- User Since
- Oct 24 2016, 8:15 AM (232 w, 5 d)
Wed, Mar 24
Mar 8 2021
discussed with sean offline, the we do not call the emitGlobalDtorWithCXAAtExit() in AIX. so we do not have the problem for the "__dso_handle"
I initially had concerns with the places clang introduces non-default visibility in codegen. The few cases I was worried we would hit were related to static init , but that was before @Xiangling_L static init related work. After those changes I have no other concerns. I would wait a day or two to ensure that there are no more comments related to the round-trip-args behaviour but otherwise LGTM.
Mar 6 2021
Mar 5 2021
Thanks for the updates Lei. LGTM.
One minor comment but otherwise LGTM.
- clang-formated lambda
- changed check-prefixes to check-prefix
- added whitespace between 32-bit and 64-bit check lines
- Changed check for further custom RegLcos to include that the ValNo of the next RegLoc is the same as the original ValNo.
Patch looks good Lei, the one thing we are missing though is a lit test that checks for the expected toc entries.
Mar 4 2021
Mar 3 2021
Fix formating on assert.
- Add asserts that the ValNo of custom RegLocs match the ValNo of the custom MemLoc.
- Assert that we are targeting 32-bit codegen when we have 4 custom RegLocs.
- Add comments to the tests to make it easier to determine what is being tested.
Mar 2 2021
- Converted loop to explicit handling of custom RegLocs
- Added const on member function.
Mar 1 2021
Feb 28 2021
Fixed a number of formatting and spelling mistakes.
Feb 25 2021
Feb 18 2021
Feb 17 2021
Could there be a simpler approach where we always emit a .got section on PPC64: In addReservedSymbols if we fail to resolve .TOC. then we create it?
Feb 16 2021
Feb 12 2021
LGTM. Please wait until David has given it the OK too though.
Feb 11 2021
Thanks for the update. LGTM.
I'll need a bit more time to finish going through ppc-shrink-wrapping.ll but the patch is looking really good. Thanks for the updates.
Feb 10 2021
Addressed review comments.
Use early return instead.
Jan 18 2021
Jan 15 2021
Rebased on trunk.
Jan 14 2021
Thanks Jinsong, LGTM.
Jan 13 2021
Thanks Jinsong, left a couple really minor comments. Will rebase the xxspltd fix on this patch now.
Jan 12 2021
clang -target powerpc64le-unknown-linux-gnu -c -o a.o modern-as.s error: instruction use requires an option to be enabled mfudscr 2
IIUC this patch is breaking compatibility with gas due to the AIX assembler not supporting the mnemonic. Similarly for my xxspltd patch, we wouldn't be able to encode the mnemonic without passing extra options on Linux. Thats why I took the approach I did in D94419. I am happy to change direction, but in my opinion we should only be disabling the mnemonics on AIX.
Thanks Jinsong, LGTM.
Jan 11 2021
Dec 7 2020
Dec 2 2020
If there is no objections, I'll go ahead and commit later this week.
Fixed one too many hyphens in man page change.
- Removed leading space in option description
- Updated the manpage doc.
On FreeBSD, the main use of this will be on the new powerpc64le arch, where we need to build a 32-bit LE bootloader for use with pseries. (it is easier to retarget LLVM than make a cross-endian bootloader, as it would involve rewriting filesystem code etc.)
Excuse my ignorance, but what are there technical limitations preventing writing n 64-bit LE boot loader and avoid having a 32-bit LE target all-together?
Dec 1 2020
- Added -mattr=+altivec to vector pop count test to keep pwr5 testing
- Added a test to show enabling altivec through function attribute.
Nov 30 2020
Nov 26 2020
Sorry Zarko, I must have deleted my comment before posting last night I've added it back now.
Nov 25 2020
1 comment, but otherwise LGTM.
Nov 24 2020
Nov 23 2020
A couple of really minor comments, but otherwise LGTM.
Nov 20 2020
Nov 19 2020
Nov 18 2020
Thanks for the updates. LGTM.
Nov 17 2020
Nov 13 2020
Nov 12 2020
- Add testing for warn-backrefs interaction.
- Spelling fixes
- Fix comments to use ## instead of # in lit test.
Nov 11 2020
Nov 9 2020
- Added a detailed comment trying to capture some of the linked discussion and distill why we want the more complicated lookup behaviour (at least by default).
- Added an option -fortran-common to toggle the new behaviour on/off.
Nov 5 2020
- Add comments about why we have the special lookup behaviour for common in archives..
- Use NonCommon as opposed to UnCommon
- switch second if to be an else if instead.
Oct 28 2020
- Added comment in the test describing why the behaviour change is needed.
- Converted a dyn_cast to a cast.
- Fixed paramater name to use lowercase.
- Fixed .size directives in the test to refer to the correct object.