Page MenuHomePhabricator

sebastian-ne (Sebastian Neubauer)
User

Projects

User does not belong to any projects.

User Details

User Since
Jan 27 2020, 1:17 AM (126 w, 2 d)

Recent Activity

Mon, Jun 20

sebastian-ne updated the diff for D127728: [BitcodeReader] Allow reading pointer types from old IR.

I found one more place where type information from dxil needs to be extracted and that’s metadata.
The patch now adds a second callback, which allows changing metadata while it’s read in.
In the test, that’s used to replace a pointer value metadata with a tuple of the original value and metadata that stores its type information.
I wasn’t able to store type info about metadata without replacing it because the values get indistinguishable after being read (e.g. an i8* and an i32* both end up as a ptr).

Mon, Jun 20, 8:08 AM · Restricted Project, Restricted Project
sebastian-ne accepted D125388: Allow type-mismatching RAUW of values in metadata, and add C API..

Since there have been no other reactions and I think this is a reasonable patch, I’ll accept it.
Please wait a day before submitting it in case others have comments.

Mon, Jun 20, 1:37 AM · Restricted Project, Restricted Project

Fri, Jun 17

sebastian-ne added a comment to D127728: [BitcodeReader] Allow reading pointer types from old IR.

I think exposing something like this is reasonable.

Fri, Jun 17, 3:52 AM · Restricted Project, Restricted Project

Tue, Jun 14

sebastian-ne requested review of D127728: [BitcodeReader] Allow reading pointer types from old IR.
Tue, Jun 14, 2:09 AM · Restricted Project, Restricted Project

May 24 2022

sebastian-ne added inline comments to D125992: [AMDGPU] gfx11 FLAT Instructions.
May 24 2022, 1:07 AM · Restricted Project, Restricted Project

May 23 2022

sebastian-ne added a comment to D125992: [AMDGPU] gfx11 FLAT Instructions.

Looks good to me, thanks for moving out the one negative test.

May 23 2022, 7:33 AM · Restricted Project, Restricted Project

May 18 2022

sebastian-ne added inline comments to D114643: [AMDGPU] Aggressively fold immediates in SIFoldOperands.
May 18 2022, 3:21 AM · Restricted Project, Restricted Project

May 17 2022

sebastian-ne added a comment to D114643: [AMDGPU] Aggressively fold immediates in SIFoldOperands.

Both patches look good to me!

May 17 2022, 5:19 AM · Restricted Project, Restricted Project

May 13 2022

sebastian-ne added a comment to D125427: [TableGen] Add a new json textmate description for syntax highlighting.

The highlighting looks nice with this.
I didn’t see anything blatantly wrong or missing, so thumbs up from me.

May 13 2022, 4:42 AM · Restricted Project, Restricted Project

May 12 2022

sebastian-ne accepted D125324: AMDGPU/GISel: Factor out AMDGPURegisterBankInfo::buildReadFirstLane.

Looks good to me

May 12 2022, 9:45 AM · Restricted Project, Restricted Project

Apr 28 2022

sebastian-ne accepted D124232: [AMDGPU] Use d16 flag for image.sample instructions.

I wanted to add this combine before, but I don’t think there is a way to add d16 to an instruction without potentially breaking the code.
The reason is, when an image_sample has the d16 flag enabled, it will use f32→f16 truncation or i32→i16 truncation, depending on the texture format in the descriptor.

Combining image_sample+fptrunc to image_sample d16 works fine for float textures, but I assume we don’t know at compile time if a texture is an integer or float texture.
The application may interpret stored values as float and does an fptrunc, but the texture is actually defined as an integer texture, so the hardware uses an integer trunc instead, giving different results.

Are you aware of any examples that would fail with this patch? At least in Vulkan, the image format must match the sampled type.

Apr 28 2022, 2:37 AM · Restricted Project, Restricted Project

Apr 27 2022

sebastian-ne requested changes to D124232: [AMDGPU] Use d16 flag for image.sample instructions.

