Compare commits
4 Commits
amd-suppor
...
amd-suppor
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
d28f49bfa1 | ||
|
|
a6856bb630 | ||
|
|
9f40cafada | ||
|
|
c9e652686a |
@@ -101,17 +101,6 @@ char* get_str_bus_width(struct gpu_info* gpu) {
|
|||||||
return string;
|
return string;
|
||||||
}
|
}
|
||||||
|
|
||||||
char* get_str_lds_size(struct gpu_info* gpu) {
|
|
||||||
// TODO: Show XX KB (XX MB Total) like in cpufetch
|
|
||||||
uint32_t size = 3+1+3+1;
|
|
||||||
assert(strlen(STRING_UNKNOWN)+1 <= size);
|
|
||||||
char* string = (char *) ecalloc(size, sizeof(char));
|
|
||||||
|
|
||||||
sprintf(string, "%d KB", gpu->mem->lds_size / 1024);
|
|
||||||
|
|
||||||
return string;
|
|
||||||
}
|
|
||||||
|
|
||||||
char* get_str_memory_clock(struct gpu_info* gpu) {
|
char* get_str_memory_clock(struct gpu_info* gpu) {
|
||||||
return get_freq_as_str_mhz(gpu->mem->freq);
|
return get_freq_as_str_mhz(gpu->mem->freq);
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -61,7 +61,6 @@ struct memory {
|
|||||||
int32_t bus_width;
|
int32_t bus_width;
|
||||||
int32_t freq;
|
int32_t freq;
|
||||||
int32_t clk_mul; // clock multiplier
|
int32_t clk_mul; // clock multiplier
|
||||||
int32_t lds_size; // HSA specific for now
|
|
||||||
};
|
};
|
||||||
|
|
||||||
struct gpu_info {
|
struct gpu_info {
|
||||||
@@ -89,7 +88,6 @@ char* get_str_freq(struct gpu_info* gpu);
|
|||||||
char* get_str_memory_size(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_memory_type(struct gpu_info* gpu);
|
||||||
char* get_str_bus_width(struct gpu_info* gpu);
|
char* get_str_bus_width(struct gpu_info* gpu);
|
||||||
char* get_str_lds_size(struct gpu_info* gpu);
|
|
||||||
char* get_str_memory_clock(struct gpu_info* gpu);
|
char* get_str_memory_clock(struct gpu_info* gpu);
|
||||||
char* get_str_l2(struct gpu_info* gpu);
|
char* get_str_l2(struct gpu_info* gpu);
|
||||||
char* get_str_peak_performance(struct gpu_info* gpu);
|
char* get_str_peak_performance(struct gpu_info* gpu);
|
||||||
|
|||||||
@@ -48,15 +48,14 @@ enum {
|
|||||||
ATTRIBUTE_FREQUENCY, // ALL
|
ATTRIBUTE_FREQUENCY, // ALL
|
||||||
ATTRIBUTE_PEAK, // ALL
|
ATTRIBUTE_PEAK, // ALL
|
||||||
ATTRIBUTE_COMPUTE_UNITS, // HSA
|
ATTRIBUTE_COMPUTE_UNITS, // HSA
|
||||||
ATTRIBUTE_LDS_SIZE, // HSA
|
|
||||||
ATTRIBUTE_STREAMINGMP, // CUDA
|
ATTRIBUTE_STREAMINGMP, // CUDA
|
||||||
ATTRIBUTE_CORESPERMP, // CUDA
|
ATTRIBUTE_CORESPERMP, // CUDA
|
||||||
ATTRIBUTE_CUDA_CORES, // CUDA
|
ATTRIBUTE_CUDA_CORES, // CUDA
|
||||||
ATTRIBUTE_TENSOR_CORES, // CUDA
|
ATTRIBUTE_TENSOR_CORES, // CUDA
|
||||||
ATTRIBUTE_L2, // CUDA
|
ATTRIBUTE_L2, // CUDA
|
||||||
ATTRIBUTE_MEMORY, // CUDA,HSA
|
ATTRIBUTE_MEMORY, // CUDA
|
||||||
ATTRIBUTE_MEMORY_FREQ, // CUDA
|
ATTRIBUTE_MEMORY_FREQ, // CUDA
