OSDN Git Service
Roman Lebedev [Mon, 6 Jul 2020 09:38:41 +0000 (12:38 +0300)]
[InstCombine] Lower infinite combine loop detection thresholds
Summary:
1000 iteratons is still kinda a lot.
Would it make sense to iteratively lower it, until it becomes `2`,
with some delay inbetween in order to let users actually potentially encounter it?
Reviewers: spatel, nikic, kuhar
Reviewed By: nikic
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83160
Stephan Herhut [Mon, 6 Jul 2020 10:18:17 +0000 (12:18 +0200)]
Revert "Revert "Revert "[libcxx] Put clang::trivial_abi on std::unique_ptr, std::shared_ptr, and std::weak_ptr"""
This reverts commit
f706b01a00676ef0e7aefb253316c6418f022fa2.
David Green [Mon, 6 Jul 2020 07:59:05 +0000 (08:59 +0100)]
[ARM] Use BaseT::getMemoryOpCost for getMemoryOpCost
This alters getMemoryOpCost to use the Base TargetTransformInfo version
that includes some additional checks for whether extending loads are
legal. This will generally have the effect of making <2 x ..> and some
<4 x ..> loads/stores more expensive, which in turn should help favour
larger vector factors.
Notably it alters the cost of a <4 x half>, which with the current
codegen will be expensive if it is not extended.
Differential Revision: https://reviews.llvm.org/D82456
Bevin Hansson [Wed, 1 Jul 2020 13:28:48 +0000 (15:28 +0200)]
Fix indentation in FixedPoint.h. NFC.
Guillaume Chatelet [Mon, 6 Jul 2020 08:47:58 +0000 (08:47 +0000)]
Fix off by one error in Bitfields
Differential Revision: https://reviews.llvm.org/D83192
Guillaume Chatelet [Mon, 6 Jul 2020 08:44:49 +0000 (08:44 +0000)]
Fix 46594 - Alignment assertion failure in instcombine
Pavel Labath [Mon, 6 Jul 2020 08:32:55 +0000 (10:32 +0200)]
[lldb/Utility] Merge Scalar::Get(Value)TypeAsCString
Kazushi (Jam) Marukawa [Mon, 6 Jul 2020 07:46:52 +0000 (16:46 +0900)]
[VE] Correct stack alignment
Summary:
Change stack alignment from 64 bits to 128 bits to follow ABI correctly.
And add a regression test for datalayout.
Reviewers: simoll, k-ishizaka
Reviewed By: simoll
Subscribers: hiraditya, cfe-commits, llvm-commits
Tags: #llvm, #ve, #clang
Differential Revision: https://reviews.llvm.org/D83173
Pavel Labath [Mon, 6 Jul 2020 08:09:13 +0000 (10:09 +0200)]
[lldb/DWARF] Look for complete array element definitions in other modules
This applies the same logic we have for incomplete class bases and
members to array element types.
Chuanqi Xu [Mon, 6 Jul 2020 05:39:29 +0000 (13:39 +0800)]
[Coroutines] Warning if return type of coroutine_handle::address is not void*
User can own a version of coroutine_handle::address() whose return type is not
void* by using template specialization for coroutine_handle<> for some
promise_type.
In this case, the codes may violate the capability with existing async C APIs
that accepted a void* data parameter which was then passed back to the
user-provided callback.
Patch by ChuanqiXu
Differential Revision: https://reviews.llvm.org/D82442
Petr Hosek [Tue, 9 Jun 2020 22:45:31 +0000 (15:45 -0700)]
[Fuchsia] Set projects and runtimes in the cache file
We make assumptions about what projects and runtimes are enabled
when configuring our toolchain build, so we should enable those in
the cache file as well rather than relying on those being set
externally.
Differential Revision: https://reviews.llvm.org/D81514
Joachim Protze [Thu, 2 Jul 2020 19:29:02 +0000 (21:29 +0200)]
[OpenMP][Tests] Flag compatibility of OpenMP runtime tests with GCC versions
If the compilation fails, the test is marked as unsupported.
-> This will never change for a specific version of gcc
If the linking fails, the test is marked as expected to fail.
-> This might change as LLVM/OpenMP implements the missing GOMP interface function
Reviewed by: Hahnfeld
Differential Revision: https://reviews.llvm.org/D83077
Joachim Protze [Sun, 5 Jul 2020 06:46:31 +0000 (08:46 +0200)]
[OpenMP][OMPT] Fix ifdefs for OMPT code
Fixes build with LIBOMP_OMPT_SUPPORT=off
Reported by: Jason Edson
Reviewed by: Hahnfeld
Differential Revision: https://reviews.llvm.org/D83171
Fangrui Song [Sun, 5 Jul 2020 19:50:31 +0000 (12:50 -0700)]
[libomptarget][test] Fix text relocations by adding -fPIC
Nikita Popov [Sun, 5 Jul 2020 19:36:01 +0000 (21:36 +0200)]
[SCCP] Add test for range metadata (NFC)
Fangrui Song [Sun, 5 Jul 2020 18:47:02 +0000 (11:47 -0700)]
[Driver] Pass -gno-column-info instead of -dwarf-column-info
Making -g[no-]column-info opt out reduces the length of a typical CC1 command line.
