Page MenuHomePhabricator

SamWot (Sam Kolton)
User

Projects

User does not belong to any projects.

User Details

User Since
Jan 29 2016, 5:37 AM (374 w, 1 d)

Recent Activity

Nov 30 2017

SamWot updated the diff for D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.
Stronger verification for UNUSED_PRESERVE
Nov 30 2017, 5:13 AM

Nov 29 2017

SamWot updated the diff for D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.

Added verification for tied register for UNUSED_PRESERVE

Nov 29 2017, 7:12 AM

Nov 24 2017

SamWot accepted D39488: [AMDGPU][MC][GFX9] Added v_interp_p2_f16 and v_interp_p2_legacy_f16.
Nov 24 2017, 6:04 AM

Nov 21 2017

SamWot updated the diff for D40172: [AMDGPU] SDWA: remove omod src operand for VOP2b instructions.

Fix for VOP2b instructions was submitted in D40088.
Added SDWA dissassembler tests for VI.

Nov 21 2017, 5:42 AM

Nov 17 2017

SamWot created D40172: [AMDGPU] SDWA: remove omod src operand for VOP2b instructions.
Nov 17 2017, 3:42 AM

Nov 10 2017

SamWot added a comment to D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.

Ping.
Matt, what do you think about latest changes in reivew?

Nov 10 2017, 1:53 AM

Nov 2 2017

SamWot added inline comments to D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.
Nov 2 2017, 3:40 AM
SamWot updated the diff for D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.

Fixed latests issues from arsenm

Nov 2 2017, 3:32 AM

Oct 13 2017

SamWot added a comment to D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.

Ping

Oct 13 2017, 2:27 AM

Oct 5 2017

SamWot updated the diff for D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.

Removed SIMergeSDWAPreserve pass.
Use tied registers to achieve same results

Oct 5 2017, 8:53 AM

Sep 14 2017

SamWot updated the diff for D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.

Resolved some issues

Sep 14 2017, 7:35 AM

Sep 13 2017

SamWot created D37817: [AMDGPU] SDWA: add support for PRESERVE into SDWA peephole. Add new merge SDWA preserve pass.
Sep 13 2017, 10:42 AM

Sep 8 2017

SamWot accepted D37594: AMDGPU: Start using !con operator.
Sep 8 2017, 9:53 AM

Aug 16 2017

SamWot accepted D36694: [AMDGPU][MC][GFX9] Added op_sel support for v_mad_*16, v_fma_f16, v_div_fixup_f16.
Aug 16 2017, 4:50 AM
SamWot added a comment to D36694: [AMDGPU][MC][GFX9] Added op_sel support for v_mad_*16, v_fma_f16, v_div_fixup_f16.

LGTM with small issue.

Aug 16 2017, 4:50 AM

Aug 14 2017

SamWot accepted D36674: [AMDGPU][MC][GFX9] Added integer clamping support for VOP3 opcodes.

LGTM

Aug 14 2017, 8:27 AM

Aug 7 2017

SamWot accepted D36322: [AMDGPU][MC][GFX9] Added 16-bit renamed and "_legacy" VALU opcodes.

LGTM

Aug 7 2017, 8:52 AM

Aug 5 2017

SamWot accepted D34716: [AMDGPU] Add pseudo "old" and "wqm_mode" source to all DPP instructions.
Aug 5 2017, 12:57 AM
SamWot added inline comments to D36322: [AMDGPU][MC][GFX9] Added 16-bit renamed and "_legacy" VALU opcodes.
Aug 5 2017, 12:00 AM

Aug 4 2017

SamWot accepted D35902: [AMDGPU][MC][GFX8] added/corrected VOP3 versions of v_interp_* instructions.
Aug 4 2017, 11:34 PM

Jul 31 2017

SamWot accepted D36091: [AMDGPU][MC] Enabled expressions as operands.
Jul 31 2017, 11:28 PM

Jul 20 2017

SamWot accepted D35662: AMDGPU: Add encoding for carryless add/sub instructions.

LGTM

Jul 20 2017, 3:08 AM

Jul 19 2017

SamWot accepted D35649: AMDGPU: Correct encoding for global instructions.

LGTM

Jul 19 2017, 7:51 PM

Jul 14 2017

SamWot added a comment to D35350: [AMDGPU] CodeGen: check dst operand type to determine if omod is supported for VOP3 instructions.

There should be an assembler test for omod being rejected with these. Doesn't the same also apply for clamp?

