Page MenuHomePhabricator
Feed Advanced Search

Fri, Aug 23

jsji created D66699: [PowerPC][Altivec] Fix constant argument for vec_dss.
Fri, Aug 23, 9:17 PM · Restricted Project
jsji added a reviewer for D66699: [PowerPC][Altivec] Fix constant argument for vec_dss: Restricted Project.
Fri, Aug 23, 9:17 PM · Restricted Project
jsji added a reviewer for D61976: [PowerPC] Reorder CSR reloads in epilogue to follow the same order as CSR saves in the prologue: Restricted Project.
Fri, Aug 23, 7:46 AM · Restricted Project

Thu, Aug 22

jsji added a comment to D63676: Early exit from Hoist() in machine licm pass based on block hotness.

machinelicm_clangformat_modv1.cpp ? I think you uploaded wrong patch.

Thu, Aug 22, 12:51 PM · Restricted Project
jsji accepted D64960: [PowerPC] Expand v1i128 smin.

LGTM with minor comments regarding testcase. Thanks.

Thu, Aug 22, 12:31 PM · Restricted Project
jsji added inline comments to D66548: [DAGCombiner] Remove explicit call to AddToWorklist in sqrt and reciprocal computations.
Thu, Aug 22, 8:24 AM · Restricted Project
jsji added inline comments to D66548: [DAGCombiner] Remove explicit call to AddToWorklist in sqrt and reciprocal computations.
Thu, Aug 22, 8:21 AM · Restricted Project
jsji committed rG545e993b8b0c: [SlotIndexes] Add print-slotindexes to disable printing slotindexes (authored by jsji).
[SlotIndexes] Add print-slotindexes to disable printing slotindexes
Thu, Aug 22, 6:51 AM
jsji added a reviewer for D64960: [PowerPC] Expand v1i128 smin: Restricted Project.
Thu, Aug 22, 6:51 AM · Restricted Project
jsji committed rL369650: [SlotIndexes] Add print-slotindexes to disable printing slotindexes.
[SlotIndexes] Add print-slotindexes to disable printing slotindexes
Thu, Aug 22, 6:44 AM
jsji closed D66500: [SlotIndexes] Add print-slotindexes to disable printing slotindexes.
Thu, Aug 22, 6:44 AM · Restricted Project

Wed, Aug 21

jsji updated subscribers of D59403: [PowerPC] Add the support for __builtin_setrnd() in clang.

Looks like you did not commit the version (Diff 190782) that was accepted!

Wed, Aug 21, 6:57 AM · Restricted Project

Tue, Aug 20

jsji retitled D66500: [SlotIndexes] Add print-slotindexes to disable printing slotindexes from [SlotIndexes] Only print slotindexes when required to [SlotIndexes] Add print-slotindexes to disable printing slotindexes.
Tue, Aug 20, 7:53 PM · Restricted Project
jsji updated the diff for D66500: [SlotIndexes] Add print-slotindexes to disable printing slotindexes.

Keep the default behavior, add option to disable.
Also add a testcase.

Tue, Aug 20, 7:52 PM · Restricted Project
jsji created D66500: [SlotIndexes] Add print-slotindexes to disable printing slotindexes.
Tue, Aug 20, 3:09 PM · Restricted Project
jsji committed rG6f833c6fe132: [llvm-extract] Update the help message for group extraction feature (authored by jsji).
[llvm-extract] Update the help message for group extraction feature
Tue, Aug 20, 1:46 PM
jsji committed rL369438: [llvm-extract] Update the help message for group extraction feature.
[llvm-extract] Update the help message for group extraction feature
Tue, Aug 20, 1:46 PM
jsji closed D66438: [llvm-extract] Update the help message for group extraction feature.
Tue, Aug 20, 1:46 PM · Restricted Project
jsji updated the diff for D66438: [llvm-extract] Update the help message for group extraction feature.

Use 'extract' instead of 'split'.