Additionally, in a non-debug compile, we won't see -dwarf-column-info.
Vy Nguyen [Sat, 4 Jul 2020 15:41:25 +0000 (11:41 -0400)]
Revert "Revert "[libcxx] Put clang::trivial_abi on std::unique_ptr, std::shared_ptr, and std::weak_ptr""
This reverts commit
dc13ac02800220a33828ac3da629f382ca7e308d.
Rolling forward + fix typos and unused variables in tests
Differential Revision: https://reviews.llvm.org/D82490
David Green [Sun, 5 Jul 2020 13:54:12 +0000 (14:54 +0100)]
[ARM] Remove hasSideEffects from FP converts
Whether an instruction is deemed to have side effects in determined by
whether it has a tblgen pattern that emits a single instruction.
Because of the way a lot of the the vcvt instructions are specified
either in dagtodag code or with patterns that emit multiple
instructions, they don't get marked as not having side effects.
This just marks them as not having side effects manually. It can help
especially with instruction scheduling, to not create artificial
barriers, but one of these tests also managed to produce fewer
instructions.
Differential Revision: https://reviews.llvm.org/D81639
Simon Pilgrim [Sun, 5 Jul 2020 15:17:56 +0000 (16:17 +0100)]
[X86][SSE] Add PACKSS/PACKUS style patterns tests
Similar to the proposed generic code generated by D61129 - there's still some shuffle combining improvements to go before that patch is ready.
Benjamin Kramer [Sun, 5 Jul 2020 15:09:08 +0000 (17:09 +0200)]
[clangd] Fix stack-use-after-scope
Found by asan.
Alexander Belyaev [Sun, 5 Jul 2020 10:33:53 +0000 (12:33 +0200)]
[llvm] Cast to (void) the unused variable.
Nathan James [Sun, 5 Jul 2020 10:16:40 +0000 (11:16 +0100)]
[NFC] Clean up braces and anon namespace
Raphael Isemann [Sun, 5 Jul 2020 08:54:58 +0000 (10:54 +0200)]
[lldb][NFC] Clarify that spaces can't be in breakpoint names
We already check for spaces but we don't mention it in the error message.
Raphael Isemann [Sun, 5 Jul 2020 08:54:15 +0000 (10:54 +0200)]
[lldb][NFC] Remove an unnecessary cast in SBCommandInterpreter
Joachim Protze [Sun, 5 Jul 2020 07:16:09 +0000 (09:16 +0200)]
[OpenMP][Tests] NFC use type macro in printf
Fangrui Song [Sun, 5 Jul 2020 06:26:57 +0000 (23:26 -0700)]
Add tests for clang -fno-zero-initialized-in-bss and llc -nozero-initialized-in-bss
And rename the CC1 option.
Kai Luo [Sun, 5 Jul 2020 03:40:40 +0000 (03:40 +0000)]
[clang][PowerPC] Enable -fstack-clash-protection option for ppc64
Differential Revision: https://reviews.llvm.org/D81355
Georgy Komarov [Wed, 1 Jul 2020 07:15:39 +0000 (10:15 +0300)]
[llvm-objcopy] Fix crash when removing symbol table at same time as adding a symbol
This patch resolves crash that occurs when user wanted to remove all
symbols and add a brand new one using:
```
llvm-objcopy -R .symtab --add-symbol foo=1234 in.o out.o
```
Before these changes the symbol table internally being null when adding
new symbols. For now we will regenerate symtab in this case.
This fixes: https://bugs.llvm.org/show_bug.cgi?id=43930
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D82935
Thomas Lively [Sun, 5 Jul 2020 01:11:24 +0000 (18:11 -0700)]
[WebAssembly] Do not assume br_table range checks will be gt_u
OSS-Fuzz and the Emscripten test suite uncovered some edge cases in
which the range check instruction seemed to be an (i32.const 0) or
other unexpected instruction, triggering an assertion. Unfortunately
the reproducers are rather complicated, so they don't make good unit
tests. This commit removes the bad assertion and conservatively
optimizes range checks only when the range check instruction is
i32.gt_u.
Differential Revision: https://reviews.llvm.org/D83169
Nico Weber [Sun, 5 Jul 2020 00:41:27 +0000 (20:41 -0400)]
Fix two -Wrange-loop-analysis warnings that Xcode 12 beta incorrectly complains about
Xcode 12 beta apparently has the Wrange-loop-analysis changes from
half a year ago, but it seems to lack https://reviews.llvm.org/D72212
which made the warning usable again.
Nico Weber [Sun, 5 Jul 2020 00:26:39 +0000 (20:26 -0400)]
[gn build] fix link of libclang_rt.asan_osx_dynamic.dylib if command line tools are not installed
Nico Weber [Fri, 3 Jul 2020 17:52:18 +0000 (13:52 -0400)]
[gn build] make stage2_unix_toolchain set clang_base_path
This fixes the build of compiler-rt on macOS when _not_ using
clang_base_path in args.gn: Xcode clang knows where to find the
SDK, but regular clang doesn't and needs a -isysroot parameter.
We correctly add that parameter when clang_base_path is set,
but else we omit it. If clang_base_path was not set, we also
didn't add the flag for stage2_unix_toolchain() when we build
compiler-rt with just-built clang.
