oota-llvm.git
9 years agoStore MCSymbols in PendingLabels.
Rafael Espindola [Fri, 29 May 2015 17:41:59 +0000 (17:41 +0000)]
Store MCSymbols in PendingLabels.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238582 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMove SymbolSize from MCSymbolData to MCSymbol.
Rafael Espindola [Fri, 29 May 2015 17:24:52 +0000 (17:24 +0000)]
Move SymbolSize from MCSymbolData to MCSymbol.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238580 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix crash in MCExpr::print.
Pete Cooper [Fri, 29 May 2015 17:19:11 +0000 (17:19 +0000)]
Fix crash in MCExpr::print.

Symbols are no longer required to be named, but this leads to a crash here if an
unnamed symbol checks that its first character is '$'.

Change the code to first check for a name, then check its first character.

No test case i'm afraid as this is debugging code, but any test case with temp labels
and 'llc --debug --filetype=obj' would have crashed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238579 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd support for VSX FMA single-precision instructions to the PPC back end
Nemanja Ivanovic [Fri, 29 May 2015 17:13:25 +0000 (17:13 +0000)]
Add support for VSX FMA single-precision instructions to the PPC back end

This patch corresponds to review:
http://reviews.llvm.org/D9941

It adds the various FMA instructions introduced in the version 2.07 of
the ISA along with the testing for them. These are operations on single
precision scalar values in VSX registers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238578 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMIR Serialization: use correct line and column numbers for LLVM IR errors.
Alex Lorenz [Fri, 29 May 2015 17:05:41 +0000 (17:05 +0000)]
MIR Serialization: use correct line and column numbers for LLVM IR errors.

This commit translates the line and column numbers for LLVM IR
errors from the numbers in the YAML block scalar to the numbers
in the MIR file so that the MIRParser users can report LLVM IR
errors with the correct line and column numbers.

Reviewers: Duncan P. N. Exon Smith

Differential Revision: http://reviews.llvm.org/D10108

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238576 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Emit EH tables for __CxxFrameHandler3 on 32-bit x86
Reid Kleckner [Fri, 29 May 2015 17:00:57 +0000 (17:00 +0000)]
[WinEH] Emit EH tables for __CxxFrameHandler3 on 32-bit x86

Small (really small!) C++ exception handling examples work on 32-bit x86
now.

This change disables the use of .seh_* directives in WinException when
CFI is not in use. It also uses absolute symbol references in the tables
instead of imagerel32 relocations.

Also fixes a cache invalidation bug in MMI personality classification.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238575 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NVPTXFavorNonGenericAddrSpaces] recursively trace into GEP and BitCast
Jingyue Wu [Fri, 29 May 2015 17:00:27 +0000 (17:00 +0000)]
[NVPTXFavorNonGenericAddrSpaces] recursively trace into GEP and BitCast

Summary:
This patch allows NVPTXFavorNonGenericAddrSpaces to remove addrspacecast
from longer chains consisting of GEPs and BitCasts. For example, it can
now optimize

  %0 = addrspacecast [10 x float] addrspace(3)* @a to [10 x float]*
  %1 = gep [10 x float]* %0, i64 0, i64 %i
  %2 = bitcast float* %1 to i32*
  %3 = load i32* %2 ; emits ld.u32

to

  %0 = gep [10 x float] addrspace(3)* @a, i64 0, i64 %i
  %1 = bitcast float addrspace(3)* %0 to i32 addrspace(3)*
  %3 = load i32 addrspace(3)* %1 ; emits ld.shared.f32

Test Plan: @ld_int_from_global_float in access-non-generic.ll

Reviewers: broune, eliben, jholewinski, meheff

Subscribers: jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D10074

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238574 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[DependenceAnalysis] Extend unifySubscriptType for handling coupled subscript groups.
Jingyue Wu [Fri, 29 May 2015 16:58:08 +0000 (16:58 +0000)]
[DependenceAnalysis] Extend unifySubscriptType for handling coupled subscript groups.

Summary:
In continuation to an earlier commit to DependenceAnalysis.cpp by jingyue (r222100), the type for all subscripts in a coupled group need to be the same since constraints from one subscript may be propagated to another during testing. During testing, new SCEVs may be created and the operands for these need to be the same.
This patch extends unifySubscriptType() to work on lists of subscript pairs, ensuring a common extended type for all of them.

Test Plan:
Added a test case to NonCanonicalizedSubscript.ll which causes dependence analysis to crash without this fix.

All regression tests pass.

Reviewers: spop, sebpop, jingyue

Reviewed By: jingyue

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D9698

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238573 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSimplify now that symbols contain the correct section.
Rafael Espindola [Fri, 29 May 2015 15:07:27 +0000 (15:07 +0000)]
Simplify now that symbols contain the correct section.

The complexity in here was because before r233995 variable symbols would report
the incorrect section.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238559 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Objdump] Removing unused parameter.
Colin LeMahieu [Fri, 29 May 2015 14:48:25 +0000 (14:48 +0000)]
[Objdump] Removing unused parameter.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238557 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] Disassembling, printing, and emitting instructions a whole-bundle at a...
Colin LeMahieu [Fri, 29 May 2015 14:44:13 +0000 (14:44 +0000)]
[Hexagon] Disassembling, printing, and emitting instructions a whole-bundle at a time which is the semantic unit for Hexagon.  Fixing tests to use the new format.  Disabling tests in the direct object emission path for a followup patch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238556 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix ELFObjectWriter::isLocal for signature symbols.
Rafael Espindola [Fri, 29 May 2015 14:20:40 +0000 (14:20 +0000)]
Fix ELFObjectWriter::isLocal for signature symbols.

