[PATCH] D92930: [Clang] Add vcmla and rotated variants for Arm ACLE.

fhahn updated this revision to Diff 310632. fhahn added a comment. Updated to use new VCMLA_ROTS multiclass to define rotated variants. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D92930/new/ https://reviews.llvm.org/D92930 Files: clang/include

[PATCH] D92930: [Clang] Add vcmla and rotated variants for Arm ACLE.

fhahn created this revision. fhahn added reviewers: SjoerdMeijer, dmgreen, t.p.northover. Herald added a subscriber: kristof.beyls. fhahn requested review of this revision. Herald added a project: clang. Herald added a subscriber: cfe-commits. This patch adds vcmla and the rotated variants as defi

[PATCH] D92436: [Time-report] Add a flag -ftime-report={per-pass,per-pass-run} to control the pass timing aggregation

fhahn added a comment. I think this also fixes https://bugs.llvm.org/show_bug.cgi?id=47397 Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D92436/new/ https://reviews.llvm.org/D92436 ___ cfe-commits mailing

[PATCH] D91631: [Matrix] Adjust matrix pointer type for inline asm arguments.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG680931af2757: [Matrix] Adjust matrix pointer type for inline asm arguments. (authored by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINC

[PATCH] D91631: [Matrix] Adjust matrix pointer type for inline asm arguments.

fhahn added a comment. There's still one remaining issue. The backend will still crash when the resulting vector does not fit into the target's vector register directly. But that's the same as with `ext_vector_type` I think. The snippet below should produce the same `asm` call as a `float x 3 x

[PATCH] D91631: [Matrix] Adjust matrix pointer type for inline asm arguments.

fhahn created this revision. Herald added subscribers: cfe-commits, tschuett. Herald added a project: clang. fhahn requested review of this revision. Matrix types in memory are represented as arrays, but accessed through vector pointers, with the alignment specified on the access operation. For i

[PATCH] D91195: Add Annotation2MD pass to add !annotate metadata from llvm.global.annotations

fhahn marked an inline comment as done. fhahn added a comment. Thanks Jessica & Francis! I committed the initial version, with a small change to only run if the `!annotation` remarks are enabled. Comment at: llvm/include/llvm/Transforms/IPO/Annotation2Metadata.h:1 +//===- SCCP

[PATCH] D91195: Add Annotation2MD pass to add !annotate metadata from llvm.global.annotations

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG8dbe44cb2936: Add pass to add !annotate metadata from @llvm.global.annotations. (authored by fhahn). Herald added a project: clang. Herald added a su

[PATCH] D91417: [IRGen] Add !annotation metadata for auto-init stores.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGca2e7e59994d: [IRGen] Add !annotation metadata for auto-init stores. (authored by fhahn). Changed prior to commit: https://reviews.llvm.org/D91417

[PATCH] D91417: [IRGen] Add !annotation metadata for auto-init stores.

fhahn added inline comments. Comment at: clang/lib/CodeGen/CGDecl.cpp:1267 assert(!isa(constant)); - emitStoresForConstant(CGM, D, Loc, isVolatile, Builder, constant); + emitStoresForConstant(CGM, D, Loc, isVolatile, Builder, constant, true); } jdoerfert wr

[PATCH] D91417: [IRGen] Add !annotation metadata for auto-init stores.

fhahn created this revision. fhahn added reviewers: thegameg, paquette, jdoerfert, rjmccall. Herald added subscribers: cfe-commits, dexonsmith, jfb, hiraditya. Herald added projects: clang, LLVM. fhahn requested review of this revision. This patch updates Clang's IRGen to add !annotation nodes wit

[PATCH] D91253: [Matrix] Update mangling to use paramterized vendor ext type syntax.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG1d5daed1997d: [Matrix] Update mangling to use paramterized vendor ext type syntax. (authored by fhahn). Repository: rG LLVM Github Monorepo CHANG

[PATCH] D91253: [Matrix] Update mangling to use paramterized vendor ext type syntax.