Make stage2_unix_toolchain() use clang_base_path instead of setting
cc / cxx. It's less code, and it gets things like this right.
Roman Lebedev [Sat, 4 Jul 2020 22:01:46 +0000 (01:01 +0300)]
[llvm-reduce] extractGVsFromModule(): don't crash when deleting instr twice
As it can be seen in newly-added (previously-crashing) test-case,
there can be a situation where multiple GV's are used in instr,
and we would schedule the same instruction to be deleted several times,
crashing when trying to delete it the second time.
We could either store WeakVH (done here), or use something set-like.
I think using WeakVH is prevalent in these cases elsewhere.
Roman Lebedev [Sat, 4 Jul 2020 21:48:05 +0000 (00:48 +0300)]
[llvm-reduce] extractArgumentsFromModule(): don't crash when deleting instr twice
As it can be seen in newly-added (previously-crashing) test-case,
there can be a situation where multiple arguments are used in instr,
and we would schedule the same instruction to be deleted several times,
crashing when trying to delete it the second time.
We could either store WeakVH (done here), or use something set-like.
I think using WeakVH is prevalent in these cases elsewhere.
Craig Topper [Sat, 4 Jul 2020 21:35:49 +0000 (14:35 -0700)]
[DAGCombiner] visitSIGN_EXTEND_INREG should fold sext_vector_inreg(undef) to 0 not undef.
We need to ensure that the sign bits of the result all match
so we can't fold to undef.
Similar to PR46585.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D83163
sstefan1 [Mon, 29 Jun 2020 22:08:37 +0000 (00:08 +0200)]
[OpenMPOpt] ICV Tracking
This is the first and most basic ICV Tracking implementation. For this
first version, we only support deduplication within the same BB.
Reviewers: jdoerfert, JonChesterfield, hamax97, jhuber6, uenoku,
baziotis
Differential Revision: https://reviews.llvm.org/D81788
Mehdi Amini [Sat, 4 Jul 2020 20:55:47 +0000 (20:55 +0000)]
Revert "[MLIR] Parallelize affine.for op to 1-D affine.parallel op"
This reverts commit
5f2843857feee6fbf755c12c21698a4987eda5d1.
This broke the build when -DDBUILD_SHARED_LIBS=ON is used.
Roman Lebedev [Sat, 4 Jul 2020 19:45:41 +0000 (22:45 +0300)]
Revert "[AssumeBundles] Use operand bundles to encode alignment assumptions"
Assume bundle can have more than one entry with the same name,
but at least AlignmentFromAssumptionsPass::extractAlignmentInfo() uses
getOperandBundle("align"), which internally assumes that it isn't the
case, and happily crashes otherwise.
Minimal reduced reproducer: run `opt -alignment-from-assumptions` on
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
%0 = type { i64, %1*, i8*, i64, %2, i32, %3*, i8* }
%1 = type opaque
%2 = type { i8, i8, i16 }
%3 = type { i32, i32, i32, i32 }
; Function Attrs: nounwind
define i32 @f(%0* noalias nocapture readonly %arg, %0* noalias %arg1) local_unnamed_addr #0 {
bb:
call void @llvm.assume(i1 true) [ "align"(%0* %arg, i64 8), "align"(%0* %arg1, i64 8) ]
ret i32 0
}
; Function Attrs: nounwind willreturn
declare void @llvm.assume(i1) #1
attributes #0 = { nounwind "reciprocal-estimates"="none" }
attributes #1 = { nounwind willreturn }
This is what we'd have with -mllvm -enable-knowledge-retention
This reverts commit
c95ffadb2474a4d8c4f598d94d35a9f31d9606cb.
peter klausler [Sat, 4 Jul 2020 18:06:28 +0000 (11:06 -0700)]
[flang] Add fixes and missing stmts to I/O runtime API handlers (ext. I/O part 8)
Complete the rework of the initial implementation of external I/O
to fix problems found in testing (tests to follow); add handlers
for hitherto unimplemented PAUSE, FLUSH, ENDFILE, BACKSPACE, and
REWIND statements.
Reviewed By: tskeith, sscalpone
Differential Revision: https://reviews.llvm.org/D83162
Craig Topper [Sat, 4 Jul 2020 18:42:00 +0000 (11:42 -0700)]
[DAGCombiner] Don't fold zext_vector_inreg/sext_vector_inreg(undef) to undef. Fold to 0.
zext_vector_inreg needs to produces 0s in the extended bits and
sext_vector_inreg needs to produce upper bits that are all the
same. So we should fold them to a 0 vector instead of undef.
Fixes PR46585.
Craig Topper [Sat, 4 Jul 2020 18:40:40 +0000 (11:40 -0700)]
[X86] Add test caes for pr46585. NFC
Roman Lebedev [Sat, 4 Jul 2020 17:04:28 +0000 (20:04 +0300)]
[Utils] Make -assume-builder/-assume-simplify actually work on Old-PM
clang w/ old-pm currently would simply crash
when -mllvm -enable-knowledge-retention=true is specified.
Clearly, these two passes had no Old-PM test coverage,
which would have shown the problem - not requiring AssumptionCacheTracker,
but then trying to always get it.
