huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 10:38:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

More casting kernels.

Browse Source
This commit is contained in:
laurent
2023-06-27 09:36:35 +01:00
parent ee3d290f8b
commit bb262ecc99
1 changed files with 19 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
kernels/src/cast.cu
View File

@ -1,4 +1,5 @@
#include "cuda_utils.cuh"
#include<stdint.h>
#define CAST_OP(SRC_TYPENAME, DST_TYPENAME, FN_NAME) \
extern "C" __global__ void FN_NAME( \
@ -25,10 +26,23 @@ extern "C" __global__ void FN_NAME( \
#if __CUDA_ARCH__ >= 530
CAST_OP(__half, __half, cast_f16_f16)
CAST_OP(__half, float, cast_f16_f32)
CAST_OP(float, __half, cast_f32_f16)
CAST_OP(__half, uint32_t, cast_f16_u32)
CAST_OP(__half, float, cast_f16_f32)
CAST_OP(__half, double, cast_f16_f64)
CAST_OP(uint32_t, __half, cast_u32_f16)
CAST_OP(float, __half, cast_f32_f16)
CAST_OP(double, __half, cast_f64_f16)
#endif
CAST_OP(float, float, cast_f32_f32)
CAST_OP(float, double, cast_f32_f64)
CAST_OP(double, float, cast_f64_f32)
CAST_OP(uint32_t, uint32_t, cast_u32_u32)
CAST_OP(uint32_t, float, cast_u32_f32)
CAST_OP(uint32_t, double, cast_u32_f64)
CAST_OP(float, uint32_t, cast_f32_u32)
CAST_OP(float, float, cast_f32_f32)
CAST_OP(float, double, cast_f32_f64)
CAST_OP(double, uint32_t, cast_f64_u32)
CAST_OP(double, float, cast_f64_f32)
CAST_OP(double, double, cast_f64_f64)