From: R0CKSTAR <redacted>
Date: Tue, 12 Aug 2025 02:02:51 +0000 (+0800)
Subject: musa: fix failures in test-backend-ops for mul_mat_id op (#15236)
X-Git-Tag: upstream/0.0.6164~29
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=25ff6f7659f6a5c47d6a73eada5813f0495331f0;p=pkg%2Fggml%2Fsources%2Fllama.cpp

musa: fix failures in test-backend-ops for mul_mat_id op (#15236)

* musa: fix failures in test-backend-ops for mul_mat_id op

Signed-off-by: Xiaodong Ye <redacted>

* Address review comments

Signed-off-by: Xiaodong Ye <redacted>

---------

Signed-off-by: Xiaodong Ye <redacted>
---

diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh
index 2e5d4879..a23da57e 100644
--- a/ggml/src/ggml-cuda/common.cuh
+++ b/ggml/src/ggml-cuda/common.cuh
@@ -312,11 +312,11 @@ static bool turing_mma_available(const int cc) {
 }
 
 static bool ampere_mma_available(const int cc) {
-    return cc < GGML_CUDA_CC_OFFSET_AMD && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_AMPERE;
+    return GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_AMPERE;
 }
 
 static bool cp_async_available(const int cc) {
-    return cc < GGML_CUDA_CC_OFFSET_AMD && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_AMPERE;
+    return GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_AMPERE;
 }
 
 static constexpr __device__ int ggml_cuda_get_physical_warp_size() {