I wanted to add this combine before, but I don’t think there is a way to add d16 to an instruction without potentially breaking the code.
The reason is, when an image_sample has the d16 flag enabled, it will use f32→f16 truncation or i32→i16 truncation, depending on the texture format in the descriptor.

Apr 27 2022, 8:42 AM · Restricted Project, Restricted Project

Apr 22 2022

sebastian-ne added a comment to D124192: [AMDGPU] Callee must always spill writelane VGPRs.

Nice change(s).
I’m not sure why the writelane registers are added as live-in to every block. Is the same happening for WWM registers and VGPRs used for SGPR spills?

Apr 22 2022, 3:28 AM · Restricted Project, Restricted Project

Apr 19 2022

sebastian-ne accepted D123938: AMDGPU: Serialize WWM registers.
Apr 19 2022, 6:26 AM · Restricted Project, Restricted Project
sebastian-ne accepted D123937: AMDGPU: Defer creation of WWM VGPR spill slots.
Apr 19 2022, 6:19 AM · Restricted Project, Restricted Project
sebastian-ne accepted D109008: [AMDGPU][NFC] Refactor AMDGPUCallingConv.td.

LGTM, nice!

Apr 19 2022, 1:36 AM · Restricted Project, Restricted Project

Mar 29 2022

sebastian-ne requested review of D122639: [CMake] Keep targets in folder in Visual Studio.
Mar 29 2022, 12:36 AM · Restricted Project, Restricted Project

Mar 16 2022

sebastian-ne added a comment to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..

Looks good, thanks, just left some small comments.

Mar 16 2022, 5:16 AM · Restricted Project, Restricted Project

Mar 14 2022

sebastian-ne added a comment to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..

Can you add lit tests for the fixes you made please?

Mar 14 2022, 12:35 AM · Restricted Project, Restricted Project

Mar 3 2022

sebastian-ne committed rG473efae3a18f: [UpdateTestChecks] Don't skip first line with --filter (authored by sebastian-ne).
[UpdateTestChecks] Don't skip first line with --filter
Mar 3 2022, 8:27 AM · Restricted Project
sebastian-ne closed D119704: [UpdateTestChecks] Don't skip first line with --filter.
Mar 3 2022, 8:27 AM · Restricted Project, Restricted Project
sebastian-ne accepted D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..
Mar 3 2022, 12:35 AM · Restricted Project, Restricted Project

Mar 2 2022

sebastian-ne added a comment to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..

Looks good to me

Mar 2 2022, 8:15 AM · Restricted Project, Restricted Project
JOE1994 awarded D119704: [UpdateTestChecks] Don't skip first line with --filter a Love token.
Mar 2 2022, 5:52 AM · Restricted Project, Restricted Project
sebastian-ne added a comment to D117694: [UpdateTestChecks] Add --filter and --filter-out options.

I see 6 non-temporal store instructions in "llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/x86-non-temporal.ll",
but the expected output ("llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/x86-non-temporal.ll.filter.expected") checks for only 5 non-temporal instructions.

Shouldn't the expected output check for 6 non-temporal stores instead of 5 ?

Thank you

Mar 2 2022, 12:30 AM · Restricted Project, Restricted Project

Mar 1 2022

sebastian-ne added a reviewer for D119704: [UpdateTestChecks] Don't skip first line with --filter: foad.
Mar 1 2022, 8:54 AM · Restricted Project, Restricted Project
sebastian-ne added a comment to D119368: [UpdateLLCTestChecks] Add support for isel debug output in update_llc_test_checks.py.

No problem.

