public_git
/
whisper.cpp
mirror of https://github.com/ggml-org/whisper.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Templating to differenciate the block_q4_0

This commit is contained in:
Swetha B S 2025-07-15 04:01:25 -07:00
parent d3e56e5be5
commit b0c631cfb6
1 changed files with 20 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
ggml/src/ggml-cpu/repack.cpp
View File

@ -1419,7 +1419,14 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
switch (src0->type) {
case GGML_TYPE_Q4_0: {
ggml_compute_forward_get_rows_q4_0x8(params, dst);
if (ggml_cpu_has_avx2() || (ggml_cpu_has_sve() && ggml_cpu_has_matmul_int8() && ggml_cpu_get_sve_cnt() == QK8_0)) {
if (src0->ne[1] % 8 == 0) {
ggml_compute_forward_get_rows_q4_0<block_q4_0x8>(params, dst, 8);
}
} else {
GGML_ABORT("Unsupported block interleaved size for get_rows function");
}
} break;
default:
GGML_ABORT("fatal error");
@ -1427,9 +1434,11 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
}
}
static void ggml_compute_forward_get_rows_q4_0x8(
template<typename BLOCK_TYPE>
static void ggml_compute_forward_get_rows_q4_0(
const ggml_compute_params * params,
ggml_tensor * dst) {
ggml_tensor * dst,
int nrows_interleaved) {
const ggml_tensor * src0 = dst->src[0];
const ggml_tensor * src1 = dst->src[1];
@ -1453,8 +1462,7 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
const int ir0 = dr * ith;
const int ir1 = MIN(ir0 + dr, nr);
constexpr int nrows_interleaved = 8;
const size_t sizeof_one_repacked_block = sizeof(block_q4_0x8);
const size_t sizeof_one_repacked_block = sizeof(BLOCK_TYPE);
const int num_repacked_blocks_per_row_width = nc / QK4_0;
@ -1473,11 +1481,11 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
const char * base_ptr_for_higher_dims_in_src0 = (const char *)src0->data + i11 * nb02 + i12 * nb03;
// Pointer to the first block_q4_0x8 of the identified row_group_idx
const block_q4_0x8 * p_first_repacked_block_of_group_x8 = (const block_q4_0x8 *)(base_ptr_for_higher_dims_in_src0 + row_group_idx * stride_between_actual_row_groups);
// Pointer to the first <BLOCK_TYPE> of the identified row_group_idx
const BLOCK_TYPE * p_first_repacked_block_of_group_block_type = (const BLOCK_TYPE *)(base_ptr_for_higher_dims_in_src0 + row_group_idx * stride_between_actual_row_groups);
dequantize_row_q4_0x8(
p_first_repacked_block_of_group_x8,
dequantize_row_q4_0<block_q4_0x8>(
p_first_repacked_block_of_group_block_type,
(float *)((char *)dst->data + i10 * nb1 + i11 * nb2 + i12 * nb3), nc, row_idx_in_group);
}
}
@ -1490,8 +1498,9 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
* @param k Total number of elements (columns) in the logical row.
* @param row_idx_in_group Index (0-7) of the logical row to dequantize.
*/
static void dequantize_row_q4_0x8(
const block_q4_0x8 * GGML_RESTRICT p_repacked_group_column_blocks,
template<typename BLOCK_TYPE>
static void dequantize_row_q4_0(
const BLOCK_TYPE * GGML_RESTRICT p_repacked_group_column_blocks,
float * GGML_RESTRICT y,
int64_t k,
int row_idx_in_group) {