fhahn marked an inline comment as done. fhahn added inline comments. Comment at: clang/lib/AST/ItaniumMangle.cpp:3523 mangleType(T->getElementType()); + Out << "I"; } rjmccall wrote: > Your `I`s and `E`s are backwards. Argh, used different orders for comment

[PATCH] D91253: [Matrix] Update mangling to use paramterized vendor ext type syntax.

fhahn updated this revision to Diff 304605. fhahn added a comment. Update comment and fix I and E order Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D91253/new/ https://reviews.llvm.org/D91253 Files: clang/lib/AST/ItaniumMangle.cpp clang/test/

[PATCH] D72281: [Matrix] Add matrix type to Clang.

fhahn added inline comments. Comment at: clang/lib/AST/ItaniumMangle.cpp:3353-3371 + auto &ASTCtx = getASTContext(); + unsigned BitWidth = ASTCtx.getTypeSize(ASTCtx.getSizeType()); + llvm::APSInt Rows(BitWidth); + Rows = T->getNumRows(); + mangleIntegerLiteral(ASTCtx.getSize

[PATCH] D91253: [Matrix] Update mangling to use paramterized vendor ext type syntax.

fhahn created this revision. fhahn added reviewers: rjmccall, rsmith, Bigcheese. Herald added a subscriber: tschuett. Herald added a project: clang. fhahn requested review of this revision. The Itanium CXX ABI grammer has been extended to support parameterized vendor extended types [1]. This patc

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2379892 , @thakis wrote: > Very belated update: We've been running with this on our clang/tot bots for a > few weeks now and turned this on (or, more accurately stopped turning it off) > for "normal" builds today. All tes

[PATCH] D89031: [SVE] Add support to vectorize_width loop pragma for scalable vectors

fhahn added inline comments. Comment at: clang/lib/Sema/SemaStmtAttr.cpp:144 + assert(ValueExpr && "Attribute must have a valid value expression."); + if (S.CheckLoopHintExpr(ValueExpr, St->getBeginLoc())) +return nullptr; sdesmalen wrote: > dav

[PATCH] D89699: [ExtVector] Make .even/.odd/.lo/.hi return vectors for single elements.

fhahn created this revision. fhahn added reviewers: rjmccall, jfb, scanon, rsmith, hfinkel. Herald added a subscriber: dexonsmith. Herald added a project: clang. fhahn requested review of this revision. Currently the 'swizzle' accessors .even/.odd/.lo/.hi return a scalar instead of a vector with a

[PATCH] D89031: [SVE] Add support to vectorize_width loop pragma for scalable vectors

fhahn added inline comments. Comment at: clang/lib/Parse/ParsePragma.cpp:1096 static_cast(Tok.getAnnotationValue()); - IdentifierInfo *PragmaNameInfo = Info->PragmaName.getIdentifierInfo(); nit: unrelated change? Comment at: clang/lib

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

fhahn added inline comments. Comment at: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll:82 + +!1 = distinct !{!1, !2, !3} +!2 = !{!"llvm.loop.vectorize.width", i32 4} fhahn wrote: > I don't think we need this. You can just pass `-force-vector

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

fhahn added inline comments. Comment at: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll:1 +; RUN: opt -vector-library=LIBMVEC -inject-tli-mappings -force-vector-width=4 -force-vector-interleave=1 -loop-vectorize -S < %s | FileCheck %s + spatel wrot

[PATCH] D88154: Initial support for vectorization using Libmvec (GLIBC vector math library).

fhahn added inline comments. Comment at: llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll:2 +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; RUN: opt -vector-library=LIBMVEC-X86 -inject-tli-mappings -loop-vectorize -S < %s | FileChe

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. I put up an impromptu round-table to discuss MemorySSA related topics, including potential new transformations like MemCpyOpt using MemorySSA: https://whova.com/portal/webapp/llvm_202010/CommunityBoard/topic/342030/ If there's any interest, I think it would be great to sy

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. FYI this has been temporarily reverted again roughly a week ago, because there seems to be a mis-compile, that is caused by LLVM introducing invalid lifetime.end calls and exposed by the more powerful DSE, which @asbirlea is currently trying to narrow down. Repository:

[PATCH] D88265: [Sema] Support Comma operator for fp16 vectors.

fhahn accepted this revision. fhahn added a comment. LGTM, thanks. I'll commit this in a minute. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D88265/new/ https://reviews.llvm.org/D88265 ___ cfe-commits m

[PATCH] D88265: [Sema] Support Comma operator for fp16 vectors.