And with that simplify the logic for inserting them in ExternalSymbolData or
LocalSymbolData.

No functionality change overall since the old code avoided the isLocal bug.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238555 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[mips] Remove 2 unused variables in MipsTargetStreamer.cpp. NFC.
Toma Tabacu [Fri, 29 May 2015 13:52:56 +0000 (13:52 +0000)]
[mips] Remove 2 unused variables in MipsTargetStreamer.cpp. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238554 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemoving a switch statement that only contains a default; NFC.
Aaron Ballman [Fri, 29 May 2015 13:00:07 +0000 (13:00 +0000)]
Removing a switch statement that only contains a default; NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238552 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen] Remove convertValue functions for UnOpInit, BinOpInit, and TernOpInit...
Craig Topper [Fri, 29 May 2015 05:51:32 +0000 (05:51 +0000)]
[TableGen] Remove convertValue functions for UnOpInit, BinOpInit, and TernOpInit as they weren't able to be called.

I don't think converting the inputs to the Ops was the right behavior anyway.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238543 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThis should have been a reference
Matthias Braun [Fri, 29 May 2015 02:59:59 +0000 (02:59 +0000)]
This should have been a reference

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238540 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoCodeGen: Use mop_iterator instead of MIOperands/ConstMIOperands
Matthias Braun [Fri, 29 May 2015 02:56:46 +0000 (02:56 +0000)]
CodeGen: Use mop_iterator instead of MIOperands/ConstMIOperands

MIOperands/ConstMIOperands are classes iterating over the MachineOperand
of a MachineInstr, however MachineInstr::mop_iterator does the same
thing.

I assume these two iterators exist to have a uniform interface to
iterate over the operands of a machine instruction bundle and a single
machine instruction. However in practice I find it more confusing to have 2
different iterator classes, so this patch transforms (nearly all) the
code to use mop_iterators.

The only exception being MIOperands::anlayzePhysReg() and
MIOperands::analyzeVirtReg() still needing an equivalent, I leave that
as an exercise for the next patch.

Differential Revision: http://reviews.llvm.org/D9932

This version is slightly modified from the proposed revision in that it
introduces MachineInstr::getOperandNo to avoid the extra counting
variable in the few loops that previously used MIOperands::getOperandNo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238539 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd a test for the MachineCopyPropagation change landed in r238518.
Quentin Colombet [Fri, 29 May 2015 01:40:00 +0000 (01:40 +0000)]
Add a test for the MachineCopyPropagation change landed in r238518.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238537 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen][AsmMatcherEmitter] Only parse isolated tokens as registers.
Ahmed Bougacha [Fri, 29 May 2015 01:03:37 +0000 (01:03 +0000)]
[TableGen][AsmMatcherEmitter] Only parse isolated tokens as registers.

Fixes PR23455, where, when TableGen generates the matcher from the
AsmString, it splits "cmp${cc}ss" into tokens, and the "ss" suffix
is recognized as the SS register.

I can't think of a situation where that's a feature, not a bug, hence:
when a token is "isolated", i.e., it is followed and preceded by
separators, it shouldn't be parsed as a register.

Differential Revision: http://reviews.llvm.org/D9844

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238536 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen][AsmMatcherEmitter] Factor out AsmOperand creation. NFC.
Ahmed Bougacha [Fri, 29 May 2015 00:55:55 +0000 (00:55 +0000)]
[TableGen][AsmMatcherEmitter] Factor out AsmOperand creation.  NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238534 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[IR] fptrunc-of-fptrunc isn't an EliminableCastPair.
Ahmed Bougacha [Fri, 29 May 2015 00:04:30 +0000 (00:04 +0000)]
[IR] fptrunc-of-fptrunc isn't an EliminableCastPair.

Double and single rounding can produce different results.
This is the IR counterpart to r228911.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238531 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMachineFrameInfo: Simplify pristine register calculation.
Matthias Braun [Thu, 28 May 2015 23:20:35 +0000 (23:20 +0000)]
MachineFrameInfo: Simplify pristine register calculation.

About pristine regsiters:
Pristine registers "hold a value that is useless to the current
function, but that must be preserved - they are callee saved registers
that have not been saved." This concept saves compile time as it frees
the prologue/epilogue inserter from adding every such register to every
basic blocks live-in list.

However the current code in getPristineRegs is formulated in a
complicated way: Inside the function prologue and epilogue all callee
saves are considered pristine, while in the rest of the code only the
non-saved ones are considered pristine.  This requires logic to
differentiate between prologue/epilogue and the rest and in the presence
of shrink-wrapping this even becomes complicated/expensive.  It's also
unnecessary because the prologue epilogue inserters already mark
callee-save registers that are saved/restores properly in the respective
blocks in the prologue/epilogue (see updateLiveness() in
PrologueEpilogueInserter.cpp). So only declaring non-saved/restored
callee saved registers as pristine just works.

