Page MenuHomePhabricator
Feed Advanced Search

Today

frasercrmck added a comment to D115161: [IR] Allow the 'align' param attr on vectors of pointers.

I think this needs some more changes down in codegen to pass them through

Mon, Dec 6, 8:38 AM · Restricted Project
frasercrmck retitled D114423: [VP] Propagate align parameter attr on VP gather/scatter to ISel from [IR][VP] Extend the align param attr to vectors of pointers to [VP] Propagate align parameter attr on VP gather/scatter to ISel .
Mon, Dec 6, 8:36 AM · Restricted Project
frasercrmck updated the diff for D114423: [VP] Propagate align parameter attr on VP gather/scatter to ISel .
Mon, Dec 6, 8:34 AM · Restricted Project
frasercrmck requested review of D115161: [IR] Allow the 'align' param attr on vectors of pointers.
Mon, Dec 6, 8:32 AM · Restricted Project
frasercrmck accepted D115081: [RISCV] Loosen some restrictions on lowering constant BUILD_VECTORs using vid.v..

LGTM, cheers. Nit: there's a typo in the description. I think it should be "further up"?

Mon, Dec 6, 7:46 AM · Restricted Project
frasercrmck updated subscribers of D114423: [VP] Propagate align parameter attr on VP gather/scatter to ISel .

Aha I just found D87304 - maybe @jdoerfert @fhahn @anna @dantrushin would be interested in this patch.

Mon, Dec 6, 3:51 AM · Restricted Project
frasercrmck added a comment to D114423: [VP] Propagate align parameter attr on VP gather/scatter to ISel .

What's our status with this patch?
I know we have been discussing among VP folks to get this approved by somebody who had worked on align before.

Mon, Dec 6, 3:13 AM · Restricted Project
frasercrmck accepted D113353: [RISCV] Add scheduling resources for Vector pseudo instructions..

LGTM from what I can see. Thanks!

Mon, Dec 6, 2:12 AM · Restricted Project

Fri, Dec 3

frasercrmck requested review of D115036: [SelectionDAG] Use UnknownSize for VP memory ops.
Fri, Dec 3, 4:07 AM · Restricted Project
frasercrmck added inline comments to D114422: [VP] Propagate align parameter attr on VP load/store to ISel.
Fri, Dec 3, 3:14 AM · Restricted Project
frasercrmck updated the diff for D114422: [VP] Propagate align parameter attr on VP load/store to ISel.
  • rebase and add missing word
Fri, Dec 3, 3:14 AM · Restricted Project
frasercrmck committed rG9eb7322748cf: [RISCV][VP] Add RVV codegen for vp.select (authored by victor-eds).
[RISCV][VP] Add RVV codegen for vp.select
Fri, Dec 3, 3:11 AM
frasercrmck closed D114629: [RISCV][VP] Add RVV codegen for vp.select.
Fri, Dec 3, 3:11 AM · Restricted Project
frasercrmck accepted D115027: [VP] [NFC] Fix vp_store signature and vp_gather examples.

Oh my, sorry about that... LGTM

Fri, Dec 3, 2:04 AM · Restricted Project

Thu, Dec 2

frasercrmck added a comment to D112531: [RFC][DataLayout] Allow vector specifications by element size.

Just to make sure that we don't end up with another incomplete solution: Are there any architectures where alignment could depend on both element size and total size? Are there any where it could depend on element type, i.e. where a vector of i32 and a vector of f32 may have different alignment?

Thu, Dec 2, 11:00 PM · Restricted Project

Wed, Dec 1

frasercrmck added inline comments to D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.
Wed, Dec 1, 12:46 AM · Restricted Project
frasercrmck updated the diff for D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.
  • Use existing Chain over Root. No test changes.
Wed, Dec 1, 12:44 AM · Restricted Project
frasercrmck accepted D113543: [RISCV] Add inline expansion for vector ftrunc/fceil/ffloor..

LGTM but there's one clang-format issue that should be addressed. Thanks, @craig.topper!

Wed, Dec 1, 12:11 AM · Restricted Project

Tue, Nov 30

frasercrmck added inline comments to D112463: [LangRef] Remove incorrect vector alignment rules.
Tue, Nov 30, 10:02 AM · Restricted Project
frasercrmck updated the diff for D112463: [LangRef] Remove incorrect vector alignment rules.