Mar 1 2022, 6:34 AM · Restricted Project
sebastian-ne committed rGc74f54f2f451: [UpdateTestChecks] Add requires asserts to tests (authored by sebastian-ne).
[UpdateTestChecks] Add requires asserts to tests
Mar 1 2022, 6:29 AM
sebastian-ne committed rG8565b6f9f278: [UpdateLLCTestChecks] Add support for isel debug output in… (authored by ningxinr).
[UpdateLLCTestChecks] Add support for isel debug output in…
Mar 1 2022, 2:00 AM
sebastian-ne closed D119368: [UpdateLLCTestChecks] Add support for isel debug output in update_llc_test_checks.py.
Mar 1 2022, 2:00 AM · Restricted Project

Feb 28 2022

sebastian-ne added inline comments to D119368: [UpdateLLCTestChecks] Add support for isel debug output in update_llc_test_checks.py.
Feb 28 2022, 8:17 AM · Restricted Project

Feb 22 2022

sebastian-ne added inline comments to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..
Feb 22 2022, 2:56 AM · Restricted Project, Restricted Project
sebastian-ne added inline comments to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..
Feb 22 2022, 2:25 AM · Restricted Project, Restricted Project

Feb 21 2022

sebastian-ne added a comment to D119704: [UpdateTestChecks] Don't skip first line with --filter.

friendly ping for review

Feb 21 2022, 1:34 AM · Restricted Project, Restricted Project

Feb 18 2022

sebastian-ne added a reviewer for D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3.: sebastian-ne.
Feb 18 2022, 6:50 AM · Restricted Project, Restricted Project
sebastian-ne committed rG6527b2a4d5fa: [AMDGPU][NFC] Fix typos (authored by sebastian-ne).
[AMDGPU][NFC] Fix typos
Feb 18 2022, 6:06 AM
sebastian-ne closed D119235: [AMDGPU][NFC] Fix typos.
Feb 18 2022, 6:05 AM · Restricted Project
sebastian-ne added a comment to D119235: [AMDGPU][NFC] Fix typos.

friendly ping for review

Feb 18 2022, 5:48 AM · Restricted Project
sebastian-ne committed rG1f0aadfa62a5: [AMDGPU] Fix kill flag on overlapping sgpr copy (authored by sebastian-ne).
[AMDGPU] Fix kill flag on overlapping sgpr copy
Feb 18 2022, 5:36 AM
sebastian-ne closed D120042: [AMDGPU] Fix kill flag on overlapping sgpr copy.
Feb 18 2022, 5:36 AM · Restricted Project

Feb 17 2022

sebastian-ne added a comment to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..

The pre-merge builds report some test failure in MC/AMDGPU and MC/Disassembler/AMDGPU. I think the assembler fix in VOPCInstructions.td and these test changes could be a separate patch.

Feb 17 2022, 5:21 AM · Restricted Project, Restricted Project
sebastian-ne added inline comments to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..
Feb 17 2022, 5:15 AM · Restricted Project, Restricted Project
sebastian-ne requested review of D120042: [AMDGPU] Fix kill flag on overlapping sgpr copy.
Feb 17 2022, 4:36 AM · Restricted Project

Feb 14 2022

sebastian-ne requested review of D119704: [UpdateTestChecks] Don't skip first line with --filter.
Feb 14 2022, 3:09 AM · Restricted Project, Restricted Project
sebastian-ne added inline comments to D119696: [AMDGPU] Improve v_cmpx usage on GFX10.3..
Feb 14 2022, 2:44 AM · Restricted Project, Restricted Project

Feb 11 2022

sebastian-ne committed rGa5d4f82b7392: [AMDGPU] Make enable-flat-scratch a subtarget feature (authored by sebastian-ne).
[AMDGPU] Make enable-flat-scratch a subtarget feature
Feb 11 2022, 9:24 AM
sebastian-ne closed D119425: [AMDGPU] Make enable-flat-scratch a subtarget feature.
Feb 11 2022, 9:23 AM · Restricted Project

Feb 10 2022

sebastian-ne retitled D119425: [AMDGPU] Make enable-flat-scratch a subtarget feature from [AMDGPU] Make enable-flat-scratch an attribute to [AMDGPU] Make enable-flat-scratch a subtarget feature.
Feb 10 2022, 6:41 AM · Restricted Project
sebastian-ne added a comment to D119425: [AMDGPU] Make enable-flat-scratch a subtarget feature.

