Page MenuHomePhabricator

yinma (Yin Ma)
User

Projects

User does not belong to any projects.

User Details

User Since
Jan 19 2015, 10:06 AM (248 w, 1 d)

Recent Activity

Nov 15 2018

yinma added inline comments to D53444: Support ARM_V4BX relocation.
Nov 15 2018, 10:47 AM
yinma updated the diff for D53444: Support ARM_V4BX relocation.
Nov 15 2018, 10:30 AM

Nov 14 2018

yinma updated the diff for D53444: Support ARM_V4BX relocation.
Nov 14 2018, 6:00 PM

Nov 8 2018

yinma updated the diff for D54173: [ARM] Add MemOp into LDRcp to enable removing.
Nov 8 2018, 10:38 AM

Nov 7 2018

yinma updated the diff for D54173: [ARM] Add MemOp into LDRcp to enable removing.
Nov 7 2018, 3:38 PM
yinma added inline comments to D53444: Support ARM_V4BX relocation.
Nov 7 2018, 2:38 PM
yinma updated the diff for D53444: Support ARM_V4BX relocation.
Nov 7 2018, 2:28 PM

Nov 6 2018

yinma updated the diff for D54173: [ARM] Add MemOp into LDRcp to enable removing.
Nov 6 2018, 1:54 PM
yinma retitled D54173: [ARM] Add MemOp into LDRcp to enable removing from Add MemOp into LDRcp to enable removing to [ARM] Add MemOp into LDRcp to enable removing.
Nov 6 2018, 11:55 AM
yinma added a reviewer for D54173: [ARM] Add MemOp into LDRcp to enable removing: eli.friedman.
Nov 6 2018, 11:49 AM
yinma created D54173: [ARM] Add MemOp into LDRcp to enable removing.
Nov 6 2018, 11:49 AM

Nov 1 2018

yinma updated the diff for D53996: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Nov 1 2018, 4:23 PM
yinma added a comment to D53998: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry.

exactly the same

Nov 1 2018, 3:47 PM
yinma added a comment to D53998: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry.

re-upload due to the depended llvm patch need rework.

Nov 1 2018, 2:33 PM
yinma added reviewers for D53998: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry: mgrang, eli.friedman, ssijaric, mstorsjo, rnk, compnerd.
Nov 1 2018, 2:31 PM
yinma created D53998: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry.
Nov 1 2018, 2:30 PM
yinma added a comment to D53996: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.

The preview revision failed in expensive build bot check due to register class check. this is the new upload.

Nov 1 2018, 2:06 PM
yinma added reviewers for D53996: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64: mgrang, ssijaric, eli.friedman, TomTan, mstorsjo, rnk, compnerd.
Nov 1 2018, 2:05 PM
yinma created D53996: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Nov 1 2018, 2:04 PM
yinma updated the diff for D53444: Support ARM_V4BX relocation.
Nov 1 2018, 11:20 AM
yinma updated the diff for D53446: Set MAttrs in LTO mode.
Nov 1 2018, 11:13 AM

Oct 31 2018

yinma updated the diff for D53684: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry.
Oct 31 2018, 3:18 PM
yinma updated the diff for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Oct 31 2018, 3:13 PM
yinma updated the diff for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Oct 31 2018, 2:51 PM
yinma updated the diff for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Oct 31 2018, 1:34 PM

Oct 30 2018

yinma updated the diff for D53843: Make sure to forward arguments from vararg to musttail vararg.
Oct 30 2018, 1:28 PM
yinma updated the diff for D53843: Make sure to forward arguments from vararg to musttail vararg.
Oct 30 2018, 11:42 AM

Oct 29 2018

yinma added reviewers for D53843: Make sure to forward arguments from vararg to musttail vararg: ssijaric, eli.friedman, TomTan, mgrang, mstorsjo, rnk, compnerd.
Oct 29 2018, 3:47 PM
yinma created D53843: Make sure to forward arguments from vararg to musttail vararg.
Oct 29 2018, 3:45 PM
yinma updated the diff for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Oct 29 2018, 2:55 PM

Oct 26 2018

yinma updated the diff for D53446: Set MAttrs in LTO mode.
Oct 26 2018, 5:29 PM
yinma updated the diff for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.

add var_arg test case

Oct 26 2018, 3:01 PM

Oct 25 2018

yinma updated the diff for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.

Please review again

Oct 25 2018, 3:37 PM
yinma added a comment to D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.

Because __security_push_cookie has extra 0x10 subtraction on SP, if you removes buf[100], you will see no security_push_cookie is used. it will be sp on entry.

Oct 25 2018, 10:54 AM

Oct 24 2018

yinma added a reviewer for D53684: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry: efriedma.
Oct 24 2018, 5:52 PM
yinma added reviewers for D53684: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry: mgrang, rnk, TomTan, compnerd, mstorsjo.
Oct 24 2018, 5:52 PM
yinma created D53684: [COFF, ARM64] Change setjmp for AArch64 Windows to use Intrinsic.sponentry.
Oct 24 2018, 5:52 PM
yinma added a comment to D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.

