Page MenuHomePhabricator

pdhaliwal (Pushpinder Singh)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 23 2020, 6:41 PM (44 w, 2 d)

Recent Activity

Thu, Feb 25

pdhaliwal updated the diff for D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.

Add extra llc step to produce assembly in the linker.

Thu, Feb 25, 7:58 AM · Restricted Project
pdhaliwal added a comment to D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.

So, neither emit-llvm-bc or emit-llvm work well with save-temps. Therefore, I feel the current approach is still valid. This does not impact nvptx or any other target in any way. And I don't see how.

Thu, Feb 25, 7:04 AM · Restricted Project

Wed, Feb 24

pdhaliwal committed rG99951aa68da3: OpenMP: Fix object clobbering issue when using save-temps (authored by pdhaliwal).
OpenMP: Fix object clobbering issue when using save-temps
Wed, Feb 24, 9:51 PM
pdhaliwal closed D97273: OpenMP: Fix object clobbering issue when using save-temps.
Wed, Feb 24, 9:51 PM · Restricted Project

Tue, Feb 23

pdhaliwal added a reviewer for D97273: OpenMP: Fix object clobbering issue when using save-temps: sdmitriev.
Tue, Feb 23, 8:07 AM · Restricted Project
pdhaliwal added a comment to D97273: OpenMP: Fix object clobbering issue when using save-temps.

Here's a bit of background,
OffloadingPrefix was not getting properly set in the dependent actions of OffloadWrapperJobAction (which are backend [11] and assemble [12]). Since backend [11] and assemble [12] host-wrapper actions have same logic to the other host actions (3 & 4), those will overwrite the previous generated files from host-only actions.

Tue, Feb 23, 8:05 AM · Restricted Project
pdhaliwal updated the summary of D97273: OpenMP: Fix object clobbering issue when using save-temps.
Tue, Feb 23, 6:11 AM · Restricted Project
pdhaliwal updated the summary of D97273: OpenMP: Fix object clobbering issue when using save-temps.
Tue, Feb 23, 5:48 AM · Restricted Project
pdhaliwal requested review of D97273: OpenMP: Fix object clobbering issue when using save-temps.
Tue, Feb 23, 5:06 AM · Restricted Project

Thu, Feb 18

pdhaliwal accepted D96998: [OPENMP][AMDGCN] Improvements to print_kernel_trace (bit mask).

Looks good to me. Thanks!

Thu, Feb 18, 5:38 PM · Unknown Object (Project)

Wed, Feb 17

pdhaliwal added inline comments to D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.
Wed, Feb 17, 2:37 AM · Restricted Project
pdhaliwal updated the diff for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Fixed the assert.

Wed, Feb 17, 2:36 AM · Restricted Project

Tue, Feb 16

pdhaliwal added a comment to D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.

It is because of how addClangTargetOptions is invoked. In case of save-temps, it is being invoked for all the actions resulting in target cc1 call. That's why all these invocations have -emit-llvm-bc. I guess we need Action as an argument to addClangTargetOptions.

Tue, Feb 16, 8:28 AM · Restricted Project
pdhaliwal added a comment to D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.

emit-llvm-bc does not correctly solve the problem. It works because [input, compile, assemble, backend] actions collapse to a single action by driver. This single command handles emit-llvm-bc properly. But when save-temps is specified, this collapsing does not happen which messes up command line flags of the jobs and hence the output, for e.g., preprocessor command also has -emit-llvm-bc.

Tue, Feb 16, 7:59 AM · Restricted Project
pdhaliwal added a comment to D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.

This does fixes the save-temps but only when -o is not specified. If -o is specified the name of host object file and host-wrapper object file (second last phase) is same, which fails the linker. This does not seem to be related to this patch.

Tue, Feb 16, 6:15 AM · Restricted Project
pdhaliwal added inline comments to D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.
Tue, Feb 16, 4:26 AM · Restricted Project
pdhaliwal requested review of D96769: [OpenMP][AMDGPU] Skip backend and assemble phases for amdgcn.
Tue, Feb 16, 4:24 AM · Restricted Project

Mon, Feb 15

pdhaliwal updated the diff for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Addressed review comments.

