Disable LSR retry by default.

Disabling aggressive LSR saves compilation time, and with the new
indvars behavior usually improves performance.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140590 91177308-0d34-0410-b5e6-96231b3b80d8

Last batch of test conversions to new atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140585 91177308-0d34-0410-b5e6-96231b3b80d8

LSR, one of the new Cost::isLoser() checks did not get merged in the previous checkin.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140583 91177308-0d34-0410-b5e6-96231b3b80d8

Convert a bunch more tests over to the new atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140582 91177308-0d34-0410-b5e6-96231b3b80d8

Remove extraneous commit garbage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140581 91177308-0d34-0410-b5e6-96231b3b80d8

LSR cost metric minor fix and verification.

The minor bug heuristic was noticed by inspection. I added the
isLoser/isValid helpers because they will become more
important with subsequent checkins.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140580 91177308-0d34-0410-b5e6-96231b3b80d8

Fix an incorrect decoder test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140579 91177308-0d34-0410-b5e6-96231b3b80d8

Remove old hack for compiling with gcc-4.0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140573 91177308-0d34-0410-b5e6-96231b3b80d8

Remove incorrect testcases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140572 91177308-0d34-0410-b5e6-96231b3b80d8

Set register class of a register according to value of HasMips64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140570 91177308-0d34-0410-b5e6-96231b3b80d8

Define variable HasMips64 in MipsTargetLowering.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140569 91177308-0d34-0410-b5e6-96231b3b80d8

In single float mode, double precision FP arguments are passed in integer
registers, so there is no need to check here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140568 91177308-0d34-0410-b5e6-96231b3b80d8

Convert more tests to new atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140567 91177308-0d34-0410-b5e6-96231b3b80d8

Convert more tests over to the new atomic instructions.

I did not convert Atomics-32.ll and Atomics-64.ll by hand; the diff is autoupgrade output.

The wmb test is gone because there isn't any way to express wmb with the new atomic instructions; if someone really needs a non-asm way to write a wmb on Alpha, a platform-specific intrisic could be added.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140566 91177308-0d34-0410-b5e6-96231b3b80d8

Fix grammar.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140564 91177308-0d34-0410-b5e6-96231b3b80d8

Remove dead table entry.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140563 91177308-0d34-0410-b5e6-96231b3b80d8

Some minor (and more involved) cleanups. No real context changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140561 91177308-0d34-0410-b5e6-96231b3b80d8

ASR #32 is not allowed on Thumb2 USAT and SSAT instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140560 91177308-0d34-0410-b5e6-96231b3b80d8

Convert more tests over to the new atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140559 91177308-0d34-0410-b5e6-96231b3b80d8

Upgrade a couple more tests to the new atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140558 91177308-0d34-0410-b5e6-96231b3b80d8

Enhance alias analysis for atomic instructions a bit.  Upgrade a couple alias-analysis tests to the new atomic instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140557 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Fix memcpy intrinsic to handle 64-bit pointers

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140556 91177308-0d34-0410-b5e6-96231b3b80d8

Fix this test so it doesn't fail on Mac.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140553 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Implement PTXSelectionDAGInfo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140549 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Implement ISD::ANY_EXTEND

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140548 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Fix detection of stack load/store vs. global load/store, as well as fix the
printing of local offsets

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140547 91177308-0d34-0410-b5e6-96231b3b80d8

Fix emission of debug data for global variables. getContext() on DIGlobalVariables is not valid any more.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140539 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Add .align tests to stack object test file

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140537 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: SM > 2.0 implies +double

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140536 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Fix some lingering issues with stack allocation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140535 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Split up the TableGen instruction definitions into logical units

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140534 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Unify handling of loads/stores

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140533 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Handle FrameIndex nodes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140532 91177308-0d34-0410-b5e6-96231b3b80d8

Only run tests in test/CodeGen/CBackend/X86 when both X86 and CBackend are supported

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140517 91177308-0d34-0410-b5e6-96231b3b80d8

PR11004: Inline memcpy to avoid generating nested call sequence. Un-XFAIL 2011-06-09-TailCallByVal and 2010-11-04-BigByval

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140516 91177308-0d34-0410-b5e6-96231b3b80d8

Fix VEX decoding in i386 mode. Fixes PR11008.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140515 91177308-0d34-0410-b5e6-96231b3b80d8

Add target hook for pseudo instruction expansion.

