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

vulkan: add REPEAT op support for f16 to f16. (llama/23298) 

* feat: extend repeat op for vulkan

* feat: add repeat_f16 vulkan pipeline

* fix: ensure same dst and src types

* fix: use type_size instead of data types

* fix: use int16 and int32 for repeat shader op

* chore: rename repeat_f* to repeat_i*

* chore: rename repeat vulkan pipelines
This commit is contained in:
l8bloom 2026-05-27 16:59:08 +02:00 committed by Georgi Gerganov
parent 98c6722fec
commit c5cde8c717
2 changed files with 14 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@ -768,7 +768,8 @@ struct vk_device_struct {
vk_pipeline pipeline_clamp_f32;
vk_pipeline pipeline_pad_f32;
vk_pipeline pipeline_roll_f32;
vk_pipeline pipeline_repeat_f32, pipeline_repeat_back_f32;
vk_pipeline pipeline_repeat_i32, pipeline_repeat_back_f32;
vk_pipeline pipeline_repeat_i16;
vk_pipeline pipeline_cpy_f32_f32, pipeline_cpy_f32_f16, pipeline_cpy_f16_f16, pipeline_cpy_f16_f32, pipeline_cpy_f32_bf16, pipeline_cpy_bf16_f32, pipeline_cpy_f32_i32, pipeline_cpy_i32_f32;
vk_pipeline pipeline_contig_cpy_f32_f32, pipeline_contig_cpy_f32_f16, pipeline_contig_cpy_f16_f16, pipeline_contig_cpy_f16_f32, pipeline_contig_cpy_f32_bf16, pipeline_contig_cpy_bf16_f32, pipeline_contig_cpy_f32_i32, pipeline_contig_cpy_i32_f32;
vk_pipeline pipeline_cpy_f32_quant[GGML_TYPE_COUNT];
@ -4708,9 +4709,11 @@ static void ggml_vk_load_shaders(vk_device& device) {
ggml_vk_create_pipeline(device, device->pipeline_roll_f32, "roll_f32", roll_f32_len, roll_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_repeat_f32, "repeat_f32", repeat_f32_len, repeat_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_repeat_i32, "repeat_i32", repeat_i32_len, repeat_i32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_repeat_back_f32, "repeat_back_f32", repeat_back_f32_len, repeat_back_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_repeat_i16, "repeat_i16", repeat_i16_len, repeat_i16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
#define CREATE_UNARY(name) \
ggml_vk_create_pipeline(device, device->pipeline_ ## name [0], #name "_f32", name ## _f32_len, name ## _f32_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1); \
ggml_vk_create_pipeline(device, device->pipeline_ ## name [1], #name "_f16", name ## _f16_len, name ## _f16_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);
@ -9738,7 +9741,10 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
return nullptr;
case GGML_OP_REPEAT:
if (ggml_type_size(src0->type) == sizeof(float) && ggml_type_size(dst->type) == sizeof(float)) {
return ctx->device->pipeline_repeat_f32;
return ctx->device->pipeline_repeat_i32;
}
if (ggml_type_size(src0->type) == 2 && ggml_type_size(dst->type) == 2) {
return ctx->device->pipeline_repeat_i16;
}
return nullptr;
case GGML_OP_REPEAT_BACK:
@ -16253,7 +16259,8 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
return false;
}
case GGML_OP_REPEAT:
return ggml_type_size(op->type) == sizeof(float) && ggml_type_size(op->src[0]->type) == sizeof(float);
return ggml_type_size(op->type) == ggml_type_size(op->src[0]->type) &&
(ggml_type_size(op->type) == sizeof(float) || ggml_type_size(op->type) == 2);
case GGML_OP_REPEAT_BACK:
return op->type == GGML_TYPE_F32 && op->src[0]->type == GGML_TYPE_F32;
case GGML_OP_ROPE:

4
ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
View File

@ -798,9 +798,11 @@ void process_shaders() {
string_to_spv("div_f32", "div.comp", {{"A_TYPE", "float"}, {"B_TYPE", "float"}, {"D_TYPE", "float"}, {"FLOAT_TYPE", "float"}});
string_to_spv("repeat_f32", "repeat.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}});
string_to_spv("repeat_i32", "repeat.comp", {{"A_TYPE", "int32_t"}, {"D_TYPE", "int32_t"}});
string_to_spv("repeat_back_f32", "repeat_back.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}});
string_to_spv("repeat_i16", "repeat.comp", {{"A_TYPE", "int16_t"}, {"D_TYPE", "int16_t"}});
string_to_spv("scale_f32", "scale.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}, {"FLOAT_TYPE", "float"}});
string_to_spv("sqr_f32", "square.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}, {"FLOAT_TYPE", "float"}});