Instcombine folds (a + b <u a) to (a ^ -1 <u b) and that does not match
the expected pattern in CodeGenPerpare via UAddWithOverflow.

This causes a regression over Clang 7 on both X86 and AArch64:
https://gcc.godbolt.org/z/juhXYV

This patch extends UAddWithOverflow to also catch the XOR case, if the
XOR is only used in the ICMP. This covers just a single case, but I'd
like to make sure I am not missing anything before tackling the other
cases.

Reviewers: nikic, RKSimon, lebedev.ri, spatel

Reviewed By: nikic, lebedev.ri

Differential Revision: https://reviews.llvm.org/D74228

I'm not sure why this isn't a pattern, but the DAG manually selects
this.

Differential Revision: https://reviews.llvm.org/D74822

All these tests can just call lldbutil.run_to_source_breakpoint
instead of reimplementing it.

TestCPPAuto was only failing on windows due to the std::string
copying (which was not related at all to 'auto' functionality).

TestStepTarget is now also passing but that seems more that we
now have by accident the right behavior in Windows. I'll remove
the x-fail just to make the bot green again.

Summary:
Previously TableGen would crash trying to print the undefined value as
an integer.

Change-Id: I3900071ceaa07c26acafb33bc49966d7d7a02828

Reviewers: nhaehnle

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74210

The `-fdeclare-opencl-builtins` option was accepting saturated
conversions for non-integer types, which contradicts both the OpenCL
specification (v2.0 s6.2.3) and Clang's opencl-c.h file.

[lldb][NFC] Pointer to reference conversion for CompilerDeclContext params in ClangExpressionDeclMap.

Follow up for f9568a95.

Summary: Commit 63bb9fee was reverted in
7603bfb4 because it broke builds that treat
warnings as errors.
This commit updates the calls to `assembleToStream()` in tests to check that
the return value is valid.

Original commit message:

Followup to D74084.
Replace the use of `report_fatal_error()` with returning the error to
`llvm-exegesis.cpp` and handling it there.

Differential Revision: https://reviews.llvm.org/D74325

The only thing needed was to account for the offset from the
debug_cu_index section when searching for the location list.

This patch also fixes a bug in the Module::ParseAllDebugSymbols
function, which meant that we would only parse the variables of the
first compile unit in the module. This function is only used from
lldb-test, so this does not fix any real issue, besides preventing me
from writing a test for this patch.

This test was getting a bit long. Before adding more checks, group the
existing checks according to the matcher used, and break it up into
smaller tests.

Check that no Q-regs are live out of the loop, unless the instruction
within the loop is predicated on the vctp.

Differential Revision: https://reviews.llvm.org/D72713

Differential revision: https://reviews.llvm.org/D74768

Similar to VADDV and VADDLV that have been added recently, this adds
lowering and patterns for VMLAV, VMLAVA, VMLALV and VMLALVA. They
perform the same roles as the add's, just folding a mul into the same
instruction (and so taking two inputs). As such, they need to be lowered
in the same way as the types are often not legal.

Differential Revision: https://reviews.llvm.org/D74390

Comparing those two `const char *` values relies on the assumption that both
strings were created by a ConstString. Let's check that assumption with an
assert as otherwise this code silently does nothing and that's not great.

Summary:
This patch adds codegen support for the ACLE builtin types added in:

  https://reviews.llvm.org/D62960

so that the ACLE builtin types are emitted as corresponding scalable
vector types in LLVM.

Reviewers: rsandifo-arm, rovka, rjmccall, efriedma

Reviewed By: efriedma

Subscribers: tschuett, kristof.beyls, hiraditya, rkruppe, psnobl, llvm-commits, cfe-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D74724

.dynsym and .dynstr are allocatable and therefore normally are placed
before non-allocatable .strtab, .shstrtab, .symtab sections.
But we are placing them after currently what creates a mix of
alloc/non-alloc sections and does not look normal.

Differential revision: https://reviews.llvm.org/D74756

Reverting because this patch is causing ~20 llvm-test-suite failures on
a number of different bots:
* http://lab.llvm.org:8011/builders/clang-cmake-armv8-lld/builds/3366
* http://lab.llvm.org:8011/builders/clang-cmake-aarch64-lld/builds/8222
* http://lab.llvm.org:8011/builders/clang-cmake-x86_64-avx2-linux/builds/13275
* http://lab.llvm.org:8011/builders/clang-s390x-linux-lnt/builds/17213

This reverts commit cd2c5af6.

This is part of the work to remove SelectionDAG::GetDemandedBits and just use SimplifyMultipleUseDemandedBits.

Recent experiments raised some v_cvt_f32_ubyte*_e32 regressions, so I've added some additional abilities to performCvtF32UByteNCombine to help unpack byte data more aggressively.

We still don't remove all OR(SHL,SRL) patterns as some of the regenerated nodes don't get combined again, but we are getting closer.

Differential Revision: https://reviews.llvm.org/D74786

Following on from the extra VADDV lowering, this extends things to
handle VADDLV which allows summing values into a pair of i32 registers,
together treated as a i64. This needs to be done in DAGCombine too as
the types are otherwise illegal, which is a fairly simple addition on
top of the existing code.

There is also a VADDLVA instruction handled here, that adds the incoming
values from the two general purpose registers. As opposed to the
non-long version where we could just add patterns for add(x, VADDV), the
long version needs to handle this early before the i64 has being split
into too many pieces.

Differential Revision: https://reviews.llvm.org/D74224

Custom legalize non-power-of-2 and unaligned load and store for MIPS32r5
and older, custom legalize non-power-of-2 load and store for MIPS32r6.

Don't attempt to combine non power of 2 loads or unaligned loads when
subtarget doesn't support them (MIPS32r5 and older).

