CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: add exp operation (#15456)

Browse Source 
Co-authored-by: aeseulgi <kim2h7903@gmail.com>
This commit is contained in:
Dong Won Kim
2025-08-22 00:00:16 +09:00
committed by GitHub
parent 96452a3fa4
commit 20c2dac8c6
3 changed files with 33 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@@ -490,6 +490,7 @@ struct vk_device_struct {
vk_pipeline pipeline_l2_norm_f32;
// [src/dst 0=fp32,1=fp16]
vk_pipeline pipeline_exp[2];
vk_pipeline pipeline_gelu[2];
vk_pipeline pipeline_gelu_erf[2];
vk_pipeline pipeline_gelu_quick[2];
@@ -3066,6 +3067,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
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);
CREATE_UNARY(exp)
CREATE_UNARY(gelu)
CREATE_UNARY(gelu_erf)
CREATE_UNARY(gelu_quick)
@@ -7133,6 +7135,8 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
}
switch (ggml_get_unary_op(dst)) {
case GGML_UNARY_OP_EXP:
return ctx->device->pipeline_exp[dst->type == GGML_TYPE_F16];
case GGML_UNARY_OP_SILU:
return ctx->device->pipeline_silu[dst->type == GGML_TYPE_F16];
case GGML_UNARY_OP_GELU:
@@ -9738,6 +9742,7 @@ static bool ggml_vk_build_graph(ggml_backend_vk_context * ctx, ggml_cgraph * cgr
return false;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(node)) {
case GGML_UNARY_OP_EXP:
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_GELU_ERF:
@@ -10015,6 +10020,7 @@ static bool ggml_vk_build_graph(ggml_backend_vk_context * ctx, ggml_cgraph * cgr
break;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(node)) {
case GGML_UNARY_OP_EXP:
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_GELU_ERF:
@@ -10251,6 +10257,7 @@ static bool ggml_vk_compute_forward(ggml_backend_vk_context * ctx, ggml_cgraph *
break;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(tensor)) {
case GGML_UNARY_OP_EXP:
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_GELU_ERF:
@@ -11166,6 +11173,7 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
switch (op->op) {
case GGML_OP_UNARY:
switch (ggml_get_unary_op(op)) {
case GGML_UNARY_OP_EXP:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_GELU_ERF:
case GGML_UNARY_OP_GELU_QUICK:
@@ -11965,6 +11973,9 @@ static void ggml_vk_check_results_0(ggml_backend_vk_context * ctx, ggml_cgraph *
}
} else if (tensor->op == GGML_OP_UNARY) {
switch (ggml_get_unary_op(tensor)) {
case GGML_UNARY_OP_EXP:
tensor_clone = ggml_exp(ggml_ctx, src_clone[0]);
break;
case GGML_UNARY_OP_SILU:
tensor_clone = ggml_silu(ggml_ctx, src_clone[0]);
break;