Page MenuHomePhabricator

Today

cchen updated the diff for D70726: [OpenMP50] Add parallel master construct.

Pass triple for all the test runs

Thu, Dec 5, 9:26 AM · Restricted Project
jfb added a reviewer for D66822: Hardware cache line size builtins: __simt__.
Thu, Dec 5, 9:26 AM · Restricted Project
miyuki added inline comments to D71066: [ARM][MVE][Intrinsics] Add VMULL[BT]Q_(INT|POLY) intrinsics..
Thu, Dec 5, 9:25 AM · Restricted Project, Restricted Project
dschuff added a comment to D70500: [WebAssembly] Enable use of wasm-opt and LTO-enabled system libraries.

I do find it odd that there is a PATH fallback in the existing code in the first place. I agree that basically no compiler other than the "system" compiler should ever use it (and also even the concept of the "system" compiler really only makes much sense on systems like Linux and BSDs where compiling things for the local system is common). I guess the other option here would be to just require that wasm-opt be in the same directory as clang, which we can arrange in wasi-sdk or wherever.

Thu, Dec 5, 9:24 AM · Restricted Project
aprantl added inline comments to D70121: [DebugInfo][LDV] Teach LDV how to identify source variables and handle fragments.
Thu, Dec 5, 9:24 AM · Restricted Project, debug-info
lebedev.ri added reviewers for D71064: [InstCombine] Invert `add A, sext(B) --> sub A, zext(B)` canonicalization (to `sub A, zext B -> add A, sext B`): t.p.northover, hfinkel.

Scalar looks same all-around. Vector shows some potential diffs:
https://godbolt.org/z/y3E-mb

If I'm seeing it correctly, we always do better on the typical case where the bool vector is produced by a compare, but we might do worse if we don't have that cmp and don't have AssertSext knowledge.

So the comment is that the undo fold needs to be adjusted first, to fire for non-cmp i1 vectors on aarch64 and powerpc64le?

Thu, Dec 5, 9:24 AM · Restricted Project
davidxl accepted D64376: [MBP] Avoid tail duplication if it can't bring benefit.

lgtm

Thu, Dec 5, 9:24 AM · Restricted Project
lebedev.ri added a comment to D71064: [InstCombine] Invert `add A, sext(B) --> sub A, zext(B)` canonicalization (to `sub A, zext B -> add A, sext B`).

Did you confirm that codegen is equal or better for these cases?
I think we have DAGCombiner reversals for this transform, but some targets that seem like they would benefit have not enabled the TLI hook.

I didn't yet. So, as far as i can tell, these are all the interesting cases:
(i'm looking at @t0_new_canon vs @t1_old_canon since that is the only change in this patch)
https://godbolt.org/z/vbb25Q - no regression for x86
https://godbolt.org/z/Htg7m5 - aarch64 also looks ok?
https://godbolt.org/z/xmP6JV - arm good?
https://godbolt.org/z/vNrNJ8 - thumb good?

So i'd say everything is already covered by backend undo folds?
Let me know if i'm missing the point here?

Scalar looks same all-around. Vector shows some potential diffs:
https://godbolt.org/z/y3E-mb

If I'm seeing it correctly, we always do better on the typical case where the bool vector is produced by a compare, but we might do worse if we don't have that cmp and don't have AssertSext knowledge.

Thu, Dec 5, 9:24 AM · Restricted Project
aprantl added inline comments to D70720: [llvm-objdump] Display locations of variables alongside disassembly.
Thu, Dec 5, 9:24 AM · debug-info, Restricted Project
wmi added a comment to D70750: [AutoFDO] Inline replay for cold/small callees from sample profile loader.

Did performance test and I saw 0.4% improvement in an internal benchmark. That is a good improvement, thanks for the change!

Thu, Dec 5, 9:14 AM · Restricted Project
jmorse added inline comments to D69178: [DebugInfo] Use DBG_VALUEs IsIndirect field for describing stack spills.
Thu, Dec 5, 9:14 AM · Restricted Project
lsandov1 added inline comments to D70866: Improve SLP code snippet.
Thu, Dec 5, 9:14 AM · Restricted Project
JDevlieghere accepted D71003: [lldb/DWARF] Switch to llvm location list parser.
Thu, Dec 5, 9:14 AM · Restricted Project
andwar updated the diff for D70812: [Aarch64][SVE] Add DAG combine rules for gather loads and sext/zext.

Remove a typo after rebase (uimm5s2 vs uim5s4)

Thu, Dec 5, 9:14 AM · Restricted Project
jasonliu added a comment to D70859: [XCOFF] fixed a bug of XCOFFObjectFile.cpp and adding new test case to verify one mergeable string for xcoffobjectfile.

I agree with David about splitting. There are 4 issues mentioned in this patch, and they are not related. (I'm not sure if 1 or 3 could be combine or not)
Putting them together makes it hard to review, and hard to determine if the test case actually covered the issue that's raised.
Let's make separate patches if possible.

Thu, Dec 5, 9:14 AM · Restricted Project
wmi accepted D70653: [AutoFDO] Properly merge context-sensitive profile of inlinee back to outlined function.

I rerun perf test and I don't see any performance change. Last run I saw very small improvement on latency in a benchmark. This is fine since the benchmark has some fluctuation by itself.

Thu, Dec 5, 9:14 AM · Restricted Project
jmorse added inline comments to D58453: [DebugInfo][CGP] Limit placeDbgValues movement of dbg.value intrinsics.
Thu, Dec 5, 9:06 AM · Restricted Project
spatel accepted D71048: [X86] Remove ProcIntelGLM/ProcIntelGLP/ProcIntelTRM and replace them with a single feature flag covers the two places they were used..

