581],_mm512_xor_si512(c2[10645],_mm512_xor_si512(c2[10644],_mm512_xor_si512(c2[14344],_mm512_xor_si512(c2[10657],_mm512_xor_si512(c2[16204],_mm512_xor_si512(c2[5643],_mm512_xor_si512(c2[6973],_mm512_xor_si512(c2[4073],_mm512_xor_si512(c2[7503],_mm512_xor_si512(c2[16224],_mm512_xor_si512(c2[8044],_mm512_xor_si512(c2[8836],_mm512_xor_si512(c2[2776],_mm512_xor_si512(c2[1189],_mm512_xor_si512(c2[3829],_mm512_xor_si512(c2[12290],_mm512_xor_si512(c2[13609],_mm512_xor_si512(c2[1200],_mm512_xor_si512(c2[10178],_mm512_xor_si512(c2[686],_mm512_xor_si512(c2[9929],_mm512_xor_si512(c2[13357],_mm512_xor_si512(c2[14429],_mm512_xor_si512(c2[14163],_mm512_xor_si512(c2[13109],_mm512_xor_si512(c2[13913],_mm512_xor_si512(c2[184],_mm512_xor_si512(c2[4935],_mm512_xor_si512(c2[2044],_mm512_xor_si512(c2[16825],_mm512_xor_si512(c2[3625],_mm512_xor_si512(c2[13131],_mm512_xor_si512(c2[4168],_mm512_xor_si512(c2[8656],_mm512_xor_si512(c2[2581],_mm512_xor_si512(c2[4176],_mm512_xor_si512(c2[3648],_mm512_xor_si512(c2[11832],_mm512_xor_si512(c2[4721],_mm512_xor_si512(c2[10789],_mm512_xor_si512(c2[11321],_mm512_xor_si512(c2[10537],_mm512_xor_si512(c2[11330],_mm512_xor_si512(c2[1299],_mm512_xor_si512(c2[14773],_mm512_xor_si512(c2[3688],_mm512_xor_si512(c2[782],c2[13716]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 6
+     d2[36]=_mm512_xor_si512(c2[5812],_mm512_xor_si512(c2[340],_mm512_xor_si512(c2[384],_mm512_xor_si512(c2[8321],_mm512_xor_si512(c2[7549],_mm512_xor_si512(c2[16311],_mm512_xor_si512(c2[4968],c2[8955])))))));
+
+//row: 7
+     d2[42]=_mm512_xor_si512(c2[2376],_mm512_xor_si512(c2[16380],_mm512_xor_si512(c2[15892],_mm512_xor_si512(c2[3521],_mm512_xor_si512(c2[9076],c2[13369])))));
+
+//row: 8
+     d2[48]=_mm512_xor_si512(c2[14261],_mm512_xor_si512(c2[7921],_mm512_xor_si512(c2[3962],_mm512_xor_si512(c2[14523],_mm512_xor_si512(c2[4228],_mm512_xor_si512(c2[14525],_mm512_xor_si512(c2[14789],_mm512_xor_si512(c2[6077],_mm512_xor_si512(c2[16368],_mm512_xor_si512(c2[16632],_mm512_xor_si512(c2[13468],_mm512_xor_si512(c2[5821],_mm512_xor_si512(c2[16382],_mm512_xor_si512(c2[16120],_mm512_xor_si512(c2[9516],_mm512_xor_si512(c2[9780],_mm512_xor_si512(c2[6878],_mm512_xor_si512(c2[280],_mm512_xor_si512(c2[544],_mm512_xor_si512(c2[13478],_mm512_xor_si512(c2[14017],_mm512_xor_si512(c2[7683],_mm512_xor_si512(c2[3986],_mm512_xor_si512(c2[14547],_mm512_xor_si512(c2[2928],_mm512_xor_si512(c2[13225],_mm512_xor_si512(c2[13489],_mm512_xor_si512(c2[13764],_mm512_xor_si512(c2[7430],_mm512_xor_si512(c2[3206],_mm512_xor_si512(c2[13767],_mm512_xor_si512(c2[829],_mm512_xor_si512(c2[11126],_mm512_xor_si512(c2[11390],_mm512_xor_si512(c2[9806],_mm512_xor_si512(c2[9293],_mm512_xor_si512(c2[2953],_mm512_xor_si512(c2[4008],_mm512_xor_si512(c2[14305],_mm512_xor_si512(c2[14569],_mm512_xor_si512(c2[5861],_mm512_xor_si512(c2[16152],_mm512_xor_si512(c2[16416],_mm512_xor_si512(c2[14847],_mm512_xor_si512(c2[8513],_mm512_xor_si512(c2[5079],_mm512_xor_si512(c2[15640],_mm512_xor_si512(c2[857],_mm512_xor_si512(c2[11148],_mm512_xor_si512(c2[11412],_mm512_xor_si512(c2[7204],_mm512_xor_si512(c2[864],_mm512_xor_si512(c2[9579],_mm512_xor_si512(c2[2981],_mm512_xor_si512(c2[3245],_mm512_xor_si512(c2[2716],_mm512_xor_si512(c2[13013],_mm512_xor_si512(c2[13277],_mm512_xor_si512(c2[3780],_mm512_xor_si512(c2[14341],_mm512_xor_si512(c2[3785],_mm512_xor_si512(c2[14076],_mm512_xor_si512(c2[14340],_mm512_xor_si512(c2[7479],_mm512_xor_si512(c2[881],_mm512_xor_si512(c2[1145],_mm512_xor_si512(c2[3792],_mm512_xor_si512(c2[14353],_mm512_xor_si512(c2[9339],_mm512_xor_si512(c2[2741],_mm512_xor_si512(c2[3005],_mm512_xor_si512(c2[15673],_mm512_xor_si512(c2[9075],_mm512_xor_si512(c2[9339],_mm512_xor_si512(c2[108],_mm512_xor_si512(c2[10669],_mm512_xor_si512(c2[14103],_mm512_xor_si512(c2[7769],_mm512_xor_si512(c2[638],_mm512_xor_si512(c2[10935],_mm512_xor_si512(c2[11199],_mm512_xor_si512(c2[9365],_mm512_xor_si512(c2[3025],_mm512_xor_si512(c2[1179],_mm512_xor_si512(c2[11476],_mm512_xor_si512(c2[11740],_mm512_xor_si512(c2[1971],_mm512_xor_si512(c2[12268],_mm512_xor_si512(c2[12532],_mm512_xor_si512(c2[12806],_mm512_xor_si512(c2[6472],_mm512_xor_si512(c2[11225],_mm512_xor_si512(c2[4885],_mm512_xor_si512(c2[13865],_mm512_xor_si512(c2[7261],_mm512_xor_si512(c2[7525],_mm512_xor_si512(c2[5425],_mm512_xor_si512(c2[15986],_mm512_xor_si512(c2[6744],_mm512_xor_si512(c2[410],_mm512_xor_si512(c2[11236],_mm512_xor_si512(c2[4632],_mm512_xor_si512(c2[4896],_mm512_xor_si512(c2[4896],_mm512_xor_si512(c2[10716],_mm512_xor_si512(c2[4382],_mm512_xor_si512(c2[3064],_mm512_xor_si512(c2[13361],_mm512_xor_si512(c2[13625],_mm512_xor_si512(c2[6492],_mm512_xor_si512(c2[16789],_mm512_xor_si512(c2[158],_mm512_xor_si512(c2[7564],_mm512_xor_si512(c2[1224],_mm512_xor_si512(c2[7298],_mm512_xor_si512(c2[700],_mm512_xor_si512(c2[964],_mm512_xor_si512(c2[6244],_mm512_xor_si512(c2[16541],_mm512_xor_si512(c2[16805],_mm512_xor_si512(c2[7048],_mm512_xor_si512(c2[708],_mm512_xor_si512(c2[10214],_mm512_xor_si512(c2[3880],_mm512_xor_si512(c2[14965],_mm512_xor_si512(c2[8367],_mm512_xor_si512(c2[8631],_mm512_xor_si512(c2[12074],_mm512_xor_si512(c2[5740],_mm512_xor_si512(c2[9960],_mm512_xor_si512(c2[3626],_mm512_xor_si512(c2[13661],_mm512_xor_si512(c2[7057],_mm512_xor_si512(c2[7321],_mm512_xor_si512(c2[10488],_mm512_xor_si512(c2[14198],_mm512_xor_si512(c2[7864],_mm512_xor_si512(c2[1791],_mm512_xor_si512(c2[12088],_mm512_xor_si512(c2[12352],_mm512_xor_si512(c2[12617],_mm512_xor_si512(c2[6013],_mm512_xor_si512(c2[6277],_mm512_xor_si512(c2[14212],_mm512_xor_si512(c2[7872],_mm512_xor_si512(c2[13684],_mm512_xor_si512(c2[7080],_mm512_xor_si512(c2[7344],_mm512_xor_si512(c2[4973],_mm512_xor_si512(c2[15264],_mm512_xor_si512(c2[15528],_mm512_xor_si512(c2[14751],_mm512_xor_si512(c2[8417],_mm512_xor_si512(c2[3924],_mm512_xor_si512(c2[14485],_mm512_xor_si512(c2[4456],_mm512_xor_si512(c2[14753],_mm512_xor_si512(c2[15017],_mm512_xor_si512(c2[8679],_mm512_xor_si512(c2[3672],_mm512_xor_si512(c2[14233],_mm512_xor_si512(c2[4465],_mm512_xor_si512(c2[14762],_mm512_xor_si512(c2[15026],_mm512_xor_si512(c2[11329],_mm512_xor_si512(c2[4731],_mm512_xor_si512(c2[4995],_mm512_xor_si512(c2[7908],_mm512_xor_si512(c2[1574],_mm512_xor_si512(c2[13718],_mm512_xor_si512(c2[7384],_mm512_xor_si512(c2[10812],_mm512_xor_si512(c2[4214],_mm512_xor_si512(c2[4478],c2[12929]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 9
+     d2[54]=_mm512_xor_si512(c2[12149],_mm512_xor_si512(c2[10575],_mm512_xor_si512(c2[389],_mm512_xor_si512(c2[1454],_mm512_xor_si512(c2[15204],_mm512_xor_si512(c2[12616],_mm512_xor_si512(c2[16848],c2[4993])))))));
+
+//row: 10
+     d2[60]=_mm512_xor_si512(c2[9781],_mm512_xor_si512(c2[5045],_mm512_xor_si512(c2[4804],_mm512_xor_si512(c2[12493],_mm512_xor_si512(c2[16733],c2[9149])))));
+
+//row: 11
+     d2[66]=_mm512_xor_si512(c2[268],_mm512_xor_si512(c2[10032],_mm512_xor_si512(c2[10296],_mm512_xor_si512(c2[6864],_mm512_xor_si512(c2[3],_mm512_xor_si512(c2[7130],_mm512_xor_si512(c2[269],_mm512_xor_si512(c2[8979],_mm512_xor_si512(c2[2112],_mm512_xor_si512(c2[12672],_mm512_xor_si512(c2[8729],_mm512_xor_si512(c2[1598],_mm512_xor_si512(c2[1862],_mm512_xor_si512(c2[2127],_mm512_xor_si512(c2[12161],_mm512_xor_si512(c2[9780],_mm512_xor_si512(c2[2919],_mm512_xor_si512(c2[10045],_mm512_xor_si512(c2[24],_mm512_xor_si512(c2[9794],_mm512_xor_si512(c2[10058],_mm512_xor_si512(c2[6888],_mm512_xor_si512(c2[27],_mm512_xor_si512(c2[5836],_mm512_xor_si512(c2[15864],_mm512_xor_si512(c2[16672],_mm512_xor_si512(c2[9541],_mm512_xor_si512(c2[9805],_mm512_xor_si512(c2[6108],_mm512_xor_si512(c2[16142],_mm512_xor_si512(c2[3737],_mm512_xor_si512(c2[13765],_mm512_xor_si512(c2[12195],_mm512_xor_si512(c2[5328],_mm512_xor_si512(c2[6916],_mm512_xor_si512(c2[49],_mm512_xor_si512(c2[8763],_mm512_xor_si512(c2[1896],_mm512_xor_si512(c2[854],_mm512_xor_si512(c2[10624],_mm512_xor_si512(c2[10888],_mm512_xor_si512(c2[7981],_mm512_xor_si512(c2[1120],_mm512_xor_si512(c2[3759],_mm512_xor_si512(c2[13793],_mm512_xor_si512(c2[10106],_mm512_xor_si512(c2[2981],_mm512_xor_si512(c2[3245],_mm512_xor_si512(c2[12481],_mm512_xor_si512(c2[5620],_mm512_xor_si512(c2[5618],_mm512_xor_si512(c2[15652],_mm512_xor_si512(c2[6688],_mm512_xor_si512(c2[16716],_mm512_xor_si512(c2[6687],_mm512_xor_si512(c2[16721],_mm512_xor_si512(c2[10381],_mm512_xor_si512(c2[3520],_mm512_xor_si512(c2[6700],_mm512_xor_si512(c2[16728],_mm512_xor_si512(c2[12241],_mm512_xor_si512(c2[5380],_mm512_xor_si512(c2[1680],_mm512_xor_si512(c2[11714],_mm512_xor_si512(c2[3016],_mm512_xor_si512(c2[12780],_mm512_xor_si512(c2[13044],_mm512_xor_si512(c2[110],_mm512_xor_si512(c2[10144],_mm512_xor_si512(c2[3540],_mm512_xor_si512(c2[13574],_mm512_xor_si512(c2[12267],_mm512_xor_si512(c2[5136],_mm512_xor_si512(c2[5400],_mm512_xor_si512(c2[4081],_mm512_xor_si512(c2[14115],_mm512_xor_si512(c2[4873],_mm512_xor_si512(c2[14907],_mm512_xor_si512(c2[15708],_mm512_xor_si512(c2[8583],_mm512_xor_si512(c2[8847],_mm512_xor_si512(c2[14127],_mm512_xor_si512(c2[7260],_mm512_xor_si512(c2[16767],_mm512_xor_si512(c2[9900],_mm512_xor_si512(c2[8333],_mm512_xor_si512(c2[1202],_mm512_xor_si512(c2[1466],_mm512_xor_si512(c2[9652],_mm512_xor_si512(c2[2785],_mm512_xor_si512(c2[14138],_mm512_xor_si512(c2[7277],_mm512_xor_si512(c2[2256],_mm512_xor_si512(c2[13624],_mm512_xor_si512(c2[6493],_mm512_xor_si512(c2[6757],_mm512_xor_si512(c2[5966],_mm512_xor_si512(c2[16000],_mm512_xor_si512(c2[9400],_mm512_xor_si512(c2[2533],_mm512_xor_si512(c2[10466],_mm512_xor_si512(c2[3605],_mm512_xor_si512(c2[10200],_mm512_xor_si512(c2[3339],_mm512_xor_si512(c2[9146],_mm512_xor_si512(c2[2285],_mm512_xor_si512(c2[9950],_mm512_xor_si512(c2[2825],_mm512_xor_si512(c2[3089],_mm512_xor_si512(c2[13116],_mm512_xor_si512(c2[6255],_mm512_xor_si512(c2[972],_mm512_xor_si512(c2[11006],_mm512_xor_si512(c2[14976],_mm512_xor_si512(c2[7851],_mm512_xor_si512(c2[8115],_mm512_xor_si512(c2[12868],_mm512_xor_si512(c2[6001],_mm512_xor_si512(c2[16563],_mm512_xor_si512(c2[9696],_mm512_xor_si512(c2[12600],_mm512_xor_si512(c2[205],_mm512_xor_si512(c2[10239],_mm512_xor_si512(c2[4693],_mm512_xor_si512(c2[14727],_mm512_xor_si512(c2[15519],_mm512_xor_si512(c2[8652],_mm512_xor_si512(c2[219],_mm512_xor_si512(c2[9989],_mm512_xor_si512(c2[10253],_mm512_xor_si512(c2[16586],_mm512_xor_si512(c2[9725],_mm512_xor_si512(c2[7875],_mm512_xor_si512(c2[1008],_mm512_xor_si512(c2[758],_mm512_xor_si512(c2[10528],_mm512_xor_si512(c2[10792],_mm512_xor_si512(c2[6832],_mm512_xor_si512(c2[16860],_mm512_xor_si512(c2[7358],_mm512_xor_si512(c2[497],_mm512_xor_si512(c2[6580],_mm512_xor_si512(c2[16344],_mm512_xor_si512(c2[16608],_mm512_xor_si512(c2[7373],_mm512_xor_si512(c2[506],_mm512_xor_si512(c2[14237],_mm512_xor_si512(c2[7370],_mm512_xor_si512(c2[10816],_mm512_xor_si512(c2[3685],_mm512_xor_si512(c2[3949],_mm512_xor_si512(c2[16620],_mm512_xor_si512(c2[9759],_mm512_xor_si512(c2[13720],_mm512_xor_si512(c2[6853],c2[16094])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 12
+     d2[72]=_mm512_xor_si512(c2[3433],_mm512_xor_si512(c2[15326],_mm512_xor_si512(c2[12266],_mm512_xor_si512(c2[10695],_mm512_xor_si512(c2[13356],c2[2857])))));
+
+//row: 13
+     d2[78]=_mm512_xor_si512(c2[9772],_mm512_xor_si512(c2[10036],_mm512_xor_si512(c2[16632],_mm512_xor_si512(c2[3],_mm512_xor_si512(c2[1852],_mm512_xor_si512(c2[15052],_mm512_xor_si512(c2[1332],_mm512_xor_si512(c2[1596],_mm512_xor_si512(c2[11895],_mm512_xor_si512(c2[2653],_mm512_xor_si512(c2[9528],_mm512_xor_si512(c2[9792],_mm512_xor_si512(c2[16656],_mm512_xor_si512(c2[15604],_mm512_xor_si512(c2[9281],_mm512_xor_si512(c2[9545],_mm512_xor_si512(c2[15876],_mm512_xor_si512(c2[13505],_mm512_xor_si512(c2[12974],_mm512_xor_si512(c2[5068],_mm512_xor_si512(c2[16684],_mm512_xor_si512(c2[1636],_mm512_xor_si512(c2[10358],_mm512_xor_si512(c2[10622],_mm512_xor_si512(c2[854],_mm512_xor_si512(c2[13527],_mm512_xor_si512(c2[2715],_mm512_xor_si512(c2[2979],_mm512_xor_si512(c2[5354],_mm512_xor_si512(c2[15386],_mm512_xor_si512(c2[16456],_mm512_xor_si512(c2[16455],_mm512_xor_si512(c2[3254],_mm512_xor_si512(c2[2728],_mm512_xor_si512(c2[16468],_mm512_xor_si512(c2[5114],_mm512_xor_si512(c2[11448],_mm512_xor_si512(c2[12520],_mm512_xor_si512(c2[12784],_mm512_xor_si512(c2[9878],_mm512_xor_si512(c2[13308],_mm512_xor_si512(c2[4876],_mm512_xor_si512(c2[5140],_mm512_xor_si512(c2[13849],_mm512_xor_si512(c2[14641],_mm512_xor_si512(c2[8317],_mm512_xor_si512(c2[8581],_mm512_xor_si512(c2[7000],_mm512_xor_si512(c2[9640],_mm512_xor_si512(c2[936],_mm512_xor_si512(c2[1200],_mm512_xor_si512(c2[2525],_mm512_xor_si512(c2[7011],_mm512_xor_si512(c2[6233],_mm512_xor_si512(c2[6497],_mm512_xor_si512(c2[15734],_mm512_xor_si512(c2[2273],_mm512_xor_si512(c2[3339],_mm512_xor_si512(c2[3073],_mm512_xor_si512(c2[2019],_mm512_xor_si512(c2[2559],_mm512_xor_si512(c2[2823],_mm512_xor_si512(c2[5989],_mm512_xor_si512(c2[10740],_mm512_xor_si512(c2[7585],_mm512_xor_si512(c2[7849],_mm512_xor_si512(c2[5741],_mm512_xor_si512(c2[9436],_mm512_xor_si512(c2[9973],_mm512_xor_si512(c2[14461],_mm512_xor_si512(c2[8392],_mm512_xor_si512(c2[9723],_mm512_xor_si512(c2[9987],_mm512_xor_si512(c2[9459],_mm512_xor_si512(c2[748],_mm512_xor_si512(c2[10262],_mm512_xor_si512(c2[10526],_mm512_xor_si512(c2[16600],_mm512_xor_si512(c2[231],_mm512_xor_si512(c2[16084],_mm512_xor_si512(c2[16348],_mm512_xor_si512(c2[240],_mm512_xor_si512(c2[7104],_mm512_xor_si512(c2[13705],_mm512_xor_si512(c2[3425],_mm512_xor_si512(c2[3689],_mm512_xor_si512(c2[9493],c2[6593])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 14
+     d2[84]=_mm512_xor_si512(c2[3698],_mm512_xor_si512(c2[14931],_mm512_xor_si512(c2[2558],_mm512_xor_si512(c2[6793],_mm512_xor_si512(c2[7337],c2[3420])))));
+
+//row: 15
+     d2[90]=_mm512_xor_si512(c2[267],_mm512_xor_si512(c2[6869],_mm512_xor_si512(c2[7129],_mm512_xor_si512(c2[8714],_mm512_xor_si512(c2[8978],_mm512_xor_si512(c2[12939],_mm512_xor_si512(c2[8728],_mm512_xor_si512(c2[2126],_mm512_xor_si512(c2[9521],_mm512_xor_si512(c2[9785],_mm512_xor_si512(c2[540],_mm512_xor_si512(c2[29],_mm512_xor_si512(c2[6893],_mm512_xor_si512(c2[5835],_mm512_xor_si512(c2[16671],_mm512_xor_si512(c2[6113],_mm512_xor_si512(c2[3472],_mm512_xor_si512(c2[3736],_mm512_xor_si512(c2[12194],_mm512_xor_si512(c2[6915],_mm512_xor_si512(c2[8498],_mm512_xor_si512(c2[8762],_mm512_xor_si512(c2[853],_mm512_xor_si512(c2[7980],_mm512_xor_si512(c2[3758],_mm512_xor_si512(c2[10105],_mm512_xor_si512(c2[12480],_mm512_xor_si512(c2[5353],_mm512_xor_si512(c2[5617],_mm512_xor_si512(c2[6687],_mm512_xor_si512(c2[6686],_mm512_xor_si512(c2[10116],_mm512_xor_si512(c2[10380],_mm512_xor_si512(c2[6699],_mm512_xor_si512(c2[12240],_mm512_xor_si512(c2[1421],_mm512_xor_si512(c2[1685],_mm512_xor_si512(c2[3015],_mm512_xor_si512(c2[109],_mm512_xor_si512(c2[3545],_mm512_xor_si512(c2[12266],_mm512_xor_si512(c2[4080],_mm512_xor_si512(c2[4608],_mm512_xor_si512(c2[4872],_mm512_xor_si512(c2[4875],_mm512_xor_si512(c2[15713],_mm512_xor_si512(c2[14126],_mm512_xor_si512(c2[16502],_mm512_xor_si512(c2[16766],_mm512_xor_si512(c2[8332],_mm512_xor_si512(c2[9651],_mm512_xor_si512(c2[13873],_mm512_xor_si512(c2[14137],_mm512_xor_si512(c2[13623],_mm512_xor_si512(c2[5965],_mm512_xor_si512(c2[9135],_mm512_xor_si512(c2[9399],_mm512_xor_si512(c2[16528],_mm512_xor_si512(c2[10465],_mm512_xor_si512(c2[10205],_mm512_xor_si512(c2[8881],_mm512_xor_si512(c2[9145],_mm512_xor_si512(c2[9949],_mm512_xor_si512(c2[13121],_mm512_xor_si512(c2[977],_mm512_xor_si512(c2[14981],_mm512_xor_si512(c2[12867],_mm512_xor_si512(c2[16298],_mm512_xor_si512(c2[16562],_mm512_xor_si512(c2[204],_mm512_xor_si512(c2[4692],_mm512_xor_si512(c2[15254],_mm512_xor_si512(c2[15518],_mm512_xor_si512(c2[218],_mm512_xor_si512(c2[16585],_mm512_xor_si512(c2[7610],_mm512_xor_si512(c2[7874],_mm512_xor_si512(c2[14736],_mm512_xor_si512(c2[757],_mm512_xor_si512(c2[6831],_mm512_xor_si512(c2[7357],_mm512_xor_si512(c2[6579],_mm512_xor_si512(c2[7372],_mm512_xor_si512(c2[13972],_mm512_xor_si512(c2[14236],_mm512_xor_si512(c2[10815],_mm512_xor_si512(c2[16625],_mm512_xor_si512(c2[13455],c2[13719]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 16
+     d2[96]=_mm512_xor_si512(c2[11621],_mm512_xor_si512(c2[1322],_mm512_xor_si512(c2[1588],_mm512_xor_si512(c2[3437],_mm512_xor_si512(c2[3181],_mm512_xor_si512(c2[13480],_mm512_xor_si512(c2[4238],_mm512_xor_si512(c2[3444],_mm512_xor_si512(c2[11377],_mm512_xor_si512(c2[1346],_mm512_xor_si512(c2[288],_mm512_xor_si512(c2[11124],_mm512_xor_si512(c2[566],_mm512_xor_si512(c2[15084],_mm512_xor_si512(c2[4793],_mm512_xor_si512(c2[6653],_mm512_xor_si512(c2[1368],_mm512_xor_si512(c2[3221],_mm512_xor_si512(c2[12207],_mm512_xor_si512(c2[2439],_mm512_xor_si512(c2[15112],_mm512_xor_si512(c2[4564],_mm512_xor_si512(c2[6939],_mm512_xor_si512(c2[76],_mm512_xor_si512(c2[1140],_mm512_xor_si512(c2[1145],_mm512_xor_si512(c2[4839],_mm512_xor_si512(c2[1152],_mm512_xor_si512(c2[6699],_mm512_xor_si512(c2[13033],_mm512_xor_si512(c2[14369],_mm512_xor_si512(c2[11463],_mm512_xor_si512(c2[14893],_mm512_xor_si512(c2[6725],_mm512_xor_si512(c2[15434],_mm512_xor_si512(c2[16226],_mm512_xor_si512(c2[10166],_mm512_xor_si512(c2[8585],_mm512_xor_si512(c2[11225],_mm512_xor_si512(c2[15180],_mm512_xor_si512(c2[2785],_mm512_xor_si512(c2[4104],_mm512_xor_si512(c2[8596],_mm512_xor_si512(c2[8076],_mm512_xor_si512(c2[424],_mm512_xor_si512(c2[3852],_mm512_xor_si512(c2[4924],_mm512_xor_si512(c2[4658],_mm512_xor_si512(c2[3604],_mm512_xor_si512(c2[4408],_mm512_xor_si512(c2[7574],_mm512_xor_si512(c2[12325],_mm512_xor_si512(c2[9434],_mm512_xor_si512(c2[7320],_mm512_xor_si512(c2[11021],_mm512_xor_si512(c2[11558],_mm512_xor_si512(c2[16046],_mm512_xor_si512(c2[9977],_mm512_xor_si512(c2[11572],_mm512_xor_si512(c2[11044],_mm512_xor_si512(c2[2333],_mm512_xor_si512(c2[12111],_mm512_xor_si512(c2[1284],_mm512_xor_si512(c2[1816],_mm512_xor_si512(c2[1032],_mm512_xor_si512(c2[1825],_mm512_xor_si512(c2[8689],_mm512_xor_si512(c2[8956],_mm512_xor_si512(c2[5268],_mm512_xor_si512(c2[11078],c2[8172]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 17
+     d2[102]=_mm512_xor_si512(c2[1060],_mm512_xor_si512(c2[12580],_mm512_xor_si512(c2[4681],_mm512_xor_si512(c2[10241],c2[14777]))));
+
+//row: 18
+     d2[108]=_mm512_xor_si512(c2[542],_mm512_xor_si512(c2[9386],_mm512_xor_si512(c2[6496],_mm512_xor_si512(c2[1274],c2[1284]))));
+
+//row: 19
+     d2[114]=_mm512_xor_si512(c2[4490],_mm512_xor_si512(c2[5559],_mm512_xor_si512(c2[6425],_mm512_xor_si512(c2[13299],c2[1443]))));
+
+//row: 20
+     d2[120]=_mm512_xor_si512(c2[12149],_mm512_xor_si512(c2[1850],_mm512_xor_si512(c2[2116],_mm512_xor_si512(c2[3965],_mm512_xor_si512(c2[15578],_mm512_xor_si512(c2[3709],_mm512_xor_si512(c2[14008],_mm512_xor_si512(c2[4766],_mm512_xor_si512(c2[11905],_mm512_xor_si512(c2[1874],_mm512_xor_si512(c2[816],_mm512_xor_si512(c2[11652],_mm512_xor_si512(c2[1094],_mm512_xor_si512(c2[15612],_mm512_xor_si512(c2[3735],_mm512_xor_si512(c2[7181],_mm512_xor_si512(c2[1896],_mm512_xor_si512(c2[3749],_mm512_xor_si512(c2[12735],_mm512_xor_si512(c2[2967],_mm512_xor_si512(c2[15640],_mm512_xor_si512(c2[5092],_mm512_xor_si512(c2[7467],_mm512_xor_si512(c2[604],_mm512_xor_si512(c2[1668],_mm512_xor_si512(c2[1673],_mm512_xor_si512(c2[5367],_mm512_xor_si512(c2[1680],_mm512_xor_si512(c2[7227],_mm512_xor_si512(c2[13561],_mm512_xor_si512(c2[14897],_mm512_xor_si512(c2[11991],_mm512_xor_si512(c2[15421],_mm512_xor_si512(c2[2224],_mm512_xor_si512(c2[7253],_mm512_xor_si512(c2[15962],_mm512_xor_si512(c2[16754],_mm512_xor_si512(c2[10694],_mm512_xor_si512(c2[9113],_mm512_xor_si512(c2[11753],_mm512_xor_si512(c2[5681],_mm512_xor_si512(c2[3313],_mm512_xor_si512(c2[4632],_mm512_xor_si512(c2[9124],_mm512_xor_si512(c2[8604],_mm512_xor_si512(c2[952],_mm512_xor_si512(c2[4380],_mm512_xor_si512(c2[5452],_mm512_xor_si512(c2[5186],_mm512_xor_si512(c2[4132],_mm512_xor_si512(c2[4936],_mm512_xor_si512(c2[8102],_mm512_xor_si512(c2[12853],_mm512_xor_si512(c2[9962],_mm512_xor_si512(c2[7848],_mm512_xor_si512(c2[11549],_mm512_xor_si512(c2[12086],_mm512_xor_si512(c2[16574],_mm512_xor_si512(c2[10505],_mm512_xor_si512(c2[12100],_mm512_xor_si512(c2[11572],_mm512_xor_si512(c2[2861],_mm512_xor_si512(c2[12639],_mm512_xor_si512(c2[1812],_mm512_xor_si512(c2[2344],_mm512_xor_si512(c2[1560],_mm512_xor_si512(c2[2353],_mm512_xor_si512(c2[9217],_mm512_xor_si512(c2[5796],_mm512_xor_si512(c2[11606],c2[8700]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 21
+     d2[126]=_mm512_xor_si512(c2[3447],_mm512_xor_si512(c2[10093],_mm512_xor_si512(c2[2309],_mm512_xor_si512(c2[5787],c2[8965]))));
+
+//row: 22
+     d2[132]=_mm512_xor_si512(c2[7920],_mm512_xor_si512(c2[3048],_mm512_xor_si512(c2[10983],c2[6012])));
+
+//row: 23
+     d2[138]=_mm512_xor_si512(c2[6348],_mm512_xor_si512(c2[6625],_mm512_xor_si512(c2[16224],c2[7344])));
+
+//row: 24
+     d2[144]=_mm512_xor_si512(c2[7658],_mm512_xor_si512(c2[14260],_mm512_xor_si512(c2[14520],_mm512_xor_si512(c2[16369],_mm512_xor_si512(c2[11092],_mm512_xor_si512(c2[16119],_mm512_xor_si512(c2[9517],_mm512_xor_si512(c2[281],_mm512_xor_si512(c2[7420],_mm512_xor_si512(c2[14284],_mm512_xor_si512(c2[13226],_mm512_xor_si512(c2[7167],_mm512_xor_si512(c2[13504],_mm512_xor_si512(c2[11127],_mm512_xor_si512(c2[7958],_mm512_xor_si512(c2[2690],_mm512_xor_si512(c2[14306],_mm512_xor_si512(c2[16153],_mm512_xor_si512(c2[11403],_mm512_xor_si512(c2[8244],_mm512_xor_si512(c2[15377],_mm512_xor_si512(c2[11149],_mm512_xor_si512(c2[601],_mm512_xor_si512(c2[2976],_mm512_xor_si512(c2[13008],_mm512_xor_si512(c2[14078],_mm512_xor_si512(c2[14077],_mm512_xor_si512(c2[876],_mm512_xor_si512(c2[14090],_mm512_xor_si512(c2[2736],_mm512_xor_si512(c2[9076],_mm512_xor_si512(c2[10406],_mm512_xor_si512(c2[7500],_mm512_xor_si512(c2[10936],_mm512_xor_si512(c2[2762],_mm512_xor_si512(c2[11477],_mm512_xor_si512(c2[12269],_mm512_xor_si512(c2[6209],_mm512_xor_si512(c2[4622],_mm512_xor_si512(c2[7262],_mm512_xor_si512(c2[5153],_mm512_xor_si512(c2[15723],_mm512_xor_si512(c2[147],_mm512_xor_si512(c2[4633],_mm512_xor_si512(c2[4119],_mm512_xor_si512(c2[13356],_mm512_xor_si512(c2[16790],_mm512_xor_si512(c2[961],_mm512_xor_si512(c2[701],_mm512_xor_si512(c2[16536],_mm512_xor_si512(c2[445],_mm512_xor_si512(c2[3617],_mm512_xor_si512(c2[8368],_mm512_xor_si512(c2[5477],_mm512_xor_si512(c2[3363],_mm512_xor_si512(c2[7058],_mm512_xor_si512(c2[7601],_mm512_xor_si512(c2[12089],_mm512_xor_si512(c2[6014],_mm512_xor_si512(c2[7609],_mm512_xor_si512(c2[7081],_mm512_xor_si512(c2[15265],_mm512_xor_si512(c2[8148],_mm512_xor_si512(c2[14222],_mm512_xor_si512(c2[14748],_mm512_xor_si512(c2[13970],_mm512_xor_si512(c2[14763],_mm512_xor_si512(c2[4732],_mm512_xor_si512(c2[1311],_mm512_xor_si512(c2[7121],c2[4215]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 25
+     d2[150]=_mm512_xor_si512(c2[2125],_mm512_xor_si512(c2[4560],_mm512_xor_si512(c2[16721],c2[14956])));
+
+//row: 26
+     d2[156]=_mm512_xor_si512(c2[1849],_mm512_xor_si512(c2[4513],_mm512_xor_si512(c2[3217],c2[2294])));
+
+//row: 27
+     d2[162]=_mm512_xor_si512(c2[543],_mm512_xor_si512(c2[603],c2[9865]));
+
+//row: 28
+     d2[168]=_mm512_xor_si512(c2[7923],_mm512_xor_si512(c2[5064],_mm512_xor_si512(c2[13959],c2[5008])));
+
+//row: 29