Mon, Feb 15, 11:11 PM · Restricted Project
pdhaliwal added a comment to D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Can you use -check-prefixes=GCN,GFX8 and GCN,GFX9 so that update_mir_test_checks will common up the identical ones?

It does not work. Script warns as WARNING: Ignoring common prefixes: {'GCN'}: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umulo.mir

Mon, Feb 15, 1:15 AM · Restricted Project
pdhaliwal updated the diff for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Addressed review comments.

Mon, Feb 15, 1:06 AM · Restricted Project

Thu, Feb 11

pdhaliwal committed rG79401b43ce4e: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode (authored by Pushpinder Singh <pushpinderdhaliwal00@gmail.com>).
[OpenMP][AMDGPU] Add support for linking libomptarget bitcode
Thu, Feb 11, 9:43 PM
pdhaliwal closed D96248: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode.
Thu, Feb 11, 9:42 PM · Restricted Project

Wed, Feb 10

pdhaliwal updated the diff for D96248: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode.

I haave removed libomptarget-device-bc-path and have added amdgcn one. For diagnostic,
instead of having one per architecture, I have used the same and added second
parameter to specify arch.

Wed, Feb 10, 3:22 AM · Restricted Project

Tue, Feb 9

pdhaliwal accepted D96330: [libomptarget][amdgcn] Tolerate deadstripped device_state variable.

LGTM. Thanks!

Tue, Feb 9, 4:50 AM · Unknown Object (Project)
pdhaliwal accepted D96329: [libomptarget][amdgcn] Tolerate deadstripped env variable.

LGTM, thanks for fixing it.

Tue, Feb 9, 3:56 AM · Unknown Object (Project)

Mon, Feb 8

pdhaliwal updated the diff for D96248: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode.
  • Added check for nogpulib
  • Fixed diagnostic message
Mon, Feb 8, 11:36 PM · Restricted Project
pdhaliwal updated the diff for D96248: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode.

Addressed review comments.

Mon, Feb 8, 5:47 AM · Restricted Project
pdhaliwal updated the diff for D96248: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode.

Accidently missed some changes,

  • Fix openmp-offload.c test failure
  • Fix amdgpu-openmp-toolchain.c test failure
Mon, Feb 8, 1:51 AM · Restricted Project
pdhaliwal added a comment to D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Ping!

Mon, Feb 8, 1:23 AM · Restricted Project
pdhaliwal requested review of D96248: [OpenMP][AMDGPU] Add support for linking libomptarget bitcode.
Mon, Feb 8, 1:15 AM · Restricted Project

Tue, Feb 2

pdhaliwal committed rGfcf03e728007: [OpenMP] Add OpenMP offloading toolchain for AMDGPU (authored by Pushpinder Singh <pushpinderdhaliwal00@gmail.com>).
[OpenMP] Add OpenMP offloading toolchain for AMDGPU
Tue, Feb 2, 9:43 PM
pdhaliwal closed D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.
Tue, Feb 2, 9:43 PM · Restricted Project
pdhaliwal added inline comments to D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.
Tue, Feb 2, 3:59 AM · Restricted Project
pdhaliwal updated the diff for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.
  • Use 0 for default -O option
  • Rename addOptLevelArgs to addLLCOptArg
Tue, Feb 2, 3:58 AM · Restricted Project
pdhaliwal added a comment to D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.

After addressing the review comments, I have internally verified changes on few simple test programs. They seem to be working fine.

Tue, Feb 2, 2:47 AM · Restricted Project
pdhaliwal updated the diff for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.

Addressed review comments.

  • Combined the toolchain creation logic for nvptx and amdgcn
  • Replaced -Xopenmp-target with -emit-llvm-bc inside AMDGPUOpenMP.cpp
  • Removed opt from pipeline
Tue, Feb 2, 2:39 AM · Restricted Project

Mon, Feb 1

pdhaliwal updated the diff for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.
  • Scalarize the vectors first
  • Using widened operation for smaller types
Mon, Feb 1, 3:11 AM · Restricted Project
pdhaliwal added a comment to D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.

