Craig Topper [Sun, 8 Sep 2019 20:56:05 +0000 (20:56 +0000)]
[X86] Remove call to getZeroVector from materializeVectorConstant. Add isel patterns for zero vectors with all types.
The change to avx512-vec-cmp.ll is a regression, but should be
easy to fix. It occurs because the getZeroVector call was
canonicalizing both sides to the same node, then SimplifySelect
was able to simplify it. But since only called getZeroVector
on some VTs this isn't a robust way to combine this.
The change to vector-shuffle-combining-ssse3.ll is more
instructions, but removes a constant pool load so its unclear
if its a regression or not.
llvm-svn: 371350
Roman Lebedev [Sun, 8 Sep 2019 20:14:15 +0000 (20:14 +0000)]
[InstSimplify] simplifyUnsignedRangeCheck(): if we know that X != 0, handle more cases (PR43246)
Summary:
This is motivated by D67122 sanitizer check enhancement.
That patch seemingly worsens `-fsanitize=pointer-overflow`
overhead from 25% to 50%, which strongly implies missing folds.
In this particular case, given
```
char* test(char& base, unsigned long offset) {
return &base + offset;
}
```
it will end up producing something like
https://godbolt.org/z/LK5-iH
which after optimizations reduces down to roughly
```
define i1 @t0(i8* nonnull %base, i64 %offset) {
%base_int = ptrtoint i8* %base to i64
%adjusted = add i64 %base_int, %offset
%non_null_after_adjustment = icmp ne i64 %adjusted, 0
%no_overflow_during_adjustment = icmp uge i64 %adjusted, %base_int
%res = and i1 %non_null_after_adjustment, %no_overflow_during_adjustment
ret i1 %res
}
```
Without D67122 there was no `%non_null_after_adjustment`,
and in this particular case we can get rid of the overhead:
Here we add some offset to a non-null pointer,
and check that the result does not overflow and is not a null pointer.
But since the base pointer is already non-null, and we check for overflow,
that overflow check will already catch the null pointer,
so the separate null check is redundant and can be dropped.
Alive proofs:
https://rise4fun.com/Alive/WRzq
There are more patterns of "unsigned-add-with-overflow", they are not handled here,
but this is the main pattern, that we currently consider canonical,
so it makes sense to handle it.
https://bugs.llvm.org/show_bug.cgi?id=43246
Reviewers: spatel, nikic, vsk
Reviewed By: spatel
Subscribers: hiraditya, llvm-commits, reames
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67332
llvm-svn: 371349
Sanjay Patel [Sun, 8 Sep 2019 19:48:47 +0000 (19:48 +0000)]
[InstCombine] add tests for icmp with srem operand; NFC
llvm-svn: 371348
Roman Lebedev [Sun, 8 Sep 2019 19:36:13 +0000 (19:36 +0000)]
[X86] X86DAGToDAGISel::combineIncDecVector(): call getSplatBuildVector() manually
As reported in post-commit review of r370327,
there is some case where the code crashes.
As discussed with Craig Topper, the problem is that getConstant()
internally calls getSplatBuildVector(), so we don't insert
the constant itself.
If we do that manually we're good.
llvm-svn: 371346
Craig Topper [Sun, 8 Sep 2019 19:24:42 +0000 (19:24 +0000)]
[X86] Use DAG.getConstant instead of getZeroVector in combinePMULDQ.
getZeroVector canonicalizes the type to vXi32, but that's a
legalization action. We should use the most correct type if
possible.
llvm-svn: 371345
Craig Topper [Sun, 8 Sep 2019 19:24:39 +0000 (19:24 +0000)]
[DAGCombiner][X86][ARM] Teach visitMULO to fold multiplies with 0 to 0 and no carry.
I modified the ARM test to use two inputs instead of 0 so the
test hopefully still tests what was intended.
llvm-svn: 371344
Craig Topper [Sun, 8 Sep 2019 19:24:29 +0000 (19:24 +0000)]
[X86] Teach materializeVectorConstant to not call getZeroVector/getOnesVector on the types we already have isel patterns for.
llvm-svn: 371343
Gabor Borsik [Sun, 8 Sep 2019 19:23:43 +0000 (19:23 +0000)]
Move prop-sink branch to monorepo.
llvm-svn: 371342
Sanjay Patel [Sun, 8 Sep 2019 19:03:01 +0000 (19:03 +0000)]
[InstCombine] fold extract+insert into identity shuffle
This is similar to the existing fold for splats added with:
rL365379
If we can adjust the shuffle mask to include another element
in an identity mask (if it changes vector length, that's an
extract/insert subvector operation in the backend), then that
can eliminate extractelement/insertelement pairs in IR.
All targets are expected to lower shuffles with identity masks
efficiently.
llvm-svn: 371340
Roman Lebedev [Sun, 8 Sep 2019 17:50:40 +0000 (17:50 +0000)]
[NFC][InstSimplify] Some tests for dropping null check after uadd.with.overflow of non-null (PR43246)
https://rise4fun.com/Alive/WRzq
Name: C <= Y && Y != 0 --> C <= Y iff C != 0
Pre: C != 0
%y_is_nonnull = icmp ne i64 %y, 0
%no_overflow = icmp ule i64 C, %y
%r = and i1 %y_is_nonnull, %no_overflow
=>
%r = %no_overflow
Name: C <= Y || Y != 0 --> Y != 0 iff C != 0
Pre: C != 0
%y_is_nonnull = icmp ne i64 %y, 0
%no_overflow = icmp ule i64 C, %y
%r = or i1 %y_is_nonnull, %no_overflow
=>
%r = %y_is_nonnull
Name: C > Y || Y == 0 --> C > Y iff C != 0
Pre: C != 0
%y_is_null = icmp eq i64 %y, 0
%overflow = icmp ugt i64 C, %y
%r = or i1 %y_is_null, %overflow
=>
%r = %overflow
Name: C > Y && Y == 0 --> Y == 0 iff C != 0
Pre: C != 0
%y_is_null = icmp eq i64 %y, 0
%overflow = icmp ugt i64 C, %y
%r = and i1 %y_is_null, %overflow
=>
%r = %y_is_null
https://bugs.llvm.org/show_bug.cgi?id=43246
llvm-svn: 371339
Kamil Rytarowski [Sun, 8 Sep 2019 17:07:28 +0000 (17:07 +0000)]
Enable LSan tests for NetBSD/i386
llvm-svn: 371338
Kamil Rytarowski [Sun, 8 Sep 2019 16:15:18 +0000 (16:15 +0000)]
Stop marking 5 ASan tests as failing on NetBSD/i386
Unexpected Passing Tests (4):
AddressSanitizer-i386-netbsd :: TestCases/Posix/coverage-reset.cpp
AddressSanitizer-i386-netbsd :: TestCases/Posix/coverage.cpp
AddressSanitizer-i386-netbsd :: TestCases/Posix/interception-in-shared-lib-test.cpp
AddressSanitizer-i386-netbsd :: TestCases/suppressions-library.cpp
llvm-svn: 371337
Jonas Hahnfeld [Sun, 8 Sep 2019 16:08:54 +0000 (16:08 +0000)]
[ASan] Only run dlopen-mixed-c-cxx.c with static runtime
This is what the original bug (http://llvm.org/PR39641) and the fix
in https://reviews.llvm.org/D63877 have been about.