+     d2[174]=_mm512_xor_si512(c2[528],_mm512_xor_si512(c2[7130],_mm512_xor_si512(c2[7396],_mm512_xor_si512(c2[8981],_mm512_xor_si512(c2[9245],_mm512_xor_si512(c2[8989],_mm512_xor_si512(c2[2393],_mm512_xor_si512(c2[9782],_mm512_xor_si512(c2[10046],_mm512_xor_si512(c2[15855],_mm512_xor_si512(c2[290],_mm512_xor_si512(c2[7154],_mm512_xor_si512(c2[6096],_mm512_xor_si512(c2[37],_mm512_xor_si512(c2[6374],_mm512_xor_si512(c2[3733],_mm512_xor_si512(c2[3997],_mm512_xor_si512(c2[12461],_mm512_xor_si512(c2[7176],_mm512_xor_si512(c2[8765],_mm512_xor_si512(c2[9029],_mm512_xor_si512(c2[1120],_mm512_xor_si512(c2[8247],_mm512_xor_si512(c2[4025],_mm512_xor_si512(c2[10372],_mm512_xor_si512(c2[12747],_mm512_xor_si512(c2[5620],_mm512_xor_si512(c2[5884],_mm512_xor_si512(c2[6948],_mm512_xor_si512(c2[6953],_mm512_xor_si512(c2[10383],_mm512_xor_si512(c2[10647],_mm512_xor_si512(c2[6960],_mm512_xor_si512(c2[12507],_mm512_xor_si512(c2[1682],_mm512_xor_si512(c2[1946],_mm512_xor_si512(c2[3276],_mm512_xor_si512(c2[376],_mm512_xor_si512(c2[3806],_mm512_xor_si512(c2[12533],_mm512_xor_si512(c2[4347],_mm512_xor_si512(c2[4875],_mm512_xor_si512(c2[5139],_mm512_xor_si512(c2[15974],_mm512_xor_si512(c2[14393],_mm512_xor_si512(c2[16769],_mm512_xor_si512(c2[132],_mm512_xor_si512(c2[8593],_mm512_xor_si512(c2[9912],_mm512_xor_si512(c2[14140],_mm512_xor_si512(c2[14404],_mm512_xor_si512(c2[13884],_mm512_xor_si512(c2[6232],_mm512_xor_si512(c2[9396],_mm512_xor_si512(c2[9660],_mm512_xor_si512(c2[10732],_mm512_xor_si512(c2[10466],_mm512_xor_si512(c2[9148],_mm512_xor_si512(c2[9412],_mm512_xor_si512(c2[1488],_mm512_xor_si512(c2[10216],_mm512_xor_si512(c2[13382],_mm512_xor_si512(c2[1238],_mm512_xor_si512(c2[15242],_mm512_xor_si512(c2[13128],_mm512_xor_si512(c2[16565],_mm512_xor_si512(c2[16829],_mm512_xor_si512(c2[471],_mm512_xor_si512(c2[4959],_mm512_xor_si512(c2[15521],_mm512_xor_si512(c2[15785],_mm512_xor_si512(c2[485],_mm512_xor_si512(c2[16852],_mm512_xor_si512(c2[7877],_mm512_xor_si512(c2[8141],_mm512_xor_si512(c2[5234],_mm512_xor_si512(c2[1024],_mm512_xor_si512(c2[7092],_mm512_xor_si512(c2[7624],_mm512_xor_si512(c2[6840],_mm512_xor_si512(c2[7633],_mm512_xor_si512(c2[14233],_mm512_xor_si512(c2[14497],_mm512_xor_si512(c2[11076],_mm512_xor_si512(c2[16886],_mm512_xor_si512(c2[13716],c2[13980]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 30
+     d2[180]=_mm512_xor_si512(c2[3696],_mm512_xor_si512(c2[10298],_mm512_xor_si512(c2[10300],_mm512_xor_si512(c2[10564],_mm512_xor_si512(c2[12149],_mm512_xor_si512(c2[12413],_mm512_xor_si512(c2[8186],_mm512_xor_si512(c2[12157],_mm512_xor_si512(c2[5297],_mm512_xor_si512(c2[5561],_mm512_xor_si512(c2[12950],_mm512_xor_si512(c2[13214],_mm512_xor_si512(c2[3458],_mm512_xor_si512(c2[10322],_mm512_xor_si512(c2[9000],_mm512_xor_si512(c2[9264],_mm512_xor_si512(c2[3205],_mm512_xor_si512(c2[9542],_mm512_xor_si512(c2[6901],_mm512_xor_si512(c2[7165],_mm512_xor_si512(c2[15629],_mm512_xor_si512(c2[10080],_mm512_xor_si512(c2[10344],_mm512_xor_si512(c2[11933],_mm512_xor_si512(c2[12197],_mm512_xor_si512(c2[4288],_mm512_xor_si512(c2[11415],_mm512_xor_si512(c2[6929],_mm512_xor_si512(c2[7193],_mm512_xor_si512(c2[13540],_mm512_xor_si512(c2[15651],_mm512_xor_si512(c2[15915],_mm512_xor_si512(c2[8788],_mm512_xor_si512(c2[9052],_mm512_xor_si512(c2[10116],_mm512_xor_si512(c2[9857],_mm512_xor_si512(c2[10121],_mm512_xor_si512(c2[13551],_mm512_xor_si512(c2[13815],_mm512_xor_si512(c2[10128],_mm512_xor_si512(c2[15411],_mm512_xor_si512(c2[15675],_mm512_xor_si512(c2[4850],_mm512_xor_si512(c2[5114],_mm512_xor_si512(c2[6444],_mm512_xor_si512(c2[3544],_mm512_xor_si512(c2[6710],_mm512_xor_si512(c2[6974],_mm512_xor_si512(c2[15701],_mm512_xor_si512(c2[7251],_mm512_xor_si512(c2[7515],_mm512_xor_si512(c2[8043],_mm512_xor_si512(c2[8307],_mm512_xor_si512(c2[9891],_mm512_xor_si512(c2[2247],_mm512_xor_si512(c2[660],_mm512_xor_si512(c2[3036],_mm512_xor_si512(c2[3300],_mm512_xor_si512(c2[11761],_mm512_xor_si512(c2[13080],_mm512_xor_si512(c2[413],_mm512_xor_si512(c2[677],_mm512_xor_si512(c2[157],_mm512_xor_si512(c2[9136],_mm512_xor_si512(c2[9400],_mm512_xor_si512(c2[12564],_mm512_xor_si512(c2[12828],_mm512_xor_si512(c2[1216],_mm512_xor_si512(c2[13900],_mm512_xor_si512(c2[13370],_mm512_xor_si512(c2[13634],_mm512_xor_si512(c2[12316],_mm512_xor_si512(c2[12580],_mm512_xor_si512(c2[13384],_mm512_xor_si512(c2[16550],_mm512_xor_si512(c2[4142],_mm512_xor_si512(c2[4406],_mm512_xor_si512(c2[1515],_mm512_xor_si512(c2[16296],_mm512_xor_si512(c2[2832],_mm512_xor_si512(c2[3096],_mm512_xor_si512(c2[3639],_mm512_xor_si512(c2[7863],_mm512_xor_si512(c2[8127],_mm512_xor_si512(c2[1788],_mm512_xor_si512(c2[2052],_mm512_xor_si512(c2[3653],_mm512_xor_si512(c2[2861],_mm512_xor_si512(c2[3125],_mm512_xor_si512(c2[11045],_mm512_xor_si512(c2[11309],_mm512_xor_si512(c2[4192],_mm512_xor_si512(c2[10260],_mm512_xor_si512(c2[10528],_mm512_xor_si512(c2[10792],_mm512_xor_si512(c2[10008],_mm512_xor_si512(c2[10537],_mm512_xor_si512(c2[10801],_mm512_xor_si512(c2[506],_mm512_xor_si512(c2[770],_mm512_xor_si512(c2[14244],_mm512_xor_si512(c2[3159],_mm512_xor_si512(c2[16884],c2[253])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 31
+     d2[186]=_mm512_xor_si512(c2[13994],_mm512_xor_si512(c2[13730],_mm512_xor_si512(c2[3701],_mm512_xor_si512(c2[3437],_mm512_xor_si512(c2[3961],_mm512_xor_si512(c2[3697],_mm512_xor_si512(c2[5810],_mm512_xor_si512(c2[5282],_mm512_xor_si512(c2[5546],_mm512_xor_si512(c2[5560],_mm512_xor_si512(c2[5296],_mm512_xor_si512(c2[15853],_mm512_xor_si512(c2[15589],_mm512_xor_si512(c2[6617],_mm512_xor_si512(c2[6089],_mm512_xor_si512(c2[6353],_mm512_xor_si512(c2[9517],_mm512_xor_si512(c2[13756],_mm512_xor_si512(c2[13492],_mm512_xor_si512(c2[3725],_mm512_xor_si512(c2[3461],_mm512_xor_si512(c2[2667],_mm512_xor_si512(c2[2403],_mm512_xor_si512(c2[13503],_mm512_xor_si512(c2[13239],_mm512_xor_si512(c2[2945],_mm512_xor_si512(c2[2681],_mm512_xor_si512(c2[568],_mm512_xor_si512(c2[40],_mm512_xor_si512(c2[304],_mm512_xor_si512(c2[9026],_mm512_xor_si512(c2[8762],_mm512_xor_si512(c2[3747],_mm512_xor_si512(c2[3483],_mm512_xor_si512(c2[5594],_mm512_xor_si512(c2[5066],_mm512_xor_si512(c2[5330],_mm512_xor_si512(c2[14580],_mm512_xor_si512(c2[14316],_mm512_xor_si512(c2[4812],_mm512_xor_si512(c2[4548],_mm512_xor_si512(c2[590],_mm512_xor_si512(c2[326],_mm512_xor_si512(c2[6937],_mm512_xor_si512(c2[6673],_mm512_xor_si512(c2[9312],_mm512_xor_si512(c2[9048],_mm512_xor_si512(c2[2449],_mm512_xor_si512(c2[1921],_mm512_xor_si512(c2[2185],_mm512_xor_si512(c2[3519],_mm512_xor_si512(c2[3255],_mm512_xor_si512(c2[3518],_mm512_xor_si512(c2[3254],_mm512_xor_si512(c2[7212],_mm512_xor_si512(c2[6684],_mm512_xor_si512(c2[6948],_mm512_xor_si512(c2[6159],_mm512_xor_si512(c2[3531],_mm512_xor_si512(c2[3267],_mm512_xor_si512(c2[9072],_mm512_xor_si512(c2[8808],_mm512_xor_si512(c2[15412],_mm512_xor_si512(c2[14884],_mm512_xor_si512(c2[15148],_mm512_xor_si512(c2[16742],_mm512_xor_si512(c2[16478],_mm512_xor_si512(c2[13836],_mm512_xor_si512(c2[13572],_mm512_xor_si512(c2[377],_mm512_xor_si512(c2[113],_mm512_xor_si512(c2[9098],_mm512_xor_si512(c2[8834],_mm512_xor_si512(c2[912],_mm512_xor_si512(c2[648],_mm512_xor_si512(c2[1704],_mm512_xor_si512(c2[1176],_mm512_xor_si512(c2[1440],_mm512_xor_si512(c2[12545],_mm512_xor_si512(c2[12281],_mm512_xor_si512(c2[10958],_mm512_xor_si512(c2[10694],_mm512_xor_si512(c2[13598],_mm512_xor_si512(c2[13070],_mm512_xor_si512(c2[13334],_mm512_xor_si512(c2[5164],_mm512_xor_si512(c2[4900],_mm512_xor_si512(c2[6483],_mm512_xor_si512(c2[6219],_mm512_xor_si512(c2[10969],_mm512_xor_si512(c2[10441],_mm512_xor_si512(c2[10705],_mm512_xor_si512(c2[10455],_mm512_xor_si512(c2[10191],_mm512_xor_si512(c2[2797],_mm512_xor_si512(c2[2533],_mm512_xor_si512(c2[6231],_mm512_xor_si512(c2[5703],_mm512_xor_si512(c2[5967],_mm512_xor_si512(c2[7297],_mm512_xor_si512(c2[7033],_mm512_xor_si512(c2[7037],_mm512_xor_si512(c2[6773],_mm512_xor_si512(c2[5977],_mm512_xor_si512(c2[5449],_mm512_xor_si512(c2[5713],_mm512_xor_si512(c2[6781],_mm512_xor_si512(c2[6517],_mm512_xor_si512(c2[9953],_mm512_xor_si512(c2[9689],_mm512_xor_si512(c2[14704],_mm512_xor_si512(c2[14440],_mm512_xor_si512(c2[11813],_mm512_xor_si512(c2[11549],_mm512_xor_si512(c2[9699],_mm512_xor_si512(c2[9435],_mm512_xor_si512(c2[13394],_mm512_xor_si512(c2[12866],_mm512_xor_si512(c2[13130],_mm512_xor_si512(c2[13937],_mm512_xor_si512(c2[13673],_mm512_xor_si512(c2[1524],_mm512_xor_si512(c2[1260],_mm512_xor_si512(c2[12350],_mm512_xor_si512(c2[11822],_mm512_xor_si512(c2[12086],_mm512_xor_si512(c2[13945],_mm512_xor_si512(c2[13681],_mm512_xor_si512(c2[13417],_mm512_xor_si512(c2[13153],_mm512_xor_si512(c2[4706],_mm512_xor_si512(c2[4178],_mm512_xor_si512(c2[4442],_mm512_xor_si512(c2[14484],_mm512_xor_si512(c2[14220],_mm512_xor_si512(c2[3663],_mm512_xor_si512(c2[3399],_mm512_xor_si512(c2[4189],_mm512_xor_si512(c2[3925],_mm512_xor_si512(c2[3411],_mm512_xor_si512(c2[3147],_mm512_xor_si512(c2[4204],_mm512_xor_si512(c2[3940],_mm512_xor_si512(c2[11068],_mm512_xor_si512(c2[10540],_mm512_xor_si512(c2[10804],_mm512_xor_si512(c2[7647],_mm512_xor_si512(c2[7383],_mm512_xor_si512(c2[13457],_mm512_xor_si512(c2[13193],_mm512_xor_si512(c2[10551],_mm512_xor_si512(c2[10023],c2[10287]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 32
+     d2[192]=_mm512_xor_si512(c2[5019],_mm512_xor_si512(c2[11621],_mm512_xor_si512(c2[11617],_mm512_xor_si512(c2[11881],_mm512_xor_si512(c2[13466],_mm512_xor_si512(c2[13730],_mm512_xor_si512(c2[10033],_mm512_xor_si512(c2[13480],_mm512_xor_si512(c2[6614],_mm512_xor_si512(c2[6878],_mm512_xor_si512(c2[14273],_mm512_xor_si512(c2[14537],_mm512_xor_si512(c2[4781],_mm512_xor_si512(c2[11645],_mm512_xor_si512(c2[10323],_mm512_xor_si512(c2[10587],_mm512_xor_si512(c2[4528],_mm512_xor_si512(c2[10865],_mm512_xor_si512(c2[8224],_mm512_xor_si512(c2[8488],_mm512_xor_si512(c2[51],_mm512_xor_si512(c2[11403],_mm512_xor_si512(c2[11667],_mm512_xor_si512(c2[13250],_mm512_xor_si512(c2[13514],_mm512_xor_si512(c2[5605],_mm512_xor_si512(c2[12732],_mm512_xor_si512(c2[8246],_mm512_xor_si512(c2[8510],_mm512_xor_si512(c2[14857],_mm512_xor_si512(c2[73],_mm512_xor_si512(c2[337],_mm512_xor_si512(c2[10105],_mm512_xor_si512(c2[10369],_mm512_xor_si512(c2[11439],_mm512_xor_si512(c2[11174],_mm512_xor_si512(c2[11438],_mm512_xor_si512(c2[14868],_mm512_xor_si512(c2[15132],_mm512_xor_si512(c2[11451],_mm512_xor_si512(c2[16728],_mm512_xor_si512(c2[97],_mm512_xor_si512(c2[6173],_mm512_xor_si512(c2[6437],_mm512_xor_si512(c2[7767],_mm512_xor_si512(c2[4861],_mm512_xor_si512(c2[8033],_mm512_xor_si512(c2[8297],_mm512_xor_si512(c2[123],_mm512_xor_si512(c2[8568],_mm512_xor_si512(c2[8832],_mm512_xor_si512(c2[9360],_mm512_xor_si512(c2[9624],_mm512_xor_si512(c2[3564],_mm512_xor_si512(c2[1983],_mm512_xor_si512(c2[4359],_mm512_xor_si512(c2[4623],_mm512_xor_si512(c2[13084],_mm512_xor_si512(c2[14403],_mm512_xor_si512(c2[1730],_mm512_xor_si512(c2[1994],_mm512_xor_si512(c2[2523],_mm512_xor_si512(c2[1480],_mm512_xor_si512(c2[10453],_mm512_xor_si512(c2[10717],_mm512_xor_si512(c2[13887],_mm512_xor_si512(c2[14151],_mm512_xor_si512(c2[15217],_mm512_xor_si512(c2[14693],_mm512_xor_si512(c2[14957],_mm512_xor_si512(c2[13633],_mm512_xor_si512(c2[13897],_mm512_xor_si512(c2[12578],_mm512_xor_si512(c2[14701],_mm512_xor_si512(c2[972],_mm512_xor_si512(c2[5465],_mm512_xor_si512(c2[5729],_mm512_xor_si512(c2[2832],_mm512_xor_si512(c2[724],_mm512_xor_si512(c2[4155],_mm512_xor_si512(c2[4419],_mm512_xor_si512(c2[4956],_mm512_xor_si512(c2[9180],_mm512_xor_si512(c2[9444],_mm512_xor_si512(c2[3111],_mm512_xor_si512(c2[3375],_mm512_xor_si512(c2[4970],_mm512_xor_si512(c2[4178],_mm512_xor_si512(c2[4442],_mm512_xor_si512(c2[12362],_mm512_xor_si512(c2[12626],_mm512_xor_si512(c2[5509],_mm512_xor_si512(c2[11583],_mm512_xor_si512(c2[11845],_mm512_xor_si512(c2[12109],_mm512_xor_si512(c2[11331],_mm512_xor_si512(c2[11860],_mm512_xor_si512(c2[12124],_mm512_xor_si512(c2[1829],_mm512_xor_si512(c2[2093],_mm512_xor_si512(c2[15567],_mm512_xor_si512(c2[4476],_mm512_xor_si512(c2[1312],c2[1576])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 33
+     d2[198]=_mm512_xor_si512(c2[809],_mm512_xor_si512(c2[2136],_mm512_xor_si512(c2[10961],c2[11077])));
+
+//row: 34
+     d2[204]=_mm512_xor_si512(c2[10035],_mm512_xor_si512(c2[5366],_mm512_xor_si512(c2[2823],c2[14992])));
+
+//row: 35
+     d2[210]=_mm512_xor_si512(c2[16106],_mm512_xor_si512(c2[5813],_mm512_xor_si512(c2[6073],_mm512_xor_si512(c2[7922],_mm512_xor_si512(c2[7672],_mm512_xor_si512(c2[1070],_mm512_xor_si512(c2[8729],_mm512_xor_si512(c2[17],_mm512_xor_si512(c2[15868],_mm512_xor_si512(c2[5837],_mm512_xor_si512(c2[4779],_mm512_xor_si512(c2[15615],_mm512_xor_si512(c2[5057],_mm512_xor_si512(c2[2680],_mm512_xor_si512(c2[11138],_mm512_xor_si512(c2[5859],_mm512_xor_si512(c2[7706],_mm512_xor_si512(c2[16692],_mm512_xor_si512(c2[6924],_mm512_xor_si512(c2[2702],_mm512_xor_si512(c2[9049],_mm512_xor_si512(c2[11424],_mm512_xor_si512(c2[4561],_mm512_xor_si512(c2[4037],_mm512_xor_si512(c2[5631],_mm512_xor_si512(c2[5630],_mm512_xor_si512(c2[9324],_mm512_xor_si512(c2[5643],_mm512_xor_si512(c2[11184],_mm512_xor_si512(c2[629],_mm512_xor_si512(c2[1959],_mm512_xor_si512(c2[15948],_mm512_xor_si512(c2[2489],_mm512_xor_si512(c2[11210],_mm512_xor_si512(c2[3024],_mm512_xor_si512(c2[3816],_mm512_xor_si512(c2[14657],_mm512_xor_si512(c2[13070],_mm512_xor_si512(c2[15710],_mm512_xor_si512(c2[7276],_mm512_xor_si512(c2[8595],_mm512_xor_si512(c2[13081],_mm512_xor_si512(c2[672],_mm512_xor_si512(c2[12567],_mm512_xor_si512(c2[4909],_mm512_xor_si512(c2[8343],_mm512_xor_si512(c2[9409],_mm512_xor_si512(c2[9149],_mm512_xor_si512(c2[8089],_mm512_xor_si512(c2[8893],_mm512_xor_si512(c2[12065],_mm512_xor_si512(c2[16816],_mm512_xor_si512(c2[13925],_mm512_xor_si512(c2[11811],_mm512_xor_si512(c2[15506],_mm512_xor_si512(c2[16049],_mm512_xor_si512(c2[3636],_mm512_xor_si512(c2[14462],_mm512_xor_si512(c2[16057],_mm512_xor_si512(c2[15529],_mm512_xor_si512(c2[6818],_mm512_xor_si512(c2[16596],_mm512_xor_si512(c2[5775],_mm512_xor_si512(c2[6301],_mm512_xor_si512(c2[5523],_mm512_xor_si512(c2[6316],_mm512_xor_si512(c2[13180],_mm512_xor_si512(c2[9759],_mm512_xor_si512(c2[15569],c2[12663])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 36
+     d2[216]=_mm512_xor_si512(c2[4755],_mm512_xor_si512(c2[15220],_mm512_xor_si512(c2[11008],c2[5760])));
+
+//row: 37
+     d2[222]=_mm512_xor_si512(c2[10032],_mm512_xor_si512(c2[10296],_mm512_xor_si512(c2[3],_mm512_xor_si512(c2[269],_mm512_xor_si512(c2[2112],_mm512_xor_si512(c2[1598],_mm512_xor_si512(c2[1862],_mm512_xor_si512(c2[12161],_mm512_xor_si512(c2[2919],_mm512_xor_si512(c2[3448],_mm512_xor_si512(c2[9794],_mm512_xor_si512(c2[10058],_mm512_xor_si512(c2[27],_mm512_xor_si512(c2[15864],_mm512_xor_si512(c2[9541],_mm512_xor_si512(c2[9805],_mm512_xor_si512(c2[16142],_mm512_xor_si512(c2[13765],_mm512_xor_si512(c2[5328],_mm512_xor_si512(c2[49],_mm512_xor_si512(c2[1896],_mm512_xor_si512(c2[10624],_mm512_xor_si512(c2[10888],_mm512_xor_si512(c2[1120],_mm512_xor_si512(c2[13793],_mm512_xor_si512(c2[2981],_mm512_xor_si512(c2[3245],_mm512_xor_si512(c2[5620],_mm512_xor_si512(c2[15652],_mm512_xor_si512(c2[16716],_mm512_xor_si512(c2[16721],_mm512_xor_si512(c2[3520],_mm512_xor_si512(c2[16728],_mm512_xor_si512(c2[5380],_mm512_xor_si512(c2[11714],_mm512_xor_si512(c2[12780],_mm512_xor_si512(c2[13044],_mm512_xor_si512(c2[10144],_mm512_xor_si512(c2[13574],_mm512_xor_si512(c2[5136],_mm512_xor_si512(c2[5400],_mm512_xor_si512(c2[14115],_mm512_xor_si512(c2[14907],_mm512_xor_si512(c2[8583],_mm512_xor_si512(c2[8847],_mm512_xor_si512(c2[7260],_mm512_xor_si512(c2[9900],_mm512_xor_si512(c2[1202],_mm512_xor_si512(c2[1466],_mm512_xor_si512(c2[2785],_mm512_xor_si512(c2[7277],_mm512_xor_si512(c2[6493],_mm512_xor_si512(c2[6757],_mm512_xor_si512(c2[16000],_mm512_xor_si512(c2[2533],_mm512_xor_si512(c2[4909],_mm512_xor_si512(c2[3605],_mm512_xor_si512(c2[3339],_mm512_xor_si512(c2[2285],_mm512_xor_si512(c2[2825],_mm512_xor_si512(c2[3089],_mm512_xor_si512(c2[6255],_mm512_xor_si512(c2[11006],_mm512_xor_si512(c2[7851],_mm512_xor_si512(c2[8115],_mm512_xor_si512(c2[6001],_mm512_xor_si512(c2[9696],_mm512_xor_si512(c2[10239],_mm512_xor_si512(c2[14727],_mm512_xor_si512(c2[8652],_mm512_xor_si512(c2[9989],_mm512_xor_si512(c2[10253],_mm512_xor_si512(c2[9725],_mm512_xor_si512(c2[1008],_mm512_xor_si512(c2[10528],_mm512_xor_si512(c2[10792],_mm512_xor_si512(c2[16860],_mm512_xor_si512(c2[497],_mm512_xor_si512(c2[16344],_mm512_xor_si512(c2[16608],_mm512_xor_si512(c2[506],_mm512_xor_si512(c2[7370],_mm512_xor_si512(c2[3685],_mm512_xor_si512(c2[3949],_mm512_xor_si512(c2[9759],c2[6853])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 38
+     d2[228]=_mm512_xor_si512(c2[15050],_mm512_xor_si512(c2[13046],_mm512_xor_si512(c2[8836],c2[5955])));
+
+//row: 39
+     d2[234]=_mm512_xor_si512(c2[544],_mm512_xor_si512(c2[7693],_mm512_xor_si512(c2[6953],c2[11056])));
+
+//row: 40
+     d2[240]=_mm512_xor_si512(c2[12410],_mm512_xor_si512(c2[9864],c2[14992]));
+
+//row: 41
+     d2[246]=_mm512_xor_si512(c2[13740],_mm512_xor_si512(c2[15352],_mm512_xor_si512(c2[3014],c2[8668])));
+
+//row: 42
+     d2[252]=_mm512_xor_si512(c2[3698],_mm512_xor_si512(c2[10300],_mm512_xor_si512(c2[10296],_mm512_xor_si512(c2[10560],_mm512_xor_si512(c2[12145],_mm512_xor_si512(c2[12409],_mm512_xor_si512(c2[12937],_mm512_xor_si512(c2[12159],_mm512_xor_si512(c2[5293],_mm512_xor_si512(c2[5557],_mm512_xor_si512(c2[12952],_mm512_xor_si512(c2[13216],_mm512_xor_si512(c2[3460],_mm512_xor_si512(c2[10324],_mm512_xor_si512(c2[9002],_mm512_xor_si512(c2[9266],_mm512_xor_si512(c2[3207],_mm512_xor_si512(c2[9544],_mm512_xor_si512(c2[6903],_mm512_xor_si512(c2[7167],_mm512_xor_si512(c2[15625],_mm512_xor_si512(c2[10082],_mm512_xor_si512(c2[10346],_mm512_xor_si512(c2[11929],_mm512_xor_si512(c2[12193],_mm512_xor_si512(c2[3744],_mm512_xor_si512(c2[4284],_mm512_xor_si512(c2[11417],_mm512_xor_si512(c2[6925],_mm512_xor_si512(c2[7189],_mm512_xor_si512(c2[13536],_mm512_xor_si512(c2[15653],_mm512_xor_si512(c2[15917],_mm512_xor_si512(c2[8784],_mm512_xor_si512(c2[9048],_mm512_xor_si512(c2[10118],_mm512_xor_si512(c2[9853],_mm512_xor_si512(c2[10117],_mm512_xor_si512(c2[13553],_mm512_xor_si512(c2[13817],_mm512_xor_si512(c2[10130],_mm512_xor_si512(c2[15413],_mm512_xor_si512(c2[15677],_mm512_xor_si512(c2[4852],_mm512_xor_si512(c2[5116],_mm512_xor_si512(c2[6446],_mm512_xor_si512(c2[3540],_mm512_xor_si512(c2[6712],_mm512_xor_si512(c2[6976],_mm512_xor_si512(c2[15697],_mm512_xor_si512(c2[7253],_mm512_xor_si512(c2[7517],_mm512_xor_si512(c2[8045],_mm512_xor_si512(c2[8309],_mm512_xor_si512(c2[2249],_mm512_xor_si512(c2[662],_mm512_xor_si512(c2[3038],_mm512_xor_si512(c2[3302],_mm512_xor_si512(c2[11763],_mm512_xor_si512(c2[13082],_mm512_xor_si512(c2[409],_mm512_xor_si512(c2[673],_mm512_xor_si512(c2[159],_mm512_xor_si512(c2[9132],_mm512_xor_si512(c2[9396],_mm512_xor_si512(c2[12566],_mm512_xor_si512(c2[12830],_mm512_xor_si512(c2[13896],_mm512_xor_si512(c2[13372],_mm512_xor_si512(c2[13636],_mm512_xor_si512(c2[12312],_mm512_xor_si512(c2[12576],_mm512_xor_si512(c2[13380],_mm512_xor_si512(c2[16552],_mm512_xor_si512(c2[4144],_mm512_xor_si512(c2[4408],_mm512_xor_si512(c2[1517],_mm512_xor_si512(c2[16298],_mm512_xor_si512(c2[2834],_mm512_xor_si512(c2[3098],_mm512_xor_si512(c2[3641],_mm512_xor_si512(c2[7865],_mm512_xor_si512(c2[8129],_mm512_xor_si512(c2[1790],_mm512_xor_si512(c2[2054],_mm512_xor_si512(c2[3649],_mm512_xor_si512(c2[2857],_mm512_xor_si512(c2[3121],_mm512_xor_si512(c2[11041],_mm512_xor_si512(c2[11305],_mm512_xor_si512(c2[4188],_mm512_xor_si512(c2[10262],_mm512_xor_si512(c2[10524],_mm512_xor_si512(c2[10788],_mm512_xor_si512(c2[10010],_mm512_xor_si512(c2[10539],_mm512_xor_si512(c2[10803],_mm512_xor_si512(c2[508],_mm512_xor_si512(c2[772],_mm512_xor_si512(c2[14246],_mm512_xor_si512(c2[3161],_mm512_xor_si512(c2[16886],c2[255]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 43
+     d2[258]=_mm512_xor_si512(c2[7393],_mm512_xor_si512(c2[13995],_mm512_xor_si512(c2[14261],_mm512_xor_si512(c2[15840],_mm512_xor_si512(c2[16104],_mm512_xor_si512(c2[15854],_mm512_xor_si512(c2[9252],_mm512_xor_si512(c2[16647],_mm512_xor_si512(c2[16],_mm512_xor_si512(c2[12949],_mm512_xor_si512(c2[7155],_mm512_xor_si512(c2[14019],_mm512_xor_si512(c2[12961],_mm512_xor_si512(c2[6902],_mm512_xor_si512(c2[13239],_mm512_xor_si512(c2[10598],_mm512_xor_si512(c2[10862],_mm512_xor_si512(c2[2425],_mm512_xor_si512(c2[14041],_mm512_xor_si512(c2[15624],_mm512_xor_si512(c2[15888],_mm512_xor_si512(c2[7985],_mm512_xor_si512(c2[15112],_mm512_xor_si512(c2[10884],_mm512_xor_si512(c2[336],_mm512_xor_si512(c2[2717],_mm512_xor_si512(c2[12485],_mm512_xor_si512(c2[12749],_mm512_xor_si512(c2[13813],_mm512_xor_si512(c2[13812],_mm512_xor_si512(c2[353],_mm512_xor_si512(c2[617],_mm512_xor_si512(c2[13825],_mm512_xor_si512(c2[2477],_mm512_xor_si512(c2[8547],_mm512_xor_si512(c2[8811],_mm512_xor_si512(c2[10141],_mm512_xor_si512(c2[7241],_mm512_xor_si512(c2[10671],_mm512_xor_si512(c2[2497],_mm512_xor_si512(c2[11212],_mm512_xor_si512(c2[11740],_mm512_xor_si512(c2[12004],_mm512_xor_si512(c2[5944],_mm512_xor_si512(c2[4357],_mm512_xor_si512(c2[6733],_mm512_xor_si512(c2[6997],_mm512_xor_si512(c2[15458],_mm512_xor_si512(c2[16777],_mm512_xor_si512(c2[4104],_mm512_xor_si512(c2[4368],_mm512_xor_si512(c2[3854],_mm512_xor_si512(c2[13097],_mm512_xor_si512(c2[16261],_mm512_xor_si512(c2[16525],_mm512_xor_si512(c2[696],_mm512_xor_si512(c2[436],_mm512_xor_si512(c2[16013],_mm512_xor_si512(c2[16277],_mm512_xor_si512(c2[180],_mm512_xor_si512(c2[3352],_mm512_xor_si512(c2[8103],_mm512_xor_si512(c2[5212],_mm512_xor_si512(c2[3098],_mm512_xor_si512(c2[6529],_mm512_xor_si512(c2[6793],_mm512_xor_si512(c2[1250],_mm512_xor_si512(c2[7336],_mm512_xor_si512(c2[11824],_mm512_xor_si512(c2[5485],_mm512_xor_si512(c2[5749],_mm512_xor_si512(c2[7344],_mm512_xor_si512(c2[6816],_mm512_xor_si512(c2[14736],_mm512_xor_si512(c2[15000],_mm512_xor_si512(c2[13417],_mm512_xor_si512(c2[7889],_mm512_xor_si512(c2[13957],_mm512_xor_si512(c2[14489],_mm512_xor_si512(c2[13705],_mm512_xor_si512(c2[14498],_mm512_xor_si512(c2[4203],_mm512_xor_si512(c2[4467],_mm512_xor_si512(c2[1046],_mm512_xor_si512(c2[6856],_mm512_xor_si512(c2[3686],c2[3950]))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 44
+     d2[264]=_mm512_xor_si512(c2[1323],_mm512_xor_si512(c2[7925],_mm512_xor_si512(c2[8185],_mm512_xor_si512(c2[10034],_mm512_xor_si512(c2[4225],_mm512_xor_si512(c2[9784],_mm512_xor_si512(c2[3182],_mm512_xor_si512(c2[10841],_mm512_xor_si512(c2[1085],_mm512_xor_si512(c2[7949],_mm512_xor_si512(c2[6891],_mm512_xor_si512(c2[832],_mm512_xor_si512(c2[7169],_mm512_xor_si512(c2[4792],_mm512_xor_si512(c2[13250],_mm512_xor_si512(c2[7971],_mm512_xor_si512(c2[9818],_mm512_xor_si512(c2[1909],_mm512_xor_si512(c2[9036],_mm512_xor_si512(c2[4814],_mm512_xor_si512(c2[11161],_mm512_xor_si512(c2[13536],_mm512_xor_si512(c2[6673],_mm512_xor_si512(c2[7743],_mm512_xor_si512(c2[7742],_mm512_xor_si512(c2[11436],_mm512_xor_si512(c2[3781],_mm512_xor_si512(c2[7755],_mm512_xor_si512(c2[13296],_mm512_xor_si512(c2[2741],_mm512_xor_si512(c2[4071],_mm512_xor_si512(c2[1165],_mm512_xor_si512(c2[4601],_mm512_xor_si512(c2[9350],_mm512_xor_si512(c2[13322],_mm512_xor_si512(c2[5136],_mm512_xor_si512(c2[5928],_mm512_xor_si512(c2[16769],_mm512_xor_si512(c2[15182],_mm512_xor_si512(c2[927],_mm512_xor_si512(c2[9388],_mm512_xor_si512(c2[10707],_mm512_xor_si512(c2[15193],_mm512_xor_si512(c2[14679],_mm512_xor_si512(c2[7021],_mm512_xor_si512(c2[10455],_mm512_xor_si512(c2[11521],_mm512_xor_si512(c2[11261],_mm512_xor_si512(c2[10201],_mm512_xor_si512(c2[11005],_mm512_xor_si512(c2[14177],_mm512_xor_si512(c2[2033],_mm512_xor_si512(c2[16037],_mm512_xor_si512(c2[13923],_mm512_xor_si512(c2[723],_mm512_xor_si512(c2[1260],_mm512_xor_si512(c2[5748],_mm512_xor_si512(c2[16574],_mm512_xor_si512(c2[1274],_mm512_xor_si512(c2[746],_mm512_xor_si512(c2[8930],_mm512_xor_si512(c2[1813],_mm512_xor_si512(c2[7887],_mm512_xor_si512(c2[8413],_mm512_xor_si512(c2[7635],_mm512_xor_si512(c2[8428],_mm512_xor_si512(c2[15292],_mm512_xor_si512(c2[11871],_mm512_xor_si512(c2[780],c2[14775])))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))));
+
+//row: 45
+     d2[270]=_mm512_xor_si512(c2[1862],_mm512_xor_si512(c2[5618],c2[4080]));
+  }
+}
--- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c
+++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encode_parity_check.c
@@ -37,7 +37,11 @@
 #include "common/utils/LOG/log.h"