Ping!

Mon, Feb 1, 2:08 AM · Restricted Project

Jan 28 2021

pdhaliwal added a reviewer for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO: foad.
Jan 28 2021, 2:27 AM · Restricted Project
pdhaliwal added a comment to D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Hi, apologies for late reply as I got sidetracked to some other work.

Jan 28 2021, 2:27 AM · Restricted Project

Jan 27 2021

pdhaliwal added a reviewer for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU: saiislam.
Jan 27 2021, 6:42 AM · Restricted Project

Jan 20 2021

pdhaliwal added a reviewer for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU: yaxunl.
Jan 20 2021, 11:54 PM · Restricted Project
pdhaliwal updated the diff for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.
  • Moved common methods of HIP and OpenMP to base AMDGPUToolChain
  • Removed unnecessary asserts
Jan 20 2021, 11:51 PM · Restricted Project
pdhaliwal updated the diff for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.

Fixed failing debian tests

Jan 20 2021, 4:09 AM · Restricted Project
pdhaliwal retitled D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU from [OpenMP] Add OpenMP offloading toolchain skeleton for AMDGPU to [OpenMP] Add OpenMP offloading toolchain for AMDGPU.
Jan 20 2021, 2:16 AM · Restricted Project
pdhaliwal updated the diff for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.

Won't this just prevent us from building clang due to the missing cmake changes?

It compiles and builds fine, however, I wasn't actually aware such sanity checking being present. It turns out
the unknown files inside llvm/ will lead cmake to report error but such reporting will not happen inside clang. Maybe such checks
were not enabled inside clang. Anyways thanks for pointing out. I will keep that in mind in future.

Jan 20 2021, 1:49 AM · Restricted Project

Jan 19 2021

pdhaliwal updated the diff for D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.

Fix clang-tidy error

Jan 19 2021, 6:56 AM · Restricted Project
pdhaliwal requested review of D94961: [OpenMP] Add OpenMP offloading toolchain for AMDGPU.
Jan 19 2021, 3:45 AM · Restricted Project

Jan 14 2021

pdhaliwal added inline comments to D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.
Jan 14 2021, 10:26 PM · Restricted Project
pdhaliwal updated the diff for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Moved ops close to ADDO

Jan 14 2021, 10:22 PM · Restricted Project

Jan 10 2021

pdhaliwal added inline comments to D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.
Jan 10 2021, 11:45 PM · Restricted Project
pdhaliwal updated the diff for D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.

Removed global-isel-abort=0

Jan 10 2021, 11:44 PM · Restricted Project

Jan 5 2021

pdhaliwal committed rG4909cb1a0fe9: [OpenMP][AMDGPU] Use AMDGPU_KERNEL calling convention for entry function (authored by pdhaliwal).
[OpenMP][AMDGPU] Use AMDGPU_KERNEL calling convention for entry function
Jan 5 2021, 11:04 PM
pdhaliwal closed D94060: [OpenMP][AMDGPU] Use AMDGPU_KERNEL calling convention for entry function.
Jan 5 2021, 11:03 PM · Restricted Project

Jan 4 2021

pdhaliwal requested review of D94060: [OpenMP][AMDGPU] Use AMDGPU_KERNEL calling convention for entry function.
Jan 4 2021, 8:54 PM · Restricted Project

Jan 1 2021

pdhaliwal requested review of D93963: [GlobalISel][AMDGPU] Lower G_UMULO/G_SMULO.
Jan 1 2021, 2:58 AM · Restricted Project

Dec 20 2020

pdhaliwal committed rGe2303a448e2f: [FastRA] Fix handling of bundled MIs (authored by pdhaliwal).
[FastRA] Fix handling of bundled MIs
Dec 20 2020, 11:11 PM
pdhaliwal closed D90369: [FastRA] Fix handling of bundled MIs.
Dec 20 2020, 11:11 PM · Restricted Project
pdhaliwal updated the diff for D90369: [FastRA] Fix handling of bundled MIs.

Update AMDGPU barrier intrinsic

Dec 20 2020, 11:08 PM · Restricted Project

Dec 7 2020

pdhaliwal added a comment to D90369: [FastRA] Fix handling of bundled MIs.

Ping!

Dec 7 2020, 2:36 AM · Restricted Project

Dec 3 2020

pdhaliwal accepted D92559: [libomptarget][amdgpu] Address compiler warnings, drive by fixes.

Looks good, thanks.

Dec 3 2020, 2:54 AM · Unknown Object (Project)

Dec 2 2020

pdhaliwal committed rGafc09c6fe44e: [libomptarget][AMDGPU] Remove MaxParallelLevel (authored by pdhaliwal).
[libomptarget][AMDGPU] Remove MaxParallelLevel
Dec 2 2020, 9:27 PM
pdhaliwal closed D92463: [libomptarget][AMDGPU] Remove MaxParallelLevel.
Dec 2 2020, 9:27 PM · Unknown Object (Project)

Dec 1 2020

pdhaliwal requested review of D92463: [libomptarget][AMDGPU] Remove MaxParallelLevel.
Dec 1 2020, 11:23 PM · Unknown Object (Project)
pdhaliwal added a comment to D90369: [FastRA] Fix handling of bundled MIs.

Ping!

Dec 1 2020, 6:19 AM · Restricted Project

Nov 23 2020

pdhaliwal added a comment to D90369: [FastRA] Fix handling of bundled MIs.

Ping!

Nov 23 2020, 11:00 PM · Restricted Project

Nov 18 2020

pdhaliwal added inline comments to D90369: [FastRA] Fix handling of bundled MIs.
Nov 18 2020, 5:56 AM · Restricted Project

Nov 4 2020

pdhaliwal added inline comments to D90369: [FastRA] Fix handling of bundled MIs.
Nov 4 2020, 8:45 PM · Restricted Project

Nov 3 2020

pdhaliwal added a comment to D90369: [FastRA] Fix handling of bundled MIs.

Ping!

Nov 3 2020, 3:59 AM · Restricted Project

Oct 28 2020

pdhaliwal added a reviewer for D90369: [FastRA] Fix handling of bundled MIs: arsenm.
Oct 28 2020, 11:59 PM · Restricted Project
pdhaliwal updated the diff for D90369: [FastRA] Fix handling of bundled MIs.

Removed redundant header

Oct 28 2020, 11:58 PM · Restricted Project
pdhaliwal requested review of D90369: [FastRA] Fix handling of bundled MIs.
Oct 28 2020, 11:55 PM · Restricted Project

Oct 21 2020

pdhaliwal accepted D89888: [libomptarget][amdgcn] Refactor memcpy to eliminate maps.

LGTM, thanks!

Oct 21 2020, 8:46 AM · Unknown Object (Project)

Oct 20 2020

pdhaliwal committed rGaa616efbb34e: [libomptarget][AMDGPU][NFC] Split atmi_memcpy for h2d and d2h (authored by pdhaliwal).
[libomptarget][AMDGPU][NFC] Split atmi_memcpy for h2d and d2h
Oct 20 2020, 3:29 AM
pdhaliwal closed D89776: [libomptarget][AMDGPU][NFC] Split atmi_memcpy for h2d and d2h.
Oct 20 2020, 3:29 AM · Unknown Object (Project)
pdhaliwal updated the diff for D89776: [libomptarget][AMDGPU][NFC] Split atmi_memcpy for h2d and d2h.

clang-format'ed the changes

Oct 20 2020, 2:39 AM · Unknown Object (Project)
pdhaliwal edited reviewers for D89776: [libomptarget][AMDGPU][NFC] Split atmi_memcpy for h2d and d2h, added: JonChesterfield, ronlieb; removed: jdoerfert.
Oct 20 2020, 2:12 AM · Unknown Object (Project)
pdhaliwal requested review of D89776: [libomptarget][AMDGPU][NFC] Split atmi_memcpy for h2d and d2h.
Oct 20 2020, 2:12 AM · Unknown Object (Project)

