/** * app.c * HST-S Host Application Source File * */ #include #include #include #include #include #include #include #include #include #include #include "../support/common.h" #include "../support/timer.h" #include "../support/params.h" // Define the DPU Binary path as DPU_BINARY here #ifndef DPU_BINARY #define DPU_BINARY "./bin/dpu_code" #endif #define XSTR(x) STR(x) #define STR(x) #x #if ENERGY #include #endif #include #include #if SDK_SINGLETHREADED #define DPU_ALLOC_PROFILE "nrThreadsPerRank=0" #else #define DPU_ALLOC_PROFILE NULL #endif // Pointer declaration static T* A; static unsigned int* histo_host; static unsigned int* histo; // Create input arrays static void read_input(T* A, const Params p) { char dctFileName[100]; FILE *File = NULL; // Open input file unsigned short temp; sprintf(dctFileName, "%s", p.file_name); if((File = fopen(dctFileName, "rb")) != NULL) { for(unsigned int y = 0; y < p.input_size; y++) { if (fread(&temp, sizeof(unsigned short), 1, File) == 1) { A[y] = (unsigned int)ByteSwap16(temp); if(A[y] >= 4096) A[y] = 4095; } else { //printf("out of bounds read at offset %d -- seeking back to 0\n", y); rewind(File); } } fclose(File); } else { printf("%s does not exist\n", dctFileName); exit(1); } } // Compute output in the host static void histogram_host(unsigned int* histo, T* A, unsigned int bins, unsigned int nr_elements, int exp, unsigned int nr_of_dpus) { if(!exp){ for (unsigned int i = 0; i < nr_of_dpus; i++) { for (unsigned int j = 0; j < nr_elements; j++) { T d = A[j]; histo[i * bins + ((d * bins) >> DEPTH)] += 1; } } } else{ for (unsigned int j = 0; j < nr_elements; j++) { T d = A[j]; histo[(d * bins) >> DEPTH] += 1; } } } // Main of the Host Application int main(int argc, char **argv) { struct Params p = input_params(argc, argv); struct dpu_set_t dpu_set, dpu; uint32_t nr_of_dpus; uint32_t nr_of_ranks; #if ENERGY struct dpu_probe_t probe; DPU_ASSERT(dpu_probe_init("energy_probe", &probe)); #endif // Timer declaration Timer timer; int numa_node_rank = -2; // Allocate DPUs and load binary #if !WITH_ALLOC_OVERHEAD DPU_ASSERT(dpu_alloc(NR_DPUS, DPU_ALLOC_PROFILE, &dpu_set)); timer.time[0] = 0; // alloc #endif #if !WITH_LOAD_OVERHEAD DPU_ASSERT(dpu_load(dpu_set, DPU_BINARY, NULL)); DPU_ASSERT(dpu_get_nr_dpus(dpu_set, &nr_of_dpus)); DPU_ASSERT(dpu_get_nr_ranks(dpu_set, &nr_of_ranks)); assert(nr_of_dpus == NR_DPUS); timer.time[1] = 0; // load #endif #if !WITH_FREE_OVERHEAD timer.time[6] = 0; // free #endif unsigned int i = 0; unsigned int input_size; // Size of input image unsigned int dpu_s = p.dpu_s; if(p.exp == 0) input_size = p.input_size * NR_DPUS; // Size of input image else if(p.exp == 1) input_size = p.input_size; // Size of input image else input_size = p.input_size * dpu_s; // Size of input image const unsigned int input_size_8bytes = ((input_size * sizeof(T)) % 8) != 0 ? roundup(input_size, 8) : input_size; // Input size per DPU (max.), 8-byte aligned const unsigned int input_size_dpu = divceil(input_size, NR_DPUS); // Input size per DPU (max.) const unsigned int input_size_dpu_8bytes = ((input_size_dpu * sizeof(T)) % 8) != 0 ? roundup(input_size_dpu, 8) : input_size_dpu; // Input size per DPU (max.), 8-byte aligned // Input/output allocation A = malloc(input_size_dpu_8bytes * NR_DPUS * sizeof(T)); T *bufferA = A; histo_host = malloc(p.bins * sizeof(unsigned int)); histo = malloc(NR_DPUS * p.bins * sizeof(unsigned int)); // Create an input file with arbitrary data read_input(A, p); if(p.exp == 0){ for(unsigned int j = 1; j < NR_DPUS; j++){ memcpy(&A[j * input_size_dpu_8bytes], &A[0], input_size_dpu_8bytes * sizeof(T)); } } else if(p.exp == 2){ for(unsigned int j = 1; j < dpu_s; j++) memcpy(&A[j * p.input_size], &A[0], p.input_size * sizeof(T)); } // Loop over main kernel for(int rep = 0; rep < p.n_warmup + p.n_reps; rep++) { memset(histo_host, 0, p.bins * sizeof(unsigned int)); memset(histo, 0, NR_DPUS * p.bins * sizeof(unsigned int)); #if WITH_ALLOC_OVERHEAD if(rep >= p.n_warmup) { start(&timer, 0, 0); } DPU_ASSERT(dpu_alloc(NR_DPUS, DPU_ALLOC_PROFILE, &dpu_set)); if(rep >= p.n_warmup) { stop(&timer, 0); } #endif #if WITH_DPUINFO printf("DPUs:"); DPU_FOREACH (dpu_set, dpu) { int rank = dpu_get_rank_id(dpu_get_rank(dpu_from_set(dpu))) & DPU_TARGET_MASK; int slice = dpu_get_slice_id(dpu_from_set(dpu)); int member = dpu_get_member_id(dpu_from_set(dpu)); printf(" %d(%d.%d)", rank, slice, member); } printf("\n"); #endif #if WITH_LOAD_OVERHEAD if(rep >= p.n_warmup) { start(&timer, 1, 0); } DPU_ASSERT(dpu_load(dpu_set, DPU_BINARY, NULL)); if(rep >= p.n_warmup) { stop(&timer, 1); } DPU_ASSERT(dpu_get_nr_dpus(dpu_set, &nr_of_dpus)); DPU_ASSERT(dpu_get_nr_ranks(dpu_set, &nr_of_ranks)); assert(nr_of_dpus == NR_DPUS); #endif // int prev_rank_id = -1; int rank_id = -1; DPU_FOREACH (dpu_set, dpu) { rank_id = dpu_get_rank_id(dpu_get_rank(dpu_from_set(dpu))) & DPU_TARGET_MASK; if ((numa_node_rank != -2) && numa_node_rank != dpu_get_rank_numa_node(dpu_get_rank(dpu_from_set(dpu)))) { numa_node_rank = -1; } else { numa_node_rank = dpu_get_rank_numa_node(dpu_get_rank(dpu_from_set(dpu))); } /* if (rank_id != prev_rank_id) { printf("/dev/dpu_rank%d @ NUMA node %d\n", rank_id, numa_node_rank); prev_rank_id = rank_id; } */ } // Compute output on CPU (performance comparison and verification purposes) if(rep >= p.n_warmup) { start(&timer, 2, 0); } histogram_host(histo_host, A, p.bins, p.input_size, 1, NR_DPUS); if(rep >= p.n_warmup) { stop(&timer, 2); } if(rep >= p.n_warmup) { start(&timer, 3, 0); } // Input arguments unsigned int kernel = 0; i = 0; dpu_arguments_t input_arguments[NR_DPUS]; for(i=0; i= p.n_warmup) { stop(&timer, 3); } // Run DPU kernel if(rep >= p.n_warmup) { start(&timer, 4, 0); #if ENERGY DPU_ASSERT(dpu_probe_start(&probe)); #endif } DPU_ASSERT(dpu_launch(dpu_set, DPU_SYNCHRONOUS)); if(rep >= p.n_warmup) { stop(&timer, 4); #if ENERGY DPU_ASSERT(dpu_probe_stop(&probe)); #endif } #if PRINT { unsigned int each_dpu = 0; printf("Display DPU Logs\n"); DPU_FOREACH (dpu_set, dpu) { printf("DPU#%d:\n", each_dpu); DPU_ASSERT(dpulog_read_for_dpu(dpu.dpu, stdout)); each_dpu++; } } #endif i = 0; if(rep >= p.n_warmup) { start(&timer, 5, 0); } // PARALLEL RETRIEVE TRANSFER DPU_FOREACH(dpu_set, dpu, i) { DPU_ASSERT(dpu_prepare_xfer(dpu, histo + p.bins * i)); } DPU_ASSERT(dpu_push_xfer(dpu_set, DPU_XFER_FROM_DPU, DPU_MRAM_HEAP_POINTER_NAME, input_size_dpu_8bytes * sizeof(T), p.bins * sizeof(unsigned int), DPU_XFER_DEFAULT)); // Final histogram merging for(i = 1; i < NR_DPUS; i++){ for(unsigned