Just curious: why is it not enabled by default on subtargets that support it? Then you would only need the attribute if you wanted to turn it off for some reason.

Feb 10 2022, 4:09 AM · Restricted Project
sebastian-ne requested review of D119425: [AMDGPU] Make enable-flat-scratch a subtarget feature.
Feb 10 2022, 3:46 AM · Restricted Project
sebastian-ne updated the diff for D119235: [AMDGPU][NFC] Fix typos.

Improve WQM comment

Feb 10 2022, 1:40 AM · Restricted Project

Feb 8 2022

sebastian-ne added inline comments to D119235: [AMDGPU][NFC] Fix typos.
Feb 8 2022, 5:30 AM · Restricted Project
sebastian-ne updated the diff for D119235: [AMDGPU][NFC] Fix typos.

Thanks! Fixed your comments

Feb 8 2022, 5:30 AM · Restricted Project
sebastian-ne updated the diff for D119235: [AMDGPU][NFC] Fix typos.

Two more typos

Feb 8 2022, 5:10 AM · Restricted Project
sebastian-ne requested review of D119235: [AMDGPU][NFC] Fix typos.
Feb 8 2022, 5:08 AM · Restricted Project

Feb 7 2022

sebastian-ne added a comment to D116832: [UpdateLLCTestChecks] Allow replacing register names with variables.

Looks good to me, with all the comments and fixes the generated tests look a load better than in the first version.

Feb 7 2022, 8:40 AM · Restricted Project

Feb 4 2022

sebastian-ne committed rG4a02562275d4: [AMDGPU] Lazily init pal metadata on first function (authored by sebastian-ne).
[AMDGPU] Lazily init pal metadata on first function
Feb 4 2022, 9:41 AM
sebastian-ne closed D118492: [AMDGPU] Lazily init pal metadata on first function.
Feb 4 2022, 9:41 AM · Restricted Project

Feb 2 2022

sebastian-ne accepted D117504: AMDGPU: Convert AMDGPUResourceUsageAnalysis to a Module pass.
Feb 2 2022, 1:13 AM · Restricted Project
sebastian-ne accepted D117364: AMDGPU: Use module level register maximums for unknown callees.

Forgot to accept as amdgpu last time.

Feb 2 2022, 1:12 AM · Restricted Project

Feb 1 2022

sebastian-ne added inline comments to D118492: [AMDGPU] Lazily init pal metadata on first function.
Feb 1 2022, 1:29 AM · Restricted Project

Jan 28 2022

sebastian-ne requested review of D118492: [AMDGPU] Lazily init pal metadata on first function.
Jan 28 2022, 10:03 AM · Restricted Project

Jan 27 2022

sebastian-ne accepted D118309: AMDGPU/GlobalISel: Fix assertions on invalid addrspacecasts.
Jan 27 2022, 8:18 AM · Restricted Project
sebastian-ne accepted D117618: AMDGPU/GlobalISel: Fix flat_scratch_init handling for shaders.
Jan 27 2022, 2:42 AM · Restricted Project
sebastian-ne added inline comments to D118309: AMDGPU/GlobalISel: Fix assertions on invalid addrspacecasts.
Jan 27 2022, 1:57 AM · Restricted Project
sebastian-ne accepted D118308: AMDGPU: Handle addrspacecast of constant 32-bit to flat.
Jan 27 2022, 1:51 AM · Restricted Project