Also, why try to get domtree only if it's cached,
but at the same time marking it as required?
peter klausler [Sat, 4 Jul 2020 00:57:28 +0000 (17:57 -0700)]
[flang] Extend & fix per-I/O-statement state (ext. I/O work part 7)
The per-I/O-statement state structures need to support missing
external I/O statements, and some bugs found in testing with
formatted input and record advancement are fixed. The effects
of these changes will not be visible until further patches to
the I/O API handlers are pushed.
Reviewed By: tskeith
Differential Revision: https://reviews.llvm.org/D83151
Amara Emerson [Sat, 4 Jul 2020 17:41:27 +0000 (10:41 -0700)]
[profile] Mark gcov test as unsupported on Darwin.
Craig Topper [Sat, 4 Jul 2020 17:26:56 +0000 (10:26 -0700)]
[X86] Teach lowerShuffleAsBlend to use bit blend for v16i8/v32i8/v16i16 when avx512vl is enabled but not avx512bw.
Probably not super important since there are no real CPUs with
avx512vl and not avx512bw. But vpternlog should be better than
vblendvb.
I do wonder if we should use vpternlog even with BWI. We
currently use vblendmb or vpblendmw by putting the mask into a GPR
and moving it to a k-register. But I don't think we hoist the
GPR to k-register copy in machine LICM. Using VPTERNLOG would use
a constant pool load, but has the advantage that we're pretty good
at hoisting and rematerializing those.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D83156
Craig Topper [Sat, 4 Jul 2020 16:59:52 +0000 (09:59 -0700)]
[X86] Disable VPBLENDVB formation in combineLogicBlendIntoPBLENDV if VPTERNLOG is supported.
VPBLENDVB is multiple uops while VPTERNLOG is a single uop. So
we should use that instead.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D83155
Vy Nguyen [Sat, 4 Jul 2020 15:29:08 +0000 (11:29 -0400)]
Revert "[libcxx] Put clang::trivial_abi on std::unique_ptr, std::shared_ptr, and std::weak_ptr"
This reverts commit
5cde3c9633fd071c90e9f9ce54a002e78fdd9df9.
The tests were reported failing on clang10
Roman Lebedev [Sat, 4 Jul 2020 15:26:06 +0000 (18:26 +0300)]
[clang] Fix two tests that are affected by llvm opt change
Sanjay Patel [Sat, 4 Jul 2020 15:09:14 +0000 (11:09 -0400)]
[InstCombine] fix miscompile from umul_with_overflow matching
As noted in PR46561:
https://bugs.llvm.org/show_bug.cgi?id=46561
...it takes something beyond a minimal IR example to trigger
this bug because it relies on matching non-canonical IR.
There are no tests that show the need for matching this
pattern, so I'm just deleting it to fix the miscompile.
Roman Lebedev [Sat, 4 Jul 2020 14:39:48 +0000 (17:39 +0300)]
[InstCombine] Always try to invert non-canonical predicate of an icmp
Summary:
The actual transform i was going after was:
https://rise4fun.com/Alive/Tp9H
```
Name: zz
Pre: isPowerOf2(C0) && isPowerOf2(C1) && C1 == C0
%t0 = and i8 %x, C0
%r = icmp eq i8 %t0, C1
=>
%t = icmp eq i8 %t0, 0
%r = xor i1 %t, -1
Name: zz
Pre: isPowerOf2(C0)
%t0 = and i8 %x, C0
%r = icmp ne i8 %t0, 0
=>
%t = icmp eq i8 %t0, 0
%r = xor i1 %t, -1
```
but as it can be seen from the current tests, we already canonicalize most of it,
and we are only missing handling multi-use non-canonical icmp predicates.
If we have both `!=0` and `==0`, even though we can CSE them,
we end up being stuck with them. We should canonicalize to the `==0`.
I believe this is one of the cleanup steps i'll need after `-scalarizer`
if i end up proceeding with my WIP alloca promotion helper pass.
Reviewers: spatel, jdoerfert, nikic
Reviewed By: nikic
Subscribers: zzheng, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83139
Sanjay Patel [Sat, 4 Jul 2020 14:44:27 +0000 (10:44 -0400)]
[InstCombine] improve debug value names; NFC
The use of 'tmp' can trigger warnings from the update_test_checks.py
script. That's evidence of a flaw in the script's logic, but we
can always do better than naming variables 'tmp' in LLVM too.
The phi test file should be updated with auto-generated regex CHECK
lines, so it isn't affected by cosmetic diffs, but I don't have
time to do that right now.
Sanjay Patel [Sat, 4 Jul 2020 14:08:44 +0000 (10:08 -0400)]
[InstCombine] add test for miscompile (PR46561); NFC
Simon Pilgrim [Sat, 4 Jul 2020 14:27:55 +0000 (15:27 +0100)]
[DAG] matchBinOpReduction - match subvector reduction patterns beyond a matched shufflevector reduction
Currently matchBinOpReduction only handles shufflevector reduction patterns, but in many cases these only occur in the final stages of a reduction, once we're down to legal vector widths.
Before this its likely that we are performing reductions using subvector extractions to repeatedly split the source vector in half and perform the binop on the halves.
Assuming we've found a non-partial reduction, this patch continues looking for subvector reductions as far as it can beyond the last shufflevector.