Differential Revision: https://reviews.llvm.org/D74625

Improve legality checks for load and store, 4 byte scalar
load and store are now legal for all subtargets.
During regbank selection 4 byte unaligned loads and stores
for MIPS32r5 and older get mapped to gprb.
Select 4 byte unaligned loads and stores for MIPS32r5.
Fix tests that unintentionally had unaligned load or store.

Differential Revision: https://reviews.llvm.org/D74624

On some targets, like SPARC, forming overflow ops is only profitable if
the math result is used: https://godbolt.org/z/DxSmdB
This patch adds a new MathUsed parameter to allow the targets
to make the decision and defaults to only allowing it
if the math result is used. That is the conservative choice.

This patch also updates AArch64ISelLowering, X86ISelLowering,
ARMISelLowering.h, SystemZISelLowering.h to allow forming overflow
ops if the math result is not used. On those targets using the
overflow intrinsic for the overflow check only generates better code.

Reviewers: nikic, RKSimon, lebedev.ri, spatel

Reviewed By: lebedev.ri

Differential Revision: https://reviews.llvm.org/D74722

Summary:
Implements the following intrinsics:
 - @llvm.aarch64.sve.eorbt
 - @llvm.aarch64.sve.eortb
 - @llvm.aarch64.sve.pmullb.pair
 - @llvm.aarch64.sve.pmullt.pair

Reviewers: sdesmalen, c-rhodes, dancgr, cameron.mcinally, efriedma, rengolin

Reviewed By: efriedma

Subscribers: tschuett, kristof.beyls, hiraditya, rkruppe, psnobl, cfe-commits, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D74769

Differential Revision: https://reviews.llvm.org/D73534

With this patch lld recognizes ARM SBREL relocations.
R_ARM*_MOVW_BREL relocations are not tested because they are not used.

Patch by Tamas Petz

Differential Revision: https://reviews.llvm.org/D74604

This should use -> instead of '.', but the fix-it functionality of
the expression evaluator saved us here. Let's use the proper syntax
in the first place as we don't want to test fix-its here.

This directory escaped the modularization effort it seems. Just adding
this to the Host module along with the other common headers, which should
make this code less likely to break under modules and speed up compilation.

We already make use of the VADDV vector reduction instruction for cases
where the input and the output start out at the same type. The MVE
instruction however will sum into an i32, so if we are summing a v16i8
into an i32, we can still use the same instructions. In terms of IR,
this looks like a sext of a legal type (v16i8) into a very illegal type
(v16i32) and a vecreduce.add of that into the result. This means we have
to catch the pattern early in a DAG combine, producing a target VADDVs/u
node, where the signedness is now important.

This is the first part, handling VADDV and VADDVA. There are also
VADDVL/VADDVLA instructions, which are interesting because they sum into
a 64bit value. And VMLAV and VMLALV, which are interesting because they
also do a multiply of two values. It may look a little odd in places as
a result.

On it's own this will probably not do very much, as the vectorizer will
not produce this IR yet.

Differential Revision: https://reviews.llvm.org/D74218

The tests added in D74425/commit a71feda2
fail with an assertion on macOS, as they seem to require ELF support.

Passing a linux triple ensures the object files are using ELF.

This fixes some GreenDragon failures.

Summary:
Currently when printing data types we include implicit scopes such as inline namespaces or anonymous namespaces.
This leads to command output like this (for `std::set<X>` with X being in an anonymous namespace):

```
(lldb) print my_set
(std::__1::set<(anonymous namespace)::X, std::__1::less<(anonymous namespace)::X>, std::__1::allocator<(anonymous namespace)::X> >) $0 = size=0 {}
```

This patch removes all the implicit scopes when printing type names in TypeSystemClang::GetDisplayTypeName
so that our output now looks like this:

```
(lldb) print my_set
(std::set<X, std::less<X>, std::allocator<X> >) $0 = size=0 {}
```

As previously GetDisplayTypeName and GetTypeName had the same output we actually often used the
two as if they are the same method (they were in fact using the same implementation), so this patch also
fixes the places where we actually want the display type name and not the actual type name.

N...

All calls to operator new in this test fail for me with:
```
expression --show-types -- *(new foo(47))`
Error output:
error: Execution was interrupted, reason: internal c++ exception breakpoint(-6)..
The process has been returned to the state before expression evaluation.
```

As calling operator new isn't the idea of this test, this patch moves that
logic to the binary with some new_* utility functions and explicitly tests
this logic in the constructor test (where we can isolate the failures and
skip them on Linux).

Differential Revision: https://reviews.llvm.org/D74708

Consider large operands in G_MERGE_VALUES and G_UNMERGE_VALUES as
Ambiguous during regbank selection.
Introducing new InstType AmbiguousWithMergeOrUnmerge which will
allow us to recognize whether to narrow scalar or use s64:fprb.

This change exposed a bug when reusing data from TypeInfoForMF.
Thus when Instr is about to get destroyed (using narrow scalar)
clear its data in TypeInfoForMF. Internal data is saved based on
Instr's address, and it will no longer be valid.
Add detailed asserts for InstType and operand size.

Generate generic instructions instead of MIPS target instructions
during argument lowering and custom legalizer.
Select G_UNMERGE_VALUES and G_MERGE_VALUES when proper banks are
selected: {s32:gprb, s32:gprb, s64:fprb} for G_UNMERGE_VALUES and
{s64:fprb, s32:gprb, s32:gprb} for G_MERGE_VALUES.
Update tests. One improvement is when floating point argument in
gpr(or two gprs) gets passed to another function through gpr
unnecessary fpr-to-gpr moves are no longer generated.

Differential Revision: https://reviews.llvm.org/D74623