Jan 26 2022

sebastian-ne accepted D117135: AMDGPU/GlobalISel: Fold wave address into mubuf addressing modes.
Jan 26 2022, 11:41 AM · Restricted Project
sebastian-ne committed rGecf7a0e33883: [CMake] Disable mvsc warning for new versions (authored by sebastian-ne).
[CMake] Disable mvsc warning for new versions
Jan 26 2022, 6:08 AM
sebastian-ne closed D118228: [CMake] Disable mvsc warning for new versions.
Jan 26 2022, 6:08 AM · Restricted Project
sebastian-ne committed rG4723f3cf03a9: [AMDGPU][GlobalISel] Combine unmerge of undef (authored by sebastian-ne).
[AMDGPU][GlobalISel] Combine unmerge of undef
Jan 26 2022, 3:31 AM
sebastian-ne committed rG668046666382: [AMDGPU][NFC] Pre-commit regenerated test (authored by sebastian-ne).
[AMDGPU][NFC] Pre-commit regenerated test
Jan 26 2022, 3:31 AM
sebastian-ne closed D118138: [AMDGPU][GlobalISel] Combine unmerge of undef.
Jan 26 2022, 3:31 AM · Restricted Project
sebastian-ne added a comment to D117537: [cmake] Partially deduplicate `{llvm,compiler_rt}_check_linker_flag` for runtime libs and llvm.

@phosek, this patch fixes a regression that was introduced with D116521.
Could we fix this regression first with a simple patch that does not risk to be reverted again and do further refactorings afterwards?
Our downstream gcc build is broken and we’d like to re-enable it rather sooner than later.

Jan 26 2022, 2:36 AM · Restricted Project, Restricted Project, Restricted Project, Restricted Project, Restricted Project
sebastian-ne requested review of D118228: [CMake] Disable mvsc warning for new versions.
Jan 26 2022, 1:42 AM · Restricted Project

Jan 25 2022

sebastian-ne updated the diff for D118138: [AMDGPU][GlobalISel] Combine unmerge of undef.

Thanks, the patch got smaller with applyBuildFn.

Jan 25 2022, 6:27 AM · Restricted Project
sebastian-ne committed rG4ed7c6eec979: [AMDGPU] Only match correct type for a16 (authored by sebastian-ne).
[AMDGPU] Only match correct type for a16
Jan 25 2022, 6:07 AM
sebastian-ne closed D118043: [AMDGPU] Only match correct type for a16.
Jan 25 2022, 6:07 AM · Restricted Project
sebastian-ne requested review of D118138: [AMDGPU][GlobalISel] Combine unmerge of undef.
Jan 25 2022, 5:54 AM · Restricted Project

Jan 24 2022

sebastian-ne updated the diff for D118043: [AMDGPU] Only match correct type for a16.

Thanks for the review, I fixed the comments.

Jan 24 2022, 9:46 AM · Restricted Project
sebastian-ne added a comment to D117876: [AMDGPU][InstCombine] Remove zero image offset.

wouldn't it be nice if there was some way to autogenerate the Offset to NoOffset Optimization Mapping table?

Jan 24 2022, 9:27 AM · Restricted Project
sebastian-ne committed rG80532ebb508d: [AMDGPU][InstCombine] Remove zero image offset (authored by sebastian-ne).
[AMDGPU][InstCombine] Remove zero image offset
Jan 24 2022, 9:08 AM
sebastian-ne closed D117876: [AMDGPU][InstCombine] Remove zero image offset.
Jan 24 2022, 9:07 AM · Restricted Project
sebastian-ne requested review of D118043: [AMDGPU] Only match correct type for a16.
Jan 24 2022, 7:13 AM · Restricted Project
sebastian-ne abandoned D117875: [AMDGPU][InstCombine] Use D16 if only f16 precision is needed.
Jan 24 2022, 5:45 AM · Restricted Project
sebastian-ne committed rGf1e36474b9e5: [AMDGPU][NFC] Fix debug prints (authored by sebastian-ne).
[AMDGPU][NFC] Fix debug prints
Jan 24 2022, 4:55 AM

Jan 21 2022

sebastian-ne added inline comments to D117135: AMDGPU/GlobalISel: Fold wave address into mubuf addressing modes.
Jan 21 2022, 9:07 AM · Restricted Project
sebastian-ne added a comment to D117504: AMDGPU: Convert AMDGPUResourceUsageAnalysis to a Module pass.