Fixes PR37890
David Zarzycki [Sat, 4 Jul 2020 14:13:46 +0000 (10:13 -0400)]
[libcxx testing] Remove ALLOW_RETRIES from another test
Simon Pilgrim [Sat, 4 Jul 2020 13:42:20 +0000 (14:42 +0100)]
[X86][SSE] Add add/fadd reduction shuffle+subvector tests
Tests based on the PR37890 test cases - the vector combine pass should leave us with a reduction chain ending in extract(add(x,shuffle(x,1,-1,...))), but the higher reduction stages will be subvector extractions not shuffles.
Yash Jain [Sat, 4 Jul 2020 13:22:46 +0000 (18:52 +0530)]
[MLIR] Parallelize affine.for op to 1-D affine.parallel op
Introduce pass to convert parallel affine.for op into 1-D
affine.parallel op. Run using --affine-parallelize. Removes
test-detect-parallel: pass for checking parallel affine.for ops.
Differential Revision: https://reviews.llvm.org/D82672
Simon Pilgrim [Sat, 4 Jul 2020 12:54:30 +0000 (13:54 +0100)]
[X86][AVX] Fold PACK(LOSUBVECTOR(SHUFFLE(X)),HISUBVECTOR(SHUFFLE(X))) -> SHUFFLE(PACK(LOSUBVECTOR(X),HISUBVECTOR(X)))
Using PACK for truncations leaves us with intermediate shuffles that can be tricky to remove while the truncation tree is being formed.
This fold helps pull out the PERMQ case which is one of the most common, avoiding some costly lane-crossing shuffles.
A future patch will begin adding more general shuffle folding, which we should be able to use for HADD/HSUB as well.
LLVM GN Syncbot [Sat, 4 Jul 2020 12:02:31 +0000 (12:02 +0000)]
[gn build] Port
b6cbe6cb039
Kirstóf Umann [Thu, 18 Jun 2020 19:40:43 +0000 (21:40 +0200)]
[analyzer][NFC] Move the data structures from CheckerRegistry to the Core library
If you were around the analyzer for a while now, you must've seen a lot of
patches that awkwardly puts code from one library to the other:
* D75360 moves the constructors of CheckerManager, which lies in the Core
library, to the Frontend library. Most the patch itself was a struggle along
the library lines.
* D78126 had to be reverted because dependency information would be utilized
in the Core library, but the actual data lied in the frontend.
D78126#inline-751477 touches on this issue as well.
This stems from the often mentioned problem: the Frontend library depends on
Core and Checkers, Checkers depends on Core. The checker registry functions
(`registerMallocChecker`, etc) lie in the Checkers library in order to keep each
checker its own module. What this implies is that checker registration cannot
take place in the Core, but the Core might still want to use the data that
results from it (which checker/package is enabled, dependencies, etc).
D54436 was the patch that initiated this. Back in the days when CheckerRegistry
was super dumb and buggy, it implemented a non-documented solution to this
problem by keeping the data in the Core, and leaving the logic in the Frontend.
At the time when the patch landed, the merger to the Frontend made sense,
because the data hadn't been utilized anywhere, and the whole workaround without
any documentation made little sense to me.
So, lets put the data back where it belongs, in the Core library. This patch
introduces `CheckerRegistryData`, and turns `CheckerRegistry` into a short lived
wrapper around this data that implements the logic of checker registration. The
data is tied to CheckerManager because it is required to parse it.
Side note: I can't help but cringe at the fact how ridiculously awkward the
library lines are. I feel like I'm thinking too much inside the box, but I guess
this is just the price of keeping the checkers so modularized.
Differential Revision: https://reviews.llvm.org/D82585
Sam McCall [Sat, 4 Jul 2020 10:03:22 +0000 (12:03 +0200)]
[clangd] Try to fix windows buildbot. NFC
http://45.33.8.238/win/19116/step_9.txt
Uday Bondhugula [Thu, 11 Jun 2020 12:40:53 +0000 (18:10 +0530)]
[MLIR][Linalg] Generate the right type of load/store when lowering max/min pooling ops
While lowering min/max pooling ops to loops, generate the right kind of
load/stores (std or affine) instead of always generating std
load/stores.
Differential Revision: https://reviews.llvm.org/D83080
Paul Walker [Sat, 4 Jul 2020 09:17:53 +0000 (09:17 +0000)]
[SVE] Fix invalid assert in expand_DestructiveOp.
AArch64ExpandPseudo::expand_DestructiveOp contains an assert to
ensure the destructive operand's register is unique. However,
this is only required when psuedo expansion emits a movprfx.
A simple example when a movprfx is not required is
Z0 = FADD_ZPZZ_UNDEF_S P0, Z0, Z0
which expands to an unprefixed FADD_ZPmZ_S instruction.
This patch moves the assert to the places where a movprfx is emitted.
Differential Revision: https://reviews.llvm.org/D83029
Sam McCall [Thu, 2 Jul 2020 21:09:25 +0000 (23:09 +0200)]
[clangd] Config: compute config in TUScheduler and BackgroundIndex
Summary:
ClangdServer owns the question of exactly which config to create, but
TUScheduler/BackgroundIndex control threads and so decide at which point
to inject it.