Differential Revision: http://reviews.llvm.org/D10101

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238524 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix typos in variable/grammar names.
Eric Christopher [Thu, 28 May 2015 23:07:39 +0000 (23:07 +0000)]
Fix typos in variable/grammar names.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238523 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRename Win64Exception.(cpp|h) to WinException.(cpp|h)
Reid Kleckner [Thu, 28 May 2015 22:47:01 +0000 (22:47 +0000)]
Rename Win64Exception.(cpp|h) to WinException.(cpp|h)

This is in preparation for reusing this for 32-bit x86 EH table
emission.  Also updates the type name for consistency. NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238521 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[x86] Move the vector popcount tests into non-ISA files, and instead
Chandler Carruth [Thu, 28 May 2015 22:46:48 +0000 (22:46 +0000)]
[x86] Move the vector popcount tests into non-ISA files, and instead
organize them by the width of vector.

This makes it a lot easier to see that we're covering all of the vector
types but not doing so excessively. This also adds tests across the
spectrum of SSE versions in addition to the AVX versions.

If you're really tired of seeing the *massive* sprawl of scalarized code
for this, don't worry, I'm just about to land Bruno's patch that
dramatically improve the situation for SSSE3 and newer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238520 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMIR Serialization: print and parse machine function names.
Alex Lorenz [Thu, 28 May 2015 22:41:12 +0000 (22:41 +0000)]
MIR Serialization: print and parse machine function names.

This commit introduces a serializable structure called
'llvm::yaml::MachineFunction' that stores the machine
function's name. This structure will mirror the machine
function's state in the future.

This commit prints machine functions as YAML documents
containing a YAML mapping that stores the state of a machine
function. This commit also parses the YAML documents
that contain the machine functions.

Reviewers: Duncan P. N. Exon Smith

Differential Revision: http://reviews.llvm.org/D9841

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238519 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[MachineCopyPropagation] Fix a bug with undef handling when the value is actualy...
Quentin Colombet [Thu, 28 May 2015 22:38:40 +0000 (22:38 +0000)]
[MachineCopyPropagation] Fix a bug with undef handling when the value is actualy alive.
Test case will follow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238518 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFixing broken bots after r238505.
Chris Bieneman [Thu, 28 May 2015 22:18:34 +0000 (22:18 +0000)]
Fixing broken bots after r238505.

Need non-const iterator inserts too. These failures seem to be due to differences in the versions of libstdc++ on various operating systems.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238516 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd testcase for r238503.
David Majnemer [Thu, 28 May 2015 22:12:27 +0000 (22:12 +0000)]
Add testcase for r238503.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238515 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Start inserting state number stores for C++ EH
Reid Kleckner [Thu, 28 May 2015 22:00:24 +0000 (22:00 +0000)]
[WinEH] Start inserting state number stores for C++ EH

This moves all the state numbering code for C++ EH to WinEHPrepare so
that we can call it from the X86 state numbering IR pass that runs
before isel.

Now we just call the same state numbering machinery and insert a bunch
of stores. It also populates MachineModuleInfo with information about
the current function.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238514 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDon't special case undefined symbol when deciding the symbol order.
Rafael Espindola [Thu, 28 May 2015 21:59:34 +0000 (21:59 +0000)]
Don't special case undefined symbol when deciding the symbol order.

ELF has no restrictions on where undefined symbols go relative to other defined
symbols. In fact, gas just sorts them together. Do the same.

This was there since r111174 probably just because the MachO writer has it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238513 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUpdate documentation for llvm-profdata.
Diego Novillo [Thu, 28 May 2015 21:57:17 +0000 (21:57 +0000)]
Update documentation for llvm-profdata.

These options have been present for a while, but I had never updated the
documentation. Fixed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238511 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFixing the polly build.
Chris Bieneman [Thu, 28 May 2015 21:51:52 +0000 (21:51 +0000)]
Fixing the polly build.

I broke the polly build in r238505. This fixes the failure by adding non-const iterator erase methods to cl::list_storage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238509 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevise test to run llc and llvm-mc separately.
Andy Ayers [Thu, 28 May 2015 21:49:50 +0000 (21:49 +0000)]
Revise test to run llc and llvm-mc separately.

Differential Revision: http://reviews.llvm.org/D10066

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238508 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoEnable exitValue rewrite only when the cost of expansion is low.
Wei Mi [Thu, 28 May 2015 21:49:07 +0000 (21:49 +0000)]
Enable exitValue rewrite only when the cost of expansion is low.

The patch evaluates the expansion cost of exitValue in indVarSimplify pass, and only does the rewriting when the expansion cost is low or loop can be deleted with the rewriting. It provides an option "-replexitval=" to control the default aggressiveness of the exitvalue rewriting. It also fixes some missing cases in SCEVExpander::isHighCostExpansionHelper to enhance the evaluation of SCEV expansion cost.

Differential Revision: http://reviews.llvm.org/D9800

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238507 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove a trivial forwarding function. NFC.
Rafael Espindola [Thu, 28 May 2015 21:36:02 +0000 (21:36 +0000)]
Remove a trivial forwarding function. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238506 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRe-landing "Refactoring cl::list_storage from "is a" to "has a" std::vector."
Chris Bieneman [Thu, 28 May 2015 21:31:22 +0000 (21:31 +0000)]
Re-landing "Refactoring cl::list_storage from "is a" to "has a" std::vector."