+#ifdef __AVX512F__
+#include "ldpc384_simd512_byte.c"
+#else
 #include "ldpc384_byte.c"
+#endif
 #include "ldpc384_byte_128.c"
 #include "ldpc352_byte.c"
 #include "ldpc352_byte_128.c"
@@ -87,7 +91,7 @@

 static void encode_parity_check_part_optim(uint8_t *cc, uint8_t *d, short BG,short Zc,short Kb, int simd_size, int ncols)
 {
-  unsigned char c[2 * 22 * Zc * simd_size] __attribute__((aligned(32))); //double size matrix of c
+  unsigned char c[2 * 22 * Zc * simd_size] __attribute__((aligned(64))); //double size matrix of c
  for (int i1 = 0; i1 < ncols; i1++)   {
    memcpy(&c[2 * i1 * Zc], &cc[i1 * Zc], Zc * sizeof(unsigned char));
    memcpy(&c[(2 * i1 + 1) * Zc], &cc[i1 * Zc], Zc * sizeof(unsigned char));
@@ -95,7 +99,6 @@ static void encode_parity_check_part_optim(uint8_t *cc, uint8_t *d, short BG,sho
  for (int i1 = 1; i1 < simd_size; i1++) {
    memcpy(&c[(2 * ncols * Zc * i1)], &c[i1], (2 * ncols * Zc * sizeof(unsigned char)) - i1);
  }
-  
  if (BG == 1) {
    switch (Zc) {
      case 176:
--- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder.c
+++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder.c
@@ -114,7 +114,13 @@ int LDPCencoder(unsigned char **inputArray, unsigned char **outputArray, encoder
    fprintf(fd,"#include \"PHY/sse_intrin.h\"\n");
    fprintf(fd2,"#include \"PHY/sse_intrin.h\"\n");