Tue, Aug 20, 1:37 PM · Restricted Project
jsji committed rGcda334ba5417: [BlockExtractor] Avoid assert with wrong line format (authored by jsji).
[BlockExtractor] Avoid assert with wrong line format
Tue, Aug 20, 7:48 AM
jsji committed rL369389: [BlockExtractor] Avoid assert with wrong line format.
[BlockExtractor] Avoid assert with wrong line format
Tue, Aug 20, 7:48 AM
jsji closed D66444: [BlockExtractor] Avoid assert with wrong line format.
Tue, Aug 20, 7:48 AM · Restricted Project

Mon, Aug 19

jsji updated the diff for D66438: [llvm-extract] Update the help message for group extraction feature.

Add missing '-'.

Mon, Aug 19, 8:04 PM · Restricted Project
jsji updated the diff for D66438: [llvm-extract] Update the help message for group extraction feature.

Agree, how about this?

Mon, Aug 19, 8:04 PM · Restricted Project
jsji created D66444: [BlockExtractor] Avoid assert with wrong line format.
Mon, Aug 19, 2:55 PM · Restricted Project
jsji created D66438: [llvm-extract] Update the help message for group extraction feature.
Mon, Aug 19, 2:08 PM · Restricted Project
jsji committed rG0776da5236e0: [PeepholeOptimizer] Don't assume bitcast def always has input (authored by jsji).
[PeepholeOptimizer] Don't assume bitcast def always has input
Mon, Aug 19, 7:20 AM
jsji committed rL369261: [PeepholeOptimizer] Don't assume bitcast def always has input.
[PeepholeOptimizer] Don't assume bitcast def always has input
Mon, Aug 19, 7:18 AM
jsji closed D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.
Mon, Aug 19, 7:18 AM · Restricted Project

Fri, Aug 16

jsji added inline comments to D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.
Fri, Aug 16, 1:39 PM · Restricted Project
jsji updated the diff for D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

Update the MIR test using script.

Fri, Aug 16, 11:36 AM · Restricted Project
jsji added inline comments to D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.
Fri, Aug 16, 11:28 AM · Restricted Project
jsji added a comment to D64665: [MachinePipeliner] Refactor schedule emission logic.

Gave it a try on SPEC2017 with -O3 -mcpu=native -mllvm --ppc-enable-pipeliner on P9,
500.perlbench_r, 525.x264_r , 531.deepsjeng_r, 557.xz_r failed with new cg, pass with old cg.

Fri, Aug 16, 8:01 AM · Restricted Project

Thu, Aug 15

jsji committed rG9106f8c87e80: [Utils][NFC] Copy bisect usage documents from commit msg into script. (authored by jsji).
[Utils][NFC] Copy bisect usage documents from commit msg into script.
Thu, Aug 15, 3:09 PM
jsji committed rL369054: [Utils][NFC] Copy bisect usage documents from commit msg into script..
[Utils][NFC] Copy bisect usage documents from commit msg into script.
Thu, Aug 15, 3:09 PM
jsji added a comment to D64665: [MachinePipeliner] Refactor schedule emission logic.

Gave it a try on test-suite, with -O3 -mcpu=pwr9 --ppc-enable-pipeliner

Thu, Aug 15, 1:43 PM · Restricted Project
jsji added a comment to D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

@arsenm Is this MIR test OK?

I think you should also keep the IR test, as it will still test the pattern change

Thu, Aug 15, 10:51 AM · Restricted Project
jsji added a reviewer for D65542: [PeepholeOptimizer] Don't assume bitcast def always has input: arsenm.

@arsenm Is this MIR test OK?

Thu, Aug 15, 10:39 AM · Restricted Project
jsji updated the diff for D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

Using MIR test instead.

Thu, Aug 15, 10:37 AM · Restricted Project
jsji added a comment to D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

@qcolombet Can you have a quick look at the testcase to see whether this is OK? Thanks.

Thu, Aug 15, 10:05 AM · Restricted Project
jsji updated the diff for D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

Rebased to latest ToT.

Thu, Aug 15, 10:05 AM · Restricted Project
jsji created D66304: [PeepholeOptimizer] Don't assume bitcast def always has input.
Thu, Aug 15, 10:05 AM · Restricted Project
jsji abandoned D66304: [PeepholeOptimizer] Don't assume bitcast def always has input.

arc command mistake.