Originally landed r238485

MSVC resolves identifiers differently from Clang and GCC, this resulted in build bot failures. This pach re-lands r238485 and fixes the build failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238505 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[SelectionDAG] Scalar shift amounts may require legalization
David Majnemer [Thu, 28 May 2015 21:29:59 +0000 (21:29 +0000)]
[SelectionDAG] Scalar shift amounts may require legalization

The shift amount may be too small to cope with promoted left hand side,
make sure to promote it as well.

This fixes PR23664.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238503 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove debug prints from r238487
Reid Kleckner [Thu, 28 May 2015 21:23:53 +0000 (21:23 +0000)]
Remove debug prints from r238487

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238501 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[llvm] Adding vdtor to fix warning.
Colin LeMahieu [Thu, 28 May 2015 20:59:08 +0000 (20:59 +0000)]
[llvm] Adding vdtor to fix warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238494 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoInline trivial method. NFC.
Rafael Espindola [Thu, 28 May 2015 20:53:09 +0000 (20:53 +0000)]
Inline trivial method. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238492 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert "Refactoring cl::list_storage from "is a" to "has a" std::vector."
Chris Bieneman [Thu, 28 May 2015 20:47:02 +0000 (20:47 +0000)]
Revert "Refactoring cl::list_storage from "is a" to "has a" std::vector."

This reverts commit 117715ca0613d3db144241499401f2ec5398f1d5.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238491 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDisable x86 tail call optimizations that jump through GOT
Reid Kleckner [Thu, 28 May 2015 20:44:28 +0000 (20:44 +0000)]
Disable x86 tail call optimizations that jump through GOT

For x86 targets, do not do sibling call optimization when materializing
the callee's address would require a GOT relocation. We can still do
tail calls to internal functions, hidden functions, and protected
functions, because they do not require this kind of relocation. It is
still possible to get GOT relocations when the user explicitly asks for
it with musttail or -tailcallopt, both of which are supposed to
guarantee TCO.

Based on a patch by Chih-hung Hsieh.

Reviewers: srhines, timmurray, danalbert, enh, void, nadav, rnk

Subscribers: joerg, davidxl, llvm-commits

Differential Revision: http://reviews.llvm.org/D9799

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238487 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRefactoring cl::list_storage from "is a" to "has a" std::vector.
Chris Bieneman [Thu, 28 May 2015 20:38:12 +0000 (20:38 +0000)]
Refactoring cl::list_storage from "is a" to "has a" std::vector.

Summary: This isn't necessarily an ideal change, and I want to at least reduce the API surface area, but for the new API we really shouldn't be relying on cl::list being a std::vector.

Reviewers: chandlerc

Reviewed By: chandlerc

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D10093

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238485 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert r238427 - [mips] Make TTypeEncoding indirect to allow .eh_frame to be read...
Daniel Sanders [Thu, 28 May 2015 20:30:32 +0000 (20:30 +0000)]
Revert r238427 - [mips] Make TTypeEncoding indirect to allow .eh_frame to be read-only.

It caused a smaller number of failures than the previous attempt at committing but still caused a couple on the llvm-linux-mips builder. Reverting while I investigate the remainder.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238483 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoObject, ELF: Use error code instead of calling report_fatal_error()
Alexey Samsonov [Thu, 28 May 2015 20:25:42 +0000 (20:25 +0000)]
Object, ELF: Use error code instead of calling report_fatal_error()

Make createELFObjectFile() return object_error::parse_failed on
encountering invalid ELF file, instead of crashing the program.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238481 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove structure field that can be computed just before use.
Rafael Espindola [Thu, 28 May 2015 20:25:29 +0000 (20:25 +0000)]
Remove structure field that can be computed just before use.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238480 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAvoid warnings when building without asserts.
Rafael Espindola [Thu, 28 May 2015 20:19:31 +0000 (20:19 +0000)]
Avoid warnings when building without asserts.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238479 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMove these vectors to the only function where they are used.
Rafael Espindola [Thu, 28 May 2015 20:11:34 +0000 (20:11 +0000)]
Move these vectors to the only function where they are used.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238477 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThumb2: Modify codegen for memcpy intrinsic to prefer LDM/STM.
Peter Collingbourne [Thu, 28 May 2015 20:02:45 +0000 (20:02 +0000)]
Thumb2: Modify codegen for memcpy intrinsic to prefer LDM/STM.

We were previously codegen'ing these as regular load/store operations and
hoping that the register allocator would allocate registers in ascending order
so that we could apply an LDM/STM combine after register allocation. According
to the commit that first introduced this code (r37179), we planned to teach
the register allocator to allocate the registers in ascending order. This
never got implemented, and up to now we've been stuck with very poor codegen.

A much simpler approach for achiveing better codegen is to create LDM/STM
instructions with identical sets of virtual registers, let the register
allocator pick arbitrary registers and order register lists when printing an
MCInst. This approach also avoids the need to repeatedly calculate offsets
which ultimately ought to be eliminated pre-RA in order to decrease register
pressure.