This revision was automatically updated to reflect the committed changes. Closed by commit rG700e63293eea: [Sema] Support Comma operator for fp16 vectors. (authored by arames, committed by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D88265/n

[PATCH] D88265: Fix comma with half vectors.

fhahn added a comment. In D88265#2303841 , @arames wrote: > In D88265#2302653 , @fhahn wrote: > >> Could you adjust the commit message to be a bit more descriptive, e.g >> something like `[Sema] Support Comma opera

[PATCH] D88265: Fix comma with half vectors.

fhahn added a comment. Could you adjust the commit message to be a bit more descriptive, e.g something like `[Sema] Support Comma operator for fp16 vectors.` Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D88265/new/ https://reviews.llvm.org/D88265

[PATCH] D88265: Fix comma with half vectors.

fhahn added inline comments. Comment at: clang/lib/Sema/SemaExpr.cpp:13936 // arm64). - assert(isVector(RHS.get()->getType(), Context.HalfTy) == - isVector(LHS.get()->getType(), Context.HalfTy) && - "both sides are half vectors or neither sides are"); + asser

[PATCH] D87972: [OldPM] Pass manager: run SROA after (simple) loop unrolling

fhahn added a comment. In D87972#2284060 , @MaskRay wrote: > I have tested this patch internally and seen gains and losses. On one > document search related benchmark 3~5% improvement. One zippy (snappy) there > is 3~5% regression. Perhaps we do need a c

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2285984 , @dmajor wrote: > I threw as many tests as I could find at 9d172c8e9c84 > , and I > don't see any regressions compared to its parent revision.

[PATCH] D87972: [OldPM] Pass manager: run SROA after (simple) loop unrolling

fhahn added a comment. Surprising this causes only such a small perf regression. I guess it should be OK given that, but here probably are some pathological cases out there where this may cause some noticeable compile-time regressions. IIUC the additional cases this catches come mainly from ful

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. All reported problems should be fixed now and we also made the fix for the issue @dmajor reported more conservative to catch more problematic cases. I just enabled DSE + MemorySSA again in 9d172c8e9c84 .

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2276899 , @dmajor wrote: > In D87163#2275896 , @asbirlea wrote: > >> I checked in a fix in https://reviews.llvm.org/rGfc8200633122, but I haven't >> yet verified it addresses all

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2275058 , @asbirlea wrote: > Yes, the load should have the Phi as the defining access. I'm still looking > into where this information should come from, but it's not hitting the phi > translation. > Thank you for the reve

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2274549 , @dmajor wrote: > Reduced a bit more: https://godbolt.org/z/j59evK (C++) and > https://godbolt.org/z/8xG688 (IR) -- the store at line 43 of `while.end` has > been removed. Thanks! I reverted the change for now i

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2273917 , @thakis wrote: > It's not blocking us because we added an explicit flag to force this off to > our build config for "normal" builds. (We still get the default-on behavior > on our bots that build with trunk clan

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2272749 , @asbirlea wrote: > I'd suggest reverting if the failures are blocking, but we do need a > reproducer so it can be recommitted after a fix is in place. +1, but I think ideally we would have reproducers for the re

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2271998 , @dmajor wrote: > In D87163#2270871 , @fhahn wrote: > >> In D87163#2270444 , @thakis wrote: >> >>> Heads-up: We're seeing fairly wid

