From 31b62276cfb47f005349f150e76375ff315e999f Mon Sep 17 00:00:00 2001
From: R0CKSTAR <xiaodong.ye@mthreads.com>
Date: Wed, 19 Mar 2025 02:28:26 +0800
Subject: [PATCH] musa: override warp_size of musa device to 32 (llama/12445)

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---
 ggml/src/ggml-cuda/ggml-cuda.cu | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
index 8fb06382..5cb56df9 100644
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -262,6 +262,8 @@ static ggml_cuda_device_info ggml_cuda_init() {
                       id, prop.name, prop.gcnArchName, info.devices[id].cc & 0xffff,
                       device_vmm ? "yes" : "no", prop.warpSize);
 #elif defined(GGML_USE_MUSA)
+        // FIXME: Ensure compatibility with varying warp sizes across different MUSA archs.
+        info.devices[id].warp_size = 32;
         // TODO: refine the .cc to reflect MUSA's actual CC capabilities
         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
         info.devices[id].cc = 100*prop.major + 10*prop.minor;