Thu, Aug 15, 10:05 AM · Restricted Project
jsji committed rG9fd81dc13902: [PowerPC] Use xxleqv to set all one vector IMM(-1). (authored by jsji).
[PowerPC] Use xxleqv to set all one vector IMM(-1).
Thu, Aug 15, 7:35 AM
jsji committed rL369006: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
[PowerPC] Use xxleqv to set all one vector IMM(-1).
Thu, Aug 15, 7:35 AM
jsji closed D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
Thu, Aug 15, 7:35 AM · Restricted Project
jsji updated the diff for D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..

Fix long line and move the patterns.

Thu, Aug 15, 7:31 AM · Restricted Project
jsji added inline comments to D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
Thu, Aug 15, 6:51 AM · Restricted Project

Wed, Aug 14

jsji updated the diff for D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..

Added patterns for other type and rebased to pick up new tests.

Wed, Aug 14, 1:45 PM · Restricted Project
jsji committed rG37cd0dd26a97: [PowerPC][NFC] Remove duplicate tests in build-vector-test.ll (authored by jsji).
[PowerPC][NFC] Remove duplicate tests in build-vector-test.ll
Wed, Aug 14, 12:07 PM
jsji committed rL368900: [PowerPC][NFC] Remove duplicate tests in build-vector-test.ll.
[PowerPC][NFC] Remove duplicate tests in build-vector-test.ll
Wed, Aug 14, 12:06 PM
jsji added inline comments to D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
Wed, Aug 14, 7:55 AM · Restricted Project
jsji committed rGaf93ca63c3ba: [PowerPC][NFC] Add test for build all one vector with different types. (authored by jsji).
[PowerPC][NFC] Add test for build all one vector with different types.
Wed, Aug 14, 7:54 AM
jsji committed rL368859: [PowerPC][NFC] Add test for build all one vector with different types..
[PowerPC][NFC] Add test for build all one vector with different types.
Wed, Aug 14, 7:52 AM
jsji added a comment to D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..

Thanks @steven.zhang and @wuzish.

Wed, Aug 14, 7:19 AM · Restricted Project
jsji committed rGe71db6584d8d: [PowerPC][NFC] Consolidate duplicate XX3Form_SetZero and XX3Form_Zero. (authored by jsji).
[PowerPC][NFC] Consolidate duplicate XX3Form_SetZero and XX3Form_Zero.
Wed, Aug 14, 7:17 AM
jsji committed rL368856: [PowerPC][NFC] Consolidate duplicate XX3Form_SetZero and XX3Form_Zero..
[PowerPC][NFC] Consolidate duplicate XX3Form_SetZero and XX3Form_Zero.
Wed, Aug 14, 7:17 AM
jsji added a comment to D65267: [MachineCopyPropagation] Remove redundant copies after TailDup via machine-cp.

@lkail Can you please update the summary to describe more details about why we might have such redundant copies after TailDup.
I know we have those in comments, it would be great if we can summarize those into summary, so that reviewers don't have to look into every single comments. Thanks.

Wed, Aug 14, 7:10 AM · Restricted Project
jsji added inline comments to D65267: [MachineCopyPropagation] Remove redundant copies after TailDup via machine-cp.
Wed, Aug 14, 7:08 AM · Restricted Project

Tue, Aug 13

jsji added a comment to D65267: [MachineCopyPropagation] Remove redundant copies after TailDup via machine-cp.

To summarize, without complex uses and defs(thus we don't need to do complex replace-uses-with after RA).

Tue, Aug 13, 8:26 PM · Restricted Project
jsji added a parent revision for D65542: [PeepholeOptimizer] Don't assume bitcast def always has input: D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
Tue, Aug 13, 2:16 PM · Restricted Project
jsji added a child revision for D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1).: D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.
Tue, Aug 13, 2:16 PM · Restricted Project
jsji updated the diff for D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

Fix comments and add a testcase.

Tue, Aug 13, 2:13 PM · Restricted Project
jsji added a comment to D65267: [MachineCopyPropagation] Remove redundant copies after TailDup via machine-cp.