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2270444 , @thakis wrote: > Heads-up: We're seeing fairly widespread test failures with this in Chromium > that go away when we force this flag off. It's possible that it's due to only > a small number of issues and they m

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2270568 , @uabelho wrote: > Hi, > > I'm seeing what I think is a miscompile with this: Thanks for the reproducer. We have to be more careful around pointers that may reference multiple locations in memory. I pushed f715d8

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2269674 , @thakis wrote: > The fix-up breaks tests: http://45.33.8.238/linux/27687/step_12.txt Thanks, should already be fixed in d85ac6d577ac Reposi

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn added inline comments. Comment at: clang/include/clang/Driver/Options.td:4287-4290 +def fpass_by_value_noalias: Flag<["-"], "fpass-by-value-noalias">, + HelpText<"Allows assuming no references to passed by value escape before " + "transferring execution to the ca

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGa874d6334409: [Clang] Add option to allow marking pass-by-value args as noalias. (authored by fhahn). Changed prior to commit: https://reviews.llv

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn updated this revision to Diff 291183. fhahn marked an inline comment as done. fhahn added a comment. Add check lines for escape examples. Note that for return-by-value we already add noalias to the pointer unconditionally. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn marked an inline comment as done. fhahn added inline comments. Comment at: clang/lib/CodeGen/CGCall.cpp:2198 +// reference to the underlying object. Mark it accordingly. +Attrs.addAttribute(llvm::Attribute::NoAlias); + rjmccall wrote: > fhah

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn updated this revision to Diff 291176. fhahn added a comment. Update code to limit adding noalias to types with `getArgPassingRestrictions() == RecordDecl::APK_CanPassInRegs`. This rules out problematic types, like ones with non-trivial constructors or ObjC structs with weak pointers. Thank

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGfb109c42d91c: [DSE] Switch to MemorySSA-backed DSE by default. (authored by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn updated this revision to Diff 291083. fhahn added a comment. Also restrict to types that are `isNonTrivialToPrimitiveCopy`, so we do not add noalias for problematic types, like ObjC structs with weak references. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://re

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2266725 , @asbirlea wrote: > I'm running additional testing in the background. I don't have anything > holding this back at this point. > Either check in and if something shows up, I'll let you know to resolve or > revert

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn updated this revision to Diff 291078. fhahn added a comment. Final rebase after new memcpy test. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D87163/new/ https://reviews.llvm.org/D87163 Files: clang/test/CodeGen/thinlto-distributed-newpm.l

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. In D87163#2264785 , @asbirlea wrote: > I'm running into a crash with this, could you please hold off until tomorrow? > I'm working on getting a reproducer. @asbirlea Thanks for the reproducer, the issue should be sorted out. The GV

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added inline comments. Comment at: llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp:312 switch (II->getIntrinsicID()) { default: llvm_unreachable("doesn't pass 'hasAnalyzableMemoryWrite' predicate"); case Intrinsic::lifetime_end: asbirlea

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. It looks like a recently introduced GVN crash is exposed by this change (probably D87061 ). I will commit once this is fixed. In D87163#2263685 , @nikic wrote: > LGTM as well, per llvm-dev discussi

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. As another data point, I have collected compile-time numbers for ARM64 with -Os -g LTO & -Oz -g LTO. Geoman changes below. Compile time here is actual execution time on a stabilized system. Compile time code size -Oz RLTO 0.67%

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn updated this revision to Diff 290658. fhahn added a comment. rebase on top of latest changes. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D87163/new/ https://reviews.llvm.org/D87163 Files: clang/test/CodeGen/thinlto-distributed-newpm.ll

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added a comment. Note that we now also have to compute the PostDominatorTree, which adds an extra bit of compile-time overhead. By adjusting the pipeline a bit more, we can re-use ADCE's PDTs in most cases, which gives a `-0.18%` geomean improvement for -O3 D87322

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn added inline comments. Comment at: clang/include/clang/Driver/Options.td:4287-4290 +def fpass_by_value_noalias: Flag<["-"], "fpass-by-value-noalias">, + HelpText<"Allows assuming no references to passed by value escape before " + "transferring execution to the ca