-    if (gen_code == 1 && (Zc&31)==0) {
+    if (gen_code == 1 && (Zc&63)==0) {
+      shift=6;
+      mask=63;
+      strcpy(data_type,"__m512i");
+      strcpy(xor_command,"_mm512_xor_si512");
+    }
+    else if (gen_code == 1 && (Zc&31)==0) {
      shift=5; // AVX2 - 256-bit SIMD
      mask=31;
      strcpy(data_type,"simde__m256i");
--- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim.c
+++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim.c
@@ -79,10 +79,13 @@ int LDPCencoder(uint8_t **test_input, uint8_t **channel_input, encoder_implempar
 #endif

  if ((Zc&31) > 0) simd_size = 16;
+#ifdef __AVX512F__
+  else if ((BG==1) && (Zc==384)) simd_size=64;
+#endif
  else simd_size = 32;

-  uint8_t c[22*Zc] __attribute__((aligned(32))); //padded input, unpacked, max size
-  uint8_t d[46*Zc] __attribute__((aligned(32))); //coded parity part output, unpacked, max size
+  uint8_t c[22*Zc] __attribute__((aligned(64))); //padded input, unpacked, max size
+  uint8_t d[46*Zc] __attribute__((aligned(64))); //coded parity part output, unpacked, max size

  // calculate number of punctured bits
  no_punctured_columns=(int)((nrows-2)*Zc+block_length-block_length*rate)/Zc;
--- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8seg.c
+++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8seg.c
@@ -97,10 +97,13 @@ int LDPCencoder(uint8_t **test_input, uint8_t **channel_input, encoder_implempar
  AssertFatal(Zc>0,"no valid Zc found for block length %d\n",block_length);

  if ((Zc&31) > 0) simd_size = 16;
-  else          simd_size = 32;
+#ifdef __AVX512F__
+  else if ((BG==1) && (Zc==384))   simd_size = 64;
+#endif
+  else simd_size = 32;

-  unsigned char c[22*Zc] __attribute__((aligned(32))); //padded input, unpacked, max size
-  unsigned char d[46*Zc] __attribute__((aligned(32))); //coded parity part output, unpacked, max size
+  unsigned char c[22*Zc] __attribute__((aligned(64))); //padded input, unpacked, max size
+  unsigned char d[46*Zc] __attribute__((aligned(64))); //coded parity part output, unpacked, max size


  // calculate number of punctured bits
@@ -131,6 +134,7 @@ int LDPCencoder(uint8_t **test_input, uint8_t **channel_input, encoder_implempar
    ((simde__m256i *)c)[i] = c256;
  }

+
  for (i=(block_length>>5)<<5;i<block_length;i++) {
    for (j=0; j<impp->n_segments; j++) {

@@ -164,6 +168,36 @@ int LDPCencoder(uint8_t **test_input, uint8_t **channel_input, encoder_implempar
  memcpy(&channel_input[0], &c[2*Zc], (block_length-2*Zc)*sizeof(unsigned char));
  memcpy(&channel_input[block_length-2*Zc], &d[0], ((nrows-no_punctured_columns) * Zc-removed_bit)*sizeof(unsigned char));
  */
+#ifdef __AVX512F__
+  uint32_t l1 = (block_length-(2*Zc));
+  uint32_t l2 = ((nrows-no_punctured_columns) * Zc-removed_bit);
+  if ((((2*Zc)&63) == 0) && (((block_length-(2*Zc))&63) == 0)) {
+    __m512i *c512p = (__m512i *)&c[2*Zc];
+    __m512i *d512p = (__m512i *)&d[0];
+    __m512i mask0 = _mm512_set1_epi8(0x1);
+    
+    for (i=0;i<l1>>6;i++)
+      for (j=0;j<impp->n_segments;j++) 
+        ((__m512i *)channel_input[j])[i] = _mm512_and_si512(_mm512_srai_epi16(c512p[i],j),mask0);
+    
+    for (i1=0;i1<l2>>6;i1++,i++)
+      for (j=0;j<impp->n_segments;j++) 
+        ((__m512i *)channel_input[j])[i] = _mm512_and_si512(_mm512_srai_epi16(d512p[i1],j),mask0);
+  } 
+  else if ((((2*Zc)&31) == 0) && (((block_length-(2*Zc))&31) == 0)) {
+    __m256i *c256p = (__m256i *)&c[2*Zc];
+    __m256i *d256p = (__m256i *)&d[0];
+    __m256i mask0 = _mm256_set1_epi8(0x1);
+    
+    for (i=0;i<l1>>6;i++)
+      for (j=0;j<impp->n_segments;j++) 
+        ((__m256i *)channel_input[j])[i] = _mm256_and_si256(_mm256_srai_epi16(c256p[i],j),mask0);
+    
+    for (i1=0;i1<l2>>6;i1++,i++)
+      for (j=0;j<impp->n_segments;j++) 
+        ((__m256i *)channel_input[j])[i] = _mm256_and_si256(_mm256_srai_epi16(d256p[i1],j),mask0);
+  }
+#else
  if ((((2*Zc)&31) == 0) && (((block_length-(2*Zc))&31) == 0)) {
    //AssertFatal(((2*Zc)&31) == 0,"2*Zc needs to be a multiple of 32 for now\n");
    //AssertFatal(((block_length-(2*Zc))&31) == 0,"block_length-(2*Zc) needs to be a multiple of 32 for now\n");
@@ -181,6 +215,7 @@ int LDPCencoder(uint8_t **test_input, uint8_t **channel_input, encoder_implempar
    for (i1=0;i1<l2;i1++,i++)
      for (j=0;j<impp->n_segments;j++) ((simde__m256i *)channel_input[j])[i] = simde_mm256_and_si256(simde_mm256_srai_epi16(d256p[i1],j),masks[0]);
  }
+#endif
  else {
 #ifdef DEBUG_LDPC
  LOG_W(PHY,"using non-optimized version\n");
@@ -194,6 +229,7 @@ int LDPCencoder(uint8_t **test_input, uint8_t **channel_input, encoder_implempar
 	channel_input[j][block_length-2*Zc+i] = (d[i]>>j)&1;
    }

+
  if(impp->toutput != NULL) stop_meas(impp->toutput);
  return 0;
 }
--- a/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c
+++ b/openair1/PHY/CODING/nrLDPC_encoder/ldpc_encoder_optim8segmulti.c
@@ -65,7 +65,22 @@ int LDPCencoder(uint8_t **input, uint8_t **output, encoder_implemparams_t *impp)
  macro_segment_end = (impp->n_segments > 8*(impp->macro_num+1)) ? 8*(impp->macro_num+1) : impp->n_segments;
  ///printf("macro_segment: %d\n", macro_segment);
  ///printf("macro_segment_end: %d\n", macro_segment_end );
+#ifdef __aarch64__
+  simde__m128i shufmask = simde_mm_set_epi64x(0x0101010101010101, 0x0000000000000000);
+  simde__m128i andmask  = simde_mm_set1_epi64x(0x0102040810204080);  // every 8 bits -> 8 bytes, pattern repeats.
+  simde__m128i zero128   = simde_mm_setzero_si128();
+  simde__m128i masks[8];
+  register simde__m128i c128;
+  masks[0] = simde_mm_set1_epi8(0x1);
+  masks[1] = simde_mm_set1_epi8(0x2);
+  masks[2] = simde_mm_set1_epi8(0x4);
+  masks[3] = simde_mm_set1_epi8(0x8);
+  masks[4] = simde_mm_set1_epi8(0x10);
+  masks[5] = simde_mm_set1_epi8(0x20);
+  masks[6] = simde_mm_set1_epi8(0x40);
+  masks[7] = simde_mm_set1_epi8(0x80);

+#else
  simde__m256i shufmask = simde_mm256_set_epi64x(0x0303030303030303, 0x0202020202020202,0x0101010101010101, 0x0000000000000000);
  simde__m256i andmask  = simde_mm256_set1_epi64x(0x0102040810204080);  // every 8 bits -> 8 bytes, pattern repeats.
  simde__m256i zero256   = simde_mm256_setzero_si256();
@@ -79,7 +94,7 @@ int LDPCencoder(uint8_t **input, uint8_t **output, encoder_implemparams_t *impp)
  masks[5] = simde_mm256_set1_epi8(0x20);
  masks[6] = simde_mm256_set1_epi8(0x40);
  masks[7] = simde_mm256_set1_epi8(0x80);
-
+#endif


  //determine number of bits in codeword
@@ -103,10 +118,12 @@ int LDPCencoder(uint8_t **input, uint8_t **output, encoder_implemparams_t *impp)