I debugged MS ARM64 setjmp/longjmp functions and ARM64 EH runtime. If you looks like assembly output from CL for ARM64, it is using SP instead of FP.

Oct 24 2018, 3:32 PM
yinma added a reviewer for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64: TomTan.
Oct 24 2018, 2:43 PM
yinma added a reviewer for D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64: mgrang.
Oct 24 2018, 2:41 PM
yinma created D53673: [COFF, ARM64] Implement Intrinsic.sponentry for AArch64.
Oct 24 2018, 2:41 PM

Oct 23 2018

yinma added inline comments to D53446: Set MAttrs in LTO mode.
Oct 23 2018, 12:18 PM
yinma added inline comments to D53446: Set MAttrs in LTO mode.
Oct 23 2018, 11:46 AM

Oct 19 2018

yinma added a reviewer for D53444: Support ARM_V4BX relocation: pcc.
Oct 19 2018, 12:18 PM
yinma updated the diff for D53446: Set MAttrs in LTO mode.

I see you point. I have removed CGOptLevel portion

Oct 19 2018, 12:16 PM
yinma added a reviewer for D53446: Set MAttrs in LTO mode: ruiu.
Oct 19 2018, 12:01 PM
yinma added a reviewer for D53446: Set MAttrs in LTO mode: MaskRay.
Oct 19 2018, 11:59 AM
yinma created D53446: Set MAttrs in LTO mode.
Oct 19 2018, 11:59 AM
yinma added a reviewer for D53444: Support ARM_V4BX relocation: MaskRay.
Oct 19 2018, 11:32 AM
yinma created D53444: Support ARM_V4BX relocation.
Oct 19 2018, 11:30 AM

Oct 12 2018

Herald added a reviewer for D35640: [lld] [COFF] Align import address chunks to the pointer size: javed.absar.
Oct 12 2018, 4:38 PM

Jul 19 2018

yinma added inline comments to D49464: [COFF, ARM64] Decide when to mark struct returns as SRet.
Jul 19 2018, 11:22 AM
yinma added inline comments to D49464: [COFF, ARM64] Decide when to mark struct returns as SRet.
Jul 19 2018, 10:50 AM

Jul 18 2018

yinma added a reviewer for D49464: [COFF, ARM64] Decide when to mark struct returns as SRet: yinma.
Jul 18 2018, 12:11 PM
yinma added inline comments to D49464: [COFF, ARM64] Decide when to mark struct returns as SRet.
Jul 18 2018, 12:10 PM

May 2 2017

yinma added a comment to D32567: [ARM] Miscompilation on arrays promoted to constant pools.

I tried this patch, I think it is a little bit too restrictive. If something like
define i8* @foo() {
entry:

store i8* getelementptr inbounds ([5 x i8], [5 x i8]* @.str, i32 0, i32 0), i8** @b, align 4
ret i8* getelementptr inbounds ([5 x i8], [5 x i8]* @.str, i32 0, i32 0)

}

May 2 2017, 3:40 PM

Apr 27 2017

yinma added a comment to D32567: [ARM] Miscompilation on arrays promoted to constant pools.

Hi, I am facing the same issue. In my case, I disable the rematerialization of all instructions that referring contant pool. I think only t2LEApcrel is rematerializable right now. Because
the safety assumption of ContantIslandPass is that the reference to an constant pool entry should be one use. ARM common subexpression elimination will make single reference to one constant pool, the
rest reference should be copy. I believe the only re-introduction of referencing instruction is the rematerialization in register allocation pass. If we disable rematerialization. the assumption of one use should hold
until constant island pass in general. So in constant island, we can simply check if the CPE->RefCount is not greater than 1, if is greater than 1, no cloning. How do you think?

Apr 27 2017, 11:28 AM

Dec 5 2016

yinma updated the diff for D27219: Add MCNoDeprecatedASM into MC ASMParser TargetOption to have an option to disable deprecate warning in asm.

rename unit test case from qc-deprecated-asm.s to deprecated-asm.s

Dec 5 2016, 3:19 PM

Dec 3 2016

yinma added a reviewer for D27219: Add MCNoDeprecatedASM into MC ASMParser TargetOption to have an option to disable deprecate warning in asm: weimingz.
Dec 3 2016, 9:38 AM

Nov 29 2016

yinma added reviewers for D27219: Add MCNoDeprecatedASM into MC ASMParser TargetOption to have an option to disable deprecate warning in asm: echristo, joey.
Nov 29 2016, 11:07 AM
yinma retitled D27219: Add MCNoDeprecatedASM into MC ASMParser TargetOption to have an option to disable deprecate warning in asm from to Add MCNoDeprecatedASM into MC ASMParser TargetOption to have an option to disable deprecate warning in asm.
Nov 29 2016, 11:03 AM

Oct 18 2016

yinma added a comment to D25634: Fix replacedSelectWithOperand in InstCombiner to handle branch having two same successors..

This kind of IR does actually exist. For our case, it happens after Inlining and Jump Threading, Jump Threading merges if else together and generates a branch with true/false target pointing to the same basic block. It is rare case but it happens. So we need test this situation.