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn updated this revision to Diff 290504. fhahn marked an inline comment as done. fhahn added a comment. Change to codegen option, adjust description for option, limit to trivially copyable types in C++, add corresponding test cases. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST A

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn added inline comments. Comment at: llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp:115 cl::desc("The number of memory instructions to scan for " "dead store elimination (default = 100)")); static cl::opt MemoryS

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn updated this revision to Diff 290140. fhahn added a comment. Adjust limit in description, thanks! Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D87163/new/ https://reviews.llvm.org/D87163 Files: clang/test/CodeGen/thinlto-distributed-newpm.

[PATCH] D87163: [DSE] Switch to MemorySSA-backed DSE by default.

fhahn created this revision. fhahn added reviewers: efriedma, dmgreen, asbirlea, reames. Herald added subscribers: cfe-commits, lxfind, nikic, kerbowa, jfb, dexonsmith, steven_wu, george.burgess.iv, modocache, hiraditya, Prazek, nhaehnle, jvesely. Herald added projects: clang, LLVM. fhahn requeste

[PATCH] D87102: [Sema] Fix a -Warc-repeated-use-of-weak false-positive by only calling CheckPlaceholderExpr once

fhahn accepted this revision. fhahn added a comment. LGTM, thanks for addressing this! Please wait a bit with committing, in case there are additional comments. Comment at: clang/lib/Sema/SemaExpr.cpp:4598 } // If the base is either a MatrixSubscriptExpr or a matrix type

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn added a comment. ping :) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D85473/new/ https://reviews.llvm.org/D85473 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn added a comment. ping :) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D85473/new/ https://reviews.llvm.org/D85473 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-

[PATCH] D86274: [Constants] Handle FNeg in getWithOperands.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rGbc72a3ab949e: [Constants] Handle FNeg in getWithOperands. (authored by fhahn). Changed prior to commit: https://reviews.llvm.org/D86274?vs=286756&

[PATCH] D86274: [Constants] Handle FNeg in getWithOperands.

fhahn created this revision. fhahn added reviewers: efriedma, aprantl, arsenm. Herald added subscribers: cfe-commits, hiraditya. Herald added projects: clang, LLVM. fhahn requested review of this revision. Herald added a subscriber: wdng. Currently ConstantExpr::getWithOperands does not handle FNe

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn added a comment. ping Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D85473/new/ https://reviews.llvm.org/D85473 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin

[PATCH] D84886: Create LoopNestPass

fhahn added a comment. In D84886#2212202 , @TaWeiTu wrote: > @ychen Again, thanks for your comment! > > 1. Currently, `LoopInterchange` returns immediately if the loop is not a > top-level one. The main purpose of the loop nest pass is to prevent > situa

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn updated this revision to Diff 283837. fhahn added a comment. Thanks for taking a look! I adjusted the LangOpts description and added the flag to f_Group, as suggested. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D85473/new/ https://reviews.

[PATCH] D85473: [Clang] Add option to allow marking pass-by-value args as noalias.

fhahn created this revision. fhahn added reviewers: rjmccall, rsmith, hfinkel, jfb. Herald added subscribers: dang, dexonsmith. Herald added a project: clang. fhahn requested review of this revision. After the recent discussion on cfe-dev 'Can indirect class parameters be noalias?' [1], it seems l

[PATCH] D74436: Change clang option -ffp-model=precise to select ffp-contract=on

fhahn updated this revision to Diff 282577. fhahn added a comment. Rebase Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74436/new/ https://reviews.llvm.org/D74436 Files: clang/docs/UsersManual.rst clang/lib/Driver/ToolChains/Clang.cpp clang/

[PATCH] D74436: Change clang option -ffp-model=precise to select ffp-contract=on

fhahn reopened this revision. fhahn added a comment. This revision is now accepted and ready to land. IIUC the patch is currently reverted (in 78654e8511cf16d49f6680d782f3771a767ba942 ), due to ~20 llvm-test-suite failures. T

