Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

CUDA: fuse muls (#21665)

Browse Source
This commit is contained in:
Aman Gupta
2026-04-10 10:24:09 +08:00
committed by GitHub
parent d132f22fc9
commit e34f042154
3 changed files with 42 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cuda/binbcast.cu
+30
View File
@@ -472,6 +472,36 @@ void ggml_cuda_op_fused_add(ggml_backend_cuda_context & ctx, ggml_tensor * dst,
}
}
void ggml_cuda_op_fused_mul(ggml_backend_cuda_context & ctx, ggml_tensor * dst, int n_fuse) {
GGML_ASSERT(2 <= n_fuse && n_fuse <= 8);
switch (n_fuse) {
case 2:
ggml_cuda_op_fused_binbcast_impl<op_mul, 2>(ctx, dst);
break;
case 3:
ggml_cuda_op_fused_binbcast_impl<op_mul, 3>(ctx, dst);
break;
case 4:
ggml_cuda_op_fused_binbcast_impl<op_mul, 4>(ctx, dst);
break;
case 5:
ggml_cuda_op_fused_binbcast_impl<op_mul, 5>(ctx, dst);
break;
case 6:
ggml_cuda_op_fused_binbcast_impl<op_mul, 6>(ctx, dst);
break;
case 7:
ggml_cuda_op_fused_binbcast_impl<op_mul, 7>(ctx, dst);
break;
case 8:
ggml_cuda_op_fused_binbcast_impl<op_mul, 8>(ctx, dst);
break;
default:
GGML_ASSERT(false && "Unsupported n_fuse value");
}
}
void ggml_cuda_op_repeat_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
const ggml_tensor * src0 = dst->src[0];