  AssertFatal(Zc > 0, "no valid Zc found for block length %d\n", block_length);
  if ((Zc&31) > 0) simd_size = 16;
-  else
-    simd_size = 32;
-  unsigned char cc[22*Zc] __attribute__((aligned(32))); //padded input, unpacked, max size
-  unsigned char dd[46*Zc] __attribute__((aligned(32))); //coded parity part output, unpacked, max size
+#ifdef __AVX512F__
+  else if ((BG==1) && Zc==384) simd_size =64;
+#endif
+  else simd_size=32;
+  unsigned char cc[22*Zc] __attribute__((aligned(64))); //padded input, unpacked, max size
+  unsigned char dd[46*Zc] __attribute__((aligned(64))); //coded parity part output, unpacked, max size

  // calculate number of punctured bits
  no_punctured_columns=(int)((nrows-2)*Zc+block_length-block_length*rate)/Zc;
@@ -118,6 +135,7 @@ int LDPCencoder(uint8_t **input, uint8_t **output, encoder_implemparams_t *impp)
  memset(dd,0,sizeof(unsigned char) * nrows * Zc);

  if(impp->tinput != NULL) start_meas(impp->tinput);
+  //interleave up to 8 transport-block segements at a time
 #if 0
  for (i=0; i<block_length; i++) {
 	//for (j=0; j<n_segments; j++) {
@@ -128,6 +146,25 @@ int LDPCencoder(uint8_t **input, uint8_t **output, encoder_implemparams_t *impp)
      c[i] |= (temp << (j-macro_segment));
    }
  }
+#elif defined(__aarch64__)
+  for (int i=0; i<block_length>>4; i++) {
+    c128 = simde_mm_and_si128(simde_mm_cmpeq_epi8(simde_mm_andnot_si128(simde_mm_shuffle_epi8(simde_mm_set1_epi16(((uint16_t*)input[macro_segment])[i]), shufmask),andmask),zero128),masks[0]);
+    //for (j=1; j<n_segments; j++) {
+    for (int j=macro_segment+1; j < macro_segment_end; j++) {    
+      c128 = simde_mm_or_si128(simde_mm_and_si128(simde_mm_cmpeq_epi8(simde_mm_andnot_si128(simde_mm_shuffle_epi8(simde_mm_set1_epi32(((uint16_t*)input[j])[i]), shufmask),andmask),zero128),masks[j-macro_segment]),c128);
+    }
+    ((simde__m128i *)cc)[i] = c128;
+  }
+
+  for (int i=(block_length>>4)<<4;i<block_length;i++) {
+    //for (j=0; j<n_segments; j++) {
+	  for (int j=macro_segment; j < macro_segment_end; j++) {
+
+	    temp = (input[j][i/8]&(128>>(i&7)))>>(7-(i&7));
+      //printf("c(%d,%d)=%d\n",j,i,temp);
+      cc[i] |= (temp << (j-macro_segment));
+    }
+  }
 #else
  for (int i=0; i<block_length>>5; i++) {
    c256 = simde_mm256_and_si256(simde_mm256_cmpeq_epi8(simde_mm256_andnot_si256(simde_mm256_shuffle_epi8(simde_mm256_set1_epi32(((uint32_t*)input[macro_segment])[i]), shufmask),andmask),zero256),masks[0]);
@@ -166,48 +203,8 @@ int LDPCencoder(uint8_t **input, uint8_t **output, encoder_implemparams_t *impp)
      return(-1);
    }
  }
-  if(impp->toutput != NULL) start_meas(impp->toutput);
-  // information part and puncture columns
-  /*
-  memcpy(&output[0], &c[2*Zc], (block_length-2*Zc)*sizeof(unsigned char));
-  memcpy(&output[block_length-2*Zc], &d[0], ((nrows-no_punctured_columns) * Zc-removed_bit)*sizeof(unsigned char));
-  */
-  if ((((2*Zc)&31) == 0) && (((block_length-(2*Zc))&31) == 0)) {
-    //AssertFatal(((2*Zc)&31) == 0,"2*Zc needs to be a multiple of 32 for now\n");
-    //AssertFatal(((block_length-(2*Zc))&31) == 0,"block_length-(2*Zc) needs to be a multiple of 32 for now\n");
-    uint32_t l1 = (block_length-(2*Zc))>>5;
-    uint32_t l2 = ((nrows-no_punctured_columns) * Zc-removed_bit)>>5;
-    simde__m256i *c256p = (simde__m256i *)&cc[2*Zc];
-    simde__m256i *d256p = (simde__m256i *)&dd[0];
-    //  if (((block_length-(2*Zc))&31)>0) l1++;
-
-    for (int i=0;i<l1;i++)
-      //for (j=0;j<n_segments;j++) ((simde__m256i *)output[j])[i] = simde_mm256_and_si256(simde_mm256_srai_epi16(c256p[i],j),masks[0]);
-    	for (int j=macro_segment; j < macro_segment_end; j++) ((simde__m256i *)output[j])[i] = simde_mm256_and_si256(simde_mm256_srai_epi16(c256p[i],j-macro_segment),masks[0]);
-
-
-    //  if ((((nrows-no_punctured_columns) * Zc-removed_bit)&31)>0) l2++;
-
-    for (int i1=0, i=l1;i1<l2;i1++,i++)
-      //for (j=0;j<n_segments;j++) ((simde__m256i *)output[j])[i] = simde_mm256_and_si256(simde_mm256_srai_epi16(d256p[i1],j),masks[0]);
-    	for (int j=macro_segment; j < macro_segment_end; j++)  ((simde__m256i *)output[j])[i] = simde_mm256_and_si256(simde_mm256_srai_epi16(d256p[i1],j-macro_segment),masks[0]);
-  }
-  else {
-#ifdef DEBUG_LDPC
-  LOG_W(PHY,"using non-optimized version\n");
-#endif
-    // do non-SIMD version
-    for (int i=0;i<(block_length-2*Zc);i++)
-      //for (j=0; j<n_segments; j++)
-      for (int j=macro_segment; j < macro_segment_end; j++)
-	output[j][i] = (cc[2*Zc+i]>>(j-macro_segment))&1;
-    for (int i=0;i<((nrows-no_punctured_columns) * Zc-removed_bit);i++)
-      //for (j=0; j<n_segments; j++)
-    	  for (int j=macro_segment; j < macro_segment_end; j++)
-	output[j][block_length-2*Zc+i] = (dd[i]>>(j-macro_segment))&1;
-    }
-
-  if(impp->toutput != NULL) stop_meas(impp->toutput);
+  memcpy(output[0],&cc[2*Zc],(block_length-(2*Zc)));
+  memcpy(output[0]+block_length-(2*Zc),dd,((nrows-no_punctured_columns) * Zc-removed_bit));
  return 0;
 }

--- a/openair1/PHY/CODING/nr_segmentation.c
+++ b/openair1/PHY/CODING/nr_segmentation.c
@@ -151,13 +151,8 @@ else
    for (r=0; r<*C; r++) {

      k = 0;
-
-      while (k<((Kprime - L)>>3)) {
-        output_buffers[r][k] = input_buffer[s];
-		//printf("encoding segment %d : byte %d (%d) => %d\n",r,k,(Kprime-L)>>3,input_buffer[s]);
-        k++;
-        s++;
-      }
+      memcpy(output_buffers[r],input_buffer+s,(Kprime-L)>>3);
+      s+=(Kprime-L)>>3;

      if (*C > 1) { // add CRC
        crc = crc24b(output_buffers[r],Kprime-L)>>8;
--- a/openair1/PHY/MODULATION/nr_modulation.c
+++ b/openair1/PHY/MODULATION/nr_modulation.c
@@ -24,8 +24,10 @@
 #include "executables/softmodem-common.h"

 // #define DEBUG_DLSCH_PRECODING_PRINT_WITH_TRIVIAL // TODO: For debug, to be removed if want to merge to develop
-
-//Table 6.3.1.5-1 Precoding Matrix W 1 layer 2 antenna ports 'n' = -1 and 'o' = -j
+//#define DEBUG_LAYER_MAPPING
+#define USE_NEON
+//#define USE_GATHER
+// Table 6.3.1.5-1 Precoding Matrix W 1 layer 2 antenna ports 'n' = -1 and 'o' = -j
 const char nr_W_1l_2p[6][2][1] = {
    {{'1'}, {'0'}}, // pmi 0
    {{'0'}, {'1'}},
@@ -35,7 +37,7 @@ const char nr_W_1l_2p[6][2][1] = {
    {{'1'}, {'o'}} // pmi 5
 };

-//Table 6.3.1.5-3 Precoding Matrix W 1 layer 4 antenna ports 'n' = -1 and 'o' = -j
+// Table 6.3.1.5-3 Precoding Matrix W 1 layer 4 antenna ports 'n' = -1 and 'o' = -j
 const char nr_W_1l_4p[28][4][1] = {
    {{'1'}, {'0'}, {'0'}, {'0'}}, // pmi 0
    {{'0'}, {'1'}, {'0'}, {'0'}},
@@ -68,14 +70,14 @@ const char nr_W_1l_4p[28][4][1] = {
    {{'1'}, {'o'}, {'o'}, {'n'}} // pmi 27
 };

-//Table 6.3.1.5-4 Precoding Matrix W 2 antenna ports layers 2  'n' = -1 and 'o' = -j
+// Table 6.3.1.5-4 Precoding Matrix W 2 antenna ports layers 2  'n' = -1 and 'o' = -j
 const char nr_W_2l_2p[3][2][2] = {
    {{'1', '0'}, {'0', '1'}}, // pmi 0
    {{'1', '1'}, {'1', 'n'}},
    {{'1', '1'}, {'j', 'o'}} // pmi 2
 };

-//Table 6.3.1.5-5 Precoding Matrix W 2 layers 4 antenna ports 'n' = -1 and 'o' = -j
+// Table 6.3.1.5-5 Precoding Matrix W 2 layers 4 antenna ports 'n' = -1 and 'o' = -j
 const char nr_W_2l_4p[22][4][2] = {
    {{'1', '0'}, {'0', '1'}, {'0', '0'}, {'0', '0'}}, // pmi 0
    {{'1', '0'}, {'0', '0'}, {'0', '1'}, {'0', '0'}}, {{'1', '0'}, {'0', '0'}, {'0', '0'}, {'0', '1'}},
@@ -94,7 +96,7 @@ const char nr_W_2l_4p[22][4][2] = {
    {{'1', '1'}, {'o', 'o'}, {'1', 'n'}, {'o', 'j'}}, {{'1', '1'}, {'o', 'o'}, {'j', 'o'}, {'1', 'n'}} // pmi 21
 };

-//Table 6.3.1.5-6 Precoding Matrix W 3 layers 4 antenna ports 'n' = -1 and 'o' = -j
+// Table 6.3.1.5-6 Precoding Matrix W 3 layers 4 antenna ports 'n' = -1 and 'o' = -j
 const char nr_W_3l_4p[7][4][3] = {{{'1', '0', '0'}, {'0', '1', '0'}, {'0', '0', '1'}, {'0', '0', '0'}}, // pmi 0
                                  {{'1', '0', '0'}, {'0', '1', '0'}, {'1', '0', '0'}, {'0', '0', '1'}},
                                  {{'1', '0', '0'}, {'0', '1', '0'}, {'n', '0', '0'}, {'0', '0', '1'}},
@@ -103,7 +105,7 @@ const char nr_W_3l_4p[7][4][3] = {{{'1', '0', '0'}, {'0', '1', '0'}, {'0', '0',
                                  {{'1', '1', '1'}, {'n', '1', 'n'}, {'1', '1', 'n'}, {'n', '1', '1'}},
                                  {{'1', '1', '1'}, {'n', '1', 'n'}, {'j', 'j', 'o'}, {'o', 'j', 'j'}}};

-//Table 6.3.1.5-7 Precoding Matrix W 4 layers 4 antenna ports 'n' = -1 and 'o' = -j
+// Table 6.3.1.5-7 Precoding Matrix W 4 layers 4 antenna ports 'n' = -1 and 'o' = -j
 const char nr_W_4l_4p[5][4][4] = {
    {{'1', '0', '0', '0'}, {'0', '1', '0', '0'}, {'0', '0', '1', '0'}, {'0', '0', '0', '1'}}, // pmi 0
    {{'1', '1', '0', '0'}, {'0', '0', '1', '1'}, {'1', 'n', '0', '0'}, {'0', '0', '1', 'n'}},
@@ -112,137 +114,133 @@ const char nr_W_4l_4p[5][4][4] = {
    {{'1', '1', '1', '1'}, {'1', 'n', '1', 'n'}, {'j', 'j', 'o', 'o'}, {'j', 'o', 'o', 'j'}} // pmi 4
 };

-void nr_modulation(const uint32_t *in,
-                   uint32_t length,
-                   uint16_t mod_order,
-                   int16_t *out)
+void nr_modulation(const uint32_t *in, uint32_t length, uint16_t mod_order, int16_t *out)
 {
-  uint16_t mask = ((1<<mod_order)-1);
-  int32_t* nr_mod_table32;
-  int32_t* out32 = (int32_t*) out;
+  uint16_t mask = ((1 << mod_order) - 1);
+  int32_t *nr_mod_table32;
+  int32_t *out32 = (int32_t *)out;
  const uint8_t *in_bytes = (const uint8_t *)in;
  const uint64_t *in64 = (const uint64_t *)in;
-  int64_t* out64 = (int64_t*) out;
-  uint32_t i=0;
+  int64_t *out64 = (int64_t *)out;
+  uint32_t i = 0;

 #if defined(__SSE2__)
  simde__m128i *nr_mod_table128;
  simde__m128i *out128;
 #endif

-  LOG_D(PHY,"nr_modulation: length %d, mod_order %d\n",length,mod_order);
+  LOG_D(PHY, "nr_modulation: length %d, mod_order %d\n", length, mod_order);

  switch (mod_order) {
-
 #if defined(__SSE2__)
-  case 2:
-    nr_mod_table128 = (simde__m128i *)nr_qpsk_byte_mod_table;
-    out128 = (simde__m128i *)out;
-    for (i=0; i<length/8; i++)
-      out128[i] = nr_mod_table128[in_bytes[i]];
-    // the bits that are left out
-    i = i*8/2;
-    nr_mod_table32 = (int32_t*) nr_qpsk_mod_table;
-    while (i<length/2) {
-      const int idx = ((in_bytes[(i * 2) / 8] >> ((i * 2) & 0x7)) & mask);
-      out32[i] = nr_mod_table32[idx];
-      i++;
-    }
-    return;
+    case 2:
+      nr_mod_table128 = (simde__m128i *)nr_qpsk_byte_mod_table;
+      out128 = (simde__m128i *)out;
+      for (i = 0; i < length / 8; i++)
+        out128[i] = nr_mod_table128[in_bytes[i]];
+      // the bits that are left out
+      i = i * 8 / 2;
+      nr_mod_table32 = (int32_t *)nr_qpsk_mod_table;
+      while (i < length / 2) {
+        const int idx = ((in_bytes[(i * 2) / 8] >> ((i * 2) & 0x7)) & mask);
+        out32[i] = nr_mod_table32[idx];
+        i++;
+      }
+      return;
 #else
-  case 2:
-    nr_mod_table32 = (int32_t*) nr_qpsk_mod_table;
-    for (i=0; i<length/mod_order; i++) {
-      const int idx = ((in[i * 2 / 32] >> ((i * 2) & 0x1f)) & mask);
-      out32[i] = nr_mod_table32[idx];
-    }
-    return;
+    case 2:
+      nr_mod_table32 = (int32_t *)nr_qpsk_mod_table;
+      for (i = 0; i < length / mod_order; i++) {
+        const int idx = ((in[i * 2 / 32] >> ((i * 2) & 0x1f)) & mask);
+        out32[i] = nr_mod_table32[idx];
+      }
+      return;
 #endif

-  case 4:
-    out64 = (int64_t*) out;
-    for (i=0; i<length/8; i++)
-      out64[i] = nr_16qam_byte_mod_table[in_bytes[i]];
-    // the bits that are left out
-    i = i*8/4;
-    while (i<length/4) {
-      const int idx = ((in_bytes[(i * 4) / 8] >> ((i * 4) & 0x7)) & mask);
-      out32[i] = nr_16qam_mod_table[idx];
-      i++;
-    }
-    return;
-
-  case 6:
-    if (length > (3*64))
-      for (i = 0; i < length - 3 * 64; i += 3 * 64) {
-        uint64_t x = *in64++;
-        uint64_t x1 = x & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x >> 12) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x >> 24) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x >> 36) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x >> 48) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        uint64_t x2 = (x >> 60);
-        x = *in64++;
-        x2 |= x<<4;
-        x1 = x2 & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 12) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 24) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 36) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 48) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x2 = ((x>>56)&0xf0) | (x2>>60);
-        x = *in64++;
-        x2 |= x<<8;
-        x1 = x2 & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 12) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 24) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 36) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x1 = (x2 >> 48) & 0xfff;
-        *out64++ = nr_64qam_mod_table[x1];
-        x2 = ((x>>52)&0xff0) | (x2>>60);
-        *out64++ = nr_64qam_mod_table[x2];
+    case 4:
+      out64 = (int64_t *)out;
+      for (i = 0; i < length / 8; i++)
+        out64[i] = nr_16qam_byte_mod_table[in_bytes[i]];
+      // the bits that are left out
+      i = i * 8 / 4;
+      while (i < length / 4) {
+        const int idx = ((in_bytes[(i * 4) / 8] >> ((i * 4) & 0x7)) & mask);
+        out32[i] = nr_16qam_mod_table[idx];
+        i++;
      }
-    
-    while (i + 24 <= length) {
-      uint32_t xx = 0;
-      memcpy(&xx, in_bytes + i / 8, 3);
-      uint64_t x1 = xx & 0xfff;
-      *out64++ = nr_64qam_mod_table[x1];
-      x1 = (xx >> 12) & 0xfff;
-      *out64++ = nr_64qam_mod_table[x1];
-      i += 24;
-    }
-    if (i != length) {
-      uint32_t xx = 0;
-      memcpy(&xx, in_bytes + i / 8, 2);
-      uint64_t x1 = xx & 0xfff;
-      *out64++ = nr_64qam_mod_table[x1];
-    }
-    return;
+      return;

-  case 8:
-    nr_mod_table32 = (int32_t*) nr_256qam_mod_table;
-    for (i=0; i<length/8; i++)
-      out32[i] = nr_mod_table32[in_bytes[i]];
-    return;
+    case 6:
+      if (length > (3 * 64))
+        for (i = 0; i < length - 3 * 64; i += 3 * 64) {
+          uint64_t x = *in64++;
+          uint64_t x1 = x & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x >> 12) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x >> 24) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x >> 36) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x >> 48) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          uint64_t x2 = (x >> 60);
+          x = *in64++;
+          x2 |= x << 4;
+          x1 = x2 & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 12) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 24) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 36) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 48) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x2 = ((x >> 56) & 0xf0) | (x2 >> 60);
+          x = *in64++;
+          x2 |= x << 8;
+          x1 = x2 & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 12) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 24) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 36) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x1 = (x2 >> 48) & 0xfff;
+          *out64++ = nr_64qam_mod_table[x1];
+          x2 = ((x >> 52) & 0xff0) | (x2 >> 60);
+          *out64++ = nr_64qam_mod_table[x2];
+        }

-  default:
-    break;
+      while (i + 24 <= length) {
+        uint32_t xx = 0;
+        memcpy(&xx, in_bytes + i / 8, 3);
+        uint64_t x1 = xx & 0xfff;
+        *out64++ = nr_64qam_mod_table[x1];
+        x1 = (xx >> 12) & 0xfff;
+        *out64++ = nr_64qam_mod_table[x1];
+        i += 24;
+      }
+      if (i != length) {
+        uint32_t xx = 0;
+        memcpy(&xx, in_bytes + i / 8, 2);
+        uint64_t x1 = xx & 0xfff;
+        *out64++ = nr_64qam_mod_table[x1];
+      }
+      return;
+
+    case 8:
+      nr_mod_table32 = (int32_t *)nr_256qam_mod_table;
+      for (i = 0; i < length / 8; i++)
+        out32[i] = nr_mod_table32[in_bytes[i]];
+      return;
+
+    default:
+      break;
  }
-  AssertFatal(false,"Invalid or unsupported modulation order %d\n",mod_order);
+  AssertFatal(false, "Invalid or unsupported modulation order %d\n", mod_order);
 }

 void nr_layer_mapping(int nbCodes,
@@ -251,36 +249,296 @@ void nr_layer_mapping(int nbCodes,
                      uint8_t n_layers,
                      int layerSz,
                      uint32_t n_symbs,
-                      c16_t tx_layer[layerSz],
-                      int layer)
+                      c16_t tx_layers[][layerSz])
 {
-  LOG_D(PHY,"Doing layer mapping for %d layers, %d symbols\n",n_layers,n_symbs);
+  LOG_D(PHY, "Doing layer mapping for %d layers, %d symbols\n", n_layers, n_symbs);

  switch (n_layers) {
-
    case 1:
-      memcpy(tx_layer, mod_symbs[0], n_symbs * sizeof(**mod_symbs));
+      memcpy(tx_layers[0], mod_symbs[0], n_symbs * sizeof(**mod_symbs));
      break;

