- User Since
- Jun 28 2018, 9:57 PM (81 w, 1 d)
Mon, Jan 13
Sun, Jan 12
Sat, Jan 11
Hi @sanjoy.google , could you please help to have another look at this? Thanks a lot.
Fri, Jan 10
use getNonEmpty() interface & add more test cases
operand 0 does not required to be NonPos/NonNeg
Thu, Jan 9
use unsigned range min as initial value
Wed, Jan 8
Tue, Jan 7
rebase after D64869
Sun, Jan 5
Thu, Jan 2
address @steven.zhang comments
Wed, Jan 1
update testing point
Mon, Dec 30
Sun, Dec 29
for signed range with nsw flag, we can only use the sign min value when constant range for START can not cross 0.
Fri, Dec 27
Thu, Dec 26
LGTM with one nit comment.
Wed, Dec 25
fix result mask MB > ME cases.
Mon, Dec 23
LGTM. Some minor comment, you can handle it as you like.
Thanks for finding out this bug. As discussed, I will take this patch.
Seems there is indeed some bug for current folding. If the wrap relationship between mb and me for folding result rlwinm is different with he wrap relationship in MI.
Thanks for finding the bug in the testcases. I have modified them in NFC patch https://reviews.llvm.org/rG79b3325be0b016fdc1a2c55bce65ec9f1e5f4eb6
Right solution should be https://reviews.llvm.org/D71690. We can convert sle to slt if get(Signed|Unsigned)RangeMin(LHS) is not min value. In https://reviews.llvm.org/D71690, we can get a more accurate min value for AddRec with flag nsw|nuw.
Sun, Dec 22
address @nikic comments
fix @sanjoy.google comments
Dec 19 2019
Seems we can not simply add nsw|nuw for start + stride as I did in https://reviews.llvm.org/D64868. Since SCEV are shared with expressions in different blocks, we must ensure all expressions mapped to same SCEV can be added with nsw|nuw.
Dec 18 2019
Hi @dblaikie seems this commit makes check llvm fail on PowerPC.
: 'RUN: at line 1'; ./llvm-project/llvm/build/bin/llc --function-sections < ./llvm-project/llvm/test/DebugInfo/X86/gmlt-empty-base-address.ll -filetype=obj | ./llvm-project/llvm/build/bin/llvm-dwarfdump -v -debug-info - | ./llvm-project/llvm/build/bin/FileCheck ./llvm-project/llvm/test/DebugInfo/X86/gmlt-empty-base-address.ll -- Exit Code: 1
Right, current patch should still be with bug.
Dec 17 2019
Dec 11 2019
Dec 9 2019
LGTM. Thanks for exploiting these instructions.
Dec 3 2019
Nov 28 2019
fix the failure cases found by buildbot.
Nov 27 2019
This patch caused failure in http://lab.llvm.org:8011/builders/clang-ppc64be-linux-multistage/builds/21918
Nov 26 2019
is it caused by the way I sent out the review? I git mv the file, git diff to get diff patch and upload the diff patch here. Not sure which step is wrong. I will abandon this patch.
Nov 25 2019
Address @jsji comments.
Nov 24 2019
Nov 21 2019
This is a very good idea. Love it!.
Nov 20 2019
Thanks for this exploitation. LGTM too.