bit endian corrections for a,b sequences in polar encoder/decoder. CRC...

bit endian corrections for a,b sequences in polar encoder/decoder. CRC computations need to be compared to reference models

openair1/PHY/CODING/nrPolar_tools/nr_polar_decoder.c

@@ -1054,8 +1054,9 @@ void init_polar_deinterleaver_table(t_nrPolar_params *polarParams) {
     if (byte<(polarParams->K>>3)) numbits=8;
     else numbits=residue;
     for (int i=0;i<numbits;i++) {
-      ip=polarParams->interleaving_pattern[(8*byte)+i];
-#if 0
+      // flip bit endian for B
+      ip=polarParams->K - 1 - polarParams->interleaving_pattern[(8*byte)+i];
+#if 1
       printf("byte %d, i %d => ip %d\n",byte,i,ip);
 #endif
       ipmod64 = ip&63;
@@ -1117,31 +1118,46 @@ uint32_t polar_decoder_int16(int16_t *input,
 
   int len=polarParams->payloadBits;
   int len_mod64=len&63;
-  int quadwpos=len>>6;
   int crclen = polarParams->crcParityBits;
-  int quadwpos2  = polarParams->K>>6;
-  uint64_t rxcrc;
+  uint64_t rxcrc=B[0]&((1<<crclen)-1);
   uint32_t crc;
-  if (len_mod64==0) rxcrc = 0;
-  else              rxcrc = B[quadwpos]>>len_mod64;
-
-  if (quadwpos2>quadwpos) { // there are extra CRC bits in the next quadword
-    rxcrc |= (B[quadwpos2]<<(64-len_mod64));
-  }
-  // clear everything but payload bits in last quadword
-  B[quadwpos]&=((((uint64_t)1)<<len_mod64)-1);
-  crc = crc24c((uint8_t*)B,polarParams->payloadBits)>>8;
-#if 0
+  uint64_t Ar;
+
+  AssertFatal(len<65,"A must be less than 65 bits\n");
+  if (len<=32) {
+    Ar = (uint32_t)(B[0]>>crclen);
+    uint8_t A32_flip[4];
+    uint32_t Aprime= (uint32_t)(Ar<<(32-len));
+    A32_flip[0]=((uint8_t*)&Aprime)[3];
+    A32_flip[1]=((uint8_t*)&Aprime)[2];
+    A32_flip[2]=((uint8_t*)&Aprime)[1];
+    A32_flip[3]=((uint8_t*)&Aprime)[0];
+    crc = (uint64_t)(crc24c(A32_flip,len)>>8);
+  }
+  else if (len<=64) {
+    Ar = (B[0]>>crclen) | (B[1]<<(64-crclen));;
+    uint8_t A64_flip[4];
+    uint64_t Aprime= (uint32_t)(Ar<<(64-len));
+    A64_flip[0]=((uint8_t*)&Aprime)[7];
+    A64_flip[1]=((uint8_t*)&Aprime)[6];
+    A64_flip[2]=((uint8_t*)&Aprime)[5];
+    A64_flip[3]=((uint8_t*)&Aprime)[4];
+    A64_flip[4]=((uint8_t*)&Aprime)[3];
+    A64_flip[5]=((uint8_t*)&Aprime)[2];
+    A64_flip[6]=((uint8_t*)&Aprime)[1];
+    A64_flip[7]=((uint8_t*)&Aprime)[0];
+    crc = (uint64_t)(crc24c(A64_flip,len)>>8);
+  }
+
+#if 1
   printf("A %llx B %llx|%llx Cprime %llx|%llx  (crc %x,rxcrc %llx %d)\n",
-	 B[quadwpos]&((((uint64_t)1)<<len_mod64)-1),
+	 Ar,
 	 B[1],B[0],Cprime[1],Cprime[0],crc,
    	 rxcrc,polarParams->payloadBits);
 #endif
-  int k=0;
-  // copy quadwords without CRC directly
-  for (k=0;k<polarParams->payloadBits/64;k++) out[k]=B[k];
-  // copy last one
-  out[k] = B[k] & (((uint64_t)1<<(polarParams->payloadBits&63))-1);
+
+  out[0]=Ar;
+
 
   return(crc^rxcrc);
   

openair1/PHY/CODING/nrPolar_tools/nr_polar_encoder.c

@@ -30,7 +30,7 @@
  * \warning
  */
 
-//#define DEBUG_POLAR_ENCODER
+#define DEBUG_POLAR_ENCODER
 //#define DEBUG_POLAR_ENCODER_DCI
 //#define DEBUG_POLAR_ENCODER_TIMING
 