-    case 2:
-    case 3:
-    case 4:
-    for (int i = 0; i < n_symbs / n_layers; i++) {
-      const c16_t *base = mod_symbs[0] + n_layers * i;
-      tx_layer[i] = base[layer];
-    }
-      break;
+    case 2: {
+#if defined(__AVX512BW__)
+      int i=0;
+      __m512i perm2a = _mm512_set_epi32(30, 28, 26, 24, 22, 20, 18, 16, 14, 12, 10, 8, 6, 4, 2, 0);
+      __m512i perm2b = _mm512_set_epi32(31, 29, 27, 25, 23, 21, 19, 17, 15, 13, 11, 9, 7, 5, 3, 1);
+#ifndef USE_GATHER
+      __m512i a, b;
+      int j;
+      for (i = 0, j = 0; i < n_symbs >> 4; i += 2, j++) {
+        a = ((__m512i *)mod_symbs[0])[i];
+        b = ((__m512i *)mod_symbs[0])[i + 1];
+        ((__m512i *)tx_layers[0])[j] = _mm512_permutex2var_epi32(a, perm2a, b);
+        ((__m512i *)tx_layers[1])[j] = _mm512_permutex2var_epi32(a, perm2b, b);
+      }
+      if (i << 4 != n_symbs) {
+        for (int i2 = ((n_symbs >> 4) << 4); i2 < n_symbs; i2 += 2) {
+          tx_layers[0][i2 >> 1] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 1] = mod_symbs[0][i2 + 1];
+        }
+      }
+#else
+  for (i=0; i <n_symbs; i+=32) {
+    *(__m512i *)(tx_layers[0]+i/2)=_mm512_i32gather_epi32(perm2a, mod_symbs[0]+i, 4);
+    *(__m512i *)(tx_layers[1]+i/2)=_mm512_i32gather_epi32(perm2b, mod_symbs[0]+i, 4);
+  }
+  for (; i < n_symbs; i += 2) {
+    tx_layers[0][i >> 1] = mod_symbs[0][i];
+    tx_layers[1][i >> 1] = mod_symbs[0][i + 1];
+  }
+#endif
+#elif defined(__aarch64__)
+      int i;
+#ifdef USE_NEON
+      // SIMDe doesn't handle this properly, gcc up to 14.2 neither
+      uint8_t const perm0[16] = {0, 1, 2, 3, 8, 9, 10, 11, 4, 5, 6, 7, 12, 13, 14, 15};
+      uint8x16_t perm = vld1q_u8(perm0);
+      uint8x16_t d;
+      for (i = 0; i < n_symbs >> 2; i++) {
+        d = vqtbl1q_u8(((uint8x16_t *)mod_symbs[0])[i], perm);
+        ((int64_t *)tx_layers[0])[i] = vgetq_lane_u64((uint64x2_t)d, 0);
+        ((int64_t *)tx_layers[1])[i] = vgetq_lane_u64((uint64x2_t)d, 1);
+      }
+      if (i << 2 != n_symbs) {
+        for (int i2 = ((n_symbs >> 3) << 3); i2 < n_symbs; i2 += 2) {
+          tx_layers[0][i2 >> 1] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 1] = mod_symbs[0][i2 + 1];
+        }
+      }
+#else
+      for (i = 0; i < n_symbs; i += 2) {
+        tx_layers[0][i >> 1] = mod_symbs[0][i];
+        tx_layers[1][i >> 1] = mod_symbs[0][i + 1];
+      }
+#endif
+#else
+      int i;
+      simde__m256i perm2 = simde_mm256_set_epi32(7, 5, 3, 1, 6, 4, 2, 0);
+      simde__m256i d;
+      for (i = 0; i < n_symbs >> 3; i++) {
+        d = simde_mm256_permutevar8x32_epi32(((simde__m256i *)mod_symbs[0])[i], perm2);
+        ((simde__m128i *)tx_layers[0])[i] = simde_mm256_extractf128_si256(d, 0);
+        ((simde__m128i *)tx_layers[1])[i] = simde_mm256_extractf128_si256(d, 1);
+      }
+      if (i << 3 != n_symbs) {
+        for (int i2 = ((n_symbs >> 3) << 3); i2 < n_symbs; i2 += 2) {
+          tx_layers[0][i2 >> 1] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 1] = mod_symbs[0][i2 + 1];
+        }
+      }
+#endif
+    } break;
+    case 3: {
+#if defined(__AVX512BW__)
+      __m512i i0, i1, i2, d0;
+      __m512i perm3_0 =
+          _mm512_set_epi32(13 + 16, 10 + 16, 7 + 16, 4 + 16, 1 + 16, 14 + 16, 11 + 16, 8 + 16, 5 + 16, 2 + 16, 15, 12, 9, 6, 3, 0);
+      __m512i perm3_0b = _mm512_set_epi32(13 + 16, 10 + 16, 7 + 16, 4 + 16, 1 + 16, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
+      __m512i perm3_1 = _mm512_set_epi32(14 + 16,
+                                         11 + 16,
+                                         8 + 16,
+                                         5 + 16,
+                                         2 + 16,
+                                         15 + 16,
+                                         12 + 16,
+                                         9 + 16,
+                                         6 + 16,
+                                         3 + 16,
+                                         0 + 16,
+                                         13,
+                                         10,
+                                         7,
+                                         4,
+                                         1);
+      __m512i perm3_1b = _mm512_set_epi32(14 + 16, 11 + 16, 8 + 16, 5 + 16, 2 + 16, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
+      __m512i perm3_2 = _mm512_set_epi32(15 + 16,
+                                         12 + 16,
+                                         9 + 16,
+                                         6 + 16,
+                                         3 + 16,
+                                         0 + 16,
+                                         13 + 16,
+                                         10 + 16,
+                                         7 + 16,
+                                         4 + 16,
+                                         1 + 16,
+                                         14,
+                                         11,
+                                         8,
+                                         5,
+                                         2);
+      __m512i perm3_2b = _mm512_set_epi32(15 + 16, 12 + 16, 9 + 16, 6 + 16, 3 + 16, 0 + 16, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
+      int i,n;
+      for (i = 0, n = 0; i < n_symbs >> 4; i += 3, n++) {
+        i0 = ((__m512i *)mod_symbs[0])[i];
+        i1 = ((__m512i *)mod_symbs[0])[i + 1];
+        i2 = ((__m512i *)mod_symbs[0])[i + 2];
+        d0 = _mm512_permutex2var_epi32(i0, perm3_0, i1);
+        ((__m512i *)tx_layers[0])[n] = _mm512_permutex2var_epi32(d0, perm3_0b, i2); // 11000000
+        d0 = _mm512_permutex2var_epi32(i0, perm3_1, i1);
+        ((__m512i *)tx_layers[1])[n] = _mm512_permutex2var_epi32(d0, perm3_1b, i2); // 11000000
+        d0 = _mm512_permutex2var_epi32(i0, perm3_2, i1);
+        ((__m512i *)tx_layers[2])[n] = _mm512_permutex2var_epi32(d0, perm3_2b, i2); // 11000000
+      }
+#elif defined(__aarch64__)
+      for (int i = 0; i < n_symbs; i += 3) {
+        tx_layers[0][i / 3] = mod_symbs[0][i];
+        tx_layers[1][i / 3] = mod_symbs[0][i + 1];
+        tx_layers[2][i / 3] = mod_symbs[0][i + 2];
+      }
+
+#else
+      simde__m256i i0, i1, i2, d0, d1, d2, d3;
+      simde__m256i perm3_0 = simde_mm256_set_epi32(5, 2, 7, 4, 1, 6, 3, 0);
+      simde__m256i perm3_1 = simde_mm256_set_epi32(6, 3, 0, 5, 2, 7, 4, 1);
+      simde__m256i perm3_2 = simde_mm256_set_epi32(7, 4, 1, 6, 3, 0, 5, 2);
+      int i,n;
+      for (i = 0, n = 0; i < n_symbs >> 3; i += 3, n++) {
+        i0 = ((simde__m256i *)mod_symbs[0])[i];
+        i1 = ((simde__m256i *)mod_symbs[0])[i + 1];
+        i2 = ((simde__m256i *)mod_symbs[0])[i + 2];
+        d0 = simde_mm256_permutevar8x32_epi32(i0, perm3_0);
+        d1 = simde_mm256_permutevar8x32_epi32(i1, perm3_0);
+        d2 = simde_mm256_permutevar8x32_epi32(i2, perm3_0);
+        d3 = simde_mm256_blend_epi32(d0, d1, 0x38); // 00111000
+        ((simde__m256i *)tx_layers[0])[n] = simde_mm256_blend_epi32(d3, d2, 0xc0); // 11000000
+        d0 = simde_mm256_permutevar8x32_epi32(i0, perm3_1);
+        d1 = simde_mm256_permutevar8x32_epi32(i1, perm3_1);
+        d2 = simde_mm256_permutevar8x32_epi32(i2, perm3_1);
+        d3 = simde_mm256_blend_epi32(d0, d1, 0x18); // 00011000
+        ((simde__m256i *)tx_layers[1])[n] = simde_mm256_blend_epi32(d3, d2, 0xe0); // 11100000
+        d0 = simde_mm256_permutevar8x32_epi32(i0, perm3_2);
+        d1 = simde_mm256_permutevar8x32_epi32(i1, perm3_2);
+        d2 = simde_mm256_permutevar8x32_epi32(i2, perm3_2);
+        d3 = simde_mm256_blend_epi32(d0, d1, 0x1c); // 00011100
+        ((simde__m256i *)tx_layers[2])[n] = simde_mm256_blend_epi32(d3, d2, 0xe0); // 11100000
+#ifdef DEBUG_LAYER_MAPPING
+        printf("\nsymb %d/%d\n", i << 3, n_symbs);
+        printf(" layer 0:\t");
+        for (int j = 0; j < 8 * 6; j += 6) {
+          printf("%d %d ", ((int16_t *)&mod_symbs[0][i << 3])[j], ((int16_t *)&mod_symbs[0][i << 3])[j + 1]);
+        }
+        printf("\n layer 1:\t");
+        for (int j = 2; j < 8 * 6; j += 6) {
+          printf("%d %d ", ((int16_t *)&mod_symbs[0][i << 3])[j], ((int16_t *)&mod_symbs[0][i << 3])[j + 1]);
+        }
+        printf("\n layer 2:\t");
+        for (int j = 4; j < 8 * 6; j += 6) {
+          printf("%d %d ", ((int16_t *)&mod_symbs[0][i << 3])[j], ((int16_t *)&mod_symbs[0][i << 3])[j + 1]);
+        }
+        printf("\n Mapping layer 0:\t");
+        for (int j = 0; j < 16; j++) {
+          printf("%d ", ((int16_t *)&tx_layers[0][n << 3])[j]);
+        }
+        printf("\n Mapping layer 1:\t");
+        for (int j = 0; j < 16; j++) {
+          printf("%d ", ((int16_t *)&tx_layers[1][n << 3])[j]);
+        }
+        printf("\n Mapping layer 2:\t");
+        for (int j = 0; j < 16; j++) {
+          printf("%d ", ((int16_t *)&tx_layers[2][n << 3])[j]);
+        }
+#endif
+      }
+      if (i << 3 != n_symbs) {
+        for (int i2 = ((n_symbs >> 3) << 3); i2 < n_symbs; i2 += 3) {
+          tx_layers[0][i2 / 3] = mod_symbs[0][i2];
+          tx_layers[1][i2 / 3] = mod_symbs[0][i2 + 1];
+          tx_layers[2][i2 / 3] = mod_symbs[0][i2 + 2];
+        }
+      }
+#endif
+    } break;
+
+    case 4: {
+#if defined(__AVX512BW__)
+      __m512i perm4 = _mm512_set_epi32(15, 11, 7, 3, 14, 10, 6, 2, 13, 9, 5, 1, 12, 8, 4, 0);
+      __m512i e;
+      int i;
+      for (i = 0; i < n_symbs >> 4; i++) {
+        e = _mm512_permutexvar_epi32(perm4, ((__m512i *)mod_symbs[0])[i]);
+        ((__m128i *)tx_layers[0])[i] = _mm512_extracti64x2_epi64(e, 0);
+        ((__m128i *)tx_layers[1])[i] = _mm512_extracti64x2_epi64(e, 1);
+        ((__m128i *)tx_layers[2])[i] = _mm512_extracti64x2_epi64(e, 2);
+        ((__m128i *)tx_layers[3])[i] = _mm512_extracti64x2_epi64(e, 3);
+      }
+      if (i << 4 != n_symbs) {
+        for (int i2 = ((n_symbs >> 4) << 4); i2 < n_symbs; i2 += 4) {
+          tx_layers[0][i2 >> 2] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 2] = mod_symbs[0][i2 + 1];
+          tx_layers[2][i2 >> 2] = mod_symbs[0][i2 + 2];
+          tx_layers[3][i2 >> 2] = mod_symbs[0][i2 + 3];
+        }
+      }
+#elif defined(__aarch64__)
+      int i;
+#ifdef USE_NEON
+      // SIMDe doesn't handle this properly, gcc up to 14.2 neither
+      uint32x4_t d4;
+      for (i = 0; i < n_symbs >> 2; i++) {
+        d4 = ((uint32x4_t *)mod_symbs[0])[i];
+        ((uint32_t *)tx_layers[0])[i] = vgetq_lane_u32(d4, 0);
+        ((uint32_t *)tx_layers[1])[i] = vgetq_lane_u32(d4, 1);
+        ((uint32_t *)tx_layers[2])[i] = vgetq_lane_u32(d4, 0);
+        ((uint32_t *)tx_layers[3])[i] = vgetq_lane_u32(d4, 1);
+      }
+      if (i << 2 != n_symbs) {
+        for (int i2 = ((n_symbs >> 2) << 2); i2 < n_symbs; i2 += 4) {
+          tx_layers[0][i2 >> 2] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 2] = mod_symbs[0][i2 + 1];
+          tx_layers[0][i2 >> 2] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 2] = mod_symbs[0][i2 + 1];
+        }
+      }
+#else
+      for (int i = 0; i < n_symbs; i += 4) {
+        tx_layers[0][i >> 2] = mod_symbs[0][i];
+        tx_layers[1][i >> 2] = mod_symbs[0][i + 1];
+        tx_layers[2][i >> 2] = mod_symbs[0][i + 2];
+        tx_layers[3][i >> 2] = mod_symbs[0][i + 3];
+      }
+#endif
+#else
+      simde__m256i perm4 = simde_mm256_set_epi32(7, 3, 6, 2, 5, 1, 4, 0);
+      simde__m256i e;
+      int i;
+      for (i = 0; i < n_symbs >> 3; i++) {
+        e = simde_mm256_permutevar8x32_epi32(((simde__m256i *)mod_symbs[0])[i], perm4);
+        ((uint64_t *)tx_layers[0])[i] = simde_mm256_extract_epi64(e, 0);
+        ((uint64_t *)tx_layers[1])[i] = simde_mm256_extract_epi64(e, 1);
+        ((uint64_t *)tx_layers[2])[i] = simde_mm256_extract_epi64(e, 2);
+        ((uint64_t *)tx_layers[3])[i] = simde_mm256_extract_epi64(e, 3);
+      }
+      if (i << 3 != n_symbs) {
+        for (int i2 = ((n_symbs >> 3) << 3); i2 < n_symbs; i2 += 3) {
+          tx_layers[0][i2 >> 1] = mod_symbs[0][i2];
+          tx_layers[1][i2 >> 1] = mod_symbs[0][i2 + 1];
+          tx_layers[2][i2 >> 1] = mod_symbs[0][i2 + 2];
+          tx_layers[3][i2 >> 1] = mod_symbs[0][i2 + 2];
+        }
+      }
+#endif
+    } break;

    case 5:
-      if (layer < 2)
-        for (int i = 0; i < n_symbs; i += 2) {
+    case 6:
+    case 7:
+    case 8:
+      /*
+      // Layer 0,1
+      for (int i = 0; i < n_symbs; i += 2) {
          const int txIdx = i / 2;
-          tx_layer[txIdx] = mod_symbs[0][i + layer];
-        }
+          tx_layer[0][txIdx] = mod_symbs[0][i];
+          tx_layer[1][txIdx] = mod_symbs[0][i + 1];
+      }
+      // layers 2,3,4
      else
        for (int i = 0; i < n_symbs; i += 3) {
          const int txIdx = i / 3;
-          tx_layer[txIdx] = mod_symbs[1][i + layer];
+          tx_layer[2][txIdx] = mod_symbs[1][i + 2];
+          tx_layer[3][txIdx] = mod_symbs[1][i + 3];
+          tx_layer[4][txIdx] = mod_symbs[1][i + 4];
        }
      break;

@@ -288,7 +546,12 @@ void nr_layer_mapping(int nbCodes,
      for (int q=0; q<2; q++)
        for (int i = 0; i < n_symbs; i += 3) {
          const int txIdx = i / 3;
-          tx_layer[txIdx] = mod_symbs[q][i + layer];
+          tx_layer[0][txIdx] = mod_symbs[q][i + layer];
+          tx_layer[1][txIdx] = mod_symbs[q][i + layer];
+          tx_layer[2][txIdx] = mod_symbs[q][i + layer];
+          tx_layer[3][txIdx] = mod_symbs[q][i + layer];
+          tx_layer[4][txIdx] = mod_symbs[q][i + layer];
+          tx_layer[5][txIdx] = mod_symbs[q][i + layer];
        }
      break;

@@ -312,7 +575,7 @@ void nr_layer_mapping(int nbCodes,
        tx_layer[txIdx] = mod_symbs[q][i + layer];
      }
      break;
-
+*/
    default:
      AssertFatal(0, "Invalid number of layers %d\n", n_layers);
  }
@@ -337,7 +600,7 @@ void nr_dft(c16_t *z, c16_t *d, uint32_t Msc_PUSCH)
  simde__m128i norm128;

  if ((Msc_PUSCH % 1536) > 0) {
-    for (i = 0, ip = 0; i < Msc_PUSCH; i++, ip+=4) {
+    for (i = 0, ip = 0; i < Msc_PUSCH; i++, ip += 4) {
      dft_in0[ip] = d[i];
    }
  }
@@ -357,19 +620,18 @@ void nr_dft(c16_t *z, c16_t *d, uint32_t Msc_PUSCH)
  }

  if ((Msc_PUSCH % 1536) > 0) {
-    for (i = 0, ip = 0; i < Msc_PUSCH; i++, ip+=4)
+    for (i = 0, ip = 0; i < Msc_PUSCH; i++, ip += 4)
      z[i] = dft_out0[ip];
  }
-
 }

 void perform_symbol_rotation(NR_DL_FRAME_PARMS *fp, double f0, c16_t *symbol_rotation)
 {
-  const int nsymb = fp->symbols_per_slot * fp->slots_per_frame/10;
-  const double Tc=(1/480e3/4096);
-  const double Nu=2048*64*(1/(float)(1<<fp->numerology_index));
-  const double Ncp0=16*64 + (144*64*(1/(float)(1<<fp->numerology_index)));
-  const double Ncp1=(144*64*(1/(float)(1<<fp->numerology_index)));
+  const int nsymb = fp->symbols_per_slot * fp->slots_per_frame / 10;
+  const double Tc = (1 / 480e3 / 4096);
+  const double Nu = 2048 * 64 * (1 / (float)(1 << fp->numerology_index));
+  const double Ncp0 = 16 * 64 + (144 * 64 * (1 / (float)(1 << fp->numerology_index)));
+  const double Ncp1 = (144 * 64 * (1 / (float)(1 << fp->numerology_index)));

  LOG_D(PHY, "Doing symbol rotation calculation for TX/RX, f0 %f Hz, Nsymb %d\n", f0, nsymb);

@@ -430,7 +692,9 @@ void init_timeshift_rotation(NR_DL_FRAME_PARMS *fp)
    fp->timeshift_symbol_rotation[i].i = (int16_t)round(exp_im * 32767);

    if (i < 10)
-      LOG_D(PHY,"Timeshift symbol rotation %d => (%d,%d) %f\n",i,
+      LOG_D(PHY,
+            "Timeshift symbol rotation %d => (%d,%d) %f\n",
+            i,
            fp->timeshift_symbol_rotation[i].r,
            fp->timeshift_symbol_rotation[i].i,
            poff);
@@ -441,13 +705,13 @@ c16_t nr_layer_precoder(int sz, c16_t datatx_F_precoding[][sz], const char *prec
 {
  c16_t precodatatx_F = {0};

