OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

Define the SYCL device version __assert_fail when the NDEBUG defined. (#58906)

Browse Source 
Summary:
## Motivation
The utils in namespace  `c10` require the `__assert_fail` when the NDEBUG is defined in kernel code.

The `__assert_fail` declaration in pytorch is not compatible to the SYCL‘s specification.

This causes compile error when use these utils in SYCL kernels.

## Solution
Add the `__assert_fail` declaration for SYCL kernels to pytorch when compiling the SYCL kernels with `c10` utils.

## Additional context
`__assert_fail` in SYCL kernel

`extern SYCL_EXTERNAL void __assert_fail(const char *expr, const char *file, unsigned int line, const char *func);`

Pull Request resolved: https://github.com/pytorch/pytorch/pull/58906

Reviewed By: anjali411

Differential Revision: D28700863

Pulled By: ezyang

fbshipit-source-id: 81896d022b35ace8cd16474128649eabedfaf138
This commit is contained in:
chengjun 2021-05-26 12:41:23 -07:00 committed by Facebook GitHub Bot
parent 2e2a75720b
commit 511979df85
1 changed files with 8 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
c10/macros/Macros.h
View File

@ -333,6 +333,13 @@ __host__ __device__
#else // __APPLE__, _MSC_VER
#if defined(NDEBUG)
extern "C" {
#if defined(__SYCL_DEVICE_ONLY__)
extern SYCL_EXTERNAL void __assert_fail(
const char* expr,
const char* file,
unsigned int line,
const char* func);
#else // __SYCL_DEVICE_ONLY__
#if (defined(__CUDA_ARCH__) && !(defined(__clang__) && defined(__CUDA__))) || \
defined(__HIP_ARCH__) || defined(__HIP__)
__host__ __device__
@ -343,6 +350,7 @@ __host__ __device__
const char* file,
unsigned int line,
const char* function) throw();
#endif
}
#endif // NDEBUG
#define CUDA_KERNEL_ASSERT(cond) \