If I implement it in machine-cp, I might write some like isMoveImmediate and invoke target hook like TII->FoldImmediate. However currently only PowerPC implements specific TII->FoldImmediate. Taking above into account, this issues seems not general enough to implement in machine-cp.

Tue, Aug 13, 7:36 AM · Restricted Project

Mon, Aug 12

jsji added a comment to D65267: [MachineCopyPropagation] Remove redundant copies after TailDup via machine-cp.

Have you considered extend MachineCopyPropagation to cover this? Looks like to me that this is a 'backward' COPY propagation.

Mon, Aug 12, 2:44 PM · Restricted Project
jsji added a comment to D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.

Ping..

Mon, Aug 12, 8:40 AM · Restricted Project
jsji added a comment to D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..

Ping..

Mon, Aug 12, 8:37 AM · Restricted Project

Fri, Aug 9

jsji added a comment to D64665: [MachinePipeliner] Refactor schedule emission logic.

@jmolly Looks like you messed up the patch again? I can't see changes in CMakeList.txt in latest diff now, and after adding it, I am seeing the compile time error again.

Fri, Aug 9, 12:22 PM · Restricted Project
jsji committed rG6349ce5ca527: [MachinePipeliner] Avoid indeterminate order in FuncUnitSorter (authored by jsji).
[MachinePipeliner] Avoid indeterminate order in FuncUnitSorter
Fri, Aug 9, 7:11 AM
jsji committed rL368441: [MachinePipeliner] Avoid indeterminate order in FuncUnitSorter.
[MachinePipeliner] Avoid indeterminate order in FuncUnitSorter
Fri, Aug 9, 7:10 AM
jsji closed D65992: [MachinePipeliner] Avoid indeterminate order in FuncUnitSorter.
Fri, Aug 9, 7:10 AM · Restricted Project
jsji added a comment to D65992: [MachinePipeliner] Avoid indeterminate order in FuncUnitSorter.

Thanks for the patch! My only concern is with all the checks in the test case. Checking for the exact code sequence can be very sensitive to other changes in the compiler that are unrelated to this patch.

Thanks,
Brendon

Fri, Aug 9, 6:53 AM · Restricted Project

Thu, Aug 8

jsji created D65992: [MachinePipeliner] Avoid indeterminate order in FuncUnitSorter.
Thu, Aug 8, 7:52 PM · Restricted Project
jsji committed rG8faf2f3c4063: Re-commit "[PowerPC][NFC][MachinePipeliner] Add some regression testcases"" (authored by jsji).
Re-commit "[PowerPC][NFC][MachinePipeliner] Add some regression testcases""
Thu, Aug 8, 10:39 AM
jsji committed rL368318: Re-commit "[PowerPC][NFC][MachinePipeliner] Add some regression testcases"".
Re-commit "[PowerPC][NFC][MachinePipeliner] Add some regression testcases""
Thu, Aug 8, 10:39 AM
jsji accepted D65630: [PowerPC] Port SSE3, SSSE3 and SSE4 intrinsics to PowerPC..

LGTM.

Thu, Aug 8, 7:56 AM · Restricted Project, Restricted Project

Fri, Aug 2

jsji added a comment to rL367741: Temporarily Revert "[PowerPC][NFC][MachinePipeliner] Add some regression….

Thanks @echristo !

Fri, Aug 2, 7:11 PM
jsji added a comment to D64665: [MachinePipeliner] Refactor schedule emission logic.

Yes, good job! All the existing lit test passed.
Haven't done run time verification, but there are a few compile time assertion & abort when compiling test-suites,
I have reduced some of them and committed in https://reviews.llvm.org/rL367732 for your further investigation.