|
||||||
ATTRIBUTE_BUS_WIDTH, // CUDA,HSA
|
ATTRIBUTE_BUS_WIDTH, // CUDA
|
||||||
ATTRIBUTE_PEAK_TENSOR, // CUDA
|
ATTRIBUTE_PEAK_TENSOR, // CUDA
|
||||||
ATTRIBUTE_EUS, // Intel
|
ATTRIBUTE_EUS, // Intel
|
||||||
ATTRIBUTE_GT, // Intel
|
ATTRIBUTE_GT, // Intel
|
||||||
@@ -70,7 +69,6 @@ static const AttributeField ATTRIBUTE_INFO[] = {
|
|||||||
{ ATTRIBUTE_FREQUENCY, "Max Frequency:", "Max Freq.:" },
|
{ ATTRIBUTE_FREQUENCY, "Max Frequency:", "Max Freq.:" },
|
||||||
{ ATTRIBUTE_PEAK, "Peak Performance:", "Peak Perf.:" },
|
{ ATTRIBUTE_PEAK, "Peak Performance:", "Peak Perf.:" },
|
||||||
{ ATTRIBUTE_COMPUTE_UNITS, "Compute Units (CUs):", "CUs" },
|
{ ATTRIBUTE_COMPUTE_UNITS, "Compute Units (CUs):", "CUs" },
|
||||||
{ ATTRIBUTE_LDS_SIZE, "LDS size:", "LDS:" },
|
|
||||||
{ ATTRIBUTE_STREAMINGMP, "SMs:", "SMs:" },
|
{ ATTRIBUTE_STREAMINGMP, "SMs:", "SMs:" },
|
||||||
{ ATTRIBUTE_CORESPERMP, "Cores/SM:", "Cores/SM:" },
|
{ ATTRIBUTE_CORESPERMP, "Cores/SM:", "Cores/SM:" },
|
||||||
{ ATTRIBUTE_CUDA_CORES, "CUDA Cores:", "CUDA Cores:" },
|
{ ATTRIBUTE_CUDA_CORES, "CUDA Cores:", "CUDA Cores:" },
|
||||||
@@ -199,6 +197,8 @@ bool ascii_fits_screen(int termw, struct ascii_logo logo, int lf) {
|
|||||||
void replace_bgbyfg_color(struct ascii_logo* logo) {
|
void replace_bgbyfg_color(struct ascii_logo* logo) {
|
||||||
// Replace background by foreground color
|
// Replace background by foreground color
|
||||||
for(int i=0; i < 2; i++) {
|
for(int i=0; i < 2; i++) {
|
||||||
|
if(logo->color_ascii[i] == NULL) break;
|
||||||
|
|
||||||
if(strcmp(logo->color_ascii[i], C_BG_BLACK) == 0) strcpy(logo->color_ascii[i], C_FG_BLACK);
|
if(strcmp(logo->color_ascii[i], C_BG_BLACK) == 0) strcpy(logo->color_ascii[i], C_FG_BLACK);
|
||||||
else if(strcmp(logo->color_ascii[i], C_BG_RED) == 0) strcpy(logo->color_ascii[i], C_FG_RED);
|
else if(strcmp(logo->color_ascii[i], C_BG_RED) == 0) strcpy(logo->color_ascii[i], C_FG_RED);
|
||||||
else if(strcmp(logo->color_ascii[i], C_BG_GREEN) == 0) strcpy(logo->color_ascii[i], C_FG_GREEN);
|
else if(strcmp(logo->color_ascii[i], C_BG_GREEN) == 0) strcpy(logo->color_ascii[i], C_FG_GREEN);
|
||||||
@@ -489,9 +489,6 @@ bool print_gpufetch_amd(struct gpu_info* gpu, STYLE s, struct color** cs, struct
|
|||||||
char* manufacturing_process = get_str_process(gpu->arch);
|
char* manufacturing_process = get_str_process(gpu->arch);
|
||||||
char* cus = get_str_cu(gpu);
|
char* cus = get_str_cu(gpu);
|
||||||
char* max_frequency = get_str_freq(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);
|
setAttribute(art, ATTRIBUTE_NAME, gpu_name);
|
||||||
if (gpu_chip != NULL) {
|
if (gpu_chip != NULL) {
|
||||||
@@ -501,9 +498,6 @@ bool print_gpufetch_amd(struct gpu_info* gpu, STYLE s, struct color** cs, struct
