tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-08 01:26:51 -04:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: add set rows for f32 and f16 (#14551)

Browse Source 
* CUDA: add set rows for f32 and f16

* Review: change kernel params, use strides from host

* Use 1-d kernel

* Review: use int64_t for blockDim.x, rename nb->s for clarity
This commit is contained in:
Aman Gupta
2025-07-12 21:31:38 +08:00
committed by GitHub
parent 8eff95544e
commit 7de5c7cab6
3 changed files with 147 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -43,6 +43,7 @@
#include "ggml-cuda/upscale.cuh"
#include "ggml-cuda/wkv.cuh"
#include "ggml-cuda/gla.cuh"
#include "ggml-cuda/set-rows.cuh"
#include "ggml.h"
#include <algorithm>
@@ -2230,6 +2231,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
case GGML_OP_GET_ROWS_BACK:
ggml_cuda_op_get_rows_back(ctx, dst);
break;
case GGML_OP_SET_ROWS:
ggml_cuda_op_set_rows(ctx, dst);
break;
case GGML_OP_DUP:
ggml_cuda_dup(ctx, dst);
break;
@@ -3216,6 +3220,12 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
{
return op->type == GGML_TYPE_F32 && op->src[0]->type == GGML_TYPE_F32 && op->ne[2] == 1 && op->ne[3] == 1;
} break;
case GGML_OP_SET_ROWS:
{
return (op->type == GGML_TYPE_F32 || op->type == GGML_TYPE_F16) &&
op->src[0]->type == GGML_TYPE_F32 &&
op->src[1]->type == GGML_TYPE_I64;
} break;
case GGML_OP_CPY:
{
ggml_type src0_type = op->src[0]->type;