Oct 6 2020

pdhaliwal committed rG3a12ff0dac5a: [OpenMP][RTL] Remove dead code (authored by pdhaliwal).
[OpenMP][RTL] Remove dead code
Oct 6 2020, 2:44 AM
pdhaliwal closed D88829: [OpenMP][RTL] Remove dead code.
Oct 6 2020, 2:44 AM · Unknown Object (Project), Restricted Project

Oct 5 2020

pdhaliwal requested review of D88829: [OpenMP][RTL] Remove dead code.
Oct 5 2020, 6:08 AM · Unknown Object (Project), Restricted Project

Sep 23 2020

pdhaliwal committed rG41d6669f1f16: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH (authored by pdhaliwal).
[GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH
Sep 23 2020, 7:26 PM
pdhaliwal closed D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.
Sep 23 2020, 7:26 PM · Restricted Project

Sep 22 2020

pdhaliwal updated the diff for D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

Formatting and removed implicit uses

Sep 22 2020, 9:33 PM · Restricted Project
pdhaliwal updated the diff for D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

Removed unused code

Sep 22 2020, 7:01 AM · Restricted Project
pdhaliwal updated the diff for D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

Added lowerFor({V2S8})

Sep 22 2020, 6:54 AM · Restricted Project

Sep 20 2020

pdhaliwal added inline comments to D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.
Sep 20 2020, 9:14 PM · Restricted Project

Sep 17 2020

pdhaliwal added inline comments to D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.
Sep 17 2020, 9:25 PM · Restricted Project
pdhaliwal updated the diff for D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

Updated tests and clamping number of elements to 2

Sep 17 2020, 4:31 AM · Restricted Project

Sep 7 2020

pdhaliwal committed rG7634c64b6121: [OpenMP][AMDGPU] Use DS_Max_Warp_Number instead of WARPSIZE (authored by pdhaliwal).
[OpenMP][AMDGPU] Use DS_Max_Warp_Number instead of WARPSIZE
Sep 7 2020, 2:15 AM
pdhaliwal closed D87084: [OpenMP][AMDGPU] Use DS_Max_Warp_Number instead of WARPSIZE.
Sep 7 2020, 2:15 AM · Unknown Object (Project)

Sep 3 2020

pdhaliwal updated the diff for D87084: [OpenMP][AMDGPU] Use DS_Max_Warp_Number instead of WARPSIZE.

Updated data_sharing_stack_init_common

Sep 3 2020, 10:51 PM · Unknown Object (Project)
pdhaliwal added a comment to D87084: [OpenMP][AMDGPU] Use DS_Max_Warp_Number instead of WARPSIZE.

Only places where it was accessed are here and here. Jon's observation is correct. The maximum number of threads on both amdgcn and nvptx is 1024. However, on amdgcn, wave size is 64 and so maximum number of waves can be 16 and on nvptx, the warp size is 32 and maximum number of warps is 32.

Sep 3 2020, 6:51 PM · Unknown Object (Project)
pdhaliwal requested review of D87084: [OpenMP][AMDGPU] Use DS_Max_Warp_Number instead of WARPSIZE.
Sep 3 2020, 5:10 AM · Unknown Object (Project)

Sep 2 2020

pdhaliwal added a comment to D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

@arsenm , let me know if it is good to land.

Sep 2 2020, 2:39 AM · Restricted Project

Aug 30 2020

pdhaliwal updated the diff for D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

Updated review comments.

Aug 30 2020, 9:10 PM · Restricted Project

Aug 11 2020

pdhaliwal added inline comments to D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.
Aug 11 2020, 8:45 AM · Restricted Project
pdhaliwal updated the diff for D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.

Added support for vector types.

Aug 11 2020, 8:45 AM · Restricted Project
pdhaliwal retitled D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH from [GlobalISel] widenScalar G_SMULH/G_UMULH to [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.
Aug 11 2020, 5:18 AM · Restricted Project

Aug 10 2020

pdhaliwal added inline comments to D85653: [GlobalISel][AMDGPU] Lower G_SMULH/G_UMULH.
Aug 10 2020, 11:41 PM · Restricted Project