Reviewers: kadircet
Subscribers: ilya-biryukov, javed.absar, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D83095
Nikita Popov [Thu, 2 Jul 2020 19:35:40 +0000 (21:35 +0200)]
[InstSimplify] Simplify comparison between zext(x) and sext(x)
This is picking up a loose thread from D69006: We can simplify
(zext x) ule (sext x) and (zext x) sge (sext x) to true, with
various permutations. Oddly, SCEV knows about this identity,
but nothing on the IR level does.
Differential Revision: https://reviews.llvm.org/D83081
Nikita Popov [Sat, 4 Jul 2020 08:59:01 +0000 (10:59 +0200)]
[InstSimplify] Add additional zext/sext comparison tests (NFC)
Add vector variants, and negative tests where the operand does
not match.
LLVM GN Syncbot [Sat, 4 Jul 2020 08:53:11 +0000 (08:53 +0000)]
[gn build] Port
8bd000a65fe
Sam McCall [Wed, 1 Jul 2020 14:30:57 +0000 (16:30 +0200)]
[clangd] Config: loading and caching config from disk.
Summary:
The Provider extension point is designed to also be implemented by
ClangdLSPServer (to inject config-over-lsp) and likely by embedders.
Reviewers: kadircet
Subscribers: mgorny, ilya-biryukov, MaskRay, jkorous, arphaman, jfb, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D82964
Craig Topper [Sat, 4 Jul 2020 05:08:28 +0000 (22:08 -0700)]
[X86] Directly emit VPTERNLOG from canonicalizeBitSelect when possible.
Seems to produce better results on some rotate tests. And is
neutral for other tests.
Kai Luo [Fri, 3 Jul 2020 05:55:33 +0000 (05:55 +0000)]
[PowerPC] Implement probing for prologue
This patch is part of supporting `-fstack-clash-protection`. Implemented
probing when emitting prologue.
Differential Revision: https://reviews.llvm.org/D81460
Craig Topper [Sat, 4 Jul 2020 00:50:35 +0000 (17:50 -0700)]
[X86] Add matching support for X86ISD::ANDNP to X86DAGToDAGISel::tryVPTERNLOG.
peter klausler [Fri, 3 Jul 2020 19:38:22 +0000 (12:38 -0700)]
[flang] Further implementation of external I/O unit operations (part 6)
Rework initial implementation of external I/O unit operations to
fix problems exposed in unit tests (in a later patch). Add flushing.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D83147
Thomas Lively [Sat, 4 Jul 2020 00:15:39 +0000 (17:15 -0700)]
[WebAssembly] Do not omit range checks for i64 switches
Summary:
Since the br_table instruction takes an i32, switches over i64s (and
larger integers) must use the i32.wrap_i64 instruction to truncate the
table index. This truncation makes numbers just over 2^32
indistinguishable from small numbers, so it was a miscompilation to
omit the range check preceding these br_tables. This change fixes the
problem by skipping the "fixing" of the br_table when the range check
is an i64 instruction.
Fixes PR46447.
Reviewers: aheejin, dschuff, kripken
Reviewed By: kripken
Subscribers: sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83017
Fangrui Song [Sat, 4 Jul 2020 00:06:54 +0000 (17:06 -0700)]
[gcov][test] Add `UNSUPPORTED: host-byteorder-big-endian` to gcov-fork.c
This test strangely failed on ppc64be
http://lab.llvm.org:8011/builders/clang-ppc64be-linux/builds/50913
Fangrui Song [Fri, 3 Jul 2020 23:17:06 +0000 (16:17 -0700)]
[gcov][test] Reorganize some compiler-rt/test/profile tests
Francis Visoiu Mistrih [Fri, 3 Jul 2020 05:09:01 +0000 (22:09 -0700)]
[LoopDeletion] Emit a remark when a dead loop is deleted
This emits a remark when LoopDeletion deletes a dead loop, using the
source location of the loop's header. There are currently two reasons
for removing the loop: invariant loop or loop that never executes.
Differential Revision: https://reviews.llvm.org/D83113
Lei Huang [Fri, 3 Jul 2020 21:47:24 +0000 (16:47 -0500)]
[PowerPC][NFC] Fix indentation
Roman Lebedev [Fri, 3 Jul 2020 20:39:24 +0000 (23:39 +0300)]
[NFCI][LoopUnroll] s/%tmp/%i/ in one test to silence update script warning
Roman Lebedev [Fri, 3 Jul 2020 20:39:01 +0000 (23:39 +0300)]
[NFCI][InstCombine] shift.ll: s/%tmp/%i/ to silence update script warning
Sanjay Patel [Fri, 3 Jul 2020 21:31:57 +0000 (17:31 -0400)]
[x86] improve codegen for bit-masked vector compare and select (PR46531)
We canonicalize patterns like:
%s = lshr i32 %a0, 1
%t = trunc i32 %s to i1
to:
%a = and i32 %a0, 2
%c = icmp ne i32 %a, 0
...in IR, but the bit-shifting original sequence may be better for x86 vector codegen.
I tried several variants of the transform, and it's tricky to not induce regressions.