|
|||||||
setAttribute(art, ATTRIBUTE_TECHNOLOGY, manufacturing_process);
|
setAttribute(art, ATTRIBUTE_TECHNOLOGY, manufacturing_process);
|
||||||
setAttribute(art, ATTRIBUTE_FREQUENCY, max_frequency);
|
setAttribute(art, ATTRIBUTE_FREQUENCY, max_frequency);
|
||||||
setAttribute(art, ATTRIBUTE_COMPUTE_UNITS, cus);
|
setAttribute(art, ATTRIBUTE_COMPUTE_UNITS, cus);
|
||||||
setAttribute(art, ATTRIBUTE_LDS_SIZE, lds_size);
|
|
||||||
setAttribute(art, ATTRIBUTE_MEMORY, mem_size);
|
|
||||||
setAttribute(art, ATTRIBUTE_BUS_WIDTH, bus_width);
|
|
||||||
|
|
||||||
bool use_short = false;
|
bool use_short = false;
|
||||||
uint32_t longest_attribute = longest_attribute_length(art, use_short);
|
uint32_t longest_attribute = longest_attribute_length(art, use_short);
|
||||||
|
|||||||
@@ -23,9 +23,6 @@ struct agent_info {
|
|||||||
char device_mkt_name[64];
|
char device_mkt_name[64];
|
||||||
uint32_t max_clock_freq;
|
uint32_t max_clock_freq;
|
||||||
uint32_t compute_unit;
|
uint32_t compute_unit;
|
||||||
uint32_t bus_width;
|
|
||||||
uint32_t lds_size;
|
|
||||||
uint64_t global_size;
|
|
||||||
};
|
};
|
||||||
|
|
||||||
#define RET_IF_HSA_ERR(err) { \
|
#define RET_IF_HSA_ERR(err) { \
|
||||||
@@ -43,46 +40,6 @@ struct agent_info {
|
|||||||
} \
|
} \
|
||||||
}
|
}
|
||||||
|
|
||||||
hsa_status_t memory_pool_callback(hsa_amd_memory_pool_t pool, void* data) {
|
|
||||||
struct agent_info* info = reinterpret_cast<struct agent_info *>(data);
|
|
||||||
|
|
||||||
hsa_amd_segment_t segment;
|
|
||||||
hsa_status_t err = hsa_amd_memory_pool_get_info(pool, HSA_AMD_MEMORY_POOL_INFO_SEGMENT, &segment);
|
|
||||||
RET_IF_HSA_ERR(err);
|
|
||||||
|
|
||||||
if (segment == HSA_AMD_SEGMENT_GROUP) {
|
|
||||||
// LDS memory
|
|
||||||
uint32_t size = 0;
|
|
||||||
|
|
||||||
err = hsa_amd_memory_pool_get_info(pool, HSA_AMD_MEMORY_POOL_INFO_SIZE, &size);
|
|
||||||
RET_IF_HSA_ERR(err);
|
|
||||||
|
|
||||||
info->lds_size = size;
|
|
||||||
}
|
|
||||||
else if (segment == HSA_AMD_SEGMENT_GLOBAL) {
|
|
||||||
// Global memory
|
|
||||||
uint32_t global_flags = 0;
|
|
||||||
|
|
||||||
err = hsa_amd_memory_pool_get_info(pool, HSA_AMD_MEMORY_POOL_INFO_GLOBAL_FLAGS, &global_flags);
|
|
||||||
RET_IF_HSA_ERR(err);
|
|
||||||
|
|
||||||
if (global_flags & HSA_AMD_MEMORY_POOL_GLOBAL_FLAG_EXTENDED_SCOPE_FINE_GRAINED) {
|
|
||||||
if (info->global_size != 0) {
|
|
||||||
printErr("Found HSA_AMD_MEMORY_POOL_GLOBAL_FLAG_EXTENDED_SCOPE_FINE_GRAINED twice!");
|
|
||||||
return HSA_STATUS_ERROR;
|
|
||||||
}
|
|
||||||
|
|
||||||
uint64_t size = 0;
|
|
||||||
|
|
||||||
err = hsa_amd_memory_pool_get_info(pool, HSA_AMD_MEMORY_POOL_INFO_SIZE, &size);
|
|
||||||
RET_IF_HSA_ERR(err);
|
|
||||||
|
|
||||||