Jul 14 2017, 6:58 AM
SamWot updated the diff for D35350: [AMDGPU] CodeGen: check dst operand type to determine if omod is supported for VOP3 instructions.

Added assembler tests

Jul 14 2017, 6:56 AM

Jul 13 2017

SamWot updated the summary of D35350: [AMDGPU] CodeGen: check dst operand type to determine if omod is supported for VOP3 instructions.
Jul 13 2017, 4:48 AM
SamWot created D35350: [AMDGPU] CodeGen: check dst operand type to determine if omod is supported for VOP3 instructions.
Jul 13 2017, 4:47 AM

Jul 6 2017

SamWot updated the diff for D35047: [AMDGPU] Assembler: refactor convert methods (VOP3 and MIMG).

Re-used Opc

Jul 6 2017, 9:15 PM
SamWot created D35047: [AMDGPU] Assembler: refactor convert methods (VOP3 and MIMG).
Jul 6 2017, 3:17 AM

Jun 27 2017

SamWot updated the diff for D34626: [AMDGPU] SDWA: several fixes for V_CVT and VOPC instructions.

Fix comments from Matt

Jun 27 2017, 4:22 AM

Jun 26 2017

SamWot updated the summary of D34626: [AMDGPU] SDWA: several fixes for V_CVT and VOPC instructions.
Jun 26 2017, 8:32 AM
SamWot created D34626: [AMDGPU] SDWA: several fixes for V_CVT and VOPC instructions.
Jun 26 2017, 8:28 AM

Jun 22 2017

SamWot added a comment to D34549: fix regression on VI with SDWA gfx9 support.

This change should in fact fix problem, but it is only workaround.
Ultimately problem is that instruction V_CVT_U32_F32 allow omod operand (see SIInstrInfo.td:1435). In fact this operand shouldn't be allowed here.
This should work as quickfix for assert if you need this. But I think it is better to provide full fix.

Jun 22 2017, 11:07 PM
SamWot added a comment to D34549: fix regression on VI with SDWA gfx9 support.

Hi,
I have no access to bug report: https://bugs.freedesktop.org/show_bug.cgi?id=101561
Can you send me failing .ll?

Jun 22 2017, 10:07 PM

Jun 21 2017

SamWot updated the diff for D34241: [AMDGPU] SDWA: add support for GFX9 in peephole pass.

Removed changes from merge pseudo instrutions

Jun 21 2017, 10:38 PM

Jun 20 2017

SamWot created D34403: [AMDGPU] SDWA: remove support for VOP2 instructions that have only 64-bit encoding.
Jun 20 2017, 8:40 AM
SamWot accepted D34360: [AMDGPU][MC][GFX9] Corrected VOP3P relevant code to fix disassembler failures.

I agree, it's better to remove assert ,rather them commenting them out.
Otherwise, LGTM

Jun 20 2017, 2:58 AM

Jun 15 2017

SamWot updated subscribers of D34241: [AMDGPU] SDWA: add support for GFX9 in peephole pass.
Jun 15 2017, 9:30 AM
SamWot created D34241: [AMDGPU] SDWA: add support for GFX9 in peephole pass.
Jun 15 2017, 9:29 AM
SamWot added a comment to D34026: [AMDGPU] SDWA: merge VI and GFX9 pseudo instructions.

Ping

Jun 15 2017, 2:39 AM

Jun 13 2017

SamWot updated the diff for D34026: [AMDGPU] SDWA: merge VI and GFX9 pseudo instructions.

Removed generation check from constant bus verification.

Jun 13 2017, 3:01 AM

Jun 9 2017

SamWot added inline comments to D34026: [AMDGPU] SDWA: merge VI and GFX9 pseudo instructions.
Jun 9 2017, 1:17 AM
SamWot updated the diff for D34026: [AMDGPU] SDWA: merge VI and GFX9 pseudo instructions.

Fixed issues in SIInstrInfo.cpp

Jun 9 2017, 1:17 AM

Jun 8 2017

SamWot created D34026: [AMDGPU] SDWA: merge VI and GFX9 pseudo instructions.
Jun 8 2017, 4:18 AM

Jun 6 2017

SamWot accepted D33927: [AMDGPU] Return correct value from SDWA pass.
Jun 6 2017, 3:40 AM

Jun 3 2017

SamWot accepted D33860: [AMDGPU] Untangle SDWA pass from SIShrinkInstructions.
Jun 3 2017, 10:24 AM
SamWot added a comment to D33860: [AMDGPU] Untangle SDWA pass from SIShrinkInstructions.