Many targets use pseudo instructions to help register allocation.  Like
the COPY instruction, these pseudos can be expanded after register
allocation.  The early expansion can make life easier for PEI and the
post-ra scheduler.

This patch adds a hook that is called for all remaining pseudo
instructions from the ExpandPostRAPseudos pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140472 91177308-0d34-0410-b5e6-96231b3b80d8

[vector-select] Address one of the issues in pr10902. EXTRACT_VECTOR_ELEMENT
SDNodes may return values which are wider than the incoming element types. In
this patch we fix the integer promotion of these nodes.

Fixes spill-q.ll when running -promote-elements.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140471 91177308-0d34-0410-b5e6-96231b3b80d8

Clean up code after renaming LowerSubregs -> ExpandPostRAPseudos.

No functional change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140470 91177308-0d34-0410-b5e6-96231b3b80d8

Rename LowerSubregs to ExpandPostRAPseudos.

I'll fix the file contents in the next commit.

This pass is currently expanding the COPY and SUBREG_TO_REG pseudos. I
am going to add a hook so targets can expand more pseudo-instructions
after register allocation.

Many targets have pseudo-instructions that assist the register
allocator.  They can be expanded after register allocation, before PEI
and PostRA scheduling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140469 91177308-0d34-0410-b5e6-96231b3b80d8

Sort CMakeLists.txt.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140465 91177308-0d34-0410-b5e6-96231b3b80d8

Implement Duncan's suggestion to use the result of getSetCCResultType if it is legal
(this is always the case for scalars), otherwise use the promoted result type.

Fix test/CodeGen/X86/vsplit-and.ll when promote-elements is enabled.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140464 91177308-0d34-0410-b5e6-96231b3b80d8

[Vector-Select] Address one of the problems in 10902.

When generating the trunc-store of i1's, we need to use the vector type and not
the scalar type.

This patch fixes the assertion in CodeGen/Generic/bool-vector.ll when
running with -promote-elements.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140463 91177308-0d34-0410-b5e6-96231b3b80d8

Add .td file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140446 91177308-0d34-0410-b5e6-96231b3b80d8

Revert change made in .gitignore.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140445 91177308-0d34-0410-b5e6-96231b3b80d8

Preparation for adding simple Mips64 instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140443 91177308-0d34-0410-b5e6-96231b3b80d8

Only run MF.verify() with EXPENSIVE_CHECKS=1.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140441 91177308-0d34-0410-b5e6-96231b3b80d8

sys::Process: Add a SetWorkingDirectory method.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140433 91177308-0d34-0410-b5e6-96231b3b80d8

LSR minor bug fix in RateRegister.

No test case. Noticed by inspection and I doubt it ever affects the
outcome of the overall heuristic, let alone final codegen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140431 91177308-0d34-0410-b5e6-96231b3b80d8

Duncan owns dragonegg too, it's all his fault :)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140430 91177308-0d34-0410-b5e6-96231b3b80d8

Verify that terminators follow non-terminators.

This exposes a -segmented-stacks bug.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140429 91177308-0d34-0410-b5e6-96231b3b80d8

PR10998: It is not legal to sink an instruction past the terminator of a block; make sure we don't do that.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140428 91177308-0d34-0410-b5e6-96231b3b80d8

Teach the Thumb2 AsmParser to accept pre-indexed loads/stores with an offset of #-0.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140426 91177308-0d34-0410-b5e6-96231b3b80d8

Also match negative offsets for addrmode3 and addrmode5.

