- User Since
- Jul 31 2020, 12:34 PM (106 w, 3 d)
Thu, Jul 28
@rampitec I was loading the pressure values at every candidate. initCandidate clears the values in Pressure so the array becomes empty for the next iteration.
- Use DAG to check if RPTracking is enabled.
- Eliminate diffs in extract-element.ll as this patch should not have altered that test.
Jul 11 2022
@rampitec Agreed, I have been examining the code around the changes I have made. I am not sure why that test was affected.
- reverted the function name back to initCandidate
- added testing for the option -misched-regpressure=false
Jul 7 2022
Apr 12 2022
Thanks for the review.
- replaced the feature to HasG16
- added missing test cases
Apr 9 2022
Apr 7 2022
Perhaps, a description could be added to this patch before committing.
Nov 29 2021
Thanks for the review, I will merge this patch in.
- replaced integers with memory order macros in SemaCUDA/atomic-ops.cu
Nov 26 2021
addressed feedback including
- correction of function signatures
- checking for order combinations
- correcting expected store atomic instruction
- using predefined macros for memory order
Nov 25 2021
Passed internal CI
- applied clang-format
- added order argument in the builtins and changed the tests accordingly
- adding Sema test
Nov 24 2021
removed changes in atomic-ops.cl, CUDA test may be sufficient.
Nov 21 2021
This patch needs a rebase.
LGTM, thanks for the bug fix!
Nov 18 2021
@yaxunl thanks for the review! My Github account is locked unfortunately so I will have to ask you to push this commit to the main branch. Thank you!
Nov 16 2021
- removed debug code
- some macro definitions need to be defined for HIP-only compilation
Nov 15 2021
Typo in the summary: "D16PreservesUnesedBits" -> "D16PreservesUnusedBits"
Oct 20 2021
Oct 19 2021
LGTM, appreciate your efforts!
Oct 18 2021
Oct 16 2021
Oct 15 2021
@thakis can you please check if this solution is sufficient? Thanks for bringing it up
added -target option in the test amdgpu-alias-undef-symbols.cu
cleaned up code for readability
@tra I see, we sure can rewrite that segment for readability.
Thanks for the review!
corrected sema test, as requested
removed unused diagnostic sema note
installed clang-format, refreshing patch
adding sema test
Oct 14 2021
Passed internal CI, still working on a Sema test.
Oct 13 2021
add a restrictions to what architecture AlwaysInliner should run on, updated the inline-calls.ll test.
adding codegen test
removed irrelevant lines in the test
Oct 12 2021
Oct 8 2021
Passed internal CI
added -nogpulib and -nogpuinc flags to amdgpu-alias-undef-symbols.cu
Oct 7 2021
Oct 4 2021
Oct 1 2021
- eliminated changes in SIISelLowering
- Since callees may alias to a function pointer, it makes sense for getCalleeFunction(...) to return a Function which is a cast of the operand of a GlobalAlias.
Sep 30 2021
inline-calls.ll failed on gfx908 due to the change in SIISelLowering.cpp, line 3015. Without the change, there is a failure in AMDGPUResourceAnalysis.cpp, line 65 because Op.getGlobal() is not a Function.
Sorry, that was a mistake.
Sep 29 2021
- corrected test, address space cast should go under unsafeAtomicAdd(...) codegen
Sep 28 2021
- declare failure when lowering an accessor of a callee which is not a function, in GlobalISel