With the dynamic runtime the test only passes when the asan library
is linked against libstdc++: In contrast to libc++abi, it does not
implement __cxa_rethrow_primary_exception so the regex matches the
line saying that asan cannot intercept this function. Indeed, there
is no message that the runtime failed to intercept __cxa_throw.
Differential Revision: https://reviews.llvm.org/D67298
llvm-svn: 371336
Kamil Rytarowski [Sun, 8 Sep 2019 15:54:48 +0000 (15:54 +0000)]
Enable leak-detection for NetBSD/amd64 in test/asan
llvm-svn: 371335
Kamil Rytarowski [Sun, 8 Sep 2019 15:00:13 +0000 (15:00 +0000)]
Do not intercept malloc_usable_size on NetBSD
llvm-svn: 371334
David Stenberg [Sun, 8 Sep 2019 14:22:06 +0000 (14:22 +0000)]
[DebugInfo][X86] Describe call site values for zero-valued imms
Summary:
Add zero-materializing XORs to X86's describeLoadedValue() hook in order
to produce call site values.
I have had to change the defs logic in collectCallSiteParameters() a bit
to be able to describe the XORs. The XORs implicitly define $eflags,
which would cause them to never be considered, due to a guard condition
that I->getNumDefs() is one. I have changed that condition so that we
now only consider instructions where a forwarded register overlaps with
the instruction's single explicit define. We still need to collect the implicit
defines of other forwarded registers to remove them from the work list.
I'm not sure how to move towards supporting instructions with multiple
explicit defines, cases where forwarded register are implicitly defined,
and/or cases where an instruction produces values for multiple forwarded
registers. Perhaps the describeLoadedValue() hook should take a register
argument, and we then leave it up to the hook to describe the loaded
value in that register? I have not yet encountered a situation where
that would be necessary though.
Reviewers: aprantl, vsk, djtodoro, NikolaPrica
Reviewed By: vsk
Subscribers: ychen, hiraditya, llvm-commits
Tags: #debug-info, #llvm
Differential Revision: https://reviews.llvm.org/D67225
llvm-svn: 371333
David Stenberg [Sun, 8 Sep 2019 14:05:10 +0000 (14:05 +0000)]
[NFC] Make the describeLoadedValue() hook return machine operand objects
Summary:
This changes the ParamLoadedValue pair which the describeLoadedValue()
hook returns so that MachineOperand objects are returned instead of
pointers.
When describing call site values we may need to describe operands which
are not part of the instruction. One such example is zero-materializing
XORs on x86, which I have implemented support for in a child revision.
Instead of having to return a pointer to an operand stored somewhere
outside the instruction, start returning objects directly instead, as
that simplifies the code.
The MachineOperand class only holds POD members, and on x86-64 it is 32
bytes large. That combined with copy elision means that the overhead of
returning a machine operand object from the hook does not become very
large.
I benchmarked this on a 8-thread i7-8650U machine with 32 GB RAM. The
benchmark consisted of building a clang 8.0 binary configured with:
-DCMAKE_BUILD_TYPE=RelWithDebInfo \
-DLLVM_TARGETS_TO_BUILD=X86 \
-DLLVM_USE_SANITIZER=Address \
-DCMAKE_CXX_FLAGS="-Xclang -femit-debug-entry-values -stdlib=libc++"
The average wall clock time increased by 4 seconds, from 62:05 to
62:09, which is an 0.1% increase.
Reviewers: aprantl, vsk, djtodoro, NikolaPrica
Reviewed By: vsk
Subscribers: hiraditya, ychen, llvm-commits
Tags: #debug-info, #llvm
Differential Revision: https://reviews.llvm.org/D67261
llvm-svn: 371332
David Green [Sun, 8 Sep 2019 13:13:15 +0000 (13:13 +0000)]
[ARM] Remove declaration of unimplemented function. NFC.
llvm-svn: 371331
Simon Pilgrim [Sun, 8 Sep 2019 12:44:22 +0000 (12:44 +0000)]
[X86][SSE] Fix out of range shift introduced in D67070/rL371328
Use APInt to create the comparison mask instead.
llvm-svn: 371330
Simon Pilgrim [Sun, 8 Sep 2019 11:56:07 +0000 (11:56 +0000)]
[X86] Add test case for PR32546
llvm-svn: 371329
Simon Pilgrim [Sun, 8 Sep 2019 11:46:21 +0000 (11:46 +0000)]
[X86][SSE] Add support for <64 x i1> bool reduction
This generalizes the existing <32 x i1> pre-AVX2 split code to support reductions from <64 x i1> as well, we can probably generalize to any larger pow2 case in the future if the (unlikely) need ever arises.
We still need to tweak combineBitcastvxi1 to improve AVX512F codegen as its assumes vXi1 types should be handled on the mask registers even when they aren't legal.