In particular, I did not find a way to cleanly handle non-splat constants, so I've left
that as a TODO item here (currently negative tests for those are included). AVX512
resulted in some diffs, but didn't look meaningful, so I left that out too. Some of
the 256-bit AVX1 diffs are questionable, but close enough that they are probably
insignificant.
Differential Revision: https://reviews.llvm.org/D83073.
Sanjay Patel [Fri, 3 Jul 2020 19:34:55 +0000 (15:34 -0400)]
[InstCombine] fold mul of sext bools to 'and'
Alive2:
define i32 @src(i1 %x, i1 %y) {
%0:
%zx = sext i1 %x to i32
%zy = sext i1 %y to i32
%r = mul i32 %zx, %zy
ret i32 %r
}
=>
define i32 @tgt(i1 %x, i1 %y) {
%0:
%a = and i1 %x, %y
%r = zext i1 %a to i32
ret i32 %r
}
Transformation seems to be correct!
https://alive2.llvm.org/ce/z/gaPQxA
Sanjay Patel [Fri, 3 Jul 2020 19:27:43 +0000 (15:27 -0400)]
[InstCombine] add more tests for mul of bools; NFC
Vy Nguyen [Wed, 24 Jun 2020 19:03:08 +0000 (15:03 -0400)]
[libcxx] Put clang::trivial_abi on std::unique_ptr, std::shared_ptr, and std::weak_ptr
Reviewers: jyknight, EricWF, #libc!
Subscribers: arphaman, libcxx-commits
Tags: #libc
Differential Revision: https://reviews.llvm.org/D82490
Kadir Cetinkaya [Fri, 3 Jul 2020 18:52:41 +0000 (20:52 +0200)]
[clangd] Fix hover crash on invalid decls
Summary: This also changes the way we display Size and Offset to be independent.
Reviewers: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D83143
Biplob Mishra [Fri, 3 Jul 2020 17:45:27 +0000 (12:45 -0500)]
[PowerPC] Implement Vector Insert Builtins in LLVM/Clang
Implements vec_insertl() and vec_inserth().
Differential Revision: https://reviews.llvm.org/D82365
Stephen Kelly [Thu, 2 Jul 2020 19:46:27 +0000 (20:46 +0100)]
Revert AST Matchers default to AsIs mode
Reviewers: aaron.ballman, klimek
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D83076
peter klausler [Fri, 3 Jul 2020 17:56:08 +0000 (10:56 -0700)]
[flang] Track known file size, add IsATerminal (ext. I/O work part 5)
Add a data member knownSize_ and an accessor to allow the size of
an external file to be tracked when known. Also add a wrapper for
::isatty() here in the filesystem encapsulation module. These
features are needed for the external I/O rework changes still
to come.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D83141
peter klausler [Fri, 3 Jul 2020 17:47:02 +0000 (10:47 -0700)]
[flang] Define new runtime error IOSTAT values (I/O runtime work part 4)
Add more IOSTAT= values for errors that can arise in external I/O.
Reviewed By: sscalpone
Differential Revision: https://reviews.llvm.org/D83140
Florian Hahn [Fri, 3 Jul 2020 19:22:51 +0000 (20:22 +0100)]
[InstCombine] Try to narrow expr if trunc cannot be removed.
Narrowing an input expression of a truncate to a type larger than the
result of the truncate won't allow removing the truncate, but it may
enable further optimizations, e.g. allowing for larger vectorization
factors.
For now this is intentionally limited to integer types only, to avoid
producing new vector ops that might not be suitable for the target.
If we know that the only user is a trunc, we can also be allow more
cases, e.g. also shortening expressions with some additional shifts.
I would appreciate feedback on the best place to do such a narrowing.
This fixes PR43580.
Reviewers: spatel, RKSimon, lebedev.ri, xbolva00
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D82973
Louis Dionne [Fri, 3 Jul 2020 17:46:41 +0000 (13:46 -0400)]
[libc++/libc++abi] Automatically detect whether exceptions are enabled
Instead of detecting it automatically (in libc++) and relying on
_LIBCXXABI_NO_EXCEPTIONS being set explicitly (in libc++abi), always
detect whether exceptions are enabled automatically.
This commit also removes support for specifying -D_LIBCPP_NO_EXCEPTIONS
and -D_LIBCXXABI_NO_EXCEPTIONS explicitly -- those should just be inferred
from using -fno-exceptions (or an equivalent flag).
Allowing both -D_FOO_NO_EXCEPTIONS to be provided explicitly and trying
to detect it automatically is just confusing, especially since we did
specify it explicitly when building libc++abi. We should have only one
way to detect whether exceptions are enabled, but it should be robust.
Eric Schweitz [Fri, 3 Jul 2020 00:44:51 +0000 (17:44 -0700)]
[flang] Add FIRBuilder.cpp
The FIR builder is a helper class that manages the creation of MLIR
operations from the bridge. The focus of the builder is the creation of
Operations, Types, etc.
Differential revision: htps://reviews.llvm.org/D83107
jasonliu [Thu, 2 Jul 2020 22:45:59 +0000 (22:45 +0000)]
[XCOFF][AIX] Use 'L..' instead of '.L' for getPrivateGlobalPrefix in DataLayout
Summary:
D80831 changed part of the prefix usage for AIX.
But there are other places getting prefix from DataLayout.
This patch intends to make prefix usage consistent on AIX.
