mirror of
https://github.com/tinygrad/tinygrad.git
synced 2026-01-24 06:18:01 -05:00
add /opt/cuda/include directory (#2920)
This commit is contained in:
@@ -20,7 +20,7 @@ def check(status):
|
||||
def cu_time_execution(cb, enable=False) -> Optional[float]: return time_execution_cuda_style(cb, cuda.CUevent, cuda.cuEventCreate, cuda.cuEventRecord, cuda.cuEventSynchronize, cuda.cuEventDestroy_v2, cuda.cuEventElapsedTime, enable=enable) if not CUDACPU else cpu_time_execution(cb, enable=enable) # noqa: E501
|
||||
|
||||
@diskcache
|
||||
def compile_cuda(prg) -> bytes: return compile_cuda_style(prg, [f'--gpu-architecture={CUDADevice.default_arch_name}', "-I/usr/local/cuda/include", "-I/usr/include"], cuda.nvrtcProgram, cuda.nvrtcCreateProgram, cuda.nvrtcCompileProgram, cuda.nvrtcGetPTX, cuda.nvrtcGetPTXSize, cuda.nvrtcGetProgramLog, cuda.nvrtcGetProgramLogSize, check) # noqa: E501
|
||||
def compile_cuda(prg) -> bytes: return compile_cuda_style(prg, [f'--gpu-architecture={CUDADevice.default_arch_name}', "-I/usr/local/cuda/include", "-I/usr/include", "-I/opt/cuda/include/"], cuda.nvrtcProgram, cuda.nvrtcCreateProgram, cuda.nvrtcCompileProgram, cuda.nvrtcGetPTX, cuda.nvrtcGetPTXSize, cuda.nvrtcGetProgramLog, cuda.nvrtcGetProgramLogSize, check) # noqa: E501
|
||||
|
||||
class CUDAProgram:
|
||||
def __init__(self, device:CUDADevice, name:str, lib:bytes):
|
||||
|
||||
Reference in New Issue
Block a user