mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 29 Packages Projects Releases Wiki Activity

vulkan: support L2_NORM with contiguous rows (#19604)

Browse Source
This commit is contained in:
Jeff Bolz
2026-02-13 21:42:04 -08:00
committed by GitHub
parent 53aef25a88
commit dbb023336b
3 changed files with 31 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/ggml-vulkan.cpp
+8 -4
View File
@@ -4082,7 +4082,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
}
ggml_vk_create_pipeline(device, device->pipeline_rms_norm_back_f32, "rms_norm_back_f32", rms_norm_back_f32_len, rms_norm_back_f32_data, "main", 3, sizeof(vk_op_push_constants), {1, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_l2_norm_f32, "l2_norm_f32", l2_norm_f32_len, l2_norm_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_l2_norm_f32, "l2_norm_f32", l2_norm_f32_len, l2_norm_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {1, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_cpy_f32_f32, "cpy_f32_f32", cpy_f32_f32_len, cpy_f32_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_cpy_f32_f16, "cpy_f32_f16", cpy_f32_f16_len, cpy_f32_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
@@ -10639,8 +10639,10 @@ static void ggml_vk_rms_norm_back(ggml_backend_vk_context * ctx, vk_context& sub
}
static void ggml_vk_l2_norm(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) {
float * op_params = (float *)dst->op_params;
ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, nullptr, dst, GGML_OP_L2_NORM, { (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], op_params[0], 0.0f, 0.0f, 0.0f });
const float * op_params = (const float *)dst->op_params;
vk_op_unary_push_constants p = vk_op_unary_push_constants_init(src0, dst);
p.param1 = op_params[0];
ggml_vk_op_f32<vk_op_unary_push_constants>(ctx, subctx, src0, nullptr, nullptr, nullptr, dst, GGML_OP_L2_NORM, std::move(p));
}
static void ggml_vk_unary(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) {
@@ -14912,8 +14914,10 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
return true;
case GGML_OP_NORM:
case GGML_OP_GROUP_NORM:
case GGML_OP_L2_NORM:
return ggml_is_contiguous(op->src[0]);
case GGML_OP_L2_NORM:
return ggml_is_contiguous_rows(op->src[0]) &&
op->src[0]->type == GGML_TYPE_F32 && op->type == GGML_TYPE_F32;
case GGML_OP_ADD:
case GGML_OP_SUB:
case GGML_OP_MUL: