From 2d5acca196c298a41aeeb1f4a1e404ca133fd501 Mon Sep 17 00:00:00 2001
From: uvos <devnull@uvos.xyz>
Date: Wed, 27 Nov 2024 14:16:46 +0100
Subject: [PATCH] ggml_cuda: set launch bounds also for GCN as it helps there
 too

---
 ggml/src/ggml-cuda/mmq.cuh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/ggml/src/ggml-cuda/mmq.cuh b/ggml/src/ggml-cuda/mmq.cuh
index a8ecd8b30..8d8867121 100644
--- a/ggml/src/ggml-cuda/mmq.cuh
+++ b/ggml/src/ggml-cuda/mmq.cuh
@@ -2570,9 +2570,9 @@ static __device__ void mul_mat_q_process_tile(
 
 template <ggml_type type, int mmq_x, int nwarps, bool need_check>
 #if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
-#if defined(RDNA3) || defined(RDNA2) || defined(CDNA)
+#if defined(RDNA3) || defined(RDNA2) || defined(CDNA) || defined(GCN)
     __launch_bounds__(WARP_SIZE*nwarps, 2)
-#endif // defined(RDNA3) || defined(RDNA2)
+#endif // defined(RDNA3) || defined(RDNA2) || defined(CDNA) || defined(GCN)
 #else
 #if __CUDA_ARCH__ >= CC_VOLTA
     __launch_bounds__(WARP_SIZE*nwarps, 1)