Commit Graph

16836 Commits

Author SHA1 Message Date
Nadav Rotem c2345cbe73 X86 CostModel: Add support for a some of the common arithmetic instructions for SSE4, AVX and AVX2.
llvm-svn: 167347
2012-11-03 00:39:56 +00:00
Nadav Rotem 13da94734c CostModel: add support for Vector Insert and Extract.
llvm-svn: 167329
2012-11-02 22:31:56 +00:00
Nadav Rotem a6b91ac307 Add a cost model analysis that allows us to estimate the cost of IR-level instructions.
llvm-svn: 167324
2012-11-02 21:48:17 +00:00
Manman Ren 3d5af279b1 OutputArg: added an index of the original argument to match the change to
InputArg in r165616.

This will enable us to get the actual type for both InputArg and OutputArg.

rdar://9932559

llvm-svn: 167265
2012-11-01 23:49:58 +00:00
Chandler Carruth ba98ef892f Add a getAddressSpace method to the GEP instruction to mirror that of
the inttoptr instruction. The conceptual model here is that
'getAddressSpace' refers to the address space of this instruction's
type. It just happens that for GEPs, that is always the same as the
pointer operand's address space. We want both names so that access
patterns can be consistent between different instruction types.

llvm-svn: 167229
2012-11-01 11:25:55 +00:00
Chandler Carruth 49d7e23b14 Add some consistent doxygen comments for the address space helpers.
These clarify that the methods called 'getPointerAddressSpace' apply to
the pointer *operand* of the instruction.

llvm-svn: 167228
2012-11-01 11:25:28 +00:00
Chandler Carruth 43120f4973 Normalize the API and doxygen comments for the ptrtoint instruction.
llvm-svn: 167227
2012-11-01 11:16:47 +00:00
Chandler Carruth 52c3a3382a Remove a weird static helper from the GEP instruction and just directly
compute the address space in the one place it was used.

Also write the getPointerAddressSpace member in terms of the
getPointerOperandType member.

llvm-svn: 167226
2012-11-01 10:59:30 +00:00
Chandler Carruth 705561159c As I'm going to be touching several comments in this file, update the
'@brief' doxygen markup to the now standard '\brief' markup form, in
conformance with the coding standards. This will let me continue to
write new comments in this form without making things inconsistent.

llvm-svn: 167225
2012-11-01 10:46:54 +00:00
Chandler Carruth 4a6c2a4b4f Teach Type::getPointerAddressSpace to look through pointer vectors
politely and document this feature.

This simple API extension then allows us to write all of the
Instructions' address space query methods much more simply. No
functionality change intended here.

llvm-svn: 167223
2012-11-01 09:37:49 +00:00
Chandler Carruth 5da3f0512e Revert the majority of the next patch in the address space series:
r165941: Resubmit the changes to llvm core to update the functions to
         support different pointer sizes on a per address space basis.

Despite this commit log, this change primarily changed stuff outside of
VMCore, and those changes do not carry any tests for correctness (or
even plausibility), and we have consistently found questionable or flat
out incorrect cases in these changes. Most of them are probably correct,
but we need to devise a system that makes it more clear when we have
handled the address space concerns correctly, and ideally each pass that
gets updated would receive an accompanying test case that exercises that
pass specificaly w.r.t. alternate address spaces.

However, from this commit, I have retained the new C API entry points.
Those were an orthogonal change that probably should have been split
apart, but they seem entirely good.

In several places the changes were very obvious cleanups with no actual
multiple address space code added; these I have not reverted when
I spotted them.

In a few other places there were merge conflicts due to a cleaner
solution being implemented later, often not using address spaces at all.
In those cases, I've preserved the new code which isn't address space
dependent.

This is part of my ongoing effort to clean out the partial address space
code which carries high risk and low test coverage, and not likely to be
finished before the 3.2 release looms closer. Duncan and I would both
like to see the above issues addressed before we return to these
changes.

llvm-svn: 167222
2012-11-01 09:14:31 +00:00
Chandler Carruth 7ec5085e01 Revert the series of commits starting with r166578 which introduced the
getIntPtrType support for multiple address spaces via a pointer type,
and also introduced a crasher bug in the constant folder reported in
PR14233.