This is implemented by lowering the memcpy intrinsic to a series of SD-only
MCOPY pseudo-instructions which performs a memory copy using a given number
of registers. During SD->MI lowering, we lower MCOPY to LDM/STM. This is a
little unusual, but it avoids the need to encode register lists in the SD,
and we can take advantage of SD use lists to decide whether to use the _UPD
variant of the instructions.

Fixes PR9199.

Differential Revision: http://reviews.llvm.org/D9508

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238473 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[WinEH] Remove debugging dump() call
Reid Kleckner [Thu, 28 May 2015 20:02:05 +0000 (20:02 +0000)]
[WinEH] Remove debugging dump() call

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238472 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMerge redundant loops. NFC.
Rafael Espindola [Thu, 28 May 2015 20:00:13 +0000 (20:00 +0000)]
Merge redundant loops. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238471 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Stop exposing underlying DIE children list, NFC
Duncan P. N. Exon Smith [Thu, 28 May 2015 19:56:34 +0000 (19:56 +0000)]
AsmPrinter: Stop exposing underlying DIE children list, NFC

Update `DIE` API to hide the implementation of `DIE::Children` so we can
swap it out.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238468 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSimplify LastLocalSymbolIndex computation. NFC.
Rafael Espindola [Thu, 28 May 2015 19:46:36 +0000 (19:46 +0000)]
Simplify LastLocalSymbolIndex computation. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238465 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUse range loops. NFC.
Rafael Espindola [Thu, 28 May 2015 19:43:20 +0000 (19:43 +0000)]
Use range loops. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238463 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd BranchProbabilityInfo::releaseMemory to clear the Weights field.
Pete Cooper [Thu, 28 May 2015 19:43:06 +0000 (19:43 +0000)]
Add BranchProbabilityInfo::releaseMemory to clear the Weights field.

BranchProbabilityInfo was leaking 3MB of memory when running 'opt -O2 verify-uselistorder.lto.bc'.  This was due to the Weights member not being cleared once the pass is no longer needed.

This adds the releaseMemory override to clear that field.  The other fields are cleared at the end of runOnFunction so can stay there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238462 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove temporary FileSymbolData. NFC.
Rafael Espindola [Thu, 28 May 2015 19:29:15 +0000 (19:29 +0000)]
Remove temporary FileSymbolData. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238461 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Objdump] Allow instruction pretty printing to be specialized by the target triple.
Colin LeMahieu [Thu, 28 May 2015 19:07:14 +0000 (19:07 +0000)]
[Objdump] Allow instruction pretty printing to be specialized by the target triple.

Differential Revision: http://reviews.llvm.org/D8427

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238457 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Rename begin_values() => values_begin(), NFC
Duncan P. N. Exon Smith [Thu, 28 May 2015 18:55:38 +0000 (18:55 +0000)]
AsmPrinter: Rename begin_values() => values_begin(), NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238456 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[llvm] Parameterizing the output stream for dumpbytes and outputting directly to...
Colin LeMahieu [Thu, 28 May 2015 18:39:50 +0000 (18:39 +0000)]
[llvm] Parameterizing the output stream for dumpbytes and outputting directly to stream.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238453 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[InstCombine] Fold IntToPtr and PtrToInt into preceding loads.
David Majnemer [Thu, 28 May 2015 18:39:17 +0000 (18:39 +0000)]
[InstCombine] Fold IntToPtr and PtrToInt into preceding loads.

Currently we only fold a BitCast into a Load when the BitCast is its
only user.

Do the same for any no-op cast.

Differential Revision: http://reviews.llvm.org/D9152

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238452 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd llvm-dwarfdump-fuzzer that uses LibFuzzer to fuzz llvm-dwarfdump tool.
Alexey Samsonov [Thu, 28 May 2015 18:35:18 +0000 (18:35 +0000)]
Add llvm-dwarfdump-fuzzer that uses LibFuzzer to fuzz llvm-dwarfdump tool.

The fuzzer is very simple, but not quite useful at the moment: it's unable
to discover "interesting" examples, as LLVMObject library is terrible at
error recovery, calling "report_fatal_error()" far too often.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238451 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd support for the convergent flag at the MC and MachineInstr levels.
Owen Anderson [Thu, 28 May 2015 18:33:39 +0000 (18:33 +0000)]
Add support for the convergent flag at the MC and MachineInstr levels.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238450 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoReuse Loc variable. NFC.
Chad Rosier [Thu, 28 May 2015 18:18:21 +0000 (18:18 +0000)]
Reuse Loc variable. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238448 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Suppress warnings on GCC from r238362, NFC
Duncan P. N. Exon Smith [Thu, 28 May 2015 18:09:13 +0000 (18:09 +0000)]
AsmPrinter: Suppress warnings on GCC from r238362, NFC

GCC seems to have some overzealous warnings about strict aliasing.
Rafael reports that this patch suppresses them on GCC 4.9, and I'm
hoping this will work for GCC 4.7 as well.  I'll watch [1] and iterate
if necessary.

