[UR][CUDA][HIP] Refactor setKernelParams #18518
Conversation
- Remove unused Context parameters - Avoid unnecessary copy in `guessLocalWorkSize` - Simplify the control flow in setKernelParams - Move cached properties fetching code to constructors - Query HIP for occupancy in `guessLocalWorkSize`
npmiller
force-pushed
the
set-kernel-params
branch
from
9598ff2 to
226e80e
Compare
| int getMaxBlockDimX() const noexcept { return MaxBlockDimX; }; | ||
|
|
||
| int getMaxBlockDimY() const noexcept { return MaxBlockDimY; }; | ||
| size_t getMaxBlockDim(int dim) const noexcept { return MaxBlockDim[dim]; }; |
There was a problem hiding this comment.
Is noexcept on something that accesses MaxBlockDim[dim] correct? Might'nt this technically throw an exception when out of bounds?
There was a problem hiding this comment.
It's a size_t C array not a vector, so I don't believe this would ever throw an exception. Maybe with extra windows bounds checking or something? But even then I'd expect that to be asserting instead.
There was a problem hiding this comment.
Yeah perhaps. Unfortunately https://en.cppreference.com/w/cpp/container/language/array.html is down so I can't see the C++ behaviour, but another (less trusted) source said that the out-of-range behaviour on a C-style array is undefined and so I don't know if we can guarantee it won't throw an exception.
|
@intel/llvm-gatekeepers this is ready to merge |
guessLocalWorkSizeguessLocalWorkSize