Skip to content

LLM inference in C/C++

C++ ggml

HIP: broaden MoE ncols_opt tile heuristic on RDNA3.5 architecture (#28935)

WenqiangJia2026fccf7166fb4c
diff --git a/ggml/src/ggml-cuda/mmq.cu b/ggml/src/ggml-cuda/mmq.cu
index 9b6038adff..b13b34ee9c 100644
--- a/ggml/src/ggml-cuda/mmq.cu
+++ b/ggml/src/ggml-cuda/mmq.cu
@@ -247,7 +247,7 @@ void ggml_cuda_mul_mat_q(
// Each expert only sees ne12*n_expert_used/ne02 tokens on average.
// On RDNA3 and RDNA4 it is faster to pick the tile size against this value instead of ne12.
int64_t ncols_opt = ne12;
- if (GGML_CUDA_CC_IS_RDNA3_0(cc) || GGML_CUDA_CC_IS_RDNA4(cc)) {
+ if (GGML_CUDA_CC_IS_RDNA3(cc) || GGML_CUDA_CC_IS_RDNA4(cc)) {
ncols_opt = (ne12*n_expert_used + ne02 - 1) / ne02;
}