- User Since
- Jun 7 2017, 5:05 AM (136 w, 3 d)
Updated the comment.
Thu, Jan 16
Thank you for the review.
@atanasyan Do you have any comment?
Tue, Jan 14
Mon, Jan 13
Removed leftover lines.
Tue, Dec 31
Added a check of IssueCounters.
Mon, Dec 30
Fri, Dec 27
Wed, Dec 25
Dec 19 2019
I have version 4.4.0+git47-g71166e2-1 so yes, I think this is the problem since the alias was added only after v4.6.0.
Dec 18 2019
Dec 17 2019
Dec 16 2019
Dec 13 2019
LGTM, thanks for the fix!
Dec 12 2019
Dec 10 2019
Dec 9 2019
Yes, other targets do have it but use it quite sparingly. X86 almost never uses it but there are some edge cases when parsing certain directives. For example, this line will trigger an assert fail because of an uninitialized TargetStreamer on X86:
echo ".cv_fpo_proc foo 4" | llvm-exegesis -mode latency -snippets-file=-
MIPS on the other hand uses TargetStreamer right at the start of the MipsAsmParser constructor and throughout the code used for parsing. The design of this MIPS code is quite different from other architectures (always relying on TargetStreamer or Assembler). I'm not sure if this was out of necessity or if it just evolved in to different path. Either way, the current code doesn't facilitate simple instruction extraction without all the necessary elements being in place.
Dec 6 2019
Nov 28 2019
Nov 21 2019
It's possible to run some tests manually but the test mentioned above, which will be part of lit.local.cfg and which checks whether lit can run llvm-exegesis tests on Mips, is encountering a segfault so I'm working on fixing that first.
I came across this while running llvm-exegesis on a mips board, e.g.:
llvm-exegesis -mode latency -snippets-file /dev/null
It gets the 'generic' CPU name with sys::getHostCPUName() and calls createTargetMachine(), which then emits a warning:
'generic' is not a recognized processor for this target (ignoring processor)
Nov 20 2019
Nov 15 2019
Nov 14 2019
Nov 13 2019
Thanks for the review. I forgot to mention that I've gotten commit access in meantime so I'll be able to commit future patches.
Nov 12 2019
Merged and expanded the test.
Shortened the tests.
Nov 8 2019
Nov 6 2019
Oct 24 2019
I agree that yes is fine. Even though it's not enabled but default it's still usable and is in the spirit that the table defines.
Added full context to the diff.
I'm not certain if the tail calls should be marked as partial or if they can be considered supported.
Oct 11 2019
Oct 10 2019
LGTM, just have some clarifying questions.
Thank you for the reviews. I don't have commit access. @petarj usually commits the patches.
Oct 9 2019
Removed obsolete llvm:: and includes, ran clang-format. Added direct testing of the instruction.
Oct 8 2019
Aug 19 2019
Jul 26 2019
Should we also add the (dis)assembler tests or are these case covered in other tests?
Jul 15 2019
Maybe we should add the tests that this patch had since they were not added in rL234401.
rL234401 also disabled printing of the alias instruction while it's enabled in here. Not sure which one is preferable.
Jun 18 2019
May 21 2019
Apr 25 2019