These commits also contained several problems that should really be
addressed before they are re-committed. I have avoided reverting various
cleanups to the DataLayout APIs that are reasonable to have moving
forward in order to reduce the amount of churn, and minimize the number
of commits that were reverted. I've also manually updated merge
conflicts and manually arranged for the getIntPtrType function to stay
in DataLayout and to be defined in a plausible way after this revert.

Thanks to Duncan for working through this exact strategy with me, and
Nick Lewycky for tracking down the really annoying crasher this
triggered. (Test case to follow in its own commit.)

After discussing with Duncan extensively, and based on a note from
Micah, I'm going to continue to back out some more of the more
problematic patches in this series in order to ensure we go into the
LLVM 3.2 branch with a reasonable story here. I'll send a note to
llvmdev explaining what's going on and why.

Summary of reverted revisions:

r166634: Fix a compiler warning with an unused variable.
r166607: Add some cleanup to the DataLayout changes requested by
         Chandler.
r166596: Revert "Back out r166591, not sure why this made it through
         since I cancelled the command. Bleh, sorry about this!
r166591: Delete a directory that wasn't supposed to be checked in yet.
r166578: Add in support for getIntPtrType to get the pointer type based
         on the address space.
llvm-svn: 167221
2012-11-01 08:07:29 +00:00
Andrew Kaylor f2c10782ce Streamlined memory manager hierarchy for MCJIT and RuntimeDyld.
Patch by Ashok Thirumurthi

llvm-svn: 167192
2012-11-01 00:46:04 +00:00
Jakob Stoklund Olesen 366bd86335 Generate a table-driven version of TRI::composeSubRegIndices().
Explicitly allow composition of null sub-register indices, and handle
that common case in an inlinable stub.

Use a compressed table implementation instead of the previous nested
switches which generated pretty bad code.

llvm-svn: 167190
2012-11-01 00:32:10 +00:00
Rafael Espindola 27783bc9c1 Remove Triple::getArchTypeForDarwinArchName. I lives on the clang driver now.
llvm-svn: 167157
2012-10-31 18:52:25 +00:00
Benjamin Kramer c914ab6e3c Fix a couple of comment typos.
llvm-svn: 167113
2012-10-31 11:25:32 +00:00
Evgeniy Stepanov ef94169938 Add IRBuilderBase::getIntPtrTy.
llvm-svn: 167111
2012-10-31 09:50:01 +00:00
James Molloy 2f728a9185 Add support for ARM segment types PT_ARM_ARCHEXT, PT_ARM_EXIDX and PT_ARM_UNWIND.
Patch by Pete Chou!

llvm-svn: 167109
2012-10-31 09:10:56 +00:00
Rafael Espindola 7edf38f7e3 xlc supports __attribute__((aligned(x))), use it.
Patch by Kai.

llvm-svn: 167087
2012-10-31 00:54:26 +00:00
Rafael Espindola c14b96898a Add extra declarations of hash_value needed to build llvm with xlc 12.1.
Patch by Kai!

llvm-svn: 167085
2012-10-31 00:46:18 +00:00
Chad Rosier 909f6a035f [inline asm] Get the mayLoad/mayStore directly from the MIOp_ExtraInfo operand.
llvm-svn: 167050
2012-10-30 20:39:19 +00:00
Chad Rosier 9e1274fb48 [inline asm] Implement mayLoad and mayStore for inline assembly. In general,
the MachineInstr MayLoad/MayLoad flags are based on the tablegen implementation.
For inline assembly, however, we need to compute these based on the constraints.

Revert r166929 as this is no longer needed, but leave the test case in place. 
rdar://12033048 and PR13504

llvm-svn: 167040
2012-10-30 19:11:54 +00:00
Quentin Colombet 5799e9f66c Change ForceSizeOpt attribute into MinSize attribute
llvm-svn: 167020
2012-10-30 16:32:52 +00:00
Duncan Sands e2395dc27b Fix isEliminableCastPair to work correctly in the presence of pointers
with different sizes.

llvm-svn: 167018
2012-10-30 16:03:32 +00:00
Duncan Sands 3ce427c039 Add a helper for telling whether a type is a pointer or vector of pointer type.
Simplify the implementation of the corresponding integer and float functions and
move them inline while there.

llvm-svn: 167014
2012-10-30 13:38:54 +00:00
Hans Wennborg f3254838e4 Use TargetTransformInfo to control switch-to-lookup table transformation
When the switch-to-lookup tables transform landed in SimplifyCFG, it
was pointed out that this could be inappropriate for some targets.
Since there was no way at the time for the pass to know anything about
the target, an awkward reverse-transform was added in CodeGenPrepare
that turned lookup tables back into switches for some targets.