[1]: http://bb.pgr.jp/builders/clang-3stage-x86_64-linux/builds/8597

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238447 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUse range loops for accessing file names. NFC.
Rafael Espindola [Thu, 28 May 2015 18:03:20 +0000 (18:03 +0000)]
Use range loops for accessing file names. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238446 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoExpand the Flags field of MCInstrDesc to 64 bits, while simultaneously
Owen Anderson [Thu, 28 May 2015 18:03:07 +0000 (18:03 +0000)]
Expand the Flags field of MCInstrDesc to 64 bits, while simultaneously
shrinking the Size and NumDefs fields to offset the size growth, and
reordering the fields to preserve a good packing.

This is necessary in the short term for adding a convergent flag, and
simultaneously future-proofs us against more flags being added in the
future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238445 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMerge computeSymbolTable and writeSymbolTable.
Rafael Espindola [Thu, 28 May 2015 17:54:01 +0000 (17:54 +0000)]
Merge computeSymbolTable and writeSymbolTable.

For now this just saves a few loops, but it will allow more simplifications
in the future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238444 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[mips] Add new format for dmtc2/dmfc2 for Octeon CPUs.
Kai Nacke [Thu, 28 May 2015 16:23:16 +0000 (16:23 +0000)]
[mips] Add new format for dmtc2/dmfc2 for Octeon CPUs.

Octeon CPUs use dmtc2 rt,imm16 and dmfcp2 rt,imm16 for the crypto coprocessor.
E.g. dmtc2 rt,0x4057 starts calculation of sha-1.

I had to introduce a new deconding namespace to avoid a decoding conflict.

Reviewed By: dsanders

Differential Revision: http://reviews.llvm.org/D10083

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238439 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDebugInfo: .debug_line DWARF64 support
Ed Maste [Thu, 28 May 2015 15:38:17 +0000 (15:38 +0000)]
DebugInfo: .debug_line DWARF64 support

This adds support for the 64-bit DWARF format, but is still limited to
less than 4GB of debug data by the DataExtractor class.  Some versions
of the GNU MIPS toolchain generate 64-Bit DWARF even though it isn't
actually necessary.

Differential Revision: http://reviews.llvm.org/D1988

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238434 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDon't create an unused _GLOBAL_OFFSET_TABLE_.
Rafael Espindola [Thu, 28 May 2015 15:20:00 +0000 (15:20 +0000)]
Don't create an unused _GLOBAL_OFFSET_TABLE_.

This was a bug for bug compatibility with gas that is completely unnecessary.
If a _GLOBAL_OFFSET_TABLE_ symbol is used, it will already be created by
the time we get to the ELF writer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238432 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARMTargetParser] Adding sub-arch information for Clang. NFC
Renato Golin [Thu, 28 May 2015 15:05:18 +0000 (15:05 +0000)]
[ARMTargetParser] Adding sub-arch information for Clang. NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238429 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[mips] Make TTypeEncoding indirect to allow .eh_frame to be read-only.
Daniel Sanders [Thu, 28 May 2015 14:52:15 +0000 (14:52 +0000)]
[mips] Make TTypeEncoding indirect to allow .eh_frame to be read-only.

Summary:
Following on from r209907 which made personality encodings indirect, do the
same for TType encodings. This fixes the case where a try/catch block needs
to generate references to, for example, std::exception in the
.gcc_except_table.

Reviewers: petarj

Reviewed By: petarj

Subscribers: srhines, joerg, tberghammer, llvm-commits

Differential Revision: http://reviews.llvm.org/D9669

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238427 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Mips64] Add support for MCJIT for MIPS64r2 and MIPS64r6
Petar Jovanovic [Thu, 28 May 2015 13:48:41 +0000 (13:48 +0000)]
[Mips64] Add support for MCJIT for MIPS64r2 and MIPS64r6

Add support for resolving MIPS64r2 and MIPS64r6 relocations in MCJIT.

Patch by Vladimir Radosavljevic.

Differential Revision: http://reviews.llvm.org/D9667

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238424 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSilencing two signed/unsigned mismatch warnings; NFC.
Aaron Ballman [Thu, 28 May 2015 12:55:59 +0000 (12:55 +0000)]
Silencing two signed/unsigned mismatch warnings; NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238419 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[AsmPrinter] Destroy allocated DIEAbbrevs on teardown.
Benjamin Kramer [Thu, 28 May 2015 12:55:43 +0000 (12:55 +0000)]
[AsmPrinter] Destroy allocated DIEAbbrevs on teardown.

DIEAbbrev contains a SmallVector that can leak for overly large abbrevs. They
used to be owned by the DIE, but after the recent refactoring DWARFFile
allocates its own abbrevs.

Leak found by asan.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238418 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ARMTargetParser] Adding a few more CPUs for Clang CPU detection. NFC.
Renato Golin [Thu, 28 May 2015 12:10:37 +0000 (12:10 +0000)]
[ARMTargetParser] Adding a few more CPUs for Clang CPU detection. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238415 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[MC] Replace custom string join function with the one from StringExtras.
Benjamin Kramer [Thu, 28 May 2015 11:45:32 +0000 (11:45 +0000)]
[MC] Replace custom string join function with the one from StringExtras.

NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238414 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoDon't call utostr in Twine/raw_ostream contexts.
Benjamin Kramer [Thu, 28 May 2015 11:24:24 +0000 (11:24 +0000)]
Don't call utostr in Twine/raw_ostream contexts.

