[v0.01] Defining a first approach of the application general structure

Makefile

@@ -16,7 +16,7 @@ CUDA_HDR = $(SRC_CUDA)cuda.hpp $(SRC_CUDA)uarch.hpp
 CUDA_PATH = /usr/local/cuda/
 
 SOURCE += $(COMMON_SRC) $(CUDA_SRC)
-HEADERS += $(COMMON_HDR) $(CUDA_HRC)
+HEADERS += $(COMMON_HDR) $(CUDA_HDR)
 
 OUTPUT=gpufetch
 

src/common/gpu.cpp

@@ -1,6 +1,30 @@
 #include "gpu.hpp"
+#include <cstddef>
 
 char* get_str_gpu_name(struct gpu_info* gpu) {
   return gpu->name;
 }
 
+char* get_str_freq(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_memory_size(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_memory_type(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_l1(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_l2(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_peak_performance(struct gpu_info* gpu) {
+  return NULL;
+}

src/common/gpu.hpp

@@ -8,15 +8,52 @@ enum {
   GPU_VENDOR_NVIDIA
 };
 
+enum {
+  MEMTYPE_GDDR6
+};
+
 typedef int32_t VENDOR;
+typedef int32_t MEMTYPE;
+
+struct cach {
+  int32_t size;
+  uint8_t num_caches;
+  bool exists;
+};
+
+struct cache {
+  struct cach* L1;
+  struct cach* L2;
+};
+
+struct topology {
+  int32_t shared_mp;
+  int32_t cores_per_mp;
+  int32_t cuda_cores;
+};
+
+struct memory {
+  int32_t size_bytes;
+  MEMTYPE type;
+};
 
 struct gpu_info {
   VENDOR vendor;
   struct uarch* arch;
   char* name;
+  int64_t freq;
+  struct topology* topo;
+  struct memory* mem;
+  struct cache* cach;
+  int64_t peak_performance;
 };
 
-
 char* get_str_gpu_name(struct gpu_info* gpu);
+char* get_str_freq(struct gpu_info* gpu);
+char* get_str_memory_size(struct gpu_info* gpu);
+char* get_str_memory_type(struct gpu_info* gpu);
+char* get_str_l1(struct gpu_info* gpu);
+char* get_str_l2(struct gpu_info* gpu);
+char* get_str_peak_performance(struct gpu_info* gpu);
 
 #endif

src/common/main.cpp

@@ -5,6 +5,7 @@
 #include "args.hpp"
 #include "global.hpp"
 #include "../cuda/cuda.hpp"
+#include "../cuda/uarch.hpp"
 
 static const char* VERSION = "0.01";
 
@@ -52,7 +53,19 @@ int main(int argc, char* argv[]) {
   if(gpu == NULL)
     return EXIT_FAILURE;
 
-  printf("Name: %s\n", gpu->name);
+  printf("Name:               %s\n", get_str_gpu_name(gpu));
+  printf("Microarchitecture:  %s\n", get_str_uarch(gpu));
+  printf("Compute Capability: %s\n", get_str_cc(gpu));
+  printf("Technology:         %s\n", get_str_process(gpu));
+  printf("Max Frequency:      %s\n", get_str_freq(gpu));
+  printf("SM:                 %s\n", get_str_sm(gpu));
+  printf("Cores/MP:           %s\n", get_str_cores_sm(gpu));
+  printf("CUDA cores:         %s\n", get_str_cuda_cores(gpu));
+  printf("Memory size:        %s\n", get_str_memory_size(gpu));
+  printf("Memory type:        %s\n", get_str_memory_type(gpu));
+  printf("L1 size:            %s\n", get_str_l1(gpu));
+  printf("L2 size:            %s\n", get_str_l2(gpu));
+  printf("Peak performance:   %s\n", get_str_peak_performance(gpu));
 
   return EXIT_FAILURE;
 }

src/cuda/cuda.cpp

@@ -2,18 +2,62 @@
 #include <cuda_runtime.h>
 
 #include "cuda.hpp"
+#include "uarch.hpp"
 #include "../common/global.hpp"
 
+struct cache* get_cache_info(struct gpu_info* gpu) {
+  struct cache* cach = (struct cache*) emalloc(sizeof(struct cache));
+
+  return cach;
+}
+
+struct topology* get_topology_info(struct gpu_info* gpu) {
+  struct topology* topo = (struct topology*) emalloc(sizeof(struct topology));
+
+  return topo;
+}
+
+struct memory* get_memory_info(struct gpu_info* gpu) {
+  struct memory* mem = (struct memory*) emalloc(sizeof(struct memory));
+
+  return mem;
+}
+
+int64_t get_peak_performance(struct gpu_info* gpu) {
+  return 1000;
+}
+
 struct gpu_info* get_gpu_info() {
   struct gpu_info* gpu = (struct gpu_info*) emalloc(sizeof(struct gpu_info));
 
+  printf("Waiting for CUDA driver to start...\n");
   int dev = 0;
   cudaSetDevice(dev);
   cudaDeviceProp deviceProp;
   cudaGetDeviceProperties(&deviceProp, dev);
 
+  gpu->vendor = GPU_VENDOR_NVIDIA;
   gpu->name = (char *) emalloc(sizeof(char) * (strlen(deviceProp.name) + 1));
   strcpy(gpu->name, deviceProp.name);
+  gpu->freq = 10000;
+
+  gpu->arch = get_uarch_from_cuda(gpu);
+  gpu->cach = get_cache_info(gpu);
+  gpu->topo = get_topology_info(gpu);
+  gpu->peak_performance = get_peak_performance(gpu);
 
   return gpu;
 }
+
+char* get_str_sm(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_cores_sm(struct gpu_info* gpu) {
+  return NULL;
+}
+
+char* get_str_cuda_cores(struct gpu_info* gpu) {
+  return NULL;
+}
+

src/cuda/cuda.hpp

@@ -4,5 +4,8 @@
 #include "../common/gpu.hpp"
 
 struct gpu_info* get_gpu_info();
+char* get_str_sm(struct gpu_info* gpu);
+char* get_str_cores_sm(struct gpu_info* gpu);
+char* get_str_cuda_cores(struct gpu_info* gpu);
 
 #endif

src/cuda/uarch.cpp

@@ -28,6 +28,10 @@ char* get_str_uarch(struct gpu_info* gpu) {
   return NULL;
 }
 
+char* get_str_cc(struct gpu_info* gpu) {
+  return NULL;
+}
+
 char* get_str_process(struct gpu_info* gpu) {
   return NULL;
 }

src/cuda/uarch.hpp

@@ -5,6 +5,7 @@ struct uarch;
 
 struct uarch* get_uarch_from_cuda(struct gpu_info* gpu);
 char* get_str_uarch(struct gpu_info* gpu);
+char* get_str_cc(struct gpu_info* gpu);
 char* get_str_process(struct gpu_info* gpu);
 void free_uarch_struct(struct uarch* arch);