From: R0CKSTAR <redacted>
Date: Tue, 19 Aug 2025 10:33:47 +0000 (+0800)
Subject: musa: handle __hgt2_mask, available starting from MUSA SDK rc4.3.0 (#15413)
X-Git-Tag: upstream/0.0.6527~323
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=67f09a3a27db443f9870aac87e163dba0d08131e;p=pkg%2Fggml%2Fsources%2Fllama.cpp

musa: handle __hgt2_mask, available starting from MUSA SDK rc4.3.0 (#15413)

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

diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh
index 2b14b30a..76ace816 100644
--- a/ggml/src/ggml-cuda/common.cuh
+++ b/ggml/src/ggml-cuda/common.cuh
@@ -78,6 +78,8 @@
 #define GGML_CUDA_CC_IS_CDNA3(cc) (cc >= GGML_CUDA_CC_CDNA3 && cc < GGML_CUDA_CC_RDNA1)
 
 // Moore Threads
+#define MUSART_HMASK 40300 // MUSA rc4.3, min. ver. for half2 -> uint mask comparisons
+
 #define GGML_CUDA_CC_QY1 (GGML_CUDA_CC_OFFSET_MTHREADS + 0x210) // MTT S80, MTT S3000
 #define GGML_CUDA_CC_QY2 (GGML_CUDA_CC_OFFSET_MTHREADS + 0x220) // MTT S4000
 #define GGML_CUDA_CC_NG  (GGML_CUDA_CC_OFFSET_MTHREADS + 0x310) // TBD
@@ -490,13 +492,14 @@ static __device__ __forceinline__ half2 warp_reduce_max(half2 x) {
 #endif // !defined(GGML_USE_HIP) && __CUDA_ARCH__ >= GGML_CUDA_CC_PASCAL || defined(GGML_USE_HIP)
 }
 
-#if CUDART_VERSION < CUDART_HMASK
+#if (defined(CUDART_VERSION) && CUDART_VERSION < CUDART_HMASK) || defined(GGML_USE_HIP) || \
+    (defined(MUSART_VERSION) && MUSART_VERSION < MUSART_HMASK)
 static __device__ __forceinline__ uint32_t __hgt2_mask(const half2 a, const half2 b) {
     const uint32_t mask_low  = 0x0000FFFF * (float( __low2half(a)) > float( __low2half(b)));
     const uint32_t mask_high = 0xFFFF0000 * (float(__high2half(a)) > float(__high2half(b)));
     return mask_low | mask_high;
 }
-#endif // CUDART_VERSION < CUDART_HMASK
+#endif // (defined(CUDART_VERSION) && CUDART_VERSION < CUDART_HMASK) || defined(GGML_USE_HIP) || (defined(MUSART_VERSION) && MUSART_VERSION < MUSART_HMASK)
 
 static __device__ __forceinline__ int ggml_cuda_dp4a(const int a, const int b, int c) {
 #if defined(GGML_USE_HIP)