Creating temporary std::strings there is unnecessary.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238412 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix comments in ARMTargetParser. NFC.
Renato Golin [Thu, 28 May 2015 08:59:03 +0000 (08:59 +0000)]
Fix comments in ARMTargetParser. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238404 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ASan] Fix previous commit. Patch by Max Ostapenko!
Yury Gribov [Thu, 28 May 2015 08:03:28 +0000 (08:03 +0000)]
[ASan] Fix previous commit. Patch by Max Ostapenko!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238403 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[ASan] New approach to dynamic allocas unpoisoning. Patch by Max Ostapenko!
Yury Gribov [Thu, 28 May 2015 07:51:49 +0000 (07:51 +0000)]
[ASan] New approach to dynamic allocas unpoisoning. Patch by Max Ostapenko!

Differential Revision: http://reviews.llvm.org/D7098

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238402 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen] Use DefInit::getAsString instead of getDef()->getName(). NFC
Craig Topper [Thu, 28 May 2015 06:38:32 +0000 (06:38 +0000)]
[TableGen] Use DefInit::getAsString instead of getDef()->getName(). NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238399 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[TableGen] Don't convert types to strings to query what they are. Just use 'isa'
Craig Topper [Thu, 28 May 2015 06:38:28 +0000 (06:38 +0000)]
[TableGen] Don't convert types to strings to query what they are. Just use 'isa'

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238398 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Reassociate] Canonicalizing 'x [+-] (-Constant * y)' isn't always a win
David Majnemer [Thu, 28 May 2015 06:16:39 +0000 (06:16 +0000)]
[Reassociate] Canonicalizing 'x [+-] (-Constant * y)' isn't always a win

Canonicalizing 'x [+-] (-Constant * y)' is not a win if we don't *know*
we will open up CSE opportunities.

If the multiply was 'nsw', then negating 'y' requires us to clear the
'nsw' flag.  If this is actually worth pursuing, it is probably more
appropriate to do so in GVN or EarlyCSE.

This fixes PR23675.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238397 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NaryReassociate] Run EarlyCSE after NaryReassociate
Jingyue Wu [Thu, 28 May 2015 04:56:52 +0000 (04:56 +0000)]
[NaryReassociate] Run EarlyCSE after NaryReassociate

Summary:
This patch made two improvements to NaryReassociate and the NVPTX pipeline

1. Run EarlyCSE/GVN after NaryReassociate to get rid of redundant common
expressions.

2. When adding an instruction to SeenExprs, maps both the SCEV before and after
reassociation to that instruction.

Test Plan: updated @reassociate_gep_nsw in nary-gep.ll

Reviewers: meheff, broune

Reviewed By: broune

Subscribers: dberlin, jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D9947

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238396 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[x86] Refactor the tests for popcnt.
Chandler Carruth [Thu, 28 May 2015 02:40:15 +0000 (02:40 +0000)]
[x86] Refactor the tests for popcnt.

Extracted from the D6531 patch by Bruno Cardoso Lopes, and re-generated
to reflect the current state of the world. This should let Bruno's D6531
actually show the delta between the approaches by running the x86 test
case update script after re-building.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238391 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[omp] Actually provide a default OpenMP runtime -- libgomp for now.
Chandler Carruth [Thu, 28 May 2015 02:17:15 +0000 (02:17 +0000)]
[omp] Actually provide a default OpenMP runtime -- libgomp for now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238390 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[omp] Add a configuration variable for the default OpenMP runtime.
Chandler Carruth [Thu, 28 May 2015 01:47:22 +0000 (01:47 +0000)]
[omp] Add a configuration variable for the default OpenMP runtime.

This will be used in my next commit to Clang.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238388 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Avoid a warning in NDEBUG, NFC
Duncan P. N. Exon Smith [Wed, 27 May 2015 23:02:36 +0000 (23:02 +0000)]
AsmPrinter: Avoid a warning in NDEBUG, NFC

Should fix the -Werror release build:
http://lab.llvm.org:8011/builders/lld-x86_64-darwin13/builds/11113

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238375 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Return added DIE from DIE::addChild()
Duncan P. N. Exon Smith [Wed, 27 May 2015 22:59:03 +0000 (22:59 +0000)]
AsmPrinter: Return added DIE from DIE::addChild()

Change `DIE::addChild()` to return a reference to the just-added node,
and update consumers to use it directly.  An upcoming commit will
abstract away (and eventually change) the underlying storage of
`DIE::Children`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238372 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRegisterPressure: fix debug prints in case of physical registers
Fiona Glaser [Wed, 27 May 2015 22:51:47 +0000 (22:51 +0000)]
RegisterPressure: fix debug prints in case of physical registers

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238371 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Stop exposing underlying DIEValue list, NFC
Duncan P. N. Exon Smith [Wed, 27 May 2015 22:44:06 +0000 (22:44 +0000)]
AsmPrinter: Stop exposing underlying DIEValue list, NFC

Change the `DIE` API to hide the implementation of the list of
`DIEValue`s.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238369 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Remove DIEHash::AttrEntry, NFC
Duncan P. N. Exon Smith [Wed, 27 May 2015 22:36:37 +0000 (22:36 +0000)]
AsmPrinter: Remove DIEHash::AttrEntry, NFC

Remove "the most boring struct ever" (thanks to review by Eric).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238366 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Store abbreviation data directly in DIE and DIEValue
Duncan P. N. Exon Smith [Wed, 27 May 2015 22:31:41 +0000 (22:31 +0000)]
AsmPrinter: Store abbreviation data directly in DIE and DIEValue

Stop storing a `DIEAbbrev` in `DIE`, since the data fits neatly inside
the `DIEValue` list.  Besides being a cleaner data structure (avoiding
the parallel arrays), this gives us more freedom to rearrange the
`DIEValue` list.