Oct 18 2016, 11:02 AM

Oct 14 2016

yinma updated D25634: Fix replacedSelectWithOperand in InstCombiner to handle branch having two same successors..
Oct 14 2016, 3:42 PM
yinma added reviewers for D25634: Fix replacedSelectWithOperand in InstCombiner to handle branch having two same successors.: apazos, Gerolf.
Oct 14 2016, 3:42 PM
yinma retitled D25634: Fix replacedSelectWithOperand in InstCombiner to handle branch having two same successors. from to Fix replacedSelectWithOperand in InstCombiner to handle branch having two same successors..
Oct 14 2016, 2:48 PM

Mar 29 2016

yinma added a comment to D18524: Dump function as the only defined function in its module. .

Our code base is huge in LTO, it usually take very very long time to dump into a file. With this function, the process would be dramatically faster. I believe this would also benefit other people.

Mar 29 2016, 10:31 AM

Mar 28 2016

yinma added a comment to D18524: Dump function as the only defined function in its module. .

the functionality, it is similar to llvm-extract. however, it provides a runtime solution to extract a function from a debugger from any point in compilation time like LTO, something after inlining. You can extract function right at the assertion point. llvm-extract only can work on a file. We need a runtime solution.

Mar 28 2016, 5:18 PM
yinma added reviewers for D18524: Dump function as the only defined function in its module. : bkramer, bogner.
Mar 28 2016, 1:54 PM
yinma retitled D18524: Dump function as the only defined function in its module. from to Dump function as the only defined function in its module. .
Mar 28 2016, 1:49 PM

Mar 24 2016

yinma added inline comments to D18336: Prevent construction of cycle in DAG store merge.
Mar 24 2016, 1:47 PM
yinma added a comment to D18336: Prevent construction of cycle in DAG store merge.

Hi Nirav, I think the problem need to be solved should be the same with and without AA. It is testing the node predecessors. For your new upload, you use different check with UseAA. For my algorithm, it only tests the latest node in the store array because it is the one that links with the successors. Do you have another test case that shows failure with UseAA? What could be the different with UseAA?

Mar 24 2016, 1:42 PM

Mar 23 2016

yinma added a comment to D18336: Prevent construction of cycle in DAG store merge.

Changing finding better chain algorithm may hide the problem fixed by this patch, but for our test case, changing the FindBetterChain algorithm doesn't help.

Mar 23 2016, 10:37 AM

Mar 22 2016

yinma added reviewers for D18336: Prevent construction of cycle in DAG store merge: spatel, ahatanak.
Mar 22 2016, 10:30 AM

Mar 21 2016

yinma updated subscribers of D18330: [ELF/AArch64] Implemented set of R_AARCH64_TLSDESC_* relocations..
Mar 21 2016, 3:08 PM · lld
yinma updated subscribers of D18331: [ELF/AArch64] Simplify AArch64 isRelRelative.
Mar 21 2016, 3:08 PM · lld
yinma updated subscribers of D18332: [ELF/AArch64] Add GD to IE TLS relax optimization.
Mar 21 2016, 3:07 PM · lld
yinma updated subscribers of D18336: Prevent construction of cycle in DAG store merge.
Mar 21 2016, 2:41 PM
yinma retitled D18336: Prevent construction of cycle in DAG store merge from to Prevent cycle in DAG store merge to vector.
Mar 21 2016, 2:40 PM

Jan 8 2016

yinma added a comment to D8056: Fix for bug 21725: wrong results with union and strict-aliasing.

Hello, I am wondering if the current llvm/clang's tbaa supports union or not?

Jan 8 2016, 2:56 PM

May 18 2015

yinma added a comment to D8475: Inliner Enhancement - Delayed Inlining.

Hi Chandler, I agree that the attributes to save cost is not the right approach. I can work on this to fix it. However, we tried this algorithm for SPEC on ARM/AArch64. we saw improved results. However, when we tried this with LTO, we experienced mess regression. Largely due to too aggressive inlining because this algorithm will make some functions which cannot be inlined before inlined now. Some functions should be huge.

May 18 2015, 3:12 PM

Mar 20 2015

yinma added a comment to D8408: Inliner Enhancement.

Hi Jiangning

Mar 20 2015, 2:02 PM

Mar 19 2015

yinma added a comment to D8408: Inliner Enhancement.

Hi Jiangning,

Mar 19 2015, 4:31 PM
yinma retitled D8475: Inliner Enhancement - Delayed Inlining from to Inliner Enhancement - Delayed Inlining.
Mar 19 2015, 4:29 PM
yinma added a comment to D8408: Inliner Enhancement.

Hi Jiangning,

Mar 19 2015, 4:04 PM

Jan 19 2015

yinma added a comment to D6996: Enable inlining of recursive functions..

I looked your code, you removed the check in call analyzer and MaxRecursionInl is about inlining history. I would like to know if you set MaxRecursionInl is 0, the inliner will behave exactly like the before without your code?

Jan 19 2015, 10:19 AM