LGTM - thanks!

Thu, Dec 5, 9:06 AM · Restricted Project
aprantl added a comment to D71042: WIP: [DebugInfo] Ensure fallback artificial location is available for cleanups.

This will definitely work; Once you got the testcase, it might be good to check whether there is a more targeted root cause that we could fix and assert on EndLoc here.

Thu, Dec 5, 9:06 AM
xazax.hun committed rG8994d632c8d3: [scan-build-py] Set of small fixes (authored by xazax.hun).
[scan-build-py] Set of small fixes
Thu, Dec 5, 9:05 AM
tstellar committed rG432bf48c08da: [profile] Fix file contention causing dropped counts on Windows under -fprofile… (authored by hans).
[profile] Fix file contention causing dropped counts on Windows under -fprofile…
Thu, Dec 5, 9:05 AM
xazax.hun closed D70693: [scan-build-py] Set of small fixes.
Thu, Dec 5, 9:05 AM · Restricted Project
JDevlieghere accepted D71068: [Dsymutil][NFC] Move NonRelocatableStringpool into common CodeGen folder..

LGTM!

Thu, Dec 5, 9:05 AM · debug-info, Restricted Project
bcain added inline comments to D70919: [Hexagon] Avoid passing unsupported options to lld when -fuse-ld=lld is used.
Thu, Dec 5, 9:05 AM · Restricted Project
spatel added a comment to D71064: [InstCombine] Invert `add A, sext(B) --> sub A, zext(B)` canonicalization (to `sub A, zext B -> add A, sext B`).

Did you confirm that codegen is equal or better for these cases?
I think we have DAGCombiner reversals for this transform, but some targets that seem like they would benefit have not enabled the TLI hook.

I didn't yet. So, as far as i can tell, these are all the interesting cases:
(i'm looking at @t0_new_canon vs @t1_old_canon since that is the only change in this patch)
https://godbolt.org/z/vbb25Q - no regression for x86
https://godbolt.org/z/Htg7m5 - aarch64 also looks ok?
https://godbolt.org/z/xmP6JV - arm good?
https://godbolt.org/z/vNrNJ8 - thumb good?

So i'd say everything is already covered by backend undo folds?
Let me know if i'm missing the point here?

Thu, Dec 5, 9:05 AM · Restricted Project
jasonliu accepted D71032: [AIX] Make sure to use QualNames for external global objects.

LGTM with minor nit.

Thu, Dec 5, 9:05 AM · Restricted Project
ABataev added inline comments to D70726: [OpenMP50] Add parallel master construct.
Thu, Dec 5, 8:56 AM · Restricted Project
quic_aankit added a comment to D65326: Fix for a dangling point bug in DeadStoreElimination pass.

I'm trying to get access to commit the patch, but it would be great if you can get someone to commit it.

Thu, Dec 5, 8:55 AM · Restricted Project
Meinersbur accepted D71025: [LoopFusion] Restrict loop fusion to rotated loops..

[suggestion] Add a test case to check for "not rotated" to be rejected.

Thu, Dec 5, 8:55 AM · Restricted Project
sidneym created D71069: [ELF][Hexagon]Add TPREL relocation support to Hexagon.
Thu, Dec 5, 8:55 AM · Restricted Project
nikic added a comment to D68717: [Codegen] More add_sat and sub_sat promotion.

Sorry for the delay. Now that I have added some extra Neon codegen (and some extra tests), this is how this looks.

The unsigned cases still look like a clear win. The signed cases looks worse (although you could argue in cases that this would simplify if part of a loop).

WDYT? Remove the signed part and keep the unsigned?

Thu, Dec 5, 8:55 AM · Restricted Project
jmorse updated the diff for D58453: [DebugInfo][CGP] Limit placeDbgValues movement of dbg.value intrinsics.

Reshuffle the loop structure, early continue rather than an indented block. I've also moved the PHI/EHpad test ahead of the domination check for efficiency.

Thu, Dec 5, 8:55 AM · Restricted Project
Carrot updated the diff for D64376: [MBP] Avoid tail duplication if it can't bring benefit.
Thu, Dec 5, 8:55 AM · Restricted Project
balazske committed rGa9f10ebffaa2: [ASTImporter] Various source location and range import fixes. (authored by balazske).
[ASTImporter] Various source location and range import fixes.
Thu, Dec 5, 8:46 AM
ABataev committed rG5c517a6b13bf: [OPENMP50]Add support for if clause for simd part in parallel master taskloop… (authored by ABataev).
[OPENMP50]Add support for if clause for simd part in parallel master taskloop…
Thu, Dec 5, 8:46 AM
balazske closed D60499: [ASTImporter] Various source location and range import fixes..
Thu, Dec 5, 8:46 AM · Restricted Project
thopre edited parent revisions for D68779: [LNT][NFC] Fix global import in function, added: 1; removed: 1.
Thu, Dec 5, 8:46 AM
thopre added a child revision for D68938: [LNT] Python 3 support: subprocess IO as text: D68779: [LNT][NFC] Fix global import in function.
Thu, Dec 5, 8:46 AM
DiggerLin added inline comments to D70859: [XCOFF] fixed a bug of XCOFFObjectFile.cpp and adding new test case to verify one mergeable string for xcoffobjectfile.
Thu, Dec 5, 8:46 AM · Restricted Project
thopre removed a child revision for D69055: [LNT] Code cleanup: consistently use space after comma: D68779: [LNT][NFC] Fix global import in function.
Thu, Dec 5, 8:46 AM