This is good change. I wanted to propose it myself:)

Jun 3 2017, 2:17 AM

May 30 2017

SamWot accepted D33583: [AMDGPU] Allow SDWA in instructions with immediates and SGPRs.
May 30 2017, 1:43 AM

May 25 2017

SamWot added inline comments to D33583: [AMDGPU] Allow SDWA in instructions with immediates and SGPRs.
May 25 2017, 10:16 PM
SamWot closed D33556: [AMDGPU] SDWA: fix encoding of flat_scratch on GFX9.
May 25 2017, 9:50 PM
SamWot accepted D33553: [AMDGPU][MC][GFX9] Corrected encoding of flat_scratch* in SDWA opcodes.

LGTM
I think it's better if you submit this change. I will close my review (https://reviews.llvm.org/D33556).

May 25 2017, 9:50 PM
SamWot created D33556: [AMDGPU] SDWA: fix encoding of flat_scratch on GFX9.
May 25 2017, 10:49 AM
SamWot updated the diff for D33545: [AMDGPU] SDWA: add disassembler support for GFX9.

Fixed linux macros

May 25 2017, 7:02 AM
SamWot created D33545: [AMDGPU] SDWA: add disassembler support for GFX9.
May 25 2017, 5:11 AM

May 23 2017

SamWot added a comment to D33132: [AMDGPU] SDWA: Add assembler support for GFX9.

Removed omod operand from pseudo instructions with integer dst

It seems that have been done for Gfx9 only. Any reason for this? Perhaps it is worth to add a comment.

May 23 2017, 1:40 AM

May 19 2017

SamWot updated the diff for D33132: [AMDGPU] SDWA: Add assembler support for GFX9.

Removed omod operand from pseudo instructions with integer dst

May 19 2017, 5:39 AM

May 17 2017

SamWot added inline comments to D33132: [AMDGPU] SDWA: Add assembler support for GFX9.
May 17 2017, 9:11 AM
SamWot updated the diff for D33132: [AMDGPU] SDWA: Add assembler support for GFX9.

Rebased on top of master.
Renamed VI9 tests to GFX89.
Added commentary for SDWA on GFX9.

May 17 2017, 8:34 AM

May 16 2017

SamWot updated the diff for D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.

Simplified process. Got rid of PulledOut vector.

May 16 2017, 5:54 AM

May 12 2017

SamWot created D33132: [AMDGPU] SDWA: Add assembler support for GFX9.
May 12 2017, 8:45 AM

May 11 2017

SamWot updated the diff for D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.

Small fix

May 11 2017, 6:16 AM

May 10 2017

SamWot added inline comments to D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.
May 10 2017, 5:59 AM
SamWot updated the diff for D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.

Got rid of shared_ptr

May 10 2017, 5:57 AM

May 4 2017

SamWot updated the diff for D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.

Resolved some issues

May 4 2017, 6:57 AM

May 3 2017

SamWot updated subscribers of D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.
May 3 2017, 7:16 AM
SamWot created D32804: [AMDGPU] SDWA operands should not intersect with potential MIs.
May 3 2017, 7:14 AM

Apr 27 2017

SamWot created D32588: [AMDGPU] DPP: add support for GFX9.
Apr 27 2017, 4:32 AM

Apr 25 2017

SamWot updated the summary of D32493: [TableGen] Add EncoderMethod to RegisterOperand.
Apr 25 2017, 9:29 AM
SamWot created D32493: [TableGen] Add EncoderMethod to RegisterOperand.
Apr 25 2017, 9:24 AM

Apr 11 2017

SamWot created D31935: [AMDGPU] SDWA: make pass global.
Apr 11 2017, 8:00 AM

Apr 7 2017

SamWot added a comment to D31671: [AMDGPU] SDWA peephole: enable by default.

Fixed and resubmitted in r299654

Apr 7 2017, 2:58 AM

Apr 5 2017

SamWot updated subscribers of D31705: [AMDGPU] Move SiShrinkInstruction and SDWAPeephole to SSAOptimization passes.
Apr 5 2017, 6:53 AM
SamWot created D31705: [AMDGPU] Move SiShrinkInstruction and SDWAPeephole to SSAOptimization passes.
Apr 5 2017, 6:03 AM

Apr 4 2017

SamWot updated subscribers of D31671: [AMDGPU] SDWA peephole: enable by default.
Apr 4 2017, 9:25 AM
SamWot created D31671: [AMDGPU] SDWA peephole: enable by default.
Apr 4 2017, 9:24 AM

Mar 31 2017

SamWot closed D31455: [AMDGPU] SDWA Peephole: improve search for immediates in SDWA patterns.

