diff options
author | Hans Goudey <h.goudey@me.com> | 2022-11-11 04:14:48 +0300 |
---|---|---|
committer | Hans Goudey <h.goudey@me.com> | 2022-11-11 04:41:13 +0300 |
commit | 9465b109af0b70244a36f9e68493e316d9f8b56f (patch) | |
tree | c4d933bf09f8313071556063156538ec3356b24c /intern/cycles/kernel/device/cuda/compat.h | |
parent | 026d21a225521670c6b5083da9da61227da69e65 (diff) | |
parent | ca1642cd0c5cdf634fe2022c955d93983de95934 (diff) |
Merge branch 'master' into refactor-mesh-position-genericrefactor-mesh-position-generic
Diffstat (limited to 'intern/cycles/kernel/device/cuda/compat.h')
-rw-r--r-- | intern/cycles/kernel/device/cuda/compat.h | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/intern/cycles/kernel/device/cuda/compat.h b/intern/cycles/kernel/device/cuda/compat.h index 51e1381d552..3a950779c11 100644 --- a/intern/cycles/kernel/device/cuda/compat.h +++ b/intern/cycles/kernel/device/cuda/compat.h @@ -30,6 +30,7 @@ typedef unsigned long long uint64_t; /* Qualifiers */ #define ccl_device __device__ __inline__ +#define ccl_device_extern extern "C" __device__ #if __CUDA_ARCH__ < 500 # define ccl_device_inline __device__ __forceinline__ # define ccl_device_forceinline __device__ __forceinline__ @@ -109,14 +110,14 @@ ccl_device_forceinline T ccl_gpu_tex_object_read_3D(const ccl_gpu_tex_object_3D typedef unsigned short half; -__device__ half __float2half(const float f) +ccl_device_forceinline half __float2half(const float f) { half val; asm("{ cvt.rn.f16.f32 %0, %1;}\n" : "=h"(val) : "f"(f)); return val; } -__device__ float __half2float(const half h) +ccl_device_forceinline float __half2float(const half h) { float val; asm("{ cvt.f32.f16 %0, %1;}\n" : "=f"(val) : "h"(h)); |