Print global memory size

This commit is contained in:
Dr-Noob
2025-10-17 08:48:15 +02:00
parent 046f8c1299
commit e1f03c4e04
2 changed files with 27 additions and 6 deletions

View File

@@ -54,7 +54,7 @@ enum {
ATTRIBUTE_CUDA_CORES, // CUDA
ATTRIBUTE_TENSOR_CORES, // CUDA
ATTRIBUTE_L2, // CUDA
ATTRIBUTE_MEMORY, // CUDA
ATTRIBUTE_MEMORY, // CUDA,HSA
ATTRIBUTE_MEMORY_FREQ, // CUDA
ATTRIBUTE_BUS_WIDTH, // CUDA,HSA
ATTRIBUTE_PEAK_TENSOR, // CUDA
@@ -490,6 +490,7 @@ bool print_gpufetch_amd(struct gpu_info* gpu, STYLE s, struct color** cs, struct
char* cus = get_str_cu(gpu);
char* max_frequency = get_str_freq(gpu);
char* bus_width = get_str_bus_width(gpu);
char* mem_size = get_str_memory_size(gpu);
char* lds_size = get_str_lds_size(gpu);
setAttribute(art, ATTRIBUTE_NAME, gpu_name);
@@ -500,8 +501,9 @@ bool print_gpufetch_amd(struct gpu_info* gpu, STYLE s, struct color** cs, struct
setAttribute(art, ATTRIBUTE_TECHNOLOGY, manufacturing_process);
setAttribute(art, ATTRIBUTE_FREQUENCY, max_frequency);
setAttribute(art, ATTRIBUTE_COMPUTE_UNITS, cus);
setAttribute(art, ATTRIBUTE_BUS_WIDTH, bus_width);
setAttribute(art, ATTRIBUTE_LDS_SIZE, lds_size);
setAttribute(art, ATTRIBUTE_MEMORY, mem);
setAttribute(art, ATTRIBUTE_BUS_WIDTH, bus_width);
bool use_short = false;
uint32_t longest_attribute = longest_attribute_length(art, use_short);

View File

@@ -25,6 +25,7 @@ struct agent_info {
uint32_t compute_unit;
uint32_t bus_width;
uint32_t lds_size;
uint32_t global_size;
};
#define RET_IF_HSA_ERR(err) { \
@@ -42,18 +43,31 @@ struct agent_info {
} \
}
hsa_status_t get_lds_size_callback(hsa_region_t region, void* data) {
hsa_status_t region_callback(hsa_region_t region, void* data) {
hsa_region_segment_t segment;
hsa_status_t err = hsa_region_get_info(region, HSA_REGION_INFO_SEGMENT, &segment);
RET_IF_HSA_ERR(err);
uint32_t *mems = reinterpret_cast<uint32_t *>(data);
if (segment == HSA_REGION_SEGMENT_GROUP) {
// LDS memory
size_t size = 0;
err = hsa_region_get_info(region, HSA_REGION_INFO_SIZE, &size);
RET_IF_HSA_ERR(err);
*(size_t*)data = size;
mems[0] = size;
}
else if (segment == HSA_REGION_SEGMENT_GLOBAL) {
// Global memory
// LDS memory
size_t size = 0;
err = hsa_region_get_info(region, HSA_REGION_INFO_SIZE, &size);
RET_IF_HSA_ERR(err);
mems[1] = size;
}
return HSA_STATUS_SUCCESS;
}
@@ -85,8 +99,12 @@ hsa_status_t agent_callback(hsa_agent_t agent, void *data) {
err = hsa_agent_get_info(agent, (hsa_agent_info_t) HSA_AMD_AGENT_INFO_MEMORY_WIDTH, &info->bus_width);
RET_IF_HSA_ERR(err);
err = hsa_agent_iterate_regions(agent, get_lds_size_callback, &info->lds_size);
uint32_t mems[2];
err = hsa_agent_iterate_regions(agent, region_callback, &mems);
RET_IF_HSA_ERR(err);
info->lds_size = mems[0];
info->global_size = mems[1];
}
return HSA_STATUS_SUCCESS;
@@ -105,6 +123,7 @@ struct memory* get_memory_info(struct gpu_info* gpu, struct agent_info info) {
mem->bus_width = info.bus_width;
mem->lds_size = info.lds_size;
mem->size_bytes = info.global_size;
return mem;
}