Submitted in r299202

Mar 31 2017, 5:44 AM

Mar 29 2017

SamWot added inline comments to D31455: [AMDGPU] SDWA Peephole: improve search for immediates in SDWA patterns.
Mar 29 2017, 5:36 AM
SamWot updated the diff for D31455: [AMDGPU] SDWA Peephole: improve search for immediates in SDWA patterns.

Changed isImm() to foldToImm() that return Optional<int64_t>

Mar 29 2017, 5:34 AM
SamWot updated the summary of D31455: [AMDGPU] SDWA Peephole: improve search for immediates in SDWA patterns.
Mar 29 2017, 2:11 AM
SamWot created D31455: [AMDGPU] SDWA Peephole: improve search for immediates in SDWA patterns.
Mar 29 2017, 2:10 AM

Mar 20 2017

SamWot added inline comments to D30038: [ADMGPU] SDWA peephole optimization pass..
Mar 20 2017, 3:56 AM
SamWot updated the diff for D30038: [ADMGPU] SDWA peephole optimization pass..

Changes for latest comments

Mar 20 2017, 3:54 AM

Mar 16 2017

SamWot updated the diff for D30038: [ADMGPU] SDWA peephole optimization pass..

Add check for single use of src sdwa operand.

Mar 16 2017, 11:29 AM
SamWot added inline comments to D30038: [ADMGPU] SDWA peephole optimization pass..
Mar 16 2017, 6:27 AM
SamWot updated the diff for D30038: [ADMGPU] SDWA peephole optimization pass..

Cleanup for remaining comments.

Mar 16 2017, 6:25 AM

Mar 14 2017

SamWot added a comment to D30038: [ADMGPU] SDWA peephole optimization pass..

I would like to see this patch fixed, submitted, and optimization enabled first, before new patterns are introduced.

I alredy either fixed or answered on comments to previous version. Currently I'm waiting for new comments or acceptance thats why I started working on new functionality.

Mar 14 2017, 2:23 AM

Mar 13 2017

SamWot updated the diff for D30038: [ADMGPU] SDWA peephole optimization pass..

Introduced new sdwa patterns

Mar 13 2017, 3:48 AM

Mar 9 2017

SamWot added inline comments to D30038: [ADMGPU] SDWA peephole optimization pass..
Mar 9 2017, 3:03 AM
SamWot updated the diff for D30038: [ADMGPU] SDWA peephole optimization pass..

Fixed isSameBB

Mar 9 2017, 3:00 AM

Mar 6 2017

SamWot updated the diff for D30038: [ADMGPU] SDWA peephole optimization pass..

Added support for subregs.
Fixed most of previous comments.

Mar 6 2017, 3:50 AM

Feb 17 2017

SamWot added inline comments to D30038: [ADMGPU] SDWA peephole optimization pass..
Feb 17 2017, 3:17 AM

Feb 16 2017

SamWot updated the summary of D30038: [ADMGPU] SDWA peephole optimization pass..
Feb 16 2017, 4:15 AM
SamWot updated the summary of D30038: [ADMGPU] SDWA peephole optimization pass..
Feb 16 2017, 4:14 AM
SamWot updated the summary of D30038: [ADMGPU] SDWA peephole optimization pass..
Feb 16 2017, 4:13 AM
SamWot created D30038: [ADMGPU] SDWA peephole optimization pass..
Feb 16 2017, 4:10 AM

Jan 19 2017

SamWot created D28900: [AMDGPU] Small changes for SDWA and DPP.
Jan 19 2017, 6:52 AM

Jan 17 2017

SamWot created D28802: [AMDGPU] Assembler: fix v_mac_f16 immediates.
Jan 17 2017, 5:14 AM

Jan 10 2017

SamWot updated the diff for D28157: [AMDGPU] Assembler: SDWA/DPP should not accept scalar registers and immediate operands.

Fixed issues

Jan 10 2017, 6:17 AM

Dec 29 2016

SamWot retitled D28157: [AMDGPU] Assembler: SDWA/DPP should not accept scalar registers and immediate operands from to [AMDGPU] Assembler: SDWA/DPP should not accept scalar registers and immediate operands.
Dec 29 2016, 4:16 AM

Dec 26 2016

SamWot added a comment to D28051: [AMDGPU] Assembler: support SDWA and DPP for VOP2b instructions.

Ping

Dec 26 2016, 2:46 AM