Choose better numbers to remove confusing use of 128

Tue, Nov 30, 10:02 AM · Restricted Project
frasercrmck added inline comments to D112463: [LangRef] Remove incorrect vector alignment rules.
Tue, Nov 30, 9:31 AM · Restricted Project
frasercrmck added a comment to D112463: [LangRef] Remove incorrect vector alignment rules.

ping, thanks

Tue, Nov 30, 8:06 AM · Restricted Project
frasercrmck added a comment to D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.

Presumably this only affects IR with wacky return types, because Clang inserts an explicit sret parameter in the IR it generates when the ABI needs it?

Tue, Nov 30, 6:25 AM · Restricted Project
frasercrmck updated the diff for D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.
  • add RV32 RUN line
Tue, Nov 30, 6:23 AM · Restricted Project
frasercrmck added a reviewer for D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter: tlively.

@tlively I would appreciate your thoughts on the WebAssembly test changes

Tue, Nov 30, 6:15 AM · Restricted Project
frasercrmck updated the summary of D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.
Tue, Nov 30, 6:13 AM · Restricted Project
frasercrmck updated the diff for D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.
  • adjust formatting
Tue, Nov 30, 6:13 AM · Restricted Project
frasercrmck requested review of D114795: [PR52475] Ensure a correct chain in copies to/from hidden sret parameter.
Tue, Nov 30, 6:12 AM · Restricted Project
frasercrmck added a comment to D113543: [RISCV] Add inline expansion for vector ftrunc/fceil/ffloor..

Thanks, Craig. From what I can see, taking the alive2 for ceil and modifying it to what you've got for floor also times out rather than finding a verification error. If it works for you locally I'm happy.

Tue, Nov 30, 2:00 AM · Restricted Project

Mon, Nov 29

frasercrmck accepted D114246: [RISCV] Fix a bug in RISCVFrameLowering..

LGTM other than nits. Thanks for the fix!

Mon, Nov 29, 2:29 AM · Restricted Project
frasercrmck accepted D114245: [RISCV] Add a test case to show the bug in RISCVFrameLowering..

LGTM. Craig's comment in D114246 has helped pinpoint the problem for me. I dunno if it's worth commenting that in the test file or not.

Mon, Nov 29, 2:27 AM · Restricted Project

Thu, Nov 25

frasercrmck updated the diff for D114423: [VP] Propagate align parameter attr on VP gather/scatter to ISel .
  • rebase and use 'operand' over 'argument'
Thu, Nov 25, 3:21 AM · Restricted Project
frasercrmck added inline comments to D114422: [VP] Propagate align parameter attr on VP load/store to ISel.
Thu, Nov 25, 3:15 AM · Restricted Project
frasercrmck updated the diff for D114422: [VP] Propagate align parameter attr on VP load/store to ISel.
  • rebase and address feedback
Thu, Nov 25, 3:15 AM · Restricted Project

Tue, Nov 23

frasercrmck accepted D113822: [RISCV] Promote f16 log/pow/exp/sin/cos/etc. to f32 libcalls..

LGTM

Tue, Nov 23, 4:21 AM · Restricted Project
frasercrmck accepted D107432: [RISCV] Emit DWARF location expression for RVV stack objects..

LGTM. Anyone else?

Tue, Nov 23, 4:18 AM · Restricted Project
frasercrmck added a comment to D113353: [RISCV] Add scheduling resources for Vector pseudo instructions..

Generally looks good to me, thanks @HsiangKai

Tue, Nov 23, 3:42 AM · Restricted Project
frasercrmck accepted D114144: [VP] Canonicalize macros of VPIntrinsics.def.

LGTM, thanks

Tue, Nov 23, 3:26 AM · Restricted Project, Restricted Project
frasercrmck requested review of D114423: [VP] Propagate align parameter attr on VP gather/scatter to ISel .
Tue, Nov 23, 2:06 AM · Restricted Project
frasercrmck requested review of D114422: [VP] Propagate align parameter attr on VP load/store to ISel.
Tue, Nov 23, 2:05 AM · Restricted Project

Fri, Nov 19

frasercrmck added a comment to D113798: Add loop unrolling and peeling preferences for RISCV.

Perhaps we should run this across a set of benchmarks we're interested in?

We've been using this internally on our SiFive 7 series and U8. Should we check the CPU?

Fri, Nov 19, 10:27 AM · Restricted Project
frasercrmck added a comment to D113798: Add loop unrolling and peeling preferences for RISCV.

Perhaps we should run this across a set of benchmarks we're interested in?