Differential Revision: https://reviews.llvm.org/D67070
llvm-svn: 371328
Xing GUO [Sun, 8 Sep 2019 11:42:51 +0000 (11:42 +0000)]
[StackMap] Current stackmap version should be 3. NFC.
llvm-svn: 371327
Leonard Chan [Sun, 8 Sep 2019 07:30:17 +0000 (07:30 +0000)]
[NewPM][Sancov] Create the Sancov Pass after building the pipelines
We're running into linker errors from missing sancov sections:
```
ld.lld: error: relocation refers to a discarded section: __sancov_guards
>>> defined in user-arm64-ubsan-sancov-full.shlib/obj/third_party/ulib/scudo/scudo.wrappers_c.cc.o
>>> referenced by common.h:26 (../../zircon/third_party/ulib/scudo/common.h:26)
... many other references
```
I believe this is due to a pass in the default pipeline that somehow discards
these sections. The ModuleSanitizerCoveragePass was initially added at the
start of the pipeline. This now adds it to the end of the pipeline for
optimized and unoptimized builds.
Differential Revision: https://reviews.llvm.org/D67323
llvm-svn: 371326
Craig Topper [Sun, 8 Sep 2019 00:43:52 +0000 (00:43 +0000)]
[X86] Make getZeroVector return floating point vectors in their native type on SSE2 and later.
isel used to require zero vectors to be canonicalized to a single
type to minimize the number of patterns needed to match. This is
no longer required.
I plan to do this to integers too, but floating point was simpler
to start with. Integer has a complication where v32i16/v64i8 aren't
legal when the other 512-bit integer types are.
llvm-svn: 371325
Dimitry Andric [Sat, 7 Sep 2019 22:18:20 +0000 (22:18 +0000)]
Remove ::gets for FreeBSD 13 and later
Summary:
In https://svnweb.freebsd.org/changeset/base/351659 @emaste removed gets() from
FreeBSD 13's libc, and our copies of libc++ and libstdc++. In that change, the
declarations were simply deleted, but I would like to propose this conditional
test instead.
Reviewers: EricWF, mclow.lists, emaste
Reviewed By: mclow.lists
Subscribers: krytarowski, christof, ldionne, emaste, libcxx-commits
Differential Revision: https://reviews.llvm.org/D67316
llvm-svn: 371324
Craig Topper [Sat, 7 Sep 2019 21:54:40 +0000 (21:54 +0000)]
[X86] Add support for unfold broadcast loads from FMA instructions.
llvm-svn: 371323
Craig Topper [Sat, 7 Sep 2019 21:54:36 +0000 (21:54 +0000)]
[X86] Add broadcast load unfolding tests for FMA instructions.
llvm-svn: 371322
Sebastian Pop [Sat, 7 Sep 2019 20:24:51 +0000 (20:24 +0000)]
[aarch64] Add combine patterns for fp16 fmla
This patch enables generation of fused multiply add/sub for instructions operating on fp16.
Tested on aarch64-linux.
Differential Revision: https://reviews.llvm.org/D67297
llvm-svn: 371321
Aaron Ballman [Sat, 7 Sep 2019 20:14:09 +0000 (20:14 +0000)]
Fixes an assertion while instantiating a template with an incomplete typo-corrected type.
Fixes PR35682. When a template in instantiated with an incomplete typo corrected type an assertion can trigger if the -ferror-limit is used to reduce the number of errors.
Patch by Mark de Wever.
llvm-svn: 371320
Craig Topper [Sat, 7 Sep 2019 19:54:22 +0000 (19:54 +0000)]
[X86] Add prefer-128-bit subtarget feature.
Summary:
Similar to the previous prefer-256-bit flag. We might want to
enable this by default some CPUs. This just starts the initial
work to implement and prove that it effects TTI's vector width.
Reviewers: RKSimon, echristo, spatel, atdt
Reviewed By: RKSimon
Subscribers: lebedev.ri, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67311
llvm-svn: 371319
George Rimar [Sat, 7 Sep 2019 19:45:27 +0000 (19:45 +0000)]
[llvm-nm] - Fix a bug and unbreak ASan BB.
BB: http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-bootstrap/builds/13820/steps/check-llvm%20asan/logs/stdio
rL371074 revealed a bug in llvm-nm.
This patch fixes it.
llvm-svn: 371318
Simon Pilgrim [Sat, 7 Sep 2019 18:09:09 +0000 (18:09 +0000)]
Fix typo. NFCI
llvm-svn: 371317
Simon Atanasyan [Sat, 7 Sep 2019 16:31:37 +0000 (16:31 +0000)]
[mips] Follow-up to r371313 - fix failed test case. NFC
llvm-svn: 371316
Simon Pilgrim [Sat, 7 Sep 2019 16:13:57 +0000 (16:13 +0000)]
[X86] Avoid uses of getZextValue(). NFCI.
Use getAPIntValue() directly - this is mainly a best practice style issue to help prevent fuzz tests blowing up when a i12345 (or whatever) is generated.
Use getConstantOperandVal/getConstantOperandAPInt wrappers where possible.
llvm-svn: 371315
Simon Pilgrim [Sat, 7 Sep 2019 16:13:48 +0000 (16:13 +0000)]
[X86][AVX] Add 'f5' v4f64 shuffle test mentioned in D66004
llvm-svn: 371314
Simon Atanasyan [Sat, 7 Sep 2019 15:44:40 +0000 (15:44 +0000)]
[mips] Make another set of test cases more tolerant to exact symbol addresses. NFC
llvm-svn: 371313
Fangrui Song [Sat, 7 Sep 2019 14:58:47 +0000 (14:58 +0000)]
[ELF][MC] Set types of aliases of IFunc to STT_GNU_IFUNC
```
.type foo,@gnu_indirect_function
.set foo,foo_resolver
.set foo2,foo
.set foo3,foo2
```
The types of foo2 and foo3 should be STT_GNU_IFUNC, but we currently
resolve them to the type of foo_resolver. This patch fixes it.
