Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 05:48:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

benchmark-matmul: fix clang-tidy issues, report results in GFLOPS (#1458)

Browse Source 
* benchmark-matmul: fix command line parsing, replace macros with functions, report results in GFLOPS
This commit is contained in:
slaren 2023-05-14 22:46:00 +02:00 committed by GitHub
parent eb363627fd
commit b5c9295eef
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 19 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
examples/benchmark/benchmark-matmult.cpp
View File

@ -15,7 +15,7 @@
#include <iterator> #include <iterator>
#include <algorithm> #include <algorithm>
float tensor_sum_elements(struct ggml_tensor * tensor) { float tensor_sum_elements(const ggml_tensor * tensor) {
float sum = 0; float sum = 0;
if (tensor->type==GGML_TYPE_F32) { if (tensor->type==GGML_TYPE_F32) {
for (int j = 0; j < tensor->ne[1]; j++) { for (int j = 0; j < tensor->ne[1]; j++) {
@ -27,21 +27,15 @@ float tensor_sum_elements(struct ggml_tensor * tensor) {
return sum; return sum;
} }
void tensor_dump(const ggml_tensor * tensor, const char * name) {
printf("%15s: type = %i (%5s) ne = %5d x %5d x %5d, nb = (%5li, %5li, %5li) - ", name,
tensor->type, ggml_type_name(tensor->type),
(int) tensor->ne[0], (int) tensor->ne[1], (int) tensor->ne[2], tensor->nb[0], tensor->nb[1], tensor->nb[2]);
float sum = tensor_sum_elements(tensor);
printf("Sum of tensor %s is %6.2f\n", name, sum);
}
/* #define TENSOR_DUMP(tensor) tensor_dump(tensor, #tensor)
These are mapping to unknown
GGML_TYPE_I8,
GGML_TYPE_I16,
GGML_TYPE_I32,
GGML_TYPE_COUNT,
*/
#define TENSOR_TYPE_AS_STR(TYPE) TYPE == GGML_TYPE_F32 ? "FP32" : TYPE == GGML_TYPE_F16 ? "FP16" : TYPE == GGML_TYPE_Q4_0 ? "Q4_0" : TYPE == GGML_TYPE_Q4_1 ? "Q4_1" : "UNKNOWN"
#define TENSOR_DUMP(TENSOR) printf("%15s: type = %i (%5s) ne = %5d x %5d x %5d, nb = (%5li, %5li, %5li) - ", #TENSOR, \
TENSOR->type,TENSOR_TYPE_AS_STR(TENSOR->type),\
(int) TENSOR->ne[0], (int) TENSOR->ne[1], (int) TENSOR->ne[2], TENSOR->nb[0], TENSOR->nb[1], TENSOR->nb[2]); \
{ float sum = tensor_sum_elements(TENSOR); printf("Sum of tensor %s is %6.2f\n",#TENSOR, sum); }
struct benchmark_params_struct { struct benchmark_params_struct {
int32_t n_threads = 1; int32_t n_threads = 1;
@ -59,8 +53,6 @@ void print_usage(int /*argc*/, char ** argv, struct benchmark_params_struct para
} }
int main(int argc, char ** argv) { int main(int argc, char ** argv) {
struct benchmark_params_struct benchmark_params; struct benchmark_params_struct benchmark_params;
bool invalid_param = false; bool invalid_param = false;
@ -84,11 +76,11 @@ int main(int argc, char ** argv) {
print_usage(argc, argv, benchmark_params); print_usage(argc, argv, benchmark_params);
exit(0); exit(0);
} }
if (invalid_param) { }
fprintf(stderr, "error: invalid parameter for argument: %s\n", arg.c_str()); if (invalid_param) {
print_usage(argc, argv, benchmark_params); fprintf(stderr, "error: invalid parameter for argument: %s\n", arg.c_str());
exit(1); print_usage(argc, argv, benchmark_params);
} exit(1);
} }
fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT); fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
@ -216,9 +208,8 @@ int main(int argc, char ** argv) {
// Let's use the F32 result from above as a reference for the q4_0 multiplication // Let's use the F32 result from above as a reference for the q4_0 multiplication
float sum_of_F32_reference = tensor_sum_elements(gf.nodes[0]); float sum_of_F32_reference = tensor_sum_elements(gf.nodes[0]);
printf("Iteration;NThreads; SizeX; SizeY; SizeZ; Required_FLOPS; Elapsed_u_Seconds; gigaFLOPS\n");
printf("Iteration;NThreads; SizeX; SizeY; SizeZ; Required_FLOPS; Elapsed_u_Seconds; FLOPS_per_u_Second\n"); printf("=====================================================================================\n");
printf("==============================================================================================\n");
for (int i=0;i<benchmark_params.n_iterations ;i++) { for (int i=0;i<benchmark_params.n_iterations ;i++) {
@ -227,12 +218,12 @@ int main(int argc, char ** argv) {
ggml_graph_compute(ctx, &gf31); ggml_graph_compute(ctx, &gf31);
long long int stop = ggml_time_us(); long long int stop = ggml_time_us();
long long int usec = stop-start; long long int usec = stop-start;
float flops_per_usec = (1.0f*flops_per_matrix)/usec; double gflops = (double)(flops_per_matrix)/usec/1000.0;
printf("%9i;%8i;%6i;%6i;%6i;%15lli;%18lli;%19.2f\n", printf("%9i;%8i;%6i;%6i;%6i;%15lli;%18lli;%10.2f\n",
i, i,
gf31.n_threads, gf31.n_threads,
sizex, sizey, sizez, flops_per_matrix, sizex, sizey, sizez, flops_per_matrix,
usec,flops_per_usec); usec,gflops);
#ifdef VERBOSE_DEBUGGING #ifdef VERBOSE_DEBUGGING
TENSOR_DUMP("res",gf31.nodes[0]) TENSOR_DUMP("res",gf31.nodes[0])
@ -256,7 +247,5 @@ int main(int argc, char ** argv) {
// Running a different graph computation to make sure we override the CPU cache lines // Running a different graph computation to make sure we override the CPU cache lines
ggml_graph_compute(ctx, &gf32); ggml_graph_compute(ctx, &gf32);
} }
} }