Page MenuHomePhabricator

David.Xu (Xu Jiangwei)
User

Projects

User does not belong to any projects.

User Details

User Since
Aug 3 2014, 10:19 PM (267 w, 2 d)

Recent Activity

Sep 25 2014

David.Xu added a comment to D5455: [AArch64]Redundant store instructions should be removed as dead code.

Line 9800 can't be moved up to maximize horizontal space and reduce vertical space, because clang-formate does not allow this. When it is moved up, it will be moved down after executing "git clang-formate" command.

Sep 25 2014, 7:00 PM

Sep 23 2014

David.Xu retitled D5455: [AArch64]Redundant store instructions should be removed as dead code from to [AArch64]Redundant store instructions should be removed as dead code.
Sep 23 2014, 12:01 AM

Sep 9 2014

David.Xu updated the diff for D5139: [AArch64] Build correct vector filled with undef nodes.

Actually, getConstant has been provided to build a BuildVector Node.

Sep 9 2014, 6:27 PM

Sep 2 2014

David.Xu added a comment to D5139: [AArch64] Build correct vector filled with undef nodes.

Yes, it is more useful as a generic getConstantSplat(all of the elements of the vector are the same) function in SelectionDAG. But it may not be used frequently as getConstant function. If we want to present a bunch of APIs to generate const splat, more work is needed. For example, we should consider whether the data type is integer or float point. In SelectionDAG, there are 3 getConstant and 3 getConstantFP, thus 6 getConstantSplat functions should be added.

Sep 2 2014, 12:45 AM
David.Xu updated D5144: [AArch64]Merge Extend and Shift into a UBFX .
Sep 2 2014, 12:11 AM
David.Xu updated D5144: [AArch64]Merge Extend and Shift into a UBFX .
Sep 2 2014, 12:11 AM
David.Xu retitled D5144: [AArch64]Merge Extend and Shift into a UBFX from to [AArch64]Merge Extend and Shift into a UBFX .
Sep 2 2014, 12:09 AM

Aug 31 2014

David.Xu updated subscribers of D5139: [AArch64] Build correct vector filled with undef nodes.
Aug 31 2014, 10:36 PM
David.Xu retitled D5139: [AArch64] Build correct vector filled with undef nodes from to [AArch64] Build correct vector filled with undef nodes.
Aug 31 2014, 10:35 PM

Aug 26 2014

David.Xu added a comment to D4966: [AArch64] Generate CMN when comparing a short int with minus.

ping ....

Aug 26 2014, 10:20 PM

Aug 21 2014

David.Xu updated the diff for D4966: [AArch64] Generate CMN when comparing a short int with minus.
Aug 21 2014, 12:27 AM
David.Xu updated the diff for D4966: [AArch64] Generate CMN when comparing a short int with minus.

Comments about fundamental property are added in the patch.

Aug 21 2014, 12:21 AM

Aug 18 2014

David.Xu retitled D4966: [AArch64] Generate CMN when comparing a short int with minus from to [AArch64] Generate CMN when comparing a short int with minus.
Aug 18 2014, 6:44 PM

Aug 7 2014

David.Xu updated the diff for D4771: [AArch64] Type conversion bug when anlyze compare .
Aug 7 2014, 3:34 AM
David.Xu updated the diff for D4771: [AArch64] Type conversion bug when anlyze compare .

I have converted all of the CmpValue to 0 or 1, and also added assertion in optimizeCompare. For that our daily performance test have broken down for a long time, please review it ASAP. Thank you!

Aug 7 2014, 3:26 AM

Aug 6 2014

David.Xu updated the diff for D4771: [AArch64] Type conversion bug when anlyze compare .
Aug 6 2014, 10:02 PM
David.Xu added a comment to D4771: [AArch64] Type conversion bug when anlyze compare .

Yes, This patch changes the meaning of CmpValue. It may not be the best way to solve this problem, but it has no effect on the other compare instructions and other backends comparing to the way of changing the interface of the function. And appropriate comments are added in optimizeCompare.

Aug 6 2014, 10:00 PM
David.Xu updated the diff for D4771: [AArch64] Type conversion bug when anlyze compare .

Test case is added in the new patch. Without the patch, an ANDS instruction will be deleted when doing peephole optimization.

Aug 6 2014, 8:25 PM

Aug 4 2014

David.Xu retitled D4771: [AArch64] Type conversion bug when anlyze compare from Type conversion bug when anlyze compare to [AArch64] Type conversion bug when anlyze compare .
Aug 4 2014, 12:43 AM
David.Xu added reviewers for D4771: [AArch64] Type conversion bug when anlyze compare : kevin.qin, HaoLiu.
Aug 4 2014, 12:39 AM
David.Xu added a reviewer for D4771: [AArch64] Type conversion bug when anlyze compare : Jiangning.
Aug 4 2014, 12:37 AM
David.Xu retitled D4771: [AArch64] Type conversion bug when anlyze compare from to Type conversion bug when anlyze compare .
Aug 4 2014, 12:35 AM