adapt ldcp calls to spport new interface required by cuda decoder in shared lib

openair1/PHY/CODING/nrLDPC_decoder_LYC/nrLDPC_decoder_LYC.cu

@@ -483,6 +483,52 @@ void init_LLR_DMA_for_CUDA(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8
 	
 }
 
+using namespace std ;
+
+/* from here: entry points in decoder shared lib */
+extern "C"
+int ldpc_autoinit(void) {   // called by the library loader 
+int devices = 0; 
+
+  cudaError_t err = cudaGetDeviceCount(&devices); 
+  AssertFatal(devices>0,"\nNo cuda GPU found\n\n");
+
+    const int kb = 1024;
+    const int mb = kb * kb;
+    wcout << "NBody.GPU" << endl << "=========" << endl << endl;
+
+    wcout << "CUDA version:   v" << CUDART_VERSION << endl;    
+    
+
+    wcout << "CUDA Devices: " << endl << endl;
+
+    for(int i = 0; i < devices; ++i)
+    {
+        cudaDeviceProp props;
+        cudaGetDeviceProperties(&props, i);
+        wcout << i << ": " << props.name << ": " << props.major << "." << props.minor << endl;
+        wcout << "  Global memory:   " << props.totalGlobalMem / mb << "mb" << endl;
+        wcout << "  Shared memory:   " << props.sharedMemPerBlock / kb << "kb" << endl;
+        wcout << "  Constant memory: " << props.totalConstMem / kb << "kb" << endl;
+        wcout << "  Block registers: " << props.regsPerBlock << endl << endl;
+
+        wcout << "  Warp size:         " << props.warpSize << endl;
+        wcout << "  Threads per block: " << props.maxThreadsPerBlock << endl;
+        wcout << "  Max block dimensions: [ " << props.maxThreadsDim[0] << ", " << props.maxThreadsDim[1]  << ", " << props.maxThreadsDim[2] << " ]" << endl;
+        wcout << "  Max grid dimensions:  [ " << props.maxGridSize[0] << ", " << props.maxGridSize[1]  << ", " << props.maxGridSize[2] << " ]" << endl;
+        wcout << endl;
+    }
+  warmup_for_GPU();
+  return 0;  
+}
+
+extern "C"
+void nrLDPC_initcall(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out) {
+	set_compact_BG(p_decParams->Z,p_decParams->BG);
+	init_LLR_DMA(p_decParams, p_llr,  p_out);
+}
+
+
 extern "C"
 int32_t nrLDPC_decoder_LYC(t_nrLDPC_dec_params* p_decParams, int8_t* p_llr, int8_t* p_out, int block_length, time_stats_t *time_decoder)
 {

openair1/PHY/CODING/nrLDPC_load.c

@@ -42,14 +42,25 @@
 
 
 /* function description array, to be used when loading the encoding/decoding shared lib */
-static loader_shlibfunc_t shlib_fdesc[2];
 
-char *arg[64]={"ldpctest","-O","cmdlineonly::dbgl0"};
+static loader_shlibfunc_t shlib_fdesc[3];
+
+/* arguments used when called from phy simulators exec's which do not use the config module */
+/* arg is used to initialize the config module so that the loader works as expected */
+char *arg[64]={"ldpctest","-O","cmdlineonly::dbgl0",NULL,NULL};
 
 int load_nrLDPClib(void) {
 	 char *ptr = (char*)config_get_if();
+	 char libname[64]="ldpc";
+	 int argc=3;
+	 if (run_cuda) {
+         arg[3]="--loader.ldpc.shlibversion";
+         argc++;
+         arg[4]="_cuda";
+         argc++;
+     }
      if ( ptr==NULL )  {// phy simulators, config module possibly not loaded
-     	 load_configmodule(3,(char **)arg,CONFIG_ENABLECMDLINEONLY) ;
+     	 load_configmodule(argc,(char **)arg,CONFIG_ENABLECMDLINEONLY) ;
      	 logInit();
      }	 
      shlib_fdesc[0].fname = "nrLDPC_decod";

openair1/PHY/NR_UE_TRANSPORT/nr_dlsch_decoding.c

@@ -497,6 +497,7 @@ uint32_t nr_dlsch_decoding(PHY_VARS_NR_UE *phy_vars_ue,
       }
 
       VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_DLSCH_LDPC, VCD_FUNCTION_IN);
+      nrLDPC_initcall(p_decParams, (int8_t*)&pl[0], llrProcBuf);
       no_iteration_ldpc = nrLDPC_decoder(p_decParams,
                                          (int8_t *)&pl[0],
                                          llrProcBuf,
@@ -956,7 +957,7 @@ uint32_t  nr_dlsch_decoding_mthread(PHY_VARS_NR_UE *phy_vars_ue,
     for (i=0, j=0; j < ((kc*harq_process->Z)>>4)+1;  i+=2, j++) {
       pl[j] = _mm_packs_epi16(pv[i],pv[i+1]);
     }
-
+    nrLDPC_initcall(p_decParams, (int8_t*)&pl[0], llrProcBuf);
     no_iteration_ldpc = nrLDPC_decoder(p_decParams,
                                        (int8_t *)&pl[0],
                                        llrProcBuf,
@@ -1340,7 +1341,7 @@ void nr_dlsch_decoding_process(void *arg) {
     for (i=0, j=0; j < ((kc*harq_process->Z)>>4)+1;  i+=2, j++) {
       pl[j] = _mm_packs_epi16(pv[i],pv[i+1]);
     }
-
+    nrLDPC_initcall(p_decParams, (int8_t*)&pl[0], llrProcBuf);
     no_iteration_ldpc = nrLDPC_decoder(p_decParams,
                                        (int8_t *)&pl[0],
                                        llrProcBuf,