Fri, Nov 19, 10:02 AM · Restricted Project
frasercrmck committed rG92d279fd6d1f: [LangRef][VP] Correct operands' types in vp.select documentation (authored by frasercrmck).
[LangRef][VP] Correct operands' types in vp.select documentation
Fri, Nov 19, 4:17 AM

Thu, Nov 18

frasercrmck added a comment to D109809: [TargetLowering][RISCV] Fixed a scalable vector issue when lowering [s|u]mul.overflow intrinsics.

Hi, the tests added in this commit seem to be failing on some buildbots. E.g. https://lab.llvm.org/buildbot/#/builders/109/builds/26584

Thu, Nov 18, 4:08 AM · Restricted Project
frasercrmck committed rGe1acbda158b3: [AArch64][RISCV] Fix expected smulo/umulo test output (authored by frasercrmck).
[AArch64][RISCV] Fix expected smulo/umulo test output
Thu, Nov 18, 4:06 AM
frasercrmck added a comment to D109809: [TargetLowering][RISCV] Fixed a scalable vector issue when lowering [s|u]mul.overflow intrinsics.

Hi, the tests added in this commit seem to be failing on some buildbots. E.g. https://lab.llvm.org/buildbot/#/builders/109/builds/26584

Thu, Nov 18, 3:36 AM · Restricted Project
frasercrmck added a comment to D109809: [TargetLowering][RISCV] Fixed a scalable vector issue when lowering [s|u]mul.overflow intrinsics.

Hi @craig.topper @frasercrmck , I don’t have commit access, can you land this patch for me? Thanks.

Thu, Nov 18, 2:27 AM · Restricted Project
frasercrmck committed rG9fe6b9e80254: [TargetLowering][RISCV] Fixed a scalable vector issue when lowering [s|u]mul. (authored by tangxingxin1008).
[TargetLowering][RISCV] Fixed a scalable vector issue when lowering [s|u]mul.
Thu, Nov 18, 2:25 AM
frasercrmck closed D109809: [TargetLowering][RISCV] Fixed a scalable vector issue when lowering [s|u]mul.overflow intrinsics.
Thu, Nov 18, 2:25 AM · Restricted Project

Wed, Nov 17

frasercrmck added a comment to D111904: [RISCV] Lower vector CTLZ_ZERO_UNDEF/CTTZ_ZERO_UNDEF by converting to FP and extracting the exponent..

LGTM too other than nits. Thanks, @craig.topper!

Wed, Nov 17, 9:56 AM · Restricted Project

Mon, Nov 15

frasercrmck added inline comments to D107432: [RISCV] Emit DWARF location expression for RVV stack objects..
Mon, Nov 15, 9:57 AM · Restricted Project
frasercrmck added a comment to D111904: [RISCV] Lower vector CTLZ_ZERO_UNDEF/CTTZ_ZERO_UNDEF by converting to FP and extracting the exponent..

Sorry for not being very prompt on this one. I'll try and think about it in more detail later.

Mon, Nov 15, 9:44 AM · Restricted Project
frasercrmck added inline comments to D113453: [RISCV][NFC] Refactor some rvv instructions' definition with foreach..
Mon, Nov 15, 9:33 AM · Restricted Project
frasercrmck accepted D113453: [RISCV][NFC] Refactor some rvv instructions' definition with foreach..

LGTM.

Mon, Nov 15, 9:31 AM · Restricted Project
frasercrmck added a comment to D113204: [RISCV] Teach needVSETVLIPHI to handle mask register instructions..

Though the clang-format thing should probably be addressed.

Mon, Nov 15, 9:30 AM · Restricted Project
frasercrmck accepted D113204: [RISCV] Teach needVSETVLIPHI to handle mask register instructions..

LGTM.

Mon, Nov 15, 9:30 AM · Restricted Project
frasercrmck added inline comments to D113798: Add loop unrolling and peeling preferences for RISCV.
Mon, Nov 15, 2:00 AM · Restricted Project

Thu, Nov 11

frasercrmck added inline comments to D113475: [llvm-tblgen][RISCV] Make llvm-tblgen RISCVCompressInstEmitter to be common infra across different targets.
Thu, Nov 11, 7:48 AM · Restricted Project
frasercrmck added a comment to D113543: [RISCV] Add inline expansion for vector ftrunc/fceil/ffloor..

To what extent is this generic code that other targets would benefit from?

