User Details
- User Since
- Oct 28 2014, 11:30 AM (438 w, 2 d)
Today
Wed, Mar 22
Tue, Mar 21
Mon, Mar 20
This is essentially a GPU only change. Stamping anyway.
Fri, Mar 17
OK from the libc structuring point of view.
The libc infra components LGTM.
Thu, Mar 16
Except of the definition of the new types, the rest LGTM. We should of course formally define our position on the locales that we care about.
LGTM modulo answer to @lntue's question.
Mostly LGTM but I have left some questions inline. About the structuring, to avoid complicated conditionals in CMake and source code both, we should structure it this way:
I have no major comments so stepping aside.
Wed, Mar 15
Sorry for the delay. Going by my godbolt experiments, the problem you seem to be addressing (you have already said it) is clang specific: https://godbolt.org/z/GxfMor1KM
I have updated the commit message to make it clear that this change is for riscv64.
Do you have any resolution on the const local variable naming convention?
We have libc RISCV64 builders now: https://lab.llvm.org/buildbot/#/workers/202
Please watch them after landing.
LGTM with one comment.
Also, does it happen at -O3 as well?
Can you share the code that results in large number of instructions and register saving?
Tue, Mar 14
Address comments.
I think you need to rebase this.
Mon, Mar 13
I have left comments in longjmp.cpp but same questions apply for setjmp.cpp as well.
Rebase
Sorry, I did not intend to push this without review. Because of a mix of branches on my local checkout, I ended up pushing it accidentally.