Skip to content

feat: add Vulkan REPEAT op support for f16 to f16. #23298

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
0cc4m merged 7 commits into from
May 27, 2026
Merged

feat: add Vulkan REPEAT op support for f16 to f16. #23298

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
70c8d5f
feat: extend repeat op for vulkan
l8bloom May 18, 2026
36c6088
feat: add repeat_f16 vulkan pipeline
l8bloom May 19, 2026
0fe360b
fix: ensure same dst and src types
l8bloom May 19, 2026
81d29f4
fix: use type_size instead of data types
l8bloom May 19, 2026
7c846a3
fix: use int16 and int32 for repeat shader op
l8bloom May 20, 2026
f93e281
chore: rename repeat_f* to repeat_i*
l8bloom May 20, 2026
b5e2932
chore: rename repeat vulkan pipelines
l8bloom May 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion ggml/src/ggml-vulkan/ggml-vulkan.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -759,6 +759,7 @@ struct vk_device_struct {
vk_pipeline pipeline_pad_f32;
vk_pipeline pipeline_roll_f32;
vk_pipeline pipeline_repeat_f32, pipeline_repeat_back_f32;
vk_pipeline pipeline_repeat_f16;
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];
Expand Down Expand Up @@ -4693,6 +4694,8 @@ static void ggml_vk_load_shaders(vk_device& device) {
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_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_f16, "repeat_f16", repeat_f16_len, repeat_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
Comment thread
l8bloom marked this conversation as resolved.
Outdated

#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);
Expand Down Expand Up @@ -9632,6 +9635,9 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
if (ggml_type_size(src0->type) == sizeof(float) && ggml_type_size(dst->type) == sizeof(float)) {
return ctx->device->pipeline_repeat_f32;
}
if (ggml_type_size(src0->type) == 2 && ggml_type_size(dst->type) == 2) {
return ctx->device->pipeline_repeat_f16;
}
return nullptr;
case GGML_OP_REPEAT_BACK:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
Expand Down Expand Up @@ -16020,7 +16026,8 @@ 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:
Expand Down
4 changes: 3 additions & 1 deletion ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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_f32", "repeat.comp", {{"A_TYPE", "int32_t"}, {"D_TYPE", "int32_t"}});
Comment thread
l8bloom marked this conversation as resolved.
Outdated
string_to_spv("repeat_back_f32", "repeat_back.comp", {{"A_TYPE", "float"}, {"D_TYPE", "float"}});

string_to_spv("repeat_f16", "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"}});
Expand Down