Math is hard, and isScaledConstantInRange() always returned false for
negative constants.  It was doing unsigned division of negative numbers
before casting back to signed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140425 91177308-0d34-0410-b5e6-96231b3b80d8

Fix incorrect disassembly test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140423 91177308-0d34-0410-b5e6-96231b3b80d8

Add more fixed bits to USAT16 encoding to filter out incorrect decodings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140422 91177308-0d34-0410-b5e6-96231b3b80d8

Post-index loads/stores in still need to print the post-indexed immediate, even if it's zero, to distinguish them from non-post-indexed instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140420 91177308-0d34-0410-b5e6-96231b3b80d8

Reapply r140412 (Thumb2 reg-reg loads cannot target SP or PC), with invalid testcases updated.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140415 91177308-0d34-0410-b5e6-96231b3b80d8

Revert r140412.  This affects more instructions than intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140413 91177308-0d34-0410-b5e6-96231b3b80d8

Thumb2 register-shifted-register loads cannot target the PC or the SP.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140412 91177308-0d34-0410-b5e6-96231b3b80d8

Add getTreeFactory() to ImmutableSet to allow construction of ImmutableSetRef from an ImmutableSet object.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140402 91177308-0d34-0410-b5e6-96231b3b80d8

Implement N32/64 calling convention. Patch by Liu.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140401 91177308-0d34-0410-b5e6-96231b3b80d8

Make FGR64RegisterClass available if target is Mips64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140397 91177308-0d34-0410-b5e6-96231b3b80d8

Add definitions of 64-bit register files. Add code for returning Mips64's sets of
callee-saved registers and reserved registers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140395 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Fix parameter order bug

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140394 91177308-0d34-0410-b5e6-96231b3b80d8

Fix a couple of 80 column violations.
patch contributed by Jia Liu!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140391 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Cleanup unused code in PTXMachineFunctionInfo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140390 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Fix another 80-column violation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140387 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Handle function call return values

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140386 91177308-0d34-0410-b5e6-96231b3b80d8

Fix 80 column violations.

Original patch by Liu.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140385 91177308-0d34-0410-b5e6-96231b3b80d8

Implement Chris's suggestion of legalizing the various SSE and AVX
hadd/hsub intrinsics into the new fhadd/fhsub X86 node.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140383 91177308-0d34-0410-b5e6-96231b3b80d8

Modified demo to use 3.0 resume instruction vs calling _Unwine_Resume.

Also conducted some reformatting. As the LLVM coding standard doc does not
seem to touch on how to align function arguments, and format code longer than
80 cols in general, the confusion persists. There is the golden rule, but as
this code has gone through several styles to deal with this, the golden rule
seems to be ignored. The latest reformatting effort tries to match the other
source files as much as possible.

Tested on OS X 10.7.1 with, and without the OLD_EXC_SYSTEM defined. Have NOT
tested on LINUX.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140379 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Start fixing function calls

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140378 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Remove PTX calling convention files

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140377 91177308-0d34-0410-b5e6-96231b3b80d8

[PATCH 2/2] PTXInstrInfo.td PTXIntrinsicInstrInfo.td 80 columns

From 5936c03172e251f12a0332d1033de5718e6e2091 Mon Sep 17 00:00:00 2001
---
 lib/Target/PTX/PTXInstrInfo.td          |  165 ++++++++++++++++++++----------
 lib/Target/PTX/PTXIntrinsicInstrInfo.td |   88 +++++++++++------
 2 files changed, 167 insertions(+), 86 deletions(-)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140376 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Generalize handling of .param types

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140375 91177308-0d34-0410-b5e6-96231b3b80d8

PTX: Cleanup unused code in the PTXMFInfoExtract pass

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140374 91177308-0d34-0410-b5e6-96231b3b80d8

Tweak the handling of MERGE_VALUES nodes: remove the need for
DecomposeMERGE_VALUES to "know" that results are legalized in
a particular order, by passing it the number of the result
being legalized (the type legalization core provides this, it
just needs to be passed on).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140373 91177308-0d34-0410-b5e6-96231b3b80d8

Vector-Select: Address one of the problems in pr10902. Add handling for the
integer-promotion of CONCAT_VECTORS.