Differential Revision: https://reviews.llvm.org/D67206
Patch by Senran Zhang
llvm-svn: 371312
Roman Lebedev [Sat, 7 Sep 2019 13:53:14 +0000 (13:53 +0000)]
[SimplifyCFG][NFC] Autogenerate PhiEliminate3.ll
llvm-svn: 371311
Roman Lebedev [Sat, 7 Sep 2019 13:35:54 +0000 (13:35 +0000)]
[SimplifyCFG][NFC] Autogenerate two tests
llvm-svn: 371310
Bjorn Pettersson [Sat, 7 Sep 2019 12:16:23 +0000 (12:16 +0000)]
[CodeGen] Handle SMULFIXSAT with scale zero in TargetLowering::expandFixedPointMul
Summary:
Normally TargetLowering::expandFixedPointMul would handle
SMULFIXSAT with scale zero by using an SMULO to compute the
product and determine if saturation is needed (if overflow
happened). But if SMULO isn't custom/legal it falls through
and uses the same technique, using MULHS/SMUL_LOHI, as used
for non-zero scales.
Problem was that when checking for overflow (handling saturation)
when not using MULO we did not expect to find a zero scale. So
we ended up in an assertion when doing
APInt::getLowBitsSet(VTSize, Scale - 1)
This patch fixes the problem by adding a new special case for
how saturation is computed when scale is zero.
Reviewers: RKSimon, bevinh, leonardchan, spatel
Reviewed By: RKSimon
Subscribers: wuzish, nemanjai, hiraditya, MaskRay, jsji, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67071
llvm-svn: 371309
Bjorn Pettersson [Sat, 7 Sep 2019 12:16:14 +0000 (12:16 +0000)]
[Intrinsic] Add the llvm.umul.fix.sat intrinsic
Summary:
Add an intrinsic that takes 2 unsigned integers with
the scale of them provided as the third argument and
performs fixed point multiplication on them. The
result is saturated and clamped between the largest and
smallest representable values of the first 2 operands.
This is a part of implementing fixed point arithmetic
in clang where some of the more complex operations
will be implemented as intrinsics.
Patch by: leonardchan, bjope
Reviewers: RKSimon, craig.topper, bevinh, leonardchan, lebedev.ri, spatel
Reviewed By: leonardchan
Subscribers: ychen, wuzish, nemanjai, MaskRay, jsji, jdoerfert, Ka-Ka, hiraditya, rjmccall, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57836
llvm-svn: 371308
Nikita Popov [Sat, 7 Sep 2019 12:13:44 +0000 (12:13 +0000)]
[X86] Fix pshuflw formation from repeated shuffle mask (PR43230)
Fix for https://bugs.llvm.org/show_bug.cgi?id=43230.
When creating PSHUFLW from a repeated shuffle mask, we have to apply
the checks to the repeated mask, not the original one. For the test
case from PR43230 the inspected part of the original mask is all undef.
Differential Revision: https://reviews.llvm.org/D67314
llvm-svn: 371307
Nikita Popov [Sat, 7 Sep 2019 12:03:59 +0000 (12:03 +0000)]
[LVI] Look through extractvalue of insertvalue
This addresses the issue mentioned on D19867. When we simplify
with.overflow instructions in CVP, we leave behind extractvalue
of insertvalue sequences that LVI no longer understands. This
means that we can not simplify any instructions based on the
with.overflow anymore (until some over pass like InstCombine
cleans them up).
This patch extends LVI extractvalue handling by calling
SimplifyExtractValueInst (which doesn't do anything more than
constant folding + looking through insertvalue) and using the block
value of the simplification.
A possible alternative would be to do something similar to
SimplifyIndVars, where we instead directly try to replace
extractvalue users of the with.overflow. This would need some
additional structural changes to CVP, as it's currently not legal
to remove anything but the current instruction -- we'd have to
introduce a worklist with instructions scheduled for deletion or similar.
Differential Revision: https://reviews.llvm.org/D67035
llvm-svn: 371306
Nikita Popov [Sat, 7 Sep 2019 12:03:48 +0000 (12:03 +0000)]
[X86] Add test for PR43230; NFC
llvm-svn: 371305
Bjorn Pettersson [Sat, 7 Sep 2019 11:40:10 +0000 (11:40 +0000)]
[DwarfExpression] Disallow some rewrites to avoid undefined behavior
Summary:
The value operand in DW_OP_plus_uconst/DW_OP_constu value can be
large (it uses uint64_t as representation internally in LLVM).
This means that in the uint64_t to int conversions, previously done
by DwarfExpression::addMachineRegExpression, could lose information.
Also, the negation done in "-Offset" was undefined behavior in case
Offset was exactly INT_MIN.
To avoid the above problems, we now avoid transformation like
[Reg, DW_OP_plus_uconst, Offset] --> [DW_OP_breg, Offset]
and
[Reg, DW_OP_constu, Offset, DW_OP_plus] --> [DW_OP_breg, Offset]
when Offset > INT_MAX.
And we avoid to transform
[Reg, DW_OP_constu, Offset, DW_OP_minus] --> [DW_OP_breg,-Offset]
when Offset > INT_MAX+1.
The patch also adjusts DwarfCompileUnit::constructVariableDIEImpl
to make sure that "DW_OP_constu, Offset, DW_OP_minus" is used
instead of "DW_OP_plus_uconst, Offset" when creating DIExpressions
with negative frame index offsets.
Notice that this might just be the tip of the iceberg. There
are lots of fishy handling related to these constants. I think both
DIExpression::appendOffset and DIExpression::extractIfOffset may
trigger undefined behavior for certain values.
Reviewers: sdesmalen, rnk, JDevlieghere
Reviewed By: JDevlieghere
Subscribers: jholewinski, aprantl, hiraditya, ychen, uabelho, llvm-commits
Tags: #debug-info, #llvm
Differential Revision: https://reviews.llvm.org/D67263
llvm-svn: 371304
Bjorn Pettersson [Sat, 7 Sep 2019 11:39:57 +0000 (11:39 +0000)]
[DebugInfo] Pre-commit of test case for DW_OP_breg/DW_OP_fbreg folds
This currently triggers undefined behavior if executed with an
ubsan build. It is just a precommit of the test case to show that
we got a problem.
