The names of the tablegen defs now match the names of the ISD nodes.
This makes the world a slightly saner place, as previously "fround" matched
ISD::FP_ROUND and not ISD::FROUND.

Differential Revision: https://reviews.llvm.org/D23597

llvm-svn: 279129

Differential Revision: http://reviews.llvm.org/D23689

llvm-svn: 279126

We abort building vectorizable trees in some cases (e.g., if the maximum
recursion depth is reached, if the region size is too large, etc.). If this
happens for a reduction, we can be left with a root entry that needs to be
gathered. For these cases, we need make sure we actually set VectorizedValue to
the resulting vector.

This patch ensures we properly set VectorizedValue, and it also ensures the
insertelement sequence generated for the gathers is inserted at the correct
location.

Reference: https://llvm.org/bugs/show_bug.cgi?id=28330
Differential Revison: https://reviews.llvm.org/D23410

llvm-svn: 279125

Re-apply r276044 with off-by-1 instruction fix for the reload placement.

This is a variant of scavengeRegister() that works for
enterBasicBlockEnd()/backward(). The benefit of the backward mode is
that it is not affected by incomplete kill flags.

This patch also changes
PrologEpilogInserter::doScavengeFrameVirtualRegs() to use the register
scavenger in backwards mode.

Differential Revision: http://reviews.llvm.org/D21885

llvm-svn: 279124

This is prep work for allowing the threshold to be different during layout,
and to enforce a single threshold between merging and duplicating during
layout. No observable change intended.

llvm-svn: 279117

When running 'opt -O2 verify-uselistorder-nodbg.lto.bc', there are 33m allocations.  8.2m
come from std::string allocations in Intrinsic::getName().  Turns out this method only
returns a std::string because it needs to handle overloads, but that is not the common case.

This adds an overload of getName which just returns a StringRef when there are no overloads
and so saves on the allocations.

llvm-svn: 279113

llvm-svn: 279111

The Xcode and Visual Studio generators always log "-- No build type selected, default to Debug". This is because CMake doesn't initialize "CMAKE_CONFIGURATION_TYPES" until the generator's EnableLanguage call gets hit.

The first place EnableLanguage gets hit in our configuration is in the project() call. Since CMAKE_BUILD_TYPE isn't used until after we call project() it is safe to just move this check down a bit.

llvm-svn: 279110

Summary: PR27453

Reviewers: eugenis

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D23647

llvm-svn: 279109

Differential revision: https://reviews.llvm.org/D23666

llvm-svn: 279106

Normally, when an AND with a constant is lowered to NILL, the constant value is truncated to 16 bits. However, since r274066, ANDs whose results are used in a shift are caught by a different pattern that does not truncate. The instruction printer expects a 16-bit unsigned immediate operand for NILL, so this results in an abort.

This patch adds code to manually truncate the constant in this situation. The rest of the bits are then set, so we will detect a case for NILL "naturally" rather than using peephole optimizations.

Differential Revision: http://reviews.llvm.org/D21854

llvm-svn: 279105

Remove an unnecessary round-trip:

    iterator => operator->() => getIterator()

In some cases, the iterator is end(), so the dereference of operator->
is invalid (UB).

The testcase only crashes with r278974 (currently reverted to
investigate this), which adds an assertion for invalid dereferences of
ilist nodes.

Fixes PR29035.

llvm-svn: 279104

Differential revision: https://reviews.llvm.org/D23675

llvm-svn: 279102

This is a sibling of:
https://reviews.llvm.org/rL278859
https://reviews.llvm.org/rL278935
https://reviews.llvm.org/rL278945
https://reviews.llvm.org/rL279066
https://reviews.llvm.org/rL279077

llvm-svn: 279101

The WebAssemly spec removing the return value from store instructions, so
remove the associated optimization from LLVM.

This patch leaves the store instruction operands in place for now, so stores
now always write to "$drop"; these will be removed in a seperate patch.

llvm-svn: 279100

number of assume intrinsics.

The classical way to have a cache-friendly vector style container when
we need queue semantics for BFS instead of stack semantics for DFS is to
use an ever-growing vector and an index. Erasing from the front requires
O(size) work, and unless we expect the worklist to grow *very* large,
its probably cheaper to just grow and race down the list.

But that makes it more bad that we're putting the assume intrinsics in
this at all. We end up looking at the (by definition empty) use list to
see if they're ephemeral (when we've already put them in that set), etc.

Instead, directly populate the worklist with the operands when we mark
the assume intrinsics as ephemeral. Also, test the visited set *before*
putting things into the worklist so we don't accumulate the same value
in the list 100s of times.

It would be nice to use a set-vector for this but I think its useful to
test the set earlier to avoid repeatedly querying whether the same
instruction is safe to speculate.

Hopefully with these changes the number of values pushed onto the
worklist is smaller, and we avoid quadratic work by letting it grow as
necessary.

