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

ggml : move ggml_nbytes_split to ggml-cuda.cu

Browse source
This commit is contained in:
slaren 2023-12-14 18:17:07 +01:00
parent ae8b4840bc
commit 0040524060
3 changed files with 6 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml-cuda.cu
View file

@ -8898,6 +8898,12 @@ static void ggml_cuda_nop(const ggml_tensor * src0, const ggml_tensor * src1, gg
(void) dst; (void) dst;
} }
static size_t ggml_nbytes_split(const struct ggml_tensor * tensor, int nrows_split) {
static_assert(GGML_MAX_DIMS == 4, "GGML_MAX_DIMS is not 4 - update this function");
return nrows_split*ggml_row_size(tensor->type, tensor->ne[0]);
}
void ggml_cuda_transform_tensor(void * data, struct ggml_tensor * tensor) { void ggml_cuda_transform_tensor(void * data, struct ggml_tensor * tensor) {
const int64_t nrows = ggml_nrows(tensor); const int64_t nrows = ggml_nrows(tensor);

7
ggml.c
View file

@ -1997,13 +1997,6 @@ size_t ggml_nbytes_pad(const struct ggml_tensor * tensor) {
return GGML_PAD(ggml_nbytes(tensor), GGML_MEM_ALIGN); return GGML_PAD(ggml_nbytes(tensor), GGML_MEM_ALIGN);
} }
// TODO: move to ggml-cuda?
size_t ggml_nbytes_split(const struct ggml_tensor * tensor, int nrows_split) {
static_assert(GGML_MAX_DIMS == 4, "GGML_MAX_DIMS is not 4 - update this function");
return nrows_split*ggml_row_size(tensor->type, tensor->ne[0]);
}
int ggml_blck_size(enum ggml_type type) { int ggml_blck_size(enum ggml_type type) {
return type_traits[type].blck_size; return type_traits[type].blck_size;
} }

1
ggml.h
View file

@ -638,7 +638,6 @@ extern "C" {
GGML_API int64_t ggml_nrows (const struct ggml_tensor * tensor); GGML_API int64_t ggml_nrows (const struct ggml_tensor * tensor);
GGML_API size_t ggml_nbytes (const struct ggml_tensor * tensor); GGML_API size_t ggml_nbytes (const struct ggml_tensor * tensor);
GGML_API size_t ggml_nbytes_pad (const struct ggml_tensor * tensor); // same as ggml_nbytes() but padded to GGML_MEM_ALIGN GGML_API size_t ggml_nbytes_pad (const struct ggml_tensor * tensor); // same as ggml_nbytes() but padded to GGML_MEM_ALIGN
GGML_API size_t ggml_nbytes_split(const struct ggml_tensor * tensor, int nrows_split);
GGML_API int ggml_blck_size(enum ggml_type type); GGML_API int ggml_blck_size(enum ggml_type type);
GGML_API size_t ggml_type_size(enum ggml_type type); // size in bytes for all elements in a block GGML_API size_t ggml_type_size(enum ggml_type type); // size in bytes for all elements in a block