I think I found the problem and I’m a little surprised it currently works.
In the llpc pipeline, a module pass (PipelineStateClearer) writes pal metadata into the amdgpu.pal.metadata.msgpack metadata.
This metadata is read by the AMDGPUAsmPrinter, extended as functions are emitted and finally written out in AMDGPUTarget[…]Streamer::finish.

Jan 21 2022, 8:54 AM · Restricted Project
sebastian-ne requested review of D117876: [AMDGPU][InstCombine] Remove zero image offset.
Jan 21 2022, 4:35 AM · Restricted Project
sebastian-ne requested review of D117875: [AMDGPU][InstCombine] Use D16 if only f16 precision is needed.
Jan 21 2022, 4:34 AM · Restricted Project
sebastian-ne committed rGae2f9c8be897: [AMDGPU] Remove lz and nomip combine from codegen (authored by sebastian-ne).
[AMDGPU] Remove lz and nomip combine from codegen
Jan 21 2022, 3:09 AM
sebastian-ne committed rG603d18033c51: [AMDGPU][InstCombine] Remove zero LOD bias (authored by sebastian-ne).
[AMDGPU][InstCombine] Remove zero LOD bias
Jan 21 2022, 3:09 AM
sebastian-ne committed rG0530fdbbbb84: [AMDGPU] Fix LOD bias in A16 combine (authored by sebastian-ne).
[AMDGPU] Fix LOD bias in A16 combine
Jan 21 2022, 3:09 AM
sebastian-ne committed rGb351ac3873db: [AMDGPU][NFC] Regenerate InstCombine test (authored by sebastian-ne).
[AMDGPU][NFC] Regenerate InstCombine test
Jan 21 2022, 3:09 AM
sebastian-ne closed D116116: [AMDGPU] Remove lz and nomip combine from codegen.
Jan 21 2022, 3:09 AM · Restricted Project
sebastian-ne closed D116042: [AMDGPU][InstCombine] Remove zero LOD bias.
Jan 21 2022, 3:09 AM · Restricted Project
sebastian-ne closed D116038: [AMDGPU] Fix LOD bias in A16 combine.
Jan 21 2022, 3:09 AM · Restricted Project

Jan 20 2022

sebastian-ne added inline comments to D117618: AMDGPU/GlobalISel: Fix flat_scratch_init handling for shaders.
Jan 20 2022, 7:14 AM · Restricted Project
sebastian-ne accepted D117483: AMDGPU: Stop reserving 36-bytes before kernel arguments for amdpal.

Under the assumption that opencl or hip on pal do not use this, it looks fine to me.

Jan 20 2022, 4:15 AM · Restricted Project
sebastian-ne added a comment to D116053: [MachineSink] Allow sinking of constant or ignorable physreg uses.

Both patches seem to fix the bug

Jan 20 2022, 2:03 AM · Restricted Project

Jan 19 2022

sebastian-ne added a comment to D116053: [MachineSink] Allow sinking of constant or ignorable physreg uses.

This change seems to sink v_cmp instructions, which creates different results if the exec mask changed and that makes several Vulkan tests fail.
I put a reproducer here: https://gist.github.com/Flakebi/fd1d91a806b60ec330e9f61e19fe62ac
Compile with llc -mtriple=amdgcn--amdpal -mcpu=gfx1010 -verify-machineinstrs -start-before=machine-sink -stop-after=machine-sink PipelineVsFs_0xDD57C231E25DA514.mir -o PipelineVsFs_0xDD57C231E25DA514-after.mir
and the %104:sreg_64 = V_CMP_NE_U32_e64 %89, %101, implicit $exec instruction will be sunk from bb.5 into bb.6. For reference, the pipeline is from the dEQP-VK.subgroups.arithmetic.framebuffer.subgroupexclusiveadd_float_vertex CTS test.

Jan 19 2022, 11:07 AM · Restricted Project