- User Since
- Oct 18 2012, 4:53 AM (476 w, 6 d)
Tue, Nov 16
Nov 4 2021
Thanks John, committed as 3d39612b3dd3.
Nov 3 2021
Nov 2 2021
Oct 25 2021
Oct 22 2021
Thanks for your thoughts too John, I've implemented your suggestion instead of the ret undef path I was thinking about.
Oct 20 2021
I've looked into this some more, and the return inserted by @llvm.coro.end is not necessarily representable in this function. If it's a @llvm.coro.id.retcon.once coroutine then the return is always a ret void, regardless of this function's prototype.
Oct 5 2021
Thanks, pushed as 5f65ee260d78.
Oct 4 2021
Sep 23 2021
Sep 22 2021
Sep 20 2021
Thanks. Committed as 13aa102e0769.
Sep 16 2021
- Added tests.
- Switched to existing SDAG LDP/STP route for volatiles.
- Simplified GISel path with matching expressions.
- Think I fixed everything else mentioned.
Sep 15 2021
Added separate feature to control it (named from ARMARM), since it's optional from v8.2 onwards.
Sep 14 2021
Sep 13 2021
Thanks. Pushed as 5d070c8259a.
Sep 3 2021
Thanks for that fix, sorry I was silent.
Aug 27 2021
Aug 20 2021
Aug 19 2021
Aug 18 2021
Aug 16 2021
Aug 10 2021
Thanks, committed as 5ad086089962.
Aug 9 2021
Adopted suggested changes. The align one didn't actually change the MIR output unfortunately so I don't tihnk it's really testable.
Aug 4 2021
Thanks Gerolf, committed as d7b0e5525a4e.
Aug 3 2021
Jul 26 2021
Thanks. Committed as a487a49acc5a.
Jul 22 2021
Jul 21 2021
Thanks Jessica, committed as 291e0daa6.
Thanks Gerolf, committed as 19d2e42be2cd.
Jul 16 2021
Jul 15 2021
Thanks. Committed as 5d7632ee72c5.
Jul 14 2021
Jul 13 2021
Thanks, 7802f62b3f2c with a better test.
Jul 9 2021
Jul 8 2021
Thanks Duncan, committed as 727e1c9be3a5
Jul 7 2021
Jul 6 2021
Jun 29 2021
Should be fixed now by c82957e79236. Thanks for your patience.
Jun 28 2021
Jun 20 2021
There was a little discussion in https://reviews.llvm.org/D102612, though if you have any suggestions I'm obviously happy to rework it now. I'll also try to remember to include you if and when something similar comes up again.
Jun 16 2021
Sorry about the late reply, I’m on a pretty remote holiday in the Outer
Hebrides. Thanks for pointing out the issue, it’s clearly a wrong
assertion. I’ll make sure the underlying code is right and update it first
thing when I get back to civilization.
Jun 4 2021
Thanks, committed as b16ddd03753d.
Jun 1 2021
- Add llvm::thread for all potential platforms, allowing stack size to be specified.
- llvm::thread based on the Threading.inc where possible, else std::thread, else in-thread synchronous execution (with progressively more failures).
- Remove llvm_execute_on_thread* in favour of this new llvm::thread.
May 28 2021
Thanks, committed as e94fada045fe with the extra tests.
I guess one option would be to never use std::thread, and repurpose the guts of llvm_execute_on_thread_impl to implement a cross-platform llvm::thread that allows specifying the stack size as an optional constructor parameter (and then change llvm_execute_on_thread to use llvm::thread). WDYT?
Thanks Jessica, committed as d88f96df (with the -1 sentinel).
Thanks Eli, committed as 9ff2eb1ea596.
May 26 2021
Actual main thread stack size is 8MB, but I'd accidentally implemented 2MB originally.
May 25 2021
Added test for sret promotion case the verifier can't handle.
May 21 2021
Add error message to x86 backend if it can't honour a musttail call. It seems from the (presumably original) musttail.ll test that it's intended to override the disable-tail-calls function attribute, so I had to teach SelectionDAGBuilder about that too.