vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix typos, use GGML_TYPE defines, improve code

Browse source
This commit is contained in:
0cc4m 2023-04-25 18:43:31 +02:00
parent daa5df51f7
commit 36bfb3c158
3 changed files with 9 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

4
Makefile
View file

@ -34,8 +34,8 @@ endif
# #
# keep standard at C11 and C++11 # keep standard at C11 and C++11
CFLAGS = -I. -O3 -DNODEBUG -std=c11 -fPIC CFLAGS = -I. -O3 -DNDEBUG -std=c11 -fPIC
CXXFLAGS = -I. -I./examples -O3 -DNODEBUG -std=c++11 -fPIC CXXFLAGS = -I. -I./examples -O3 -DNDEBUG -std=c++11 -fPIC
LDFLAGS = LDFLAGS =
# warnings # warnings

14
ggml-opencl.cpp
View file

@ -111,11 +111,7 @@ void ggml_cl_malloc(size_t req_size, size_t* cur_size, cl_mem_flags flags, cl_me
void ggml_cl_sgemm_wrapper(const CLBlastLayout order, const CLBlastTranspose trans_a, const CLBlastTranspose trans_b, const int m, const int n, const int k, const float alpha, const void *host_a, const int lda, const float *host_b, const int ldb, const float beta, float *host_c, const int ldc, const int btype) { void ggml_cl_sgemm_wrapper(const CLBlastLayout order, const CLBlastTranspose trans_a, const CLBlastTranspose trans_b, const int m, const int n, const int k, const float alpha, const void *host_a, const int lda, const float *host_b, const int ldb, const float beta, float *host_c, const int ldc, const int btype) {
cl_int err = 0; cl_int err = 0;
cl_event events[4]; cl_event events[4] = { NULL };
events[0] = NULL;
events[1] = NULL;
events[2] = NULL;
events[3] = NULL;
cl_kernel kernel; cl_kernel kernel;
size_t global, local, size_qb; size_t global, local, size_qb;
@ -124,22 +120,22 @@ void ggml_cl_sgemm_wrapper(const CLBlastLayout order, const CLBlastTranspose tra
global = n * k; global = n * k;
switch (btype) { switch (btype) {
case 2: case GGML_TYPE_Q4_0:
kernel = kernel_q4_0; kernel = kernel_q4_0;
local = 16; local = 16;
size_qb = global * (sizeof(float) + local) / 32; size_qb = global * (sizeof(float) + local) / 32;
break; break;
case 3: case GGML_TYPE_Q4_1:
kernel = kernel_q4_1; kernel = kernel_q4_1;
local = 16; local = 16;
size_qb = global * (sizeof(float) * 2 + local) / 32; size_qb = global * (sizeof(float) * 2 + local) / 32;
break; break;
case 4: case GGML_TYPE_Q4_2:
kernel = kernel_q4_2; kernel = kernel_q4_2;
local = 8; local = 8;
size_qb = global * (sizeof(short) + local) / 16; size_qb = global * (sizeof(short) + local) / 16;
break; break;
case 5: case GGML_TYPE_Q4_3:
kernel = kernel_q4_3; kernel = kernel_q4_3;
local = 8; local = 8;
size_qb = global * (sizeof(short) * 2 + local) / 16; size_qb = global * (sizeof(short) * 2 + local) / 16;

4
ggml.c
View file

@ -7580,7 +7580,7 @@ static void ggml_compute_forward_mul_mat_f32(
1.0f, y, ne10, 1.0f, y, ne10,
x, ne10, x, ne10,
0.0f, d, ne01, 0.0f, d, ne01,
params->type); GGML_TYPE_F32);
#else #else
cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasTrans, cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasTrans,
ne11, ne01, ne10, ne11, ne01, ne10,
@ -7814,7 +7814,7 @@ static void ggml_compute_forward_mul_mat_f16_f32(
1.0f, y, ne10, 1.0f, y, ne10,
x, ne10, x, ne10,
0.0f, d, ne01, 0.0f, d, ne01,
params->type); GGML_TYPE_F32);
#else #else
const float * x = wdata; const float * x = wdata;
const float * y = (float *) ((char *) src1->data + i02*nb12 + i03*nb13); const float * y = (float *) ((char *) src1->data + i02*nb12 + i03*nb13);