-  for (int al = 0; al<n_layers; al++) {
+  for (int al = 0; al < n_layers; al++) {
    c16_t antenna = datatx_F_precoding[al][re_offset];
    switch (prec_matrix[al]) {
-      case '0': //multiply by zero
+      case '0': // multiply by zero
        break;

-      case '1': //multiply by 1
+      case '1': // multiply by 1
        precodatatx_F = c16add(precodatatx_F, antenna);
        break;

@@ -492,15 +756,15 @@ c16_t nr_layer_precoder_cm(int n_layers,
 }

 void nr_layer_precoder_simd(const int n_layers,
-                           const int n_symbols,
-                           const int symSz,
-                           const c16_t txdataF_res_mapped[n_layers][n_symbols][symSz],
-                           const int ant,
-                           const nfapi_nr_pm_pdu_t *pmi_pdu,
-                           const int symbol,
-                           const int sc_offset,
-                           const int re_cnt,
-                           c16_t *txdataF_precoded)
+                            const int n_symbols,
+                            const int symSz,
+                            const c16_t txdataF_res_mapped[n_layers][n_symbols][symSz],
+                            const int ant,
+                            const nfapi_nr_pm_pdu_t *pmi_pdu,
+                            const int symbol,
+                            const int sc_offset,
+                            const int re_cnt,
+                            c16_t *txdataF_precoded)
 {
  uint32_t sc = sc_offset;
  c16_t prec_weight = {0};
@@ -510,22 +774,25 @@ void nr_layer_precoder_simd(const int n_layers,
  // 256 SIMD: Do 8 RE in one iteration, 3 iterations for 2 RB
 #ifdef __AVX2__
  const uint32_t re_cnt_align8 = re_cnt & ~7;
-  for(; sc < sc_offset + (re_cnt_align8); sc += sizeof(simde__m256i) / sizeof(prec_weight)) {
+  for (; sc < sc_offset + (re_cnt_align8); sc += sizeof(simde__m256i) / sizeof(prec_weight)) {
    // Matrix multiplication for 4 elements of the result (sizeof(simde__m256i) / sizeof(*prec_matrix) = 8)
    simde__m256i y = simde_mm256_set1_epi16(0); // Y = W[0]*X[0] + W[1]*X[1] + ... + W[nrOfLayers-1]*X[nrOfLayers-1]
-    for(int nl = 0; nl < n_layers; nl++) {
+    for (int nl = 0; nl < n_layers; nl++) {
      prec_weight.r = pmi_pdu->weights[nl][ant].precoder_weight_Re;
      prec_weight.i = pmi_pdu->weights[nl][ant].precoder_weight_Im;

      const simde__m256i x = simde_mm256_loadu_epi32(&txdataF_res_mapped[nl][symbol][sc]);

      // Rearrange precoding matrix weight to match complex multiplication and broadcast it to match SIMD size
-      const simde__m256i w_c   = simde_mm256_set1_epi32(c16toI32(c16conj(prec_weight)));   // broadcast conjugate of w
-      const simde__m256i w_s   = simde_mm256_set1_epi32(c16toI32(c16swap(prec_weight)));   // broadcast swapped real and img of w
+      const simde__m256i w_c = simde_mm256_set1_epi32(c16toI32(c16conj(prec_weight))); // broadcast conjugate of w
+      const simde__m256i w_s = simde_mm256_set1_epi32(c16toI32(c16swap(prec_weight))); // broadcast swapped real and img of w

      // Multiplication and shift
-      const simde__m256i reals = simde_mm256_srai_epi32(simde_mm256_madd_epi16(x, w_c), 15); // (int32_t) .r = (x.r * w.r - x.i * w.i) >> 15
-      const simde__m256i imags = simde_mm256_slli_epi32(simde_mm256_madd_epi16(x, w_s), 1);  // (int32_t) .i = (x.r * w.i + x.i * w.r) << 1, since higher 16 bit of each 32 bit is taken by blend_epi16
+      const simde__m256i reals =
+          simde_mm256_srai_epi32(simde_mm256_madd_epi16(x, w_c), 15); // (int32_t) .r = (x.r * w.r - x.i * w.i) >> 15
+      const simde__m256i imags = simde_mm256_slli_epi32(
+          simde_mm256_madd_epi16(x, w_s),
+          1); // (int32_t) .i = (x.r * w.i + x.i * w.r) << 1, since higher 16 bit of each 32 bit is taken by blend_epi16

      // Re-arrange to match c16_t format
      const simde__m256i produ = simde_mm256_blend_epi16(reals, imags, 0xAA);
@@ -540,43 +807,40 @@ void nr_layer_precoder_simd(const int n_layers,

  // 128 SIMD: Do 4 RE in one iteration, 3 iterations for 1 RB
  const uint32_t re_cnt_align4 = re_cnt & ~3;
-  for(; sc < sc_offset+re_cnt_align4; sc += sizeof(simde__m128i) / sizeof(prec_weight)) {
-    #ifdef DEBUG_DLSCH_PRECODING_PRINT_WITH_TRIVIAL // Get result with trivial solution, TODO: To be removed
-      c16_t y_triv[4];
-      for(int i = 0; i < 4; i++)
-        y_triv[i] = nr_layer_precoder_cm(n_layers,
-                                         NR_SYMBOLS_PER_SLOT,
-                                         symSz,
-                                         txdataF_res_mapped,
-                                         ant,
-                                         pmi_pdu,
-                                         symbol,
-                                         sc + i);
-      memcpy(&txdataF_precoded[sc], y_triv, sizeof(y_triv));
-    #endif
+  for (; sc < sc_offset + re_cnt_align4; sc += sizeof(simde__m128i) / sizeof(prec_weight)) {
+#ifdef DEBUG_DLSCH_PRECODING_PRINT_WITH_TRIVIAL // Get result with trivial solution, TODO: To be removed
+    c16_t y_triv[4];
+    for (int i = 0; i < 4; i++)
+      y_triv[i] = nr_layer_precoder_cm(n_layers, NR_SYMBOLS_PER_SLOT, symSz, txdataF_res_mapped, ant, pmi_pdu, symbol, sc + i);
+    memcpy(&txdataF_precoded[sc], y_triv, sizeof(y_triv));
+#endif

    // Matrix multiplication for 4 elements of the result (sizeof(simde__m128i) / sizeof(c16_t) = 4)
    simde__m128i y = simde_mm_set1_epi16(0); // Y = W[0]*X[0] + W[1]*X[1] + ... + W[nrOfLayers-1]*X[nrOfLayers-1]
-    for(int nl = 0; nl < n_layers; nl++) {
+    for (int nl = 0; nl < n_layers; nl++) {
      prec_weight.r = pmi_pdu->weights[nl][ant].precoder_weight_Re;
      prec_weight.i = pmi_pdu->weights[nl][ant].precoder_weight_Im;

      const simde__m128i x = simde_mm_loadu_epi32(&txdataF_res_mapped[nl][symbol][sc]);

      // Rearrange precoding matrix weight to match complex multiplication and broadcast it to match SIMD size
-      const simde__m128i w_c   = simde_mm_set1_epi32(c16toI32(c16conj(prec_weight)));   // broadcast conjugate of w
-      const simde__m128i w_s   = simde_mm_set1_epi32(c16toI32(c16swap(prec_weight)));   // broadcast swapped real and img of w
+      const simde__m128i w_c = simde_mm_set1_epi32(c16toI32(c16conj(prec_weight))); // broadcast conjugate of w
+      const simde__m128i w_s = simde_mm_set1_epi32(c16toI32(c16swap(prec_weight))); // broadcast swapped real and img of w

      // Multiplication and shift
-      const simde__m128i reals = simde_mm_srai_epi32(simde_mm_madd_epi16(x, w_c), 15); // (int32_t) .r = (x.r * w.r - x.i * w.i) >> 15
-      const simde__m128i imags = simde_mm_slli_epi32(simde_mm_madd_epi16(x, w_s), 1);  // (int32_t) .i = (x.r * w.i + x.i * w.r) << 1, since higher 16 bit of each 32 bit is taken by blend_epi16
+      const simde__m128i reals =
+          simde_mm_srai_epi32(simde_mm_madd_epi16(x, w_c), 15); // (int32_t) .r = (x.r * w.r - x.i * w.i) >> 15
+      const simde__m128i imags = simde_mm_slli_epi32(
+          simde_mm_madd_epi16(x, w_s),
+          1); // (int32_t) .i = (x.r * w.i + x.i * w.r) << 1, since higher 16 bit of each 32 bit is taken by blend_epi16

      /* Re-arrange to match c16_t format
-        bit index: 0            | 16              | 32           | 48              | 64           | 80              | 96           | 112
-        reals =   {R0.r[15..30] | R0.r[31] (0)*15 | R1.r[15..30] | R1.r[31] (0)*15 | R2.r[15..30] | R2.r[31] (0)*15 | R3.r[15..30] | R3.r[31] (0)*15}
-        imags =   {0 R0.i[0..14]| R0.i[15..30]    | 0 R1.i[0..14]| R1.i[15..30]    | 0 R2.i[0..14]| R2.i[15..30]    | 0 R3.i[0..14]| R3.i[15..30]   }
-        16b from  {reals        | imags           | reals        | imags           | reals        | imags           | reals        | imags          }
-        produ =   {R0.r[15..30] | R0.i[15..30]    | R1.r[15..30] | R1.i[15..30]    | R2.r[15..30] | R2.i[15..30]    | R3.r[15..30] | R3.i[15..30]   }
+        bit index: 0            | 16              | 32           | 48              | 64           | 80              | 96           |
+        112 reals =   {R0.r[15..30] | R0.r[31] (0)*15 | R1.r[15..30] | R1.r[31] (0)*15 | R2.r[15..30] | R2.r[31] (0)*15 |
+        R3.r[15..30] | R3.r[31] (0)*15} imags =   {0 R0.i[0..14]| R0.i[15..30]    | 0 R1.i[0..14]| R1.i[15..30]    | 0 R2.i[0..14]|
+        R2.i[15..30]    | 0 R3.i[0..14]| R3.i[15..30]   } 16b from  {reals        | imags           | reals        | imags | reals
+        | imags           | reals        | imags          } produ =   {R0.r[15..30] | R0.i[15..30]    | R1.r[15..30] | R1.i[15..30]
+        | R2.r[15..30] | R2.i[15..30]    | R3.r[15..30] | R3.i[15..30]   }
      */
      const simde__m128i produ = simde_mm_blend_epi16(reals, imags, 0xAA);

@@ -586,30 +850,30 @@ void nr_layer_precoder_simd(const int n_layers,
    // Store the result to txdataF
    simde_mm_storeu_si128(&txdataF_precoded[sc], y);

-    #ifdef DEBUG_DLSCH_PRECODING_PRINT_WITH_TRIVIAL // Print simd and trivial result, TODO: To be removed
-      c16_t *y_simd = (c16_t*) &y;
-      printf("debug_to_be_removed re_cnt=%d, sc=%u, y_simd=(%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d)\n",
-             re_cnt,
-             sc,
-             y_simd[0].r,
-             y_simd[0].i,
-             y_simd[1].r,
-             y_simd[1].i,
-             y_simd[2].r,
-             y_simd[2].i,
-             y_simd[3].r,
-             y_simd[3].i);
-      printf("debug_to_be_removed re_cnt=%d, sc=%u, y_triv=(%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d)\n",
-             re_cnt,
-             sc,
-             y_triv[0].r,
-             y_triv[0].i,
-             y_triv[1].r,
-             y_triv[1].i,
-             y_triv[2].r,
-             y_triv[2].i,
-             y_triv[3].r,
-             y_triv[3].i);
-    #endif
+#ifdef DEBUG_DLSCH_PRECODING_PRINT_WITH_TRIVIAL // Print simd and trivial result, TODO: To be removed
+    c16_t *y_simd = (c16_t *)&y;
+    printf("debug_to_be_removed re_cnt=%d, sc=%u, y_simd=(%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d)\n",
+           re_cnt,
+           sc,
+           y_simd[0].r,
+           y_simd[0].i,
+           y_simd[1].r,
+           y_simd[1].i,
+           y_simd[2].r,
+           y_simd[2].i,
+           y_simd[3].r,
+           y_simd[3].i);
+    printf("debug_to_be_removed re_cnt=%d, sc=%u, y_triv=(%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d), (%+4d,%+4d)\n",
+           re_cnt,
+           sc,
+           y_triv[0].r,
+           y_triv[0].i,
+           y_triv[1].r,
+           y_triv[1].i,
+           y_triv[2].r,
+           y_triv[2].i,
+           y_triv[3].r,
+           y_triv[3].i);
+#endif
  }
 }
--- a/openair1/PHY/MODULATION/nr_modulation.h
+++ b/openair1/PHY/MODULATION/nr_modulation.h
@@ -60,8 +60,7 @@ void nr_layer_mapping(int nbCodes,
                      uint8_t n_layers,
                      int layerSz,
                      uint32_t n_symbs,
-                      c16_t tx_layers[layerSz],
-                      int l);
+                      c16_t tx_layers[][layerSz]);

 /*! \brief Perform NR layer mapping. TS 38.211 V15.4.0 subclause 7.3.1.3
  @param[in] ulsch_ue, double Pointer to NR_UE_ULSCH_t struct
--- a/openair1/PHY/NR_TRANSPORT/nr_dci.c
+++ b/openair1/PHY/NR_TRANSPORT/nr_dci.c
@@ -248,9 +248,11 @@ void nr_generate_dci_top(processingData_L1tx_t *msgTx, int slot, int txdataF_off
 {
  PHY_VARS_gNB *gNB = msgTx->gNB;
  NR_DL_FRAME_PARMS *frame_parms = &gNB->frame_parms;
+  start_meas(&gNB->dci_generation_stats);
  for (int i = 0; i < msgTx->num_ul_pdcch; i++)
    nr_generate_dci(msgTx->gNB, &msgTx->ul_pdcch_pdu[i].pdcch_pdu.pdcch_pdu_rel15, txdataF_offset, frame_parms, slot);
  for (int i = 0; i < msgTx->num_dl_pdcch; i++)
    nr_generate_dci(msgTx->gNB, &msgTx->pdcch_pdu[i].pdcch_pdu_rel15, txdataF_offset, frame_parms, slot);
+  stop_meas(&gNB->dci_generation_stats);
 }

--- a/openair1/PHY/NR_TRANSPORT/nr_dlsch.c
+++ b/openair1/PHY/NR_TRANSPORT/nr_dlsch.c
--- a/openair1/PHY/NR_TRANSPORT/nr_scrambling.c
+++ b/openair1/PHY/NR_TRANSPORT/nr_scrambling.c
@@ -33,12 +33,33 @@ void nr_codeword_scrambling(uint8_t *in,
 {
  const int roundedSz = (size + 31) / 32;
  uint32_t *seq = gold_cache((n_RNTI << 15) + (q << 14) + Nid, roundedSz);
+  for (int i = 0; i < roundedSz; i++) 
+    out[i] = ((uint32_t*)in)[i] ^ seq[i];
+#if 0
+//#elif __AVX2__
  for (int i = 0; i < roundedSz; i++) {
-    simde__m256i c = ((simde__m256i*)in)[i];
-    uint32_t in32 = simde_mm256_movemask_epi8(simde_mm256_slli_epi16(c, 7));
+    __m256i c = ((__m256i*)in)[i];
+    uint32_t in32 = _mm256_movemask_epi8(_mm256_slli_epi16(c, 7));
    out[i] = in32 ^ seq[i];
    DEBUG_SCRAMBLING(LOG_D(PHY, "in[%d] %x => %x\n", i, in32, out[i]));
  }
+//#elif defined (__aarch64__)
+  const int8_t __attribute__ ((aligned (16))) ucShift[] = {0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7};
+  int8x16_t vshift = vld1q_s8(ucShift);
+  uint8x16_t *c = (uint8x16_t*)in;		     
+  uint8x16_t cshift; 
+  uint32_t in32;
+  for (int i = 0; i < roundedSz; i++) {
+    cshift = vshlq_u8(*c++, vshift);
+    in32 = vaddv_u8(vget_low_u8(cshift));
+    in32 += (vaddv_u8(vget_high_u8(cshift)) << 8);
+    cshift = vshlq_u8(*c++, vshift);
+    in32 += (vaddv_u8(vget_low_u8(cshift)) << 16);
+    in32 += (vaddv_u8(vget_high_u8(cshift)) << 24);
+    out[i] = in32 ^ seq[i];
+    DEBUG_SCRAMBLING(LOG_D(PHY, "in[%d] %x => %x\n", i, in32, out[i]));
+  }
+#endif
 }

 void nr_codeword_unscrambling(int16_t* llr, uint32_t size, uint8_t q, uint32_t Nid, uint32_t n_RNTI)
--- a/openair1/PHY/TOOLS/cmult_sv.c
+++ b/openair1/PHY/TOOLS/cmult_sv.c
@@ -156,9 +156,8 @@ void rotate_cpx_vector(const c16_t *const x, const c16_t *const alpha, c16_t *y,
    // log2_amp - increase the output amplitude by a factor 2^log2_amp (default is 0)
    //            WARNING: log2_amp>0 can cause overflow!!

-    uint32_t i;                 // loop counter
-

+#if 0
    simd_q15_t *y_128,alpha_128;
    int32_t *xd=(int32_t *)x; 

@@ -194,6 +193,39 @@ void rotate_cpx_vector(const c16_t *const x, const c16_t *const alpha, c16_t *y,
      );
      //print_ints("y_128[0]=", &y_128[0]);
    }
+#else
+
+    const c16_t for_re = {alpha->r, -alpha->i};
+    const simde__m128i alpha_for_real = simde_mm_set1_epi32(*(uint32_t *)&for_re);
+    const c16_t for_im = {alpha->i, alpha->r};
+    const simde__m128i alpha_for_im = simde_mm_set1_epi32(*(uint32_t *)&for_im);
+    const simde__m128i perm_mask = simde_mm_set_epi8( 15,
+                                                      14,
+                                                      7,
+                                                      6,
+                                                      13,
+                                                      12,
+                                                      5,
+                                                      4,
+                                                      11,
+                                                      10,
+                                                      3,
+                                                      2,
+                                                      9,
+                                                      8,
+                                                      1,
+                                                      0);
+    simde__m128i *xd = (simde__m128i *)x;
+    const simde__m128i *end = xd + N / 8;
+    for (simde__m128i *yd = (simde__m128i *)y; xd < end; yd++, xd++) {
+      const simde__m128i y128 = simde_mm_lddqu_si128(xd);
+      const simde__m128i xre = simde_mm_srai_epi32(simde_mm_madd_epi16(y128, alpha_for_real), output_shift);
+      const simde__m128i xim = simde_mm_srai_epi32(simde_mm_madd_epi16(y128, alpha_for_im), output_shift);
+      // a bit faster than unpacklo+unpackhi+packs
+      const simde__m128i tmp = simde_mm_packs_epi32(xre, xim);
+      simde_mm_storeu_si128(yd, simde_mm_shuffle_epi8(tmp, perm_mask));
+    }
+#endif
 #if defined(__x86__) || defined(__x86_64__)
  }
 #endif
--- a/openair1/PHY/defs_gNB.h
+++ b/openair1/PHY/defs_gNB.h
@@ -521,6 +521,8 @@ typedef struct PHY_VARS_gNB_s {
  time_stats_t dlsch_interleaving_stats;
  time_stats_t dlsch_segmentation_stats;

+  time_stats_t dci_generation_stats;
+  time_stats_t phase_comp_stats;
  time_stats_t rx_pusch_stats;
  time_stats_t rx_pusch_init_stats;
  time_stats_t rx_pusch_symbol_processing_stats;
--- a/openair1/PHY/sse_intrin.h
+++ b/openair1/PHY/sse_intrin.h
@@ -55,6 +55,8 @@

 /* x86 processors */

+#define oai_mm_sign_epi8(a,b) simde_mm_sign_epi8(a,b)
+
 #if defined(__AVX512BW__) || defined(__AVX512F__)
 #include <immintrin.h>
 #endif
@@ -66,6 +68,7 @@
 #include <simde/x86/clmul.h>

 #include <simde/arm/neon.h>
+#define oai_mm_sign_epi8(a,b) (simde__m128i)vbslq_s8(vcltzq_s8((int8x16_t) (b)), vnegq_s8((int8x16_t) (a)), (int8x16_t)(a))
 #endif // x86_64 || i386
 #include <stdbool.h>
 #include "assertions.h"
--- a/openair1/SCHED_NR/phy_procedures_nr_gNB.c
+++ b/openair1/SCHED_NR/phy_procedures_nr_gNB.c
@@ -286,6 +286,7 @@ void phy_procedures_gNB_TX(processingData_L1tx_t *msgTx,