[PATCH] D84180: [Matrix] Add LowerMatrixIntrinsics to the NPM

fhahn added a comment. Awesome, thanks! Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D84180/new/ https://reviews.llvm.org/D84180 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.ll

[PATCH] D67833: [OpenMP 5.0] Codegen support to pass user-defined mapper functions to runtime

fhahn added a comment. I think this change introduced the following warnings, where `auto &` is used for types that are always copied. It would be great if you could take a look. lvm-project/clang/lib/CodeGen/CGOpenMPRuntime.cpp:8011:24: warning: loop variable 'L' is always a copy because the

[PATCH] D80911: AArch64+ARM: make LLVM consider system registers volatile to prevent unsound optimizations.

fhahn accepted this revision. fhahn added a comment. This revision is now accepted and ready to land. As mentioned earlier, I think the patch is a nice, isolated fix for the `read_register` issue and it should not make addressing the bigger issue more difficult. LGTM, thanks! Repository: rG

[PATCH] D80911: AArch64+ARM: make LLVM consider system registers volatile to prevent unsound optimizations.

fhahn added reviewers: fhahn, efriedma, hfinkel. fhahn added a comment. This seems similar to the problem described in https://bugs.llvm.org/show_bug.cgi?id=46210 : for intrinsics that are marked as reading memory, we do not account for the fact that the memory could be modified outside of the

[PATCH] D82928: [Coroutines] Fix code coverage for coroutine

fhahn added a comment. Looks like this causes a bunch of build bot failures, e.g http://lab.llvm.org:8011/builders/clang-x86_64-debian-fast/builds/31465 b It would be great if you could take a look. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D8

[PATCH] D82473: [Matrix] Use 1st/2nd instead of first/second in matrix diags.

This revision was automatically updated to reflect the committed changes. Closed by commit rG043b60839955: [Matrix] Use 1st/2nd instead of first/second in matrix diags. (authored by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D82473/new/ ht

[PATCH] D82473: [Matrix] Use 1st/2nd instead of first/second in matrix diags.

fhahn created this revision. fhahn added reviewers: rjmccall, jfb, Bigcheese. Herald added subscribers: cfe-commits, tschuett, dexonsmith. Herald added a project: clang. This was suggested in D72782 and brings the diagnostics more in line with how argument refere

[PATCH] D80279: [libclang] Extend clang_Cursor_Evaluate().

This revision was automatically updated to reflect the committed changes. Closed by commit rG72131423cc95: [libclang] Extend clang_Cursor_Evaluate(). (authored by ckandeler, committed by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D80279/new

[PATCH] D82213: [Remarks] Add callsite locations to inline remarks

fhahn added a comment. That's interesting. We are also using something similar for the matrix lowering remarks [1]: we traverse the inlining chain bottom up and emit a remark at each step which contains the expression available at that level. I think those approaches could be useful in general

[PATCH] D72782: [Matrix] Add __builtin_matrix_column_store to Clang.

This revision was automatically updated to reflect the committed changes. Closed by commit rGb5e082e72891: [Matrix] Add __builtin_matrix_column_store to Clang. (authored by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D72782/new/ https://rev

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

This revision was automatically updated to reflect the committed changes. Closed by commit rG934bcaf10b57: [Matrix] Add __builtin_matrix_column_load to Clang. (authored by fhahn). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D72781/new/ https://revi

[PATCH] D72782: [Matrix] Add __builtin_matrix_column_store to Clang.

fhahn marked an inline comment as done. fhahn added inline comments. Comment at: clang/include/clang/Basic/DiagnosticSemaKinds.td:10792 def err_builtin_matrix_pointer_arg: Error< - "%select{first|second}0 argument must be a pointer to a valid matrix element type">; + "%0 argu

[PATCH] D72782: [Matrix] Add __builtin_matrix_column_store to Clang.