int j = 0; j < p.bins; j++){ histo[j] += histo[j + i * p.bins]; } } if(rep >= p.n_warmup) { stop(&timer, 5); } #if WITH_ALLOC_OVERHEAD #if WITH_FREE_OVERHEAD if(rep >= p.n_warmup) { start(&timer, 6, 0); } #endif DPU_ASSERT(dpu_free(dpu_set)); #if WITH_FREE_OVERHEAD if(rep >= p.n_warmup) { stop(&timer, 6); } #endif #endif if (rep >= p.n_warmup) { printf("[::] HST-S-UPMEM | n_dpus=%d n_ranks=%d n_tasklets=%d e_type=%s n_elements=%d n_bins=%d", nr_of_dpus, nr_of_ranks, NR_TASKLETS, XSTR(T), input_size, p.bins); printf(" b_with_alloc_overhead=%d b_with_load_overhead=%d b_with_free_overhead=%d numa_node_rank=%d ", WITH_ALLOC_OVERHEAD, WITH_LOAD_OVERHEAD, WITH_FREE_OVERHEAD, numa_node_rank); printf("| latency_alloc_us=%f latency_load_us=%f latency_cpu_us=%f latency_write_us=%f latency_kernel_us=%f latency_read_us=%f latency_free_us=%f", timer.time[0], timer.time[1], timer.time[2], timer.time[3], timer.time[4], timer.time[5], timer.time[6]); printf(" throughput_cpu_MBps=%f throughput_upmem_kernel_MBps=%f throughput_upmem_total_MBps=%f", input_size * sizeof(T) / timer.time[2], input_size * sizeof(T) / (timer.time[4]), input_size * sizeof(T) / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5] + timer.time[6])); printf(" throughput_upmem_wxr_MBps=%f throughput_upmem_lwxr_MBps=%f throughput_upmem_alwxr_MBps=%f", input_size * sizeof(T) / (timer.time[3] + timer.time[4] + timer.time[5]), input_size * sizeof(T) / (timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5]), input_size * sizeof(T) / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5])); printf(" throughput_cpu_MOpps=%f throughput_upmem_kernel_MOpps=%f throughput_upmem_total_MOpps=%f", input_size / timer.time[2], input_size / (timer.time[4]), input_size / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5] + timer.time[6])); printf(" throughput_upmem_wxr_MOpps=%f throughput_upmem_lwxr_MOpps=%f throughput_upmem_alwxr_MOpps=%f\n", input_size / (timer.time[3] + timer.time[4] + timer.time[5]), input_size / (timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5]), input_size / (timer.time[0] + timer.time[1] + timer.time[3] + timer.time[4] + timer.time[5])); } } #if ENERGY double energy; DPU_ASSERT(dpu_probe_get(&probe, DPU_ENERGY, DPU_AVERAGE, &energy)); printf("DPU Energy (J): %f\t", energy); #endif // Check output bool status = true; if(p.exp == 1) for (unsigned int j = 0; j < p.bins; j++) { if(histo_host[j] != histo[j]){ status = false; #if PRINT printf("%u - %u: %u -- %u\n", j, j, histo_host[j], histo[j]); #endif } } else if(p.exp == 2) for (unsigned int j = 0; j < p.bins; j++) { if(dpu_s * histo_host[j] != histo[j]){ status = false; #if PRINT printf("%u - %u: %u -- %u\n", j, j, dpu_s * histo_host[j], histo[j]); #endif } } else for (unsigned int j = 0; j < p.bins; j++) { if(NR_DPUS * histo_host[j] != histo[j]){ status = false; #if PRINT printf("%u - %u: %u -- %u\n", j, j, NR_DPUS * histo_host[j], histo[j]); #endif } } if (status) { printf("[" ANSI_COLOR_GREEN "OK" ANSI_COLOR_RESET "] Outputs are equal\n"); } else { printf("[" ANSI_COLOR_RED "ERROR" ANSI_COLOR_RESET "] Outputs differ!\n"); } // Deallocation free(A); free(histo_host); free(histo); #if !WITH_ALLOC_OVERHEAD DPU_ASSERT(dpu_free(dpu_set)); #endif return status ? 0 : -1; }