CUDA: fix softmax compile for old CUDA versions

Author: JohannesGaessler

ggml-cuda.cu

@@ -116,6 +116,8 @@
 #include "ggml.h"
 #include "ggml-backend-impl.h"
 
+#define CUDART_CI     11070 // CUDA 11.7, version used for the Github CI
+
 #define CC_PASCAL     600
 #define MIN_CC_DP4A   610 // minimum compute capability for __dp4a, an intrinsic for byte-wise dot products
 #define CC_VOLTA      700
@@ -596,16 +598,16 @@ static __device__ __forceinline__ float2 warp_reduce_sum(float2 a) {
 }
 
 static __device__ __forceinline__ half2 warp_reduce_sum(half2 a) {
-#if __CUDA_ARCH__ < CC_PASCAL || (defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__))
-    (void) a;
-    bad_arch();
-#else
+#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         a = __hadd2(a, __shfl_xor_sync(0xffffffff, a, mask, 32));
     }
     return a;
-#endif // __CUDA_ARCH__ < CC_PASCAL || (defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__))
+#else
+    (void) a;
+    bad_arch();
+#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL
 }
 
 static __device__ __forceinline__ float warp_reduce_max(float x) {
@@ -617,16 +619,16 @@ static __device__ __forceinline__ float warp_reduce_max(float x) {
 }
 
 static __device__ __forceinline__ half2 warp_reduce_max(half2 x) {
-#if __CUDA_ARCH__ < CC_PASCAL || (defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__))
-    (void) x;
-    bad_arch();
-#else
+#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL && CUDART_VERSION >= CUDART_CI
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         x = __hmax2(x, __shfl_xor_sync(0xffffffff, x, mask, 32));
     }
     return x;
-#endif // __CUDA_ARCH__ < CC_PASCAL || (defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__))
+#else
+    (void) x;
+    bad_arch();
+#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL && CUDART_VERSION >= CUDART_CI
 }
 
 static __device__ __forceinline__ float op_repeat(const float a, const float b) {
@@ -5415,7 +5417,7 @@ static __global__ void diag_mask_inf_f32(const float * x, float * dst, const int
 
 template <bool vals_smem, int ncols_template, int block_size_template, bool need_check>
 static __global__ void soft_max_f16(const float * x, const float * y, float * dst, const int ncols_par, const int nrows_y, const float scale) {
-#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL
+#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL && CUDART_VERSION >= CUDART_CI
     const int ncols_data = ncols_template == 0 ? ncols_par : ncols_template;
     const int ncols_smem = GGML_PAD(ncols_data, 2*WARP_SIZE)/2;
 
@@ -5540,7 +5542,7 @@ static __global__ void soft_max_f16(const float * x, const float * y, float * ds
 #else
     (void) x; (void) y; (void) dst; (void) ncols_par; (void) nrows_y; (void) scale;
     bad_arch();
-#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL
+#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_PASCAL && CUDART_VERSION >= CUDART_CI
 }
 
 template <bool vals_smem, int ncols_template, int block_size_template>
@@ -8352,15 +8354,15 @@ static void ggml_cuda_op_soft_max(
     float scale = 1.0f;
     memcpy(&scale, dst->op_params, sizeof(float));
 
-#if defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
-    const bool use_f16_soft_max = false;
-#else
+#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && CUDART_VERSION >= CUDART_CI
 #ifdef GGML_CUDA_F16
     const bool use_f16_soft_max = true;
 #else
     const bool use_f16_soft_max = false;
 #endif // GGML_CUDA_F16
-#endif // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
+#else
+    const bool use_f16_soft_max = false;
+#endif // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__) && CUDART_VERSION >= CUDART_CI
 
     if (use_f16_soft_max) {
         soft_max_f16_cuda(src0_dd, src1 ? src1_dd : nullptr, dst_dd, ne00, nrows_x, nrows_y, scale, main_stream);