Thu, Nov 11, 7:21 AM · Restricted Project
frasercrmck added inline comments to D113453: [RISCV][NFC] Refactor some rvv instructions' definition with foreach..
Thu, Nov 11, 6:45 AM · Restricted Project

Wed, Nov 10

frasercrmck committed rG4001fea62190: [CGSCC][LazyCallGraph][NFC] Fix typos in code comments (authored by frasercrmck).
[CGSCC][LazyCallGraph][NFC] Fix typos in code comments
Wed, Nov 10, 10:23 AM
frasercrmck committed rGb1d8d70b9dea: [SelectionDAG] Replace the Chain in LOAD->VP_LOAD widening (authored by frasercrmck).
[SelectionDAG] Replace the Chain in LOAD->VP_LOAD widening
Wed, Nov 10, 9:58 AM
frasercrmck closed D113561: [SelectionDAG] Replace the Chain in LOAD->VP_LOAD widening.
Wed, Nov 10, 9:58 AM · Restricted Project
frasercrmck requested review of D113561: [SelectionDAG] Replace the Chain in LOAD->VP_LOAD widening.
Wed, Nov 10, 4:30 AM · Restricted Project
frasercrmck accepted D113219: [RISCV] Prevent crashes when bitcasting between fixed vectors and scalars..

LGTM, good catch!

Wed, Nov 10, 4:09 AM · Restricted Project
frasercrmck committed rG332318ffb613: [SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE (authored by frasercrmck).
[SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE
Wed, Nov 10, 2:04 AM
frasercrmck closed D111248: [SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE.
Wed, Nov 10, 2:04 AM · Restricted Project

Tue, Nov 9

frasercrmck added inline comments to D111248: [SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE.
Tue, Nov 9, 8:42 AM · Restricted Project
frasercrmck updated the diff for D111248: [SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE.
Tue, Nov 9, 8:42 AM · Restricted Project
frasercrmck added a comment to D111248: [SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE.

ping

Tue, Nov 9, 2:47 AM · Restricted Project

Mon, Nov 8

frasercrmck accepted D113325: [RISCV] SMUL_LOHI/UMUL_LOHI should expand for RVV..

Do we want to do the same for fixed-length vectors?

Fixed length vectors has a loop that sets every opcode to Expand.

Mon, Nov 8, 8:40 AM · Restricted Project
frasercrmck added a comment to D113325: [RISCV] SMUL_LOHI/UMUL_LOHI should expand for RVV..

Do we want to do the same for fixed-length vectors?

Mon, Nov 8, 2:15 AM · Restricted Project

Nov 5 2021

frasercrmck committed rG3a11fb572cf5: [LangRef][VP] Document vp.gather and vp.scatter intrinsics (authored by frasercrmck).
[LangRef][VP] Document vp.gather and vp.scatter intrinsics
Nov 5 2021, 4:45 AM
frasercrmck closed D112997: [LangRef][VP] Document vp.gather and vp.scatter intrinsics.
Nov 5 2021, 4:44 AM · Restricted Project
frasercrmck added a comment to D113209: [LV] Use VScaleForTuning to fine-tune the cost per lane..

How does this interact with vscale_range? Could it perhaps automatically infer getVScaleForTuning using that? Or is the idea the target ultimately chooses?

The two are actually quite different; vscale_range specifies the range of vscale that the compiled binary is compatible with. LLVM guarantees that the compiled binary is correct for that vscale_range. VScaleForTuning can be set separately by mcpu/mtune and will purely tune the cost-model without changing the requirements on vscale. This means it doesn't change the compatibility of the binary, it just helps choose a better VF for the CPU it compiles for.

Nov 5 2021, 4:13 AM · Restricted Project
frasercrmck updated the diff for D112997: [LangRef][VP] Document vp.gather and vp.scatter intrinsics.
  • rebase
Nov 5 2021, 3:59 AM · Restricted Project
frasercrmck added a comment to D113209: [LV] Use VScaleForTuning to fine-tune the cost per lane..

How does this interact with vscale_range? Could it perhaps automatically infer getVScaleForTuning using that? Or is the idea the target ultimately chooses?

Nov 5 2021, 3:53 AM · Restricted Project
frasercrmck committed rG93e1802af332: [LangRef][VP] Document vp.load and vp.store intrinsics (authored by frasercrmck).
[LangRef][VP] Document vp.load and vp.store intrinsics
Nov 5 2021, 3:48 AM
frasercrmck closed D112930: [LangRef][VP] Document vp.load and vp.store intrinsics.
Nov 5 2021, 3:48 AM · Restricted Project
frasercrmck accepted D113217: [RISCV] Enable FP extensions and ABI on fixed-vectors-bitcast.ll..