Reviewed by: hubert.reinterpretcast, daltenty
Differential Revision: https://reviews.llvm.org/D81270
sameerarora101 [Fri, 3 Jul 2020 17:57:32 +0000 (10:57 -0700)]
[llvm-ar][test] Unsupport error-opening-directory.test on FreeBSD
Differential Revision: https://reviews.llvm.org/D82786
Sanjay Patel [Fri, 3 Jul 2020 17:08:59 +0000 (13:08 -0400)]
[InstCombine] fold mul of zext bools to 'and'
The base case only works because we are relying on a
poison-unsafe select transform; if that is fixed, we
would regress on patterns like this.
The extra use tests show that the select transform can't
be applied consistently. So it may be a regression to have
an extra instruction on 1 test, but that result was not
created safely and does not happen reliably.
Sanjay Patel [Fri, 3 Jul 2020 16:52:34 +0000 (12:52 -0400)]
[InstCombine] add tests for mul of bools; NFC
Roman Lebedev [Fri, 3 Jul 2020 17:12:46 +0000 (20:12 +0300)]
[NFC][InstCombine] Add some more tests for select based on non-canonical bit-test
Nikita Popov [Sun, 28 Jun 2020 13:59:56 +0000 (15:59 +0200)]
[InstSimplify] Fold icmp with dominating assume
If we assume(x > y), then we should be able to fold the basic
implications of that, like x >= y. This already happens if either
one of the operands is constant (LVI) or if the conditions are
exactly the same (GVN), but not if we have an implication with
non-constant operands. Support this by querying AssumptionCache.
Fixes https://bugs.llvm.org/show_bug.cgi?id=40149.
Differential Revision: https://reviews.llvm.org/D82717
Fangrui Song [Fri, 3 Jul 2020 16:50:30 +0000 (09:50 -0700)]
[ELF] Resolve R_DTPREL in .debug_* referencing discarded symbols to -1
The location of a TLS variable is encoded as a DW_OP_const4u/DW_OP_const8u
followed by a DW_OP_push_tls_address (or DW_OP_GNU_push_tls_address https://sourceware.org/bugzilla/show_bug.cgi?id=11616 ).
This change follows up to D81784 and makes relocations types generalized as
R_DTPREL (e.g. R_X86_64_DTPOFF{32,64}, R_PPC64_DTPREL64) use -1 as the
tombstone value as well. This works for both TLS Variant I and Variant II
architectures.
* arm: .long tls(tlsldo) # not working currently (R_ARM_TLS_LDO32 is R_ABS)
* mips64: .dtpreldword tls+32768
* ppc64: .quad tls@DTPREL+0x8000
* riscv: neither GCC nor clang has implemented DW_AT_location. It is likely .long/.quad tls@dtprel+0x800
* x86-32: .long tls@DTPOFF
* x86-64: .long tls@DTPOFF; .quad tls@DTPOFF
tls has a non-negative st_value, so such relocations (st_value+addend)
never resolve to -1 in a normal (not discarded) case.
```
// clang -fuse-ld=lld -g -ffunction-sections a.c -Wl,--gc-sections
// foo and tls will be discarded by --gc-sections.
// DW_AT_location [DW_FORM_exprloc] (DW_OP_const8u 0xffffffffffffffff, DW_OP_GNU_push_tls_address)
thread_local int tls;
int foo() { return ++tls; }
int main() {}
```
Also, drop logic added in D26201 intended to address PR30793. It added a test
(gc-debuginfo-tls.s) using a non-SHF_ALLOC section and a local symbol, which
does not reflect the intended scenario: a relocation in a SHF_ALLOC section
referencing a discarded non-local symbol. For such a non .debug_* section, just
emit an error.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D82899
Florian Hahn [Fri, 3 Jul 2020 16:28:45 +0000 (17:28 +0100)]
[SLP] Make sure instructions are ordered when computing spill cost.
The entries in VectorizableTree are not necessarily ordered by their
position in basic blocks. Collect them and order them by dominance so
later instructions are guaranteed to be visited first. For instructions
in different basic blocks, we only scan to the beginning of the block,
so their order does not matter, as long as all instructions in a basic
block are grouped together. Using dominance ensures a deterministic order.
The modified test case contains an example where we compute a wrong
spill cost (2) without this patch, even though there is no call between
any instruction in the bundle.
This seems to have limited practical impact, .e.g on X86 with a recent
Intel Xeon CPU with -O3 -march=native -flto on MultiSource,SPEC2000,SPEC2006
there are no binary changes.
Reviewers: craig.topper, RKSimon, xbolva00, ABataev, spatel
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D82444
David Green [Fri, 3 Jul 2020 13:18:32 +0000 (14:18 +0100)]
[ARM][HWLoops] Create hardware loops for sibling loops
Given a loop with two subloops, it should be possible for both to be
converted to hardware loops. That's what this patch does, simply enough.
It slightly alters the loop iterating order to try and convert all
subloops. If one (or more) succeeds, it stops as before.
Differential Revision: https://reviews.llvm.org/D78502
Florian Hahn [Tue, 23 Jun 2020 18:07:30 +0000 (19:07 +0100)]
[SLP] Precommit test for which spill cost is computed incorrectly.
Test for D82444.