Fix is proposed in https://reviews.llvm.org/D67263 and plan is to
commit the fix directly after this patch.
llvm-svn: 371303
Simon Pilgrim [Sat, 7 Sep 2019 11:04:04 +0000 (11:04 +0000)]
Fix MSVC "32-bit shift implicitly converted to 64 bits" warnings. NFCI.
llvm-svn: 371302
Roman Lebedev [Sat, 7 Sep 2019 10:55:04 +0000 (10:55 +0000)]
[SimplifyCFG][NFC] Make merge-cond-stores-cost.ll X86-specific, and rewrite it
We clearly perform store-merging, even though div is really costly.
llvm-svn: 371300
Fangrui Song [Sat, 7 Sep 2019 10:42:11 +0000 (10:42 +0000)]
[ELF][test] Improve tests
Add file-level comments
Delete insignificant addresses to make them more tolerant to layout changes
Simplify test output
Delete weak-undef-val.s - covered by relocation-undefined-weak.s
Delete weak-undef-export.s - covered by additional test added to weak-undef.s
Delete version-undef-sym.s - covered by undefined-versioned-symbol.s => version-symbol-undef.s
Delete symbol-ordering-file2.s - covered by symbol-ordering-file.s
Delete gotpcrelx.s - covered by gotpc-relax-und-dso.s => x86-64-gotpc-relax-und-dso.s
llvm-svn: 371299
Benjamin Kramer [Sat, 7 Sep 2019 10:27:13 +0000 (10:27 +0000)]
[Attributor] Make unimplemented method pure virtual.
Otherwise the compiler mistakes it for a vtable anchor.
llvm-svn: 371298
Roman Lebedev [Sat, 7 Sep 2019 09:25:26 +0000 (09:25 +0000)]
[SimplifyCFG][NFC] Show that we don't consider the cost when merging cond stores
We count instruction count in each BB's separately, not their cost.
llvm-svn: 371297
Roman Lebedev [Sat, 7 Sep 2019 09:25:18 +0000 (09:25 +0000)]
[SimplifyCFG][NFC] Regenerate merge-cond-stores* tests
llvm-svn: 371296
Roman Lebedev [Sat, 7 Sep 2019 09:06:06 +0000 (09:06 +0000)]
[SimplifyCFG] SpeculativelyExecuteBB(): It's SpeculatedInstructions, not SpeculationCost
It counts the number of instructions we are ok speculating
(at most 1 there), not their cost, so rename accordingly.
llvm-svn: 371294
Sylvestre Ledru [Sat, 7 Sep 2019 08:38:46 +0000 (08:38 +0000)]
Replicate the change "[Alignment][NFC] Use Align with TargetLowering::setMinFunctionAlignment"
on AVR to avoid a breakage.
See r371200 / https://reviews.llvm.org/D67229
llvm-svn: 371293
Fangrui Song [Sat, 7 Sep 2019 08:20:09 +0000 (08:20 +0000)]
[ELF][test] Improve LTO tests
Add file-level comments
Delete insignificant addresses to make them more tolerant to layout changes
Simplify test output
llvm-svn: 371292
Hideto Ueno [Sat, 7 Sep 2019 07:03:05 +0000 (07:03 +0000)]
[Attributor] ValueSimplify Abstract Attribute
Summary:
This patch introduces initial `AAValueSimplify` which simplifies a value in a context.
example
- (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant.
- If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant.
Reviewers: jdoerfert, sstefan1
Reviewed By: jdoerfert
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66967
llvm-svn: 371291
David Zarzycki [Sat, 7 Sep 2019 06:44:52 +0000 (06:44 +0000)]
[git-llvm] Do not reinvent `@{upstream}`
Make `git-llvm` more robust when used with a nontrivial repository.
https://reviews.llvm.org/D67262
llvm-svn: 371290
Xing GUO [Sat, 7 Sep 2019 05:14:47 +0000 (05:14 +0000)]
Revert [CodeGen] Fix typos to run tests. NFC.
This reverts r371286 (git commit
b38105bbd0f7dfef424a4f096aa6a6b7b467fe99)
r371286 caused build bots' failure. I'll check it.
llvm-svn: 371289
Xing GUO [Sat, 7 Sep 2019 04:57:53 +0000 (04:57 +0000)]
[CodeGen] Fix typos to run tests. NFC.
llvm-svn: 371286
Teresa Johnson [Sat, 7 Sep 2019 03:09:46 +0000 (03:09 +0000)]
Remove stale TLI Module level pass registration
Clang patch to adapt to LLVM changes in D66428 that make the TLI
require a Function. There is no longer a module-level
TargetLibraryAnalysis, so remove its registration
llvm-svn: 371285
Teresa Johnson [Sat, 7 Sep 2019 03:09:36 +0000 (03:09 +0000)]
Change TargetLibraryInfo analysis passes to always require Function
Summary:
This is the first change to enable the TLI to be built per-function so
that -fno-builtin* handling can be migrated to use function attributes.
See discussion on D61634 for background. This is an enabler for fixing
handling of these options for LTO, for example.
This change should not affect behavior, as the provided function is not
yet used to build a specifically per-function TLI, but rather enables
that migration.
Most of the changes were very mechanical, e.g. passing a Function to the
legacy analysis pass's getTLI interface, or in Module level cases,
adding a callback. This is similar to the way the per-function TTI
analysis works.
There was one place where we were looking for builtins but not in the
context of a specific function. See FindCXAAtExit in
lib/Transforms/IPO/GlobalOpt.cpp. I'm somewhat concerned my workaround
could provide the wrong behavior in some corner cases. Suggestions
welcome.
Reviewers: chandlerc, hfinkel
Subscribers: arsenm, dschuff, jvesely, nhaehnle, mehdi_amini, javed.absar, sbc100, jgravelle-google, eraman, aheejin, steven_wu, george.burgess.iv, dexonsmith, jfb, asbirlea, gchatelet, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66428
llvm-svn: 371284
Craig Topper [Sat, 7 Sep 2019 01:39:21 +0000 (01:39 +0000)]
[X86] Add tests for fp128 frem, sqrt, sin, and cos.
llvm-svn: 371283
Craig Topper [Sat, 7 Sep 2019 01:39:12 +0000 (01:39 +0000)]
[X86] Autogenerate fp128-libcalls.ll
llvm-svn: 371282
Fangrui Song [Sat, 7 Sep 2019 01:38:56 +0000 (01:38 +0000)]
[llvm-objcopy] Rename SHOffset (e_shoff) field to SHOff. NFC
Similar to D67254.