Makes sense, LGTM. Cheers.

Nov 5 2021, 2:31 AM · Restricted Project

Nov 4 2021

frasercrmck added a comment to D112463: [LangRef] Remove incorrect vector alignment rules.

ping

Nov 4 2021, 3:13 AM · Restricted Project

Nov 2 2021

frasercrmck added inline comments to D112930: [LangRef][VP] Document vp.load and vp.store intrinsics.
Nov 2 2021, 10:35 AM · Restricted Project
frasercrmck committed rGd065b03801d9: [RISCV] Optimize vp.load with an all-ones mask (authored by frasercrmck).
[RISCV] Optimize vp.load with an all-ones mask
Nov 2 2021, 10:32 AM
frasercrmck closed D113022: [RISCV] Optimize vp.load with an all-ones mask.
Nov 2 2021, 10:32 AM · Restricted Project
frasercrmck committed rG9b591078cf67: [RISCV][VP] Mangle pointers in vp.load and vp.store tests (authored by frasercrmck).
[RISCV][VP] Mangle pointers in vp.load and vp.store tests
Nov 2 2021, 9:55 AM
frasercrmck closed D112939: [RISCV][VP] Mangle pointers in vp.load and vp.store tests.
Nov 2 2021, 9:55 AM · Restricted Project
frasercrmck added a comment to D112939: [RISCV][VP] Mangle pointers in vp.load and vp.store tests.

Thanks, Craig. Good point about opt. Is that official acceptance or would you like to wait for other comments?

Nov 2 2021, 9:51 AM · Restricted Project
frasercrmck updated the diff for D112939: [RISCV][VP] Mangle pointers in vp.load and vp.store tests.
  • rebase on top of 'allones' tests pushed earlier
Nov 2 2021, 9:11 AM · Restricted Project
frasercrmck added inline comments to D112930: [LangRef][VP] Document vp.load and vp.store intrinsics.
Nov 2 2021, 8:02 AM · Restricted Project
frasercrmck committed rG6fb41c3deaed: [LangRef][VP] Correct mask type in vp.slice documentation (authored by frasercrmck).
[LangRef][VP] Correct mask type in vp.slice documentation
Nov 2 2021, 7:57 AM
frasercrmck closed D112924: [LangRef][VP] Correct mask type in vp.slice documentation.
Nov 2 2021, 7:57 AM · Restricted Project
frasercrmck requested review of D113022: [RISCV] Optimize vp.load with an all-ones mask.
Nov 2 2021, 7:54 AM · Restricted Project
frasercrmck committed rG1fc80ffc921a: [RISCV][VP] Add tests for "unmasked" VP loads (authored by frasercrmck).
[RISCV][VP] Add tests for "unmasked" VP loads
Nov 2 2021, 7:44 AM
frasercrmck requested review of D112997: [LangRef][VP] Document vp.gather and vp.scatter intrinsics.
Nov 2 2021, 3:32 AM · Restricted Project

Nov 1 2021

frasercrmck requested review of D112939: [RISCV][VP] Mangle pointers in vp.load and vp.store tests.
Nov 1 2021, 10:00 AM · Restricted Project
frasercrmck updated the diff for D112930: [LangRef][VP] Document vp.load and vp.store intrinsics.

fix typos
add missing pointer mangling to examples

Nov 1 2021, 9:51 AM · Restricted Project
frasercrmck requested review of D112930: [LangRef][VP] Document vp.load and vp.store intrinsics.
Nov 1 2021, 7:52 AM · Restricted Project
frasercrmck requested review of D112924: [LangRef][VP] Correct mask type in vp.slice documentation.
Nov 1 2021, 4:51 AM · Restricted Project
frasercrmck updated the diff for D111248: [SelectionDAG] Widen scalable-vector loads/stores via VP_LOAD/VP_STORE.
  • rebase
  • fix accidental unreachable due to bad rebasing
Nov 1 2021, 3:52 AM · Restricted Project
frasercrmck accepted D108129: [DAGCombiner] Teach combineShiftToMULH to handle constant and const splat vector..

LGTM but either you need to land D110675 and rebase, or add D110675 as a parent revision. That way it's easier to see how this patch improves things.

Nov 1 2021, 3:03 AM · Restricted Project