fhahn updated this revision to Diff 271617. fhahn added a comment. Rebased after recent changes landed upstream. Will commit shortly Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D72782/new/ https://reviews.llvm.org/D72782 Files: clang/include/cl

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn updated this revision to Diff 271613. fhahn added a comment. Rebased after recent parent patches landed. Will commit shortly. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D72781/new/ https://reviews.llvm.org/D72781 Files: clang/include/cla

[PATCH] D72782: [Matrix] Add __builtin_matrix_column_store to Clang.

fhahn updated this revision to Diff 271401. fhahn added a comment. Ping :) Rebased and applied feedback from D72781 . Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D72782/new/ https://reviews.llvm.org/D72782 Files:

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn added a comment. In D72781#2084077 , @rjmccall wrote: > LGTM. Thank you very much again John! This patch is pending on a few smallish improvements to the load/store intrinsics (D81472 ) and I'll land once that one

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn updated this revision to Diff 269556. fhahn marked 2 inline comments as done. fhahn added a comment. Adjust naming as suggested, pass through volatile flag. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D72781/new/ https://reviews.llvm.org/D72

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn marked 3 inline comments as done. fhahn added inline comments. Comment at: clang/include/clang/Sema/Sema.h:12126 + ExprResult SemaBuiltinMatrixColumnMajorLoadOverload(CallExpr *TheCall, + ExprResult CallResult); --

[PATCH] D72778: [Matrix] Add __builtin_matrix_transpose to Clang.

fhahn added a comment. In D72778#2080928 , @rjmccall wrote: > Yeah, that's fine. Thank you very much John! I filed https://bugs.llvm.org/show_bug.cgi?id=46251 to keep track of the initializer support, with a note on enabling the commented out test. R

[PATCH] D72778: [Matrix] Add __builtin_matrix_transpose to Clang.

This revision was automatically updated to reflect the committed changes. Closed by commit rG3323a628ec82: [Matrix] Add __builtin_matrix_transpose to Clang. (authored by fhahn). Changed prior to commit: https://reviews.llvm.org/D72778?vs=269050&id=269463#toc Repository: rG LLVM Github Monore

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn marked 9 inline comments as done. fhahn added inline comments. Comment at: clang/include/clang/Basic/DiagnosticSemaKinds.td:10789 +def err_builtin_matrix_scalar_constant_unsigned_expr_arg: Error< + "%0 argument must be a constant unsigned integer expression">; + --

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn updated this revision to Diff 269378. fhahn marked 5 inline comments as done. fhahn added a comment. Herald added a project: LLVM. Herald added a subscriber: llvm-commits. Simplified code as suggested, check if matrix type extensions is enabled (and add test) and set align attribute for poi

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn marked 12 inline comments as done. fhahn added inline comments. Comment at: clang/lib/Sema/SemaChecking.cpp:15107 + { +ExprResult PtrConv = DefaultLvalueConversion(PtrExpr); +if (PtrConv.isInvalid()) rjmccall wrote: > You should be doing `DefaultFu

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn updated this revision to Diff 269363. fhahn added a comment. Updated to - use DefaultFunctionArrayLvalueConversion for pointer conversion, use getAs subsequently - return Context.DependentTy if any part of the result matrix type is still type-dependent - add assertion & todo for volatile

[PATCH] D72778: [Matrix] Add __builtin_matrix_transpose to Clang.

fhahn marked 2 inline comments as done. fhahn added inline comments. Comment at: clang/lib/Sema/SemaChecking.cpp:15051 + + Expr *Arg = TheCall->getArg(0); + if (!Arg->getType()->isConstantMatrixType()) { rjmccall wrote: > fhahn wrote: > > rjmccall wrote: > > >

[PATCH] D72781: [Matrix] Add __builtin_matrix_column_load to Clang.

fhahn updated this revision to Diff 269249. fhahn added a comment. Ping. Applied feedback from D72778 to this patch, improved tests, support conversions/placeholders. One thing I am not sure is how to properly handle template substitutions for the pointer expr

<    1   2   3   4   5   6   7   8   >