`struct Elf*_Shdr` has a field `sh_offset`. Rename SHOffset to SHOff to
avoid confusion.
llvm-svn: 371281
Jason Molenda [Sat, 7 Sep 2019 01:38:37 +0000 (01:38 +0000)]
Long timeouts for the MacOSX SystemRuntime plugins under ASAN; else quick.
In April via r357829, Adrian unified timeouts across lldb and set the
default value high so that we wouldn't get timeouts on ASAN bots that
were running under load.
The library that the MacOSX SystemRuntime has functions that need
to take a lock, and if that lock is held already, those functions
will never complete; we're seeing the 15 second timeout being hit
with inferiors that are doing a lot of enqueuing and dequeuing of
libdispatch work items causing this deadlocking behavior.
This patch reverts to a very short timeout for these SystemRuntime
function calls, given the behavior of this library that they are
calling into. When lldb is built with AddressSanitizer enabled,
they will use the default 15 second timeout.
tl;dr: this reverts to the previous timeouts for these SystemRuntime
inf func calls.
<rdar://problem/
54538149>
llvm-svn: 371280
Jan Korous [Sat, 7 Sep 2019 00:59:13 +0000 (00:59 +0000)]
[clang][DependencyFileGenerator] Fix missing -MT option handling
Targets in DependencyFileGenerator don't necessarily come from -MT option.
Differential Revision: https://reviews.llvm.org/D67308
llvm-svn: 371279
Reid Kleckner [Sat, 7 Sep 2019 00:41:08 +0000 (00:41 +0000)]
Fix thunks.cpp test, don't FileCheck for anon namespace id
The anon namespace id is a hash of the main input path to the compiler,
which varies in the test suite because the input path is absolute.
llvm-svn: 371277
Akira Hatanaka [Sat, 7 Sep 2019 00:34:47 +0000 (00:34 +0000)]
[Sema][ObjC] Mark C union fields that have non-trivial ObjC ownership
qualifications as unavailable if the union is declared in a system
header
r365985 stopped marking those fields as unavailable, which caused the
union's NonTrivialToPrimitive* bits to be set to true. This patch
restores the behavior prior to r365985, except that users can explicitly
specify the ownership qualification of the field to instruct the
compiler not to mark it as unavailable.
rdar://problem/
53420753
Differential Revision: https://reviews.llvm.org/D65256
llvm-svn: 371276
Akira Hatanaka [Sat, 7 Sep 2019 00:34:43 +0000 (00:34 +0000)]
[Sema] Diagnose default-initialization, destruction, and copying of
non-trivial C union types
This recommits r365985, which was reverted because it broke a few
projects using unions containing non-trivial ObjC pointer fields in
system headers. We now have a patch to fix the problem (see
https://reviews.llvm.org/D65256).
Original commit message:
This patch diagnoses uses of non-trivial C unions and structs/unions
containing non-trivial C unions in the following contexts, which require
default-initialization, destruction, or copying of the union objects,
instead of disallowing fields of non-trivial types in C unions, which is
what we currently do:
- function parameters.
- function returns.
- assignments.
- compound literals.
- block captures except capturing of `__block` variables by non-escaping blocks.
- local and global variable definitions.
- lvalue-to-rvalue conversions of volatile types.
See the discussion in https://reviews.llvm.org/D62988 for more background.
rdar://problem/
50679094
Differential Revision: https://reviews.llvm.org/D63753
llvm-svn: 371275
Richard Smith [Sat, 7 Sep 2019 00:11:53 +0000 (00:11 +0000)]
Synchronize LLVM's copy of libc++abi's demangler with the libc++abi
version after r371273.
Also fix a minor issue in r371273 that only surfaced after template
instantiation from LLVM's use of the demangler.
llvm-svn: 371274
Richard Smith [Fri, 6 Sep 2019 23:53:21 +0000 (23:53 +0000)]
Implement demangling support for C++20 lambda expression extensions.
This implements demangling support for the mangling extensions specified
in https://github.com/itanium-cxx-abi/cxx-abi/pull/85, much of which is
implemented in Clang r359967 and r371004.
Specifically, this provides demangling for:
* <template-param-decl> in <lambda-sig>
* <template-param> with non-zero level
* lambda-expression literals (not emitted by Clang yet)
* nullptr literals
* string literals
(The final two seem unrelated, but handling them was necessary in order
to disambiguate between lambda expressions and the other forms of
literal for which we have a type but no value.)
When demangling a <lambda-sig>, we form template parameters with no
corresponding argument, so we cannot substitute in the argument in the
demangling. Instead we invent synthetic names for the template
parameters (eg, '[]<typename $T>($T *x)').
llvm-svn: 371273
Reid Kleckner [Fri, 6 Sep 2019 22:55:26 +0000 (22:55 +0000)]
Use musttail for variadic method thunks when possible
This avoids cloning variadic virtual methods when the target supports
musttail and the return type is not covariant. I think we never
implemented this previously because it doesn't handle the covariant
case. But, in the MS ABI, there are some cases where vtable thunks must
be emitted even when the variadic method defintion is not available, so
it looks like we need to implement this. Do it for both ABIs, since it's
a nice size improvement and simplification for Itanium.
Emit an error when emitting thunks for variadic methods with a covariant
return type. This case is essentially not implementable unless the ABI
provides a way to perfectly forward variadic arguments without a tail
call.
Fixes PR43173.
Differential Revision: https://reviews.llvm.org/D67028
llvm-svn: 371269
Matthew Voss [Fri, 6 Sep 2019 22:51:46 +0000 (22:51 +0000)]
Update lld tests dynamic-list.s and symbol-override.s to use llvm-nm
The following tests failed on Windows bots due to nm not being
available:
lld/test/ELF/dynamic-list.s
lld/test/ELF/symbol-override.s
llvm-svn: 371267
Amara Emerson [Fri, 6 Sep 2019 22:27:09 +0000 (22:27 +0000)]
[AArch64][GlobalISel] Enable the localizer for optimized builds.
