tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-11 11:05:39 -04:00
Code Issues Packages Projects Releases Wiki Activity

HIP: Add support for RDNA4 targets (#12372)

Browse Source
This commit is contained in:
Slobodan Josic
2025-03-26 23:46:30 +01:00
committed by GitHub
parent b3298fa47a
commit bd40678df7
7 changed files with 25 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@@ -1216,7 +1216,7 @@ static void ggml_cuda_op_mul_mat_cublas(
CUBLAS_CHECK(cublasSetStream(ctx.cublas_handle(id), stream));
if (GGML_CUDA_CC_IS_CDNA(cc)) {
if (GGML_CUDA_CC_IS_CDNA(cc) || GGML_CUDA_CC_IS_RDNA4(cc)) {
const float alpha = 1.0f;
const float beta = 0.0f;
CUBLAS_CHECK(
@@ -1759,7 +1759,9 @@ static void ggml_cuda_mul_mat_batched_cublas(ggml_backend_cuda_context & ctx, co
beta = &beta_f32;
}
if (GGML_CUDA_CC_IS_CDNA(ggml_cuda_info().devices[ctx.device].cc)) {
int id = ggml_cuda_get_device();
const int cc = ggml_cuda_info().devices[id].cc;
if (GGML_CUDA_CC_IS_CDNA(cc) || GGML_CUDA_CC_IS_RDNA4(cc)) {
cu_compute_type = CUBLAS_COMPUTE_32F;
alpha = &alpha_f32;
beta = &beta_f32;
@@ -1836,7 +1838,7 @@ static void ggml_cuda_mul_mat_batched_cublas(ggml_backend_cuda_context & ctx, co
}
#endif
if (dst->op_params[0] == GGML_PREC_DEFAULT) {
if (dst->op_params[0] == GGML_PREC_DEFAULT && cu_data_type == CUDA_R_16F) {
const to_fp32_cuda_t to_fp32_cuda = ggml_get_to_fp32_cuda(GGML_TYPE_F16);
to_fp32_cuda(dst_f16.get(), dst_ddf, ne_dst, main_stream);
}