This fixes the temporary memory regression from 845 MB up to 879 MB, and
drops it further to 829 MB for a net memory decrease of around 1.9%
(incremental decrease around 5.7%).

(I'm looking at `llc` memory usage on `verify-uselistorder.lto.opt.bc`;
see r236629 for details.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238364 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoReapply "AsmPrinter: Change DIEValue to be stored by value"
Duncan P. N. Exon Smith [Wed, 27 May 2015 22:14:58 +0000 (22:14 +0000)]
Reapply "AsmPrinter: Change DIEValue to be stored by value"

This reverts commit r238350, effectively reapplying r238349 after fixing
(all?) the problems, all somehow related to how I was using
`AlignedArrayCharUnion<>` inside `DIEValue`:

  - MSVC can only handle `sizeof()` on types, not values.  Change the
    assert.
  - GCC doesn't know the `is_trivially_copyable` type trait.  Instead of
    asserting it, add destructors.
  - Call placement new even when constructing POD (i.e., the pointers).
  - Instead of copying the char buffer, copy the casted classes.

I've left in a couple of `static_assert`s that I think both MSVC and GCC
know how to handle.  If the bots disagree with me, I'll remove them.

  - Check that the constructed type is either standard layout or a
    pointer.  This protects against a programming error: we really want
    the "small" `DIEValue`s to be small and simple, so don't
    accidentally change them not to be.
  - Similarly, check that the size of the buffer is no bigger than a
    `uint64_t` or a pointer.  (I thought checking against
    `sizeof(uint64_t)` would be good enough, but Chandler suggested that
    pointers might sometimes be bigger than that in the context of
    sanitizers.)

I've also committed r238359 in the meantime, which introduces a
DIEValue.def to simplify dispatching between the various types (thanks
to a review comment by David Blaikie).  Without that, this commit would
be almost unintelligible.

Here's the original commit message:
--
Change `DIEValue` to be stored/passed/etc. by value, instead of
reference.  It's now a discriminated union, with a `Val` field storing
the actual type.  The classes that used to inherit from `DIEValue` no
longer do.  There are two categories of these:

  - Small values fit in a single pointer and are stored by value.
  - Large values require auxiliary storage, and are stored by reference.

The only non-mechanical change is to tools/dsymutil/DwarfLinker.cpp.  It
was relying on `DIEInteger`s being passed around by reference, so I
replaced that assumption with a `PatchLocation` type that stores a safe
reference to where the `DIEInteger` lives instead.

This commit causes a temporary regression in memory usage, since I've
left merging `DIEAbbrevData` into `DIEValue` for a follow-up commit.  I
measured an increase from 845 MB to 879 MB, around 3.9%.  The follow-up
drops it lower than the starting point, and I've only recently brought
the memory this low anyway, so I'm committing these changes separately
to keep them incremental.  (I also considered swapping the commits, but
the other one first would cause a lot more code churn.)

(I'm looking at `llc` memory usage on `verify-uselistorder.lto.opt.bc`;
see r236629 for details.)
--

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238362 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsmPrinter: Introduce DIEValue.def, NFC
Duncan P. N. Exon Smith [Wed, 27 May 2015 21:15:43 +0000 (21:15 +0000)]
AsmPrinter: Introduce DIEValue.def, NFC

Use a .def macro file to iterate through the various subclasses of
`DIEValue`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238359 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRename and move getCurrentSectionData.
Rafael Espindola [Wed, 27 May 2015 21:04:14 +0000 (21:04 +0000)]
Rename and move getCurrentSectionData.

I think this concludes the merge of MCSectionData and MCSection.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238358 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoThere is only one current section.
Rafael Espindola [Wed, 27 May 2015 20:52:32 +0000 (20:52 +0000)]
There is only one current section.

Both MCStreamer and MCObjectStreamer were maintaining a current section
variable and they were slightly out of sync. I don't think this was observable,
but was inefficient and error prone.

Changing this requires a few cascading changes:

* SwitchSection has to call ChangeSection earlier for ChangeSection to see
  the old section.
* With that change, ChangeSection cannot call EmitLabel, since during
  ChangeSection we are still in the old section.
* When the object streamer requires a begin label, just reused the existing
  generic support for begin labels instead of calling EmitLabel directly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238357 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[RuntimeDyld] Fix MachO i386 SECTDIFF relocation to support non-zero addends.
Lang Hames [Wed, 27 May 2015 20:50:01 +0000 (20:50 +0000)]
[RuntimeDyld] Fix MachO i386 SECTDIFF relocation to support non-zero addends.

Previously, relocations of the form 'A - B + C' would fail on i386 when C was
non-zero.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238356 91177308-0d34-0410-b5e6-96231b3b80d8