Test: test/CodeGen/X86/widen_shuffle-1.ll

This patch fixes the above tests (when running in with -promote-elements).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140372 91177308-0d34-0410-b5e6-96231b3b80d8

Don't allow 32-bit only instructions to be disassembled in 64-bit mode. Fixes part of PR10700.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140370 91177308-0d34-0410-b5e6-96231b3b80d8

Add definitions of 64-bit int registers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140366 91177308-0d34-0410-b5e6-96231b3b80d8

Do not rely on the enum values of argument registers A0-A3 being consecutive.

Define function getNextIntArgReg, which takes a register as a parameter and
returns the next O32 argument integer register. Use this function when double
precision floating point arguments are passed in two integer registers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140363 91177308-0d34-0410-b5e6-96231b3b80d8

We're no longer going to bother supporting platforms that don't
support C89.

We probably didn't support them anyways.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140361 91177308-0d34-0410-b5e6-96231b3b80d8

PR10989: Don't print .hidden on Windows.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140356 91177308-0d34-0410-b5e6-96231b3b80d8

PR10991: make fast-isel correctly check whether accessing a global through an alias involves thread-local storage.  (I'm not entirely sure how this is supposed to work, but this patch makes fast-isel consistent with the normal isel path.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140355 91177308-0d34-0410-b5e6-96231b3b80d8

Make changes in instruction and pattern definitions so that tablegen does not
complain it cannot infer types in patterns. Fix a mistake in definition of
SDT_MipsExtractElementF64.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140354 91177308-0d34-0410-b5e6-96231b3b80d8

Add new files to CMake.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140352 91177308-0d34-0410-b5e6-96231b3b80d8

Fix SimplifySelectCC to add newly created nodes to the DAGCombiner
worklist, as it may be possible to perform further optimization on them.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140349 91177308-0d34-0410-b5e6-96231b3b80d8

Add support for GR32 <-> FR32 cross class copies.

We already support GR64 <-> VR128 copies.  All of these copies break
partial register dependencies by zeroing the high part of the target
register.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140348 91177308-0d34-0410-b5e6-96231b3b80d8

Update CMake build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140347 91177308-0d34-0410-b5e6-96231b3b80d8

Start stubbing out MCModule and MCAtom, which provide an API for accessing the rich disassembly of a complete object or executable.
These are very much a work in progress, and not really useful yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140345 91177308-0d34-0410-b5e6-96231b3b80d8

Constrain register classes instead of emitting copies.

Sometimes register class constraints are trivial, like GR32->GR32_NOSP,
or GPR->rGPR.  Teach InstrEmitter to simply constrain the virtual
register instead of emitting a copy in these cases.

Normally, these copies are handled by the coalescer.  This saves some
coalescer work.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140340 91177308-0d34-0410-b5e6-96231b3b80d8

Add a MinNumRegs argument to MRI::constrainRegClass().

The function will refuse to use a register class with fewer registers
than MinNumRegs.  This can be used by clients to avoid accidentally
increase register pressure too much.

The default value of MinNumRegs=0 doesn't affect how constrainRegClass()
works.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140339 91177308-0d34-0410-b5e6-96231b3b80d8

Synthesize SSE3/AVX 128 bit horizontal add/sub instructions from
floating point add/sub of appropriate shuffle vectors.  Does not
synthesize the 256 bit AVX versions because they work differently.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140332 91177308-0d34-0410-b5e6-96231b3b80d8

PR10987: add a missed safety check to isSafePHIToSpeculate in scalarrepl.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140327 91177308-0d34-0410-b5e6-96231b3b80d8

Resynch intro to section with copyright section.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140326 91177308-0d34-0410-b5e6-96231b3b80d8

Print parentheses in next line.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140325 91177308-0d34-0410-b5e6-96231b3b80d8

Change subreg index of AFPR64 from sub_fpeven to sub_32 per Jakob's comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140324 91177308-0d34-0410-b5e6-96231b3b80d8

Define a new sub-register index sub_32 for accessing the 32-bit sub-register of
a 64-bit integer register. Move the subreg index definitions to the beginning
of the file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140319 91177308-0d34-0410-b5e6-96231b3b80d8

Use the C personality function instead of the C++ personality function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@140318 91177308-0d34-0410-b5e6-96231b3b80d8