Despite the fact that the localizer's original motivation was to fix horrendous
constant spilling at -O0, shortening live ranges still has net benefits even
with optimizations enabled.
On an -Os build of CTMark, doing this improves code size by 0.5% geomean.
There are a few regressions, bullet increasing in size by 0.5%. One example from
bullet where code size increased slightly was due to GlobalISel actually now
generating the same code as SelectionDAG. So we actually have an opportunity
in future to implement better heuristics for localization and therefore be
*better* than SDAG in some cases. In relation to other optimizations though that
one is relatively minor.
Differential Revision: https://reviews.llvm.org/D67303
llvm-svn: 371266
Evandro Menezes [Fri, 6 Sep 2019 22:07:11 +0000 (22:07 +0000)]
[InstCombine] Refactor substitution of instruction in the parent BB (NFC)
Add the new method `LibCallSimplifier::substituteInParent()` that calls
`LibCallSimplifier::replaceAllUsesWith()' and
`LibCallSimplifier::eraseFromParent()` back to back, simplifying the
resulting code.
llvm-svn: 371264
Lang Hames [Fri, 6 Sep 2019 21:55:43 +0000 (21:55 +0000)]
[ORC][RPC] Join server thread before checking condition in unit test.
Otherwise we have a race on the sent-messages count.
llvm-svn: 371263
Nick Desaulniers [Fri, 6 Sep 2019 21:50:11 +0000 (21:50 +0000)]
[IR] CallBrInst: scan+update arg list when indirect dest list changes
Summary:
There's an unspoken invariant of callbr that the list of BlockAddress
Constants in the "function args" list match the BasicBlocks in the
"other labels" list. (This invariant is being added to the LangRef in
https://reviews.llvm.org/D67196).
When modifying the any of the indirect destinations of a callbr
instruction (possible jump targets), we need to update the function
arguments if the argument is a BlockAddress whose BasicBlock refers to
the indirect destination BasicBlock being replaced. Otherwise, many
transforms that modify successors will end up violating that invariant.
A recent change to the arm64 Linux kernel exposed this bug, which
prevents the kernel from booting.
I considered maintaining a mapping from indirect destination BasicBlock
to argument operand BlockAddress, but this ends up being a one to
potentially many (though usually one) mapping. Also, the list of
arguments to a function (or more typically inline assembly) ends up
being less than 10. The implementation is significantly simpler to just
rescan the full list of arguments. Because of the one to potentially
many relationship, the full arg list must be scanned (we can't stop at
the first instance).
Thanks to the following folks that reported the issue and helped debug
it:
* Nathan Chancellor
* Will Deacon
* Andrew Murray
* Craig Topper
Link: https://bugs.llvm.org/show_bug.cgi?id=43222
Link: https://github.com/ClangBuiltLinux/linux/issues/649
Link: https://lists.infradead.org/pipermail/linux-arm-kernel/2019-September/678330.html
Reviewers: craig.topper, chandlerc
Reviewed By: craig.topper
Subscribers: void, javed.absar, kristof.beyls, hiraditya, llvm-commits, nathanchance, srhines
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67252
llvm-svn: 371262
Craig Topper [Fri, 6 Sep 2019 21:49:01 +0000 (21:49 +0000)]
[X86] Add a AVX512VBMI command line to min-legal-vector-width.ll. Always enable fast-variable-shuffle
Trying to minimize the features we need to manipulate when this
is updated for D67259.
The VBMI is interesting because it enables some improved combining
for truncates.
I enabled fast-variable-shuffle because all the CPUs we're going
to add implicitly enable it. So they can share check lines.
llvm-svn: 371261
Craig Topper [Fri, 6 Sep 2019 21:48:44 +0000 (21:48 +0000)]
[X86] Replace -mcpu with -mattr on some tests.
llvm-svn: 371260
Jonas Devlieghere [Fri, 6 Sep 2019 21:43:33 +0000 (21:43 +0000)]
[test] Add a FIXME test for stop-command-source-on-error
Modifying the interpreter settings is tricky because they don't take
effect until we create a new command interpreter, which should be merely
an implementation detail. This leads to confusing and unexpected
scenarios.
This adds a test cases with FIXMEs for some of the odd scenarios I
encountered. I didn't XFAIL the test because I don't think there's a way
to get an unexpected PASS if any of the commands succeeds and splitting
up the file in multiple tests seems excessive.
llvm-svn: 371259
Alex Langford [Fri, 6 Sep 2019 21:05:21 +0000 (21:05 +0000)]
[Core] Remove use of ClangASTContext in DumpDataExtractor
Summary:
DumpDataExtractor uses ClangASTContext in order to get the proper llvm
fltSemantics for the type it needs so that it can dump floats in a more
precise way. However, there's no reason that this behavior needs to be
specific ClangASTContext. Instead, I think it makes sense to ask
TypeSystems for the float semantics for a type of a given size.
Differential Revision: https://reviews.llvm.org/D67239
llvm-svn: 371258
Artem Dergachev [Fri, 6 Sep 2019 20:55:29 +0000 (20:55 +0000)]
[analyzer] Add minimal support for fix-it hints.
Allow attaching fixit hints to Static Analyzer BugReports.
Fixits are attached either to the bug report itself or to its notes
(path-sensitive event notes or path-insensitive extra notes).
Add support for fixits in text output (including the default text output that
goes without notes, as long as the fixit "belongs" to the warning).
Add support for fixits in the plist output mode.
Implement a fixit for the path-insensitive DeadStores checker. Only dead
initialization warning is currently covered.
Implement a fixit for the path-sensitive VirtualCall checker when the virtual
method is not pure virtual (in this case the "fix" is to suppress the warning
by qualifying the call).
Both fixits are under an off-by-default flag for now, because they
require more careful testing.
Differential Revision: https://reviews.llvm.org/D65182
llvm-svn: 371257
Artem Dergachev [Fri, 6 Sep 2019 20:55:24 +0000 (20:55 +0000)]
[analyzer] pr43179: Make CallDescription defensive against C variadic functions.