Differential Revision: https://reviews.llvm.org/D23396

llvm-svn: 279099

llvm-svn: 279095

1. Better variable names
2. Remove unnecessary check of ConstantInt

llvm-svn: 279094

This reverts commit r279086, reapplying r279084.  I'm not sure what I
ran before, because the compile failure for ADTTests reproduced locally.

The problem is that TestRev is calling BidirectionalVector::rbegin()
when the BidirectionalVector is const, but rbegin() is always non-const.
I've updated BidirectionalVector::rbegin() to be callable from const.

Original commit message follows.

--

As a follow-up to r278991, add some tests that check that
decltype(reverse(R).begin()) == decltype(R.rbegin()), and get them
passing by adding std::remove_reference to has_rbegin.

I'm using static_assert instead of EXPECT_TRUE (and updated the other
has_rbegin check from r278991 in the same way) since I figure that's
more helpful.

llvm-svn: 279091

The original patch was breaking some buildbots due to an
incorrect ordering of function definitions which caused some
compilers to recognize a definition but others to not.

llvm-svn: 279089

This adds behaviour similar to binutils' objdump which can show symbols in an
import library.  Differences from that stem around the fact that we do not
create section symbols nor the all import import descriptor symbol reference.
However, this does mean that the tool can serve as a possible replacement for
the existing tool.

llvm-svn: 279088

This reverts commit r279084, since it failed on a bot:
  http://bb.pgr.jp/builders/cmake-llvm-x86_64-linux/builds/41733

llvm-svn: 279086

As a follow-up to r278991, add some tests that check that
decltype(reverse(R).begin()) == decltype(R.rbegin()), and get them
passing by adding std::remove_reference to has_rbegin.

I'm using static_assert instead of EXPECT_TRUE (and updated the other
has_rbegin check from r278991 in the same way) since I figure that's
more helpful.

llvm-svn: 279084

It causes a regression on our internal benchmark. Introduce cvp-dont-process flag and set it off by default while investigating the regression.

llvm-svn: 279082

There is no FREM instruction, but the others are straightforward.

llvm-svn: 279081

llvm-svn: 279080

llvm-svn: 279079

The ARMv8*-A descriptions in the ARM and AArch64 TargetParsers are incorrect
architecturally and mismatched to the backend descriptions.

RAS is an optional extension to ARMv8-A and ARMv8.1-A and mandatory in
ARMv8.2-A. Correct the ARMTargetParser descriptions which had this as enabled
by default in the earlier versions.

The FP16 and SPE extensions are optional in ARMv8.2-A and the backend defaults
them as off. They are not available as extensions to earlier ARMv8-A versions.
Correct the AArch64TargetParser which had these as enabled by default in all
ARMv8-A definitions.

These macros are only used to define preprocessor macros. There are no macros
yet as ACLE has not caught up with ARMv8.2-A so not possible to add a test.

Differential Revision: https://reviews.llvm.org/D23500

llvm-svn: 279078

This is a sibling of:
https://reviews.llvm.org/rL278859
https://reviews.llvm.org/rL278935
https://reviews.llvm.org/rL278945
https://reviews.llvm.org/rL279066

llvm-svn: 279077

This patch changes the code structure of
WebAssemblyLowerEmscriptenException pass to support both exception
handling and setjmp/longjmp. It also changes the name of the pass and
the source file.

1. Change the file/pass name to WebAssemblyLowerEmscriptenExceptions ->
WebAssemblyLowerEmscriptenEHSjLj to make it clear that it supports both
EH and SjLj
2. List function / global variable names at the top so they
can be changed easily
3. Some cosmetic changes

Patch by Heejin Ahn

Differential Revision: https://reviews.llvm.org/D23588

llvm-svn: 279075

There is no REM instruction; that will require an expansion.
It's not obvious that should be done in select, rather than as a
(custom?) legalization.

llvm-svn: 279074

llvm-svn: 279073

This will enable vector splat folding, but NFC until the callers
have their ConstantInt restrictions removed.

llvm-svn: 279072

`link -dump -exports` lists exported symbols from import libraries as well as
normal dlls.  Ensure that we can handle import libraries as well in
llvm-readobj.

llvm-svn: 279069

llvm-svn: 279068

llvm-svn: 279067

This is a sibling of:
https://reviews.llvm.org/rL278859
https://reviews.llvm.org/rL278935
https://reviews.llvm.org/rL278945

llvm-svn: 279066

llvm-svn: 279065

r277708 enabled tails calls for MIPS but used the 'jr' instruction when the 
jump target was held in a register. For MIPSR6, 'jalr $zero, $reg' should
have been used. Additionally, add missing patterns for external and global
symbols for tail calls.

Reviewers: dsanders, vkalintiris

Differential Review: https://reviews.llvm.org/D23301

llvm-svn: 279064

llvm-svn: 279063