@@ -316,7 +316,8 @@ void build_polar_tables(t_nrPolar_paramsPtr polarParams) {
       polarParams->cprime_tab0[byte][val] = 0;
       polarParams->cprime_tab1[byte][val] = 0;
       for (int i=0;i<numbits;i++) {
-	ip=polarParams->deinterleaving_pattern[(8*byte)+i];
+	// flip bit endian of B bitstring
+	ip=polarParams->deinterleaving_pattern[polarParams->K-1-((8*byte)+i)];
 	AssertFatal(ip<128,"ip = %d\n",ip);
 	bit_i=(val>>i)&1;
 	if (ip<64) polarParams->cprime_tab0[byte][val] |= (((uint64_t)bit_i)<<ip);				
@@ -402,16 +403,47 @@ void polar_encoder_fast(uint64_t *A,
 
   int bitlen0=bitlen;
 
-  uint64_t tcrc = (uint64_t)((crcmask^(crc24c((uint8_t*)A,bitlen)>>8)));
-  int n;
-  for (n=0;n<(1+(bitlen>>6));n++)
-    if (bitlen0<64) B[n] = ((A[n])&((((uint64_t)1)<<bitlen0)-1)) | (tcrc<<bitlen0);
-    else {
-      B[n] = A[n];
-      bitlen0-=64;
+  uint64_t tcrc=0;
+
+  // A bitstring should be stored as a_{N-1} a_{N-2} ... a_{N-A} 0 .... 0, where N=64,128,192,..., N is smallest multiple of 64 greater than or equal to A
+
+  // First flip A bitstring byte endian for CRC routines (optimized for DLSCH/ULSCH, not PBCH/PDCCH)
+  // CRC reads in each byte in bit positions 7 downto 0, for PBCH/PDCCH we need to read in a_{A-1} downto a_{0}, A = length of bit string (e.g. 32 for PBCH)
+  if (bitlen<=32) {
+    uint8_t A32_flip[4];
+    uint32_t Aprime= (uint32_t)(((uint32_t)*A)<<(32-bitlen));
+    A32_flip[0]=((uint8_t*)&Aprime)[3];
+    A32_flip[1]=((uint8_t*)&Aprime)[2];
+    A32_flip[2]=((uint8_t*)&Aprime)[1];
+    A32_flip[3]=((uint8_t*)&Aprime)[0];
+    tcrc = (uint64_t)((crcmask^(crc24c(A32_flip,bitlen)>>8)));
   }
-  // handle residual part of CRC in next quadword
-  if (polarParams->crcParityBits > (64-bitlen0)) B[n] = tcrc>>(64-bitlen0);
+  else if (bitlen<=64) {
+    uint8_t A64_flip[4];
+    uint64_t Aprime= (uint32_t)(((uint64_t)*A)<<(64-bitlen));
+    A64_flip[0]=((uint8_t*)&Aprime)[7];
+    A64_flip[1]=((uint8_t*)&Aprime)[6];
+    A64_flip[2]=((uint8_t*)&Aprime)[5];
+    A64_flip[3]=((uint8_t*)&Aprime)[4];
+    A64_flip[4]=((uint8_t*)&Aprime)[3];
+    A64_flip[5]=((uint8_t*)&Aprime)[2];
+    A64_flip[6]=((uint8_t*)&Aprime)[1];
+    A64_flip[7]=((uint8_t*)&Aprime)[0];
+    tcrc = (uint64_t)((crcmask^(crc24c(A64_flip,bitlen)>>8)));
+  }
+
+  int n;
+  // this is number of quadwords in the bit string
+  int quadwlen = (polarParams->K>>6);
+  if ((polarParams->K&63) > 0) quadwlen++;
+
+  // Create the B bitstring as
+  // b_{N'-1} b_{N'-2} ... b_{N'-A} b_{N'-A-1} ... b_{N'-A-Nparity} = a_{N-1} a_{N-2} ... a_{N-A} p_{N_parity-1} ... p_0
+
+  for (n=0;n<quadwlen;n++) if (n==0) B[n] = (A[n] << polarParams->crcParityBits) | tcrc;
+                           else      B[n] = (A[n] << polarParams->crcParityBits) | (A[n-1]>>64-polarParams->crcParityBits);
+  
+    
   uint8_t *Bbyte = (uint8_t*)B;
   // for each byte of B, lookup in corresponding table for 64-bit word corresponding to that byte and its position
   if (polarParams->K<65) 
@@ -486,6 +518,7 @@ void polar_encoder_fast(uint64_t *A,
     for (int i=0;i<((len>63) ? 64 : len);i++) {
 
       Cprime_i = -((Cprime[j]>>i)&1); // this converts bit 0 as, 0 => 0000x00, 1 => 1111x11
+      /*
 #ifdef DEBUG_POLAR_ENCODER
       printf("%llx Cprime_%d (%llx) G %llx,%llx,%llx,%llx,%llx,%llx,%llx,%llx\n",
 	     Cprime_i,off+i,(Cprime[j]>>i) &1,
@@ -498,6 +531,7 @@ void polar_encoder_fast(uint64_t *A,
 	     polarParams->G_N_tab[off+i][6],
 	     polarParams->G_N_tab[off+i][7]);
 #endif
+      */
       uint64_t *Gi=polarParams->G_N_tab[off+i];
       D[0] ^= (Cprime_i & Gi[0]);
       D[1] ^= (Cprime_i & Gi[1]);