|
@@ -78,13 +78,13 @@
|
|
|
// Moore Threads
|
|
// Moore Threads
|
|
|
#define GGML_CUDA_MUSA_ARCH_IS_QY1 (__MUSA_ARCH__ <= 210)
|
|
#define GGML_CUDA_MUSA_ARCH_IS_QY1 (__MUSA_ARCH__ <= 210)
|
|
|
|
|
|
|
|
-#define GGML_CUDA_CC_QY1 (GGML_MUSA_CC_OFFSET_MTHREADS + 0x210) // MTT S80, MTT S3000
|
|
|
|
|
-#define GGML_CUDA_CC_QY2 (GGML_MUSA_CC_OFFSET_MTHREADS + 0x220) // MTT S4000
|
|
|
|
|
-#define GGML_CUDA_CC_NG (GGML_MUSA_CC_OFFSET_MTHREADS + 0x310) // TBD
|
|
|
|
|
|
|
+#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
|
|
|
|
|
|
|
|
#define GGML_CUDA_CC_IS_MTHREADS(cc) (cc >= GGML_CUDA_CC_OFFSET_MTHREADS && cc < GGML_CUDA_CC_OFFSET_AMD)
|
|
#define GGML_CUDA_CC_IS_MTHREADS(cc) (cc >= GGML_CUDA_CC_OFFSET_MTHREADS && cc < GGML_CUDA_CC_OFFSET_AMD)
|
|
|
#define GGML_CUDA_CC_IS_QY1(cc) (cc >= GGML_CUDA_CC_QY1 && cc < GGML_CUDA_CC_QY2)
|
|
#define GGML_CUDA_CC_IS_QY1(cc) (cc >= GGML_CUDA_CC_QY1 && cc < GGML_CUDA_CC_QY2)
|
|
|
-#define GGML_CUDA_CC_IS_QY2(cc) (cc >= GGML_CUDA_CC_QY2 && cc < GGML_CUDA_CC_NEXT)
|
|
|
|
|
|
|
+#define GGML_CUDA_CC_IS_QY2(cc) (cc >= GGML_CUDA_CC_QY2 && cc < GGML_CUDA_CC_NG)
|
|
|
#define GGML_CUDA_CC_IS_NG(cc) (cc >= GGML_CUDA_CC_NG)
|
|
#define GGML_CUDA_CC_IS_NG(cc) (cc >= GGML_CUDA_CC_NG)
|
|
|
|
|
|
|
|
#ifdef __CUDA_ARCH_LIST__
|
|
#ifdef __CUDA_ARCH_LIST__
|