  //apply the OFDM symbol rotation here
  if (gNB->phase_comp) {
+    start_meas(&gNB->phase_comp_stats);
    for(int i = 0; i < gNB->common_vars.num_beams_period; ++i) {
      for (int aa = 0; aa < cfg->carrier_config.num_tx_ant.value; aa++) {
        apply_nr_rotation_TX(fp,
@@ -300,6 +301,7 @@ void phy_procedures_gNB_TX(processingData_L1tx_t *msgTx,
          T_INT(aa), T_BUFFER(&gNB->common_vars.txdataF[aa][txdataF_offset], fp->samples_per_slot_wCP*sizeof(int32_t)));
      }
    }
+    stop_meas(&gNB->phase_comp_stats);
  }

  VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_PHY_PROCEDURES_gNB_TX + gNB->CC_id, 0);
--- a/openair1/SIMULATION/NR_PHY/dlschsim.c
+++ b/openair1/SIMULATION/NR_PHY/dlschsim.c
@@ -511,7 +511,7 @@ int main(int argc, char **argv)

 	//printf("crc32: [0]->0x%08x\n",crc24c(test_input, 32));
 	// generate signal
-        unsigned char output[rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS] __attribute__((aligned(32)));
+        unsigned char output[rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS] __attribute__((aligned(64)));
        bzero(output,rel15->rbSize * NR_SYMBOLS_PER_SLOT * NR_NB_SC_PER_RB * 8 * NR_MAX_NB_LAYERS);
 	if (input_fd == NULL) {
    msgDataTx.num_pdsch_slot = 1;
@@ -532,12 +532,12 @@ int main(int argc, char **argv)
 				//if (i<16)
 				//   printf("encoder output f[%d] = %d\n",i,dlsch->harq_processes[0]->f[i]);

-				if (output[i] == 0)
+				if ((output[i]&1) == 0)
 					modulated_input[i] = 1.0;        ///sqrt(2);  //QPSK
 				else
 					modulated_input[i] = -1.0;        ///sqrt(2);

-				//if (i<16) printf("modulated_input[%d] = %d\n",i,modulated_input[i]);
+				//if (i<16) printf("modulated_input[%d] = %f (%d)\n",i,modulated_input[i],output[i]);
 				//SNR =10;
 				SNR_lin = pow(10, SNR / 10.0);
 				sigma = 1.0 / sqrt(2 * SNR_lin);
--- a/openair1/SIMULATION/NR_PHY/dlsim.c
+++ b/openair1/SIMULATION/NR_PHY/dlsim.c
@@ -962,10 +962,12 @@ int main(int argc, char **argv)
    reset_meas(&gNB->dlsch_segmentation_stats);
    reset_meas(&gNB->dlsch_modulation_stats);
    reset_meas(&gNB->dlsch_encoding_stats);
+    reset_meas(&gNB->dci_generation_stats);
    reset_meas(&gNB->tinput);
    reset_meas(&gNB->tprep);
    reset_meas(&gNB->tparity);
    reset_meas(&gNB->toutput);
+    reset_meas(&gNB->phase_comp_stats);

    uint32_t errors_scrambling[16] = {0};
    int n_errors[16] = {0};
@@ -1263,6 +1265,7 @@ int main(int argc, char **argv)
             UE->dl_harq_processes[0][slot].C,
             msgDataTx->dlsch[0][0].harq_process.pdsch_pdu.pdsch_pdu_rel15.TBSize[0] << 3);
      printDistribution(&gNB->phy_proc_tx,table_tx,"PHY proc tx");
+      printStatIndent2(&gNB->dci_generation_stats,"DCI encoding time");
      printStatIndent2(&gNB->dlsch_encoding_stats,"DLSCH encoding time");
      printStatIndent3(&gNB->dlsch_segmentation_stats,"DLSCH segmentation time");
      printStatIndent3(&gNB->tinput,"DLSCH LDPC input processing time");
@@ -1275,6 +1278,7 @@ int main(int argc, char **argv)
      printStatIndent2(&gNB->dlsch_scrambling_stats, "DLSCH scrambling time");
      printStatIndent2(&gNB->dlsch_resource_mapping_stats, "DLSCH Resource Mapping time");
      printStatIndent2(&gNB->dlsch_precoding_stats,"DLSCH Layer Precoding time");
+      if (gNB->phase_comp) printStatIndent2(&gNB->phase_comp_stats,"Phase Compensation");

      printf("\nUE function statistics (per %d us slot)\n", 1000 >> *scc->ssbSubcarrierSpacing);
      for (int i = RX_PDSCH_STATS; i <= DLSCH_PROCEDURES_STATS; i++) {
--- a/openair1/SIMULATION/NR_PHY/ulsim.c
+++ b/openair1/SIMULATION/NR_PHY/ulsim.c
@@ -239,6 +239,7 @@ int main(int argc, char *argv[])
  int threadCnt=0;
  int max_ldpc_iterations = 5;
  int num_antennas_per_thread = 1;
+  char gNBthreads[128]="n";
  if ((uniqCfg = load_configmodule(argc, argv, CONFIG_ENABLECMDLINEONLY)) == 0) {
    exit_fun("[NR_ULSIM] Error, configuration module init failed\n");
  }
@@ -250,7 +251,7 @@ int main(int argc, char *argv[])
  InitSinLUT();

  int c;
-  while ((c = getopt(argc, argv, "--:O:a:b:c:d:ef:g:h:i:k:m:n:p:q:r:s:t:u:v:w:y:z:A:C:F:G:H:I:M:N:PR:S:T:U:L:ZW:E:X:")) != -1) {
+  while ((c = getopt(argc, argv, "--:O:a:b:c:d:ef:g:h:i:k:m:n:p:q:r:s:t:u:v:w:y:z:A:C:F:G:H:I:M:N:PQR:S:T:U:L:ZW:E:X:Y:")) != -1) {

    /* ignore long options starting with '--', option '-O' and their arguments that are handled by configmodule */
    /* with this opstring getopt returns 1 for non-option arguments, refer to 'man 3 getopt' */
@@ -384,6 +385,11 @@ int main(int argc, char *argv[])
      threadCnt = atoi(optarg);
      break;

+    case 'Y':
+      strncpy(gNBthreads,optarg, sizeof(gNBthreads)-1);
+      gNBthreads[sizeof(gNBthreads)-1]=0;
+      break;
+
    case 'u':
      mu = atoi(optarg);
      break;
@@ -589,7 +595,9 @@ int main(int argc, char *argv[])
  gNB->RU_list[0] = calloc(1, sizeof(**gNB->RU_list));
  gNB->RU_list[0]->rfdevice.openair0_cfg = openair0_cfg;

-  initFloatingCoresTpool(threadCnt, &gNB->threadPool, false, "gNB-tpool");
+  if (gNBthreads[0]=='n') initFloatingCoresTpool(threadCnt, &gNB->threadPool, false, "gNB-tpool");
+  else initNamedTpool(gNBthreads, &gNB->threadPool, true, "gNB-tpool");
+
  initNotifiedFIFO(&gNB->respDecode);

  initNotifiedFIFO(&gNB->respPuschSymb);
Author	SHA1	Message	Date
Raymond Knopp	59404cc5e8	removed buffer read overflow (harmless but flagged by ASAN)	2025-01-31 21:37:52 +01:00
Raymond Knopp	73fa7d416a	Merge remote-tracking branch 'origin/ldpc_enc_avx512' into rk_integration1224	2025-01-30 20:41:37 +01:00
Raymond Knopp	59123ce85d	aarch64 testing and corrections	2025-01-30 09:27:26 +00:00
Raymond Knopp	4b3c3c2dac	fixes after testing with AVX2	2025-01-30 06:21:40 +01:00
Raymond Knopp	c380dc4fce	bit-packing done in nrLDPC_coding_segment_encoder.c now instead of nr_scrambling. Interestingly, it improves performance significantly too. Tested for AVX512 only for now	2025-01-29 22:32:03 +01:00
Raymond Knopp	40873d264d	small fix after resolving conflicts during merge with develop	2025-01-21 21:15:39 +01:00
Raymond Knopp	d8291147b2	Merge remote-tracking branch 'origin/develop' into ldpc_enc_avx512	2025-01-21 15:54:09 +01:00
Raymond Knopp	5dc22f219a	fix after ASAN test in CI (stack array size)	2025-01-03 17:12:11 +01:00
Raymond Knopp	f1a4e69386	better 128-bit implementation in rotate_cpx_vector	2025-01-03 08:14:07 +00:00
Raymond Knopp	367675c29b	Merge branch 'ldpc_enc_avx512' into rk_integration1224	2025-01-03 08:17:02 +01:00
Raymond Knopp	540da76f1c	Merge branch 'ldpc_enc_avx512' of https://gitlab.eurecom.fr/oai/openairinterface5g into ldpc_enc_avx512	2025-01-03 08:07:44 +01:00
Raymond Knopp	4340c229b1	logging	2025-01-03 07:50:48 +01:00
Raymond Knopp	5d2c881b91	3GPP Rate-matching/interleaving on 8-segment blocks, deinterleaving at output of multi-segment processing	2025-01-03 06:37:36 +00:00
Raymond Knopp	ad0d02f922	Merge branch 'ldpc_enc_avx512' into rk_integration1224	2024-12-30 07:25:42 +00:00
Raymond Knopp	a06f80e0bd	fix for AVX2 in nr_ldpc_encoder_optim8segmulti.c	2024-12-29 19:06:25 +01:00
Raymond Knopp	6bb8bf62a3	neoverse-V2 fix	2024-12-29 14:54:59 +00:00
Jason	059501fa4b	fixes for Neoverse selection in CMakeLists.txt	2024-12-29 14:44:21 +00:00
Raymond Knopp	c85bf0c1b6	changes to select neoverse-n1/n2 in CMakeLists.txt	2024-12-29 12:26:47 +00:00
Raymond Knopp	57677629ac	modifications for aarch64 (mm_movemask_epi8 in scrambling)	2024-12-29 12:17:00 +00:00
Raymond Knopp	ead75c20fb	changes for neoverse-v2	2024-12-29 09:50:02 +00:00
Raymond Knopp	58ea54f842	fix tali condition for 256-QAM with 128-bit SIMD	2024-12-21 21:38:09 +01:00
Raymond Knopp	c9536b6fa0	Merge branch 'ldpc_enc_avx512' of https://gitlab.eurecom.fr/oai/openairinterface5g into ldpc_enc_avx512	2024-12-21 18:28:04 +01:00
Raymond Knopp	c84a30650d	fix build on aarch64	2024-12-21 17:26:50 +00:00
Raymond Knopp	b7f664dcfa	fix after rebase	2024-12-18 09:34:37 +01:00
Raymond Knopp	e2cde7e679	Merge remote-tracking branch 'origin/develop' into ldpc_enc_avx512	2024-12-17 22:46:02 +01:00
Raymond Knopp	dd33d2af0f	alignment issues in new rate-matching implementation in CI (with ASAN UBsan)	2024-12-17 18:04:02 +01:00
Raymond Knopp	4b76076d23	__AVX512BW__ -> __AVX512VBMI__ in nr_rate_matching.c	2024-12-16 17:04:28 +01:00
Raymond Knopp	018c79017f	some AVX512F -> AVX512BW for permutex2var	2024-12-16 15:36:25 +01:00
Raymond Knopp	654c56b4c8	put back AVX512 in ldpc_encode_parity	2024-12-16 10:20:13 +01:00
Raymond Knopp	37421b91e9	removed used of AVX512 in ldpc_encode_parity	2024-12-16 09:52:00 +01:00
Raymond Knopp	c4bf2d5c1a	timing measuremens for interleaving/rate matching	2024-12-16 01:29:04 +01:00
Raymond Knopp	cb465b4d56	AVX512 modifications for LDPC encoding: interleaving, ZC384 BG1, output formatting for encoder, segmentation (memcpy)	2024-12-16 01:09:38 +01:00
Raymond Knopp	966325c5b0	Merge branch 'resource_mapping_optim' into rk_integration1224	2024-12-16 00:51:33 +01:00
Raymond Knopp	20d665cc6b	Merge branch 'develop' into resource_mapping_optim	2024-12-16 00:41:25 +01:00
Raymond Knopp	ed8c668ab4	another bugfix for AVX512	2024-12-16 00:22:55 +01:00
Raymond Knopp	4d51c51fd2	bugfix for AVX512 LDPC encoder optimization	2024-12-16 00:18:02 +01:00
Raymond Knopp	e3abcc9d27	AVX512 LDPC encoder optimizations for Zc=384 BG=1	2024-12-16 00:02:47 +01:00
Raymond Knopp	2676c43173	AVX512 fix for LDPC encoder	2024-12-15 21:04:50 +01:00
Raymond Knopp	435bf0aae6	AVX512 for output processing for LDPC encoder	2024-12-15 20:38:14 +01:00
Raymond Knopp	85f7c96925	memcpy in nr_segmentation.c	2024-12-15 19:03:59 +01:00
Raymond Knopp	a0a6e78d20	intermediate commit 3	2024-12-15 18:59:14 +01:00
Raymond Knopp	54cb5c0a93	intermediate commit 2	2024-12-15 18:41:46 +01:00
Raymond Knopp	e2ee18d92e	intermediate commit	2024-12-15 18:16:52 +01:00
Raymond Knopp	ede36f272c	simd optimizations for LDPC interleaving	2024-12-15 18:03:50 +01:00
Raymond Knopp	baff396b59	testing with "gather"	2024-12-04 14:48:14 +01:00
Raymond Knopp	f9799c4991	corrections for 4-layer transmission (CI tests)	2024-12-01 23:49:05 +01:00
Jason	13559a8144	fix for unaligned reads with aarch64	2024-12-01 20:59:52 +00:00
Jason	0ed7b9165e	unaligned reads in nr_dlsch.c for aarch64	2024-12-01 20:57:35 +00:00
Raymond Knopp	3999e1f292	fix for 256-bit unaligned reads in nr_dlsch.c	2024-12-01 21:03:17 +01:00
Raymond Knopp	ec615c8435	unaligned reads in tx_layers array in resource mapping	2024-12-01 20:28:13 +01:00
Raymond Knopp	1e68a0ff0c	fixed issues with new resource mapping with PTRS formats	2024-12-01 18:19:31 +01:00
Raymond Knopp	99182533e8	check for alignment in nr_dlsch.c	2024-12-01 10:13:19 +01:00
Raymond Knopp	149e53ce05	removal of "clang" warning	2024-11-28 11:27:35 +01:00
Raymond Knopp	9f94daa251	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-28 09:17:34 +01:00
Raymond Knopp	5285267770	_mmX_xor_siY(a,a) -> _mmX_setzero_siY() to avoid warnings with clang	2024-11-28 09:16:34 +01:00
Jason	03a18d24aa	yet another error, this type aarch64. hopefully the last	2024-11-27 19:53:33 +00:00
Raymond Knopp	64393a9d2d	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-27 19:45:00 +01:00
Raymond Knopp	9c10743748	avx2 build fix	2024-11-27 19:44:44 +01:00
Jason	9248a80d03	fixed for aarch64 build	2024-11-27 15:46:45 +00:00
Raymond Knopp	456e80f4ca	compilation issues for multi-platforms	2024-11-27 15:18:40 +01:00
Raymond Knopp	cc8450628d	removal of warning	2024-11-27 14:30:39 +01:00
Raymond Knopp	b44bd9319e	removed incomplete optimization for LDPC encoder for AARCH64	2024-11-26 14:42:45 +01:00
Raymond Knopp	06d34836d0	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-26 14:29:56 +01:00
Raymond Knopp	74225f2d88	fixes for ARM	2024-11-26 14:29:32 +01:00
Jason	0e70590958	fixes for ARM	2024-11-25 21:33:26 +00:00
Raymond Knopp	fd1276c535	clang-formta	2024-11-25 22:16:25 +01:00
Karim Boutiba	99f089b9af	case statements on stupix	2024-11-25 21:47:50 +01:00
Raymond Knopp	ffc3ff0b57	fixed AVX512 issues	2024-11-25 21:37:15 +01:00
Raymond Knopp	166c519011	temporary	2024-11-25 19:33:21 +01:00
Raymond Knopp	41e215bc1a	AVX512 additions for "resource mapping" in TX DLSCH procedures	2024-11-25 14:59:10 +01:00
Jason	7f407ee19d	changes for __aarch64__	2024-11-24 14:28:46 +00:00
Raymond Knopp	1a6c69893c	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-24 07:37:44 +01:00
Raymond Knopp	68b99dd81f	commit before merge	2024-11-23 21:41:26 +01:00
Raymond Knopp	2f42628a20	bug fixes for 3-layer DL	2024-11-23 21:35:30 +01:00
Raymond Knopp	942648f628	Merge remote-tracking branch 'origin/develop' into resource_mapping_optim	2024-11-23 09:47:02 +01:00
Raymond Knopp	a163d8c609	Merge remote-tracking branch 'origin/develop' into resource_mapping_optim	2024-11-23 09:44:04 +01:00
Raymond Knopp	afe610d97c	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-23 09:18:27 +01:00
Raymond Knopp	8601a5e446	commit before merge	2024-11-23 09:18:07 +01:00
Raymond Knopp	392438cf50	bugfix for 3/4 layers	2024-11-23 09:15:16 +01:00
Raymond Knopp	ac4b540001	dmrs_offset in new mapping	2024-11-23 01:09:19 +01:00
Raymond Knopp	aee48ccaaf	debugging 3	2024-11-22 23:35:37 +01:00
Raymond Knopp	7edb74ae62	debugging3	2024-11-22 23:23:43 +01:00
Raymond Knopp	771e1c674c	debugging2	2024-11-22 23:22:31 +01:00
Raymond Knopp	41d7570b82	debuggin	2024-11-22 23:20:57 +01:00
Raymond Knopp	f79ca60ea4	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-22 11:25:23 +01:00
Raymond Knopp	fd03d288f9	debugging of 3-layer resource mapping	2024-11-22 09:21:25 +01:00
Raymond Knopp	2a28e2fd82	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-11-20 21:04:50 +01:00
Raymond Knopp	5f05a3bf66	merged all layers in nr_layer_mapping() instead of running each individually	2024-11-20 15:44:54 +01:00
Raymond Knopp	84e87248dd	Merge remote-tracking branch 'origin/develop' into resource_mapping_optim	2024-11-19 20:13:06 +01:00
Raymond Knopp	afd6cc6ade	SIMD processing for nr_layer_mapping	2024-11-19 18:52:17 +01:00
Raymond Knopp	9a278e8d2d	Merge remote-tracking branch 'origin/develop' into resource_mapping_optim	2024-11-17 23:07:20 +01:00
Karim Boutiba	29a324c37d	Merge remote-tracking branch 'origin/fix-4layers-nr_dlsim' into resource_mapping_optim	2024-11-04 20:54:29 +01:00
Karim Boutiba	ed102689ed	Merge remote-tracking branch 'origin/fix-4layers-nr_dlsim' into resource_mapping_optim	2024-11-04 20:53:58 +01:00
Laurent THOMAS	1a214beaa5	change the alignment of OAI mutli-dim arrays from 16 to 32 bytes	2024-11-04 16:07:26 +01:00
Laurent THOMAS	fd1a33d7f6	fix blocking bugs in command nr_dlsim -n100 -e27 -s35 -b 273 -R273 -x4 -y4 -z4 use a better complex vector multiplication function	2024-11-04 15:13:32 +01:00
Laurent THOMAS	2f65813a31	small fix of rotate vector unit test	2024-10-31 14:51:51 +01:00
Laurent THOMAS	5fc41e7681	add a optimzed complex vector multiplication function	2024-10-31 14:50:25 +01:00
Karim Boutiba	69ac466b4a	resolved issue after merge with develop	2024-10-30 14:11:08 +01:00
Karim Boutiba	74adddd5b1	Merge remote-tracking branch 'origin/develop' into resource_mapping_optim	2024-10-30 12:18:39 +01:00
Karim Boutiba	10a7daede4	Merge branch 'resource_mapping_optim' of https://gitlab.eurecom.fr/oai/openairinterface5g into resource_mapping_optim	2024-10-17 07:36:03 +02:00
Raymond Knopp	3f5ff5b0ef	dlsim memory allocations for 256QAM (fixes issue for large TBS)	2024-10-17 07:34:42 +02:00
Karim Boutiba	fda3db9ff3	dlsim memory allocations for 256QAM (fixes issue for large TBS)	2024-10-17 07:29:00 +02:00
Raymond Knopp	19e0d94685	testing avx2 for resource mapping	2024-10-11 11:32:07 +02:00