Fri, Aug 2, 3:30 PM · Restricted Project
jsji committed rG81a05cb9bcc5: [PowerPC][NFC][MachinePipeliner] Add some regression testcases (authored by jsji).
[PowerPC][NFC][MachinePipeliner] Add some regression testcases
Fri, Aug 2, 3:28 PM
jsji committed rL367732: [PowerPC][NFC][MachinePipeliner] Add some regression testcases.
[PowerPC][NFC][MachinePipeliner] Add some regression testcases
Fri, Aug 2, 3:28 PM
jsji committed rG490292c3e042: [PowerPC][NFC] Enable ADT BitVectorTest (authored by jsji).
[PowerPC][NFC] Enable ADT BitVectorTest
Fri, Aug 2, 12:59 PM
jsji committed rL367713: [PowerPC][NFC] Enable ADT BitVectorTest.
[PowerPC][NFC] Enable ADT BitVectorTest
Fri, Aug 2, 12:58 PM

Thu, Aug 1

jsji accepted D64905: [NFC][CodeGen] Modify the type element of TailCalls to simplify the dupRetToEnableTailCallOpts().

LGTM.

Thu, Aug 1, 11:30 AM · Restricted Project
jsji accepted D65315: [PowerPC][Peephole] Check if `extsw`'s second operand is a virtual register.

LGTM.

Thu, Aug 1, 8:05 AM · Restricted Project
jsji added a comment to D64665: [MachinePipeliner] Refactor schedule emission logic.

How do you build and test? I can't config & build it with clang.

Thu, Aug 1, 7:47 AM · Restricted Project
jsji added inline comments to D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
Thu, Aug 1, 6:31 AM · Restricted Project

Wed, Jul 31

Herald updated subscribers of D65544: [RLEV] Rewrite loop exit values for multiple exit loops w/o overall loop exit count.
Wed, Jul 31, 8:01 PM · Restricted Project
jsji created D65542: [PeepholeOptimizer] Don't assume bitcast def always has input.
Wed, Jul 31, 2:46 PM · Restricted Project
jsji created D65529: [PowerPC] Use xxleqv to set all one vector IMM(-1)..
Wed, Jul 31, 11:59 AM · Restricted Project
jsji accepted D65063: [PowerPC] Eliminate loads/swap feeding swap/store for vector type by using big-endian load/store.

LGTM.

Wed, Jul 31, 6:59 AM · Restricted Project

Tue, Jul 30

jsji accepted D65063: [PowerPC] Eliminate loads/swap feeding swap/store for vector type by using big-endian load/store.

LGTM. Thanks for exploiting.

Tue, Jul 30, 8:06 AM · Restricted Project

Mon, Jul 29

jsji added a comment to D59883: Lower generic MASSV entries to PowerPC subtarget-specific entries.

Why we need a new pass to do this simple lowering?
Why can't we do it similar to InitLibcalls and change the suffixes for subtarget using setLibcallName?

Mon, Jul 29, 3:00 PM · Restricted Project
jsji added a comment to D65063: [PowerPC] Eliminate loads/swap feeding swap/store for vector type by using big-endian load/store.

Mostly good to me.

Mon, Jul 29, 12:53 PM · Restricted Project
jsji committed rG5bb6202c44e0: [PowerPC][NFC]Fix a typo in comment. (authored by jsji).
[PowerPC][NFC]Fix a typo in comment.
Mon, Jul 29, 12:31 PM
jsji committed rL367252: [PowerPC][NFC]Fix a typo in comment..
[PowerPC][NFC]Fix a typo in comment.
Mon, Jul 29, 12:27 PM
jsji accepted D64849: Add platform guards to PPC vector intrinsics headers..

LGTM.

Mon, Jul 29, 7:54 AM · Restricted Project, Restricted Project

Jul 22 2019

jsji accepted D64814: [NFC][PowerPC]Change ADDIStocHA to ADDIStocHA8 to follow 64-bit naming convention.

LGTM, sorry for late response.

Jul 22 2019, 8:36 AM · Restricted Project
jsji added a reviewer for D65094: [PowerPC] Combine address computation to favour selecting DForm instructions: shchenz.
Jul 22 2019, 8:28 AM · Restricted Project
jsji accepted D64195: [PowerPC] Replace float load/store pair with integer load/store pair when it's only used in load/store.

LGTM.

Jul 22 2019, 8:28 AM · Restricted Project

Jul 18 2019

jsji added inline comments to rL366339: Fix MSVC 'implicitly capture' compile error. NFCI..
Jul 18 2019, 7:01 AM