- User Since
- Jan 27 2014, 9:36 AM (233 w, 3 d)
I'm sure there are other bugs around _Atomic, please file a bug an CC me if that's the case. I'll commit this fix for now.
Tue, Jul 17
Fri, Jul 13
- Simplify CreateStore.
Thu, Jul 12
- Fix silly naming and lookup.
I updated the patch to use Address, and also use inbounds. This was a bit tricky because of the GEPs. I also updated tests which run into this codegen. Only the following tests actually hit this code, and not all check these stores:
- Use Address as suggested in review.
Wed, Jul 11
Tue, Jul 10
Thu, Jul 5
LGTM after a few questions.
Fri, Jun 22
Jun 8 2018
Jun 4 2018
Jun 1 2018
May 31 2018
- Add back directory test
Note that I don't touch the OpenCL __constant stuff, because the separate address space means this can be actually read-only, which means you can't cmpxchg for very wide reads.
- Remove access time checks, simplify existing check, after talking to EricWF on IRC.
Wow this is amazing!
Making it the behavior we choose is fine with me. I just understood your comment to mean that you wanted to give this semantics, and that's more work than just stating the semantics you want.
Looks fine in general, though I know ~nothing of RISCV.
When the A extension is supported, __atomic libcalls will be generated for any atomic that isn't the native word size or has less than natural alignment.
May 30 2018
Could you detail what type of code will generate this? In C++ the standard library is responsible for guaranteeing properly sized and aligned atomics. Size will always be accurate, but I'm worried that your change wants perfect alignment information and you might not have that information. Say two pieces of code communicate through the same atomic location, but they have different alignment information, and one gets a libcall and the other doesn't... You're in trouble.
May 29 2018
May 26 2018
- Fix variable capitalization.
May 25 2018
GCC in libcxx-libcxxabi-x86_64-linux-ubuntu-cxx03 seems to mis-handle ATOMIC_VAR_INIT:
Fixed by r333290.
May 24 2018
Ping! clang side landed in https://reviews.llvm.org/rL333246
- Address nit.
- Change suggested by Richard
- Merge format-size-spec-nsinteger
- Address nit.
Can you also add a test for _Bool _Accum.
May 23 2018
Actually, scratch that. We will be enabling it since GCC does. Will update this and other relevant C++ related code appropriately.