info->global_size = size;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
return HSA_STATUS_SUCCESS;
|
|
||||||
}
|
|
||||||
|
|
||||||
hsa_status_t agent_callback(hsa_agent_t agent, void *data) {
|
hsa_status_t agent_callback(hsa_agent_t agent, void *data) {
|
||||||
struct agent_info* info = reinterpret_cast<struct agent_info *>(data);
|
struct agent_info* info = reinterpret_cast<struct agent_info *>(data);
|
||||||
|
|
||||||
@@ -105,14 +62,6 @@ 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_COMPUTE_UNIT_COUNT, &info->compute_unit);
|
err = hsa_agent_get_info(agent, (hsa_agent_info_t) HSA_AMD_AGENT_INFO_COMPUTE_UNIT_COUNT, &info->compute_unit);
|
||||||
RET_IF_HSA_ERR(err);
|
RET_IF_HSA_ERR(err);
|
||||||
|
|
||||||
// According to the documentation, this is deprecated. But what should I be using then?
|
|
||||||
err = hsa_agent_get_info(agent, (hsa_agent_info_t) HSA_AMD_AGENT_INFO_MEMORY_WIDTH, &info->bus_width);
|
|
||||||
RET_IF_HSA_ERR(err);
|
|
||||||
|
|
||||||
info->global_size = 0;
|
|
||||||
err = hsa_amd_agent_iterate_memory_pools(agent, memory_pool_callback, data);
|
|
||||||
RET_IF_HSA_ERR(err);
|
|
||||||
}
|
}
|
||||||
|
|
||||||
return HSA_STATUS_SUCCESS;
|
return HSA_STATUS_SUCCESS;
|
||||||
@@ -126,16 +75,6 @@ struct topology_h* get_topology_info(struct agent_info info) {
|
|||||||
return topo;
|
return topo;
|
||||||
}
|
}
|
||||||
|
|
||||||
struct memory* get_memory_info(struct gpu_info* gpu, struct agent_info info) {
|
|
||||||
struct memory* mem = (struct memory*) emalloc(sizeof(struct memory));
|
|
||||||
|
|
||||||
mem->bus_width = info.bus_width;
|
|
||||||
mem->lds_size = info.lds_size;
|
|
||||||
mem->size_bytes = info.global_size;
|
|
||||||
|
|
||||||
return mem;
|
|
||||||
}
|
|
||||||
|
|
||||||
struct gpu_info* get_gpu_info_hsa(int gpu_idx) {
|
struct gpu_info* get_gpu_info_hsa(int gpu_idx) {
|
||||||
struct gpu_info* gpu = (struct gpu_info*) emalloc(sizeof(struct gpu_info));
|
struct gpu_info* gpu = (struct gpu_info*) emalloc(sizeof(struct gpu_info));
|
||||||
gpu->pci = NULL;
|
gpu->pci = NULL;
|
||||||
@@ -179,7 +118,6 @@ struct gpu_info* get_gpu_info_hsa(int gpu_idx) {
|
|||||||
gpu->name = (char *) emalloc(sizeof(char) * (strlen(info.device_mkt_name) + 1));
|
gpu->name = (char *) emalloc(sizeof(char) * (strlen(info.device_mkt_name) + 1));
|
||||||
strcpy(gpu->name, info.device_mkt_name);
|
strcpy(gpu->name, info.device_mkt_name);
|
||||||
gpu->arch = get_uarch_from_hsa(gpu, info.gpu_name);
|
gpu->arch = get_uarch_from_hsa(gpu, info.gpu_name);
|
||||||
gpu->mem = get_memory_info(gpu, info);
|
|
||||||
|
|
||||||
if (gpu->arch == NULL) {
|
if (gpu->arch == NULL) {
|
||||||
return NULL;
|
return NULL;
|
||||||
|
|||||||
Reference in New Issue
Block a user