Most functions that our checkers react upon are not C-style variadic functions,
and therefore they have as many actual arguments as they have formal parameters.
However, it's not impossible to define a variadic function with the same name.
This will crash any checker that relies on CallDescription to check the number
of arguments but silently assumes that the number of parameters is the same.
Change CallDescription to check both the number of arguments and the number of
parameters by default.
If we're intentionally trying to match variadic functions, allow specifying
arguments and parameters separately (possibly omitting any of them).
For now we only have one CallDescription which would make use of those,
namely __builtin_va_start itself.
Differential Revision: https://reviews.llvm.org/D67019
llvm-svn: 371256
Lang Hames [Fri, 6 Sep 2019 20:50:00 +0000 (20:50 +0000)]
[ORC] Add a missing #include atomic.
Hopefully this will fix the bot build failures from r371245.
llvm-svn: 371255
Matt Arsenault [Fri, 6 Sep 2019 20:49:10 +0000 (20:49 +0000)]
GlobalISel: Add G_FMAD instruction
llvm-svn: 371254
Matt Arsenault [Fri, 6 Sep 2019 20:32:37 +0000 (20:32 +0000)]
GlobalISel: Support physical register inputs in patterns
llvm-svn: 371253
Richard Smith [Fri, 6 Sep 2019 20:32:01 +0000 (20:32 +0000)]
Update comment for <template-param-decl> mangling to match extended mangling rule in r371004.
llvm-svn: 371252
Reid Kleckner [Fri, 6 Sep 2019 20:12:44 +0000 (20:12 +0000)]
Remove dead .seh_stackalloc parsing method in X86AsmParser
The shared COFF asm parser code handles this directive, since it is
shared with AArch64. Spotted by Alexandre Ganea in review.
llvm-svn: 371251
Jan Korous [Fri, 6 Sep 2019 20:08:32 +0000 (20:08 +0000)]
[clang][Index][NFC] Put IndexingOptions to a separate header
llvm-svn: 371250
Matt Arsenault [Fri, 6 Sep 2019 20:00:22 +0000 (20:00 +0000)]
AMDGPU: Fix typo
llvm-svn: 371249
Puyan Lotfi [Fri, 6 Sep 2019 19:59:59 +0000 (19:59 +0000)]
[llvm-ifs] Improving detection of PlatformKind from triple for TBD generation.
It was pointed out that I had hard-coded PlatformKind. This is rectifying that.
Differential Revision: https://reviews.llvm.org/D67255
llvm-svn: 371248
Sean Fertile [Fri, 6 Sep 2019 19:55:44 +0000 (19:55 +0000)]
[PowerPC][XCOFF] Remove basic test. [NFC]
Test verified that we could compile an empty module and produce an XCOFF
object file. Newer tests superssed this coverage, its safe to remove.
llvm-svn: 371247
Evandro Menezes [Fri, 6 Sep 2019 19:41:49 +0000 (19:41 +0000)]
[ConstantFolding] Add new test cases for transcendentals (NFC)
llvm-svn: 371246
Lang Hames [Fri, 6 Sep 2019 19:21:59 +0000 (19:21 +0000)]
[ORC] Make sure RPC channel-send is called in blocking calls and responses.
ORC-RPC batches calls by default, and the channel's send method must be called
to transfer any buffered calls to the remote. The call to send was missing on
responses and blocking calls in the SingleThreadedRPCEndpoint. This patch adds
the necessary calls and modifies the RPC unit test to check for them.
llvm-svn: 371245
Lang Hames [Fri, 6 Sep 2019 19:21:55 +0000 (19:21 +0000)]
[llvm-jitlink] Add optional slab allocator for testing locality optimizations.
The llvm-jitlink utility now accepts a '-slab-allocate <size>' option. If given,
llvm-jitlink will use a slab-based memory manager rather than the default
InProcessMemoryManager. Using a slab allocator will allow reliable testing of
future locality based optimizations (e.g. PLT and GOT elimination) in JITLink.
The <size> argument is a number, optionally followed by a units specifier (Kb,
Mb, or Gb). If the units are not given then the number is assumed to be in Kb.
llvm-svn: 371244
Craig Topper [Fri, 6 Sep 2019 19:17:02 +0000 (19:17 +0000)]
[X86] Use MOVSX by default instead of CBW to extend i8 to AX for i8 sdivrem.
We can use a MOVSX16 here then rely on FixupBWInst to change to
MOVSX32 if the upper bits are dead. With a special case to
not promote if it could be turned into CBW.
Then we can rely on X86MCInstLower to turn the MOVSX into CBW
very late if register allocation worked out.
Using MOVSX gives an opportunity to use the MOVSX as a both a
copy and a sign extend since the input and output register aren't
tied together.
Differential Revision: https://reviews.llvm.org/D67192
llvm-svn: 371243
Craig Topper [Fri, 6 Sep 2019 19:15:04 +0000 (19:15 +0000)]
[X86] Use MOVZX16rr8/MOVZXrm8 when extending input for i8 udivrem.
We can rely on X86FixupBWInsts to turn these into MOVZX32. This
simplifies a follow up commit to use MOVSX for i8 sdivrem with
a late optimization to use CBW when register allocation works out.
llvm-svn: 371242
Matthias Gehre [Fri, 6 Sep 2019 19:15:02 +0000 (19:15 +0000)]
[LifetimeAnalysis] don't use raw string literals in macros
They broke the AArch64 bots (gcc does not support it)
llvm-svn: 371241
Craig Topper [Fri, 6 Sep 2019 19:14:49 +0000 (19:14 +0000)]
[X86] Teach FixupBWInsts to turn MOVSX16rr8/MOVZX16rr8/MOVSX16rm8/MOVZX16rm8 into their 32-bit dest equivalents when the upper part of the register is dead.
llvm-svn: 371240
Sean Fertile [Fri, 6 Sep 2019 18:56:14 +0000 (18:56 +0000)]
[PowerPC][XCOFF] Verify symbol table in xcoff object files. [NFC]
Extend the common/local-common testing for object files to also verify the
symbol table now that the needed functionality has landed in llvm-readobj.
Differential Revision: https://reviews.llvm.org/D66944
llvm-svn: 371237