youngkingdom/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

CUTLASS 3.1 (#915)

Browse Source 
Co-authored-by: Aniket Shivam <ashivam@nvidia.com>
This commit is contained in:
ANIKET SHIVAM
2023-04-14 20:19:34 -07:00
committed by GitHub
parent 9b8166e3f0
commit d572cc1aab
482 changed files with 37184 additions and 16419 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
include/cute/util/debug.hpp
View File

@ -99,8 +99,12 @@ namespace cute
// A dummy function that uses compilation failure to print a type
template <class T>
CUTE_HOST_DEVICE
void
CUTE_HOST_DEVICE void
print_type() {
static_assert(sizeof(T) < 0, "Printing type T.");
}
template <class T>
CUTE_HOST_DEVICE void
print_type(T&&) {
static_assert(sizeof(T) < 0, "Printing type T.");
}
@ -113,13 +117,23 @@ print_type(T&&) {
// if (block0()) print(...);
// if (thread(42)) print(...);
CUTE_HOST_DEVICE
bool
block(int bid)
{
#if defined(__CUDA_ARCH__)
return blockIdx.x + blockIdx.y*gridDim.x + blockIdx.z*gridDim.x*gridDim.y == bid;
#else
return true;
#endif
}
CUTE_HOST_DEVICE
bool
thread(int tid, int bid)
{
#if defined(__CUDA_ARCH__)
return (threadIdx.x + threadIdx.y*blockDim.x + threadIdx.z*blockDim.x*blockDim.y == tid)
&& ( blockIdx.x + blockIdx.y* gridDim.x + blockIdx.z* gridDim.x* gridDim.y == bid);
return (threadIdx.x + threadIdx.y*blockDim.x + threadIdx.z*blockDim.x*blockDim.y == tid) && block(bid);
#else
return true;
#endif
@ -129,7 +143,7 @@ CUTE_HOST_DEVICE
bool
thread(int tid)
{
return thread(tid, 0);
return thread(tid,0);
}
CUTE_HOST_DEVICE
@ -143,11 +157,7 @@ CUTE_HOST_DEVICE
bool
block0()
{
#if defined(__CUDA_ARCH__)
return !(blockIdx.x | blockIdx.y | blockIdx.z);
#else
return true;
#endif
return block(0);
}
} // end namespace cute