This patch uses the new TargetTransformInfo to determine if a
switch should be transformed, and removes
CodeGenPrepare::ConvertLoadToSwitch.

llvm-svn: 167011
2012-10-30 11:23:25 +00:00
Sid Manning 67211fb6fb * Add e_flags enum for Hexagon
* Add Hexagon specific section indexes for small data
- Reviewed by Michael Spencer

llvm-svn: 166997
2012-10-30 02:26:15 +00:00
Ulrich Weigand 3abb34389d In various places throughout the code generator, there were special
checks to avoid performing compile-time arithmetic on PPCDoubleDouble.

Now that APFloat supports arithmetic on PPCDoubleDouble, those checks
are no longer needed, and we can treat the type like any other.

llvm-svn: 166958
2012-10-29 18:35:49 +00:00
Ulrich Weigand e1d62f9c0a APFloat cleanup: Remove now unused fields "sign2" and "exponent2".
llvm-svn: 166952
2012-10-29 18:17:42 +00:00
Nadav Rotem 39aab03be3 Rename the BB-vectorize flag to match the dragonegg name
llvm-svn: 166948
2012-10-29 18:01:14 +00:00
Duncan Sands 5bdd9dda48 Remove a wrapper around getIntPtrType added to GVN by Hal in commit 166624 (the
wrapper returns a vector of integers when passed a vector of pointers) by having
getIntPtrType itself return a vector of integers in this case.  Outside of this
wrapper, I didn't find anywhere in the codebase that was relying on the old
behaviour for vectors of pointers, so give this a whirl through the buildbots.

llvm-svn: 166939
2012-10-29 17:31:46 +00:00
Bob Wilson 09d16aa87e Remove code to saturate profile counts.
We may need to change the way profile counter values are stored, but
saturation is the wrong thing to do.  Just remove it for now.

Patch by Alastair Murray!

llvm-svn: 166938
2012-10-29 17:27:39 +00:00
Nadav Rotem c59ae207ef Change the PassManagerBuilder (used by -O3) loop vectorizer flag from -vectorize to -vectorize-loops because we dont want to share the same flag as the bb-vectorizer.
llvm-svn: 166937
2012-10-29 16:36:25 +00:00
Hans Wennborg aad8ad1c36 Minor style fixes for TargetTransformationInfo and TargetTransformImpl
llvm-svn: 166936
2012-10-29 16:26:52 +00:00
NAKAMURA Takumi 83163a2ff3 llvm/Config/config.h.cmake: Good bye, Kevin! We won't honor authors in comments.
llvm-svn: 166934
2012-10-29 16:07:28 +00:00
Duncan Sands 59f4646614 Rather than duplicating the getPointerSize code just call getPointerSize.
llvm-svn: 166923
2012-10-29 12:19:04 +00:00
Tim Northover 4f223bf7c4 Add interface for querying object files for symbol values.
Currently only implemented for ELF.

Patch by Amara Emerson.

llvm-svn: 166918
2012-10-29 10:47:00 +00:00
Evgeniy Stepanov 5e9f055a4a va_start, va_end, va_copy: InstrinsicInst subclasses and InstVisitor support.
llvm-svn: 166916
2012-10-29 09:39:03 +00:00
Rafael Espindola d957cb2584 Remove TargetELFWriterInfo.
All the credit goes to Jan Voung for noticing it was dead!

llvm-svn: 166902
2012-10-28 21:34:43 +00:00
Nadav Rotem 859366f93f 1. Fix a bug in getTypeConversion. When a *simple* type is split, we need to return the type of the split result.
2. Change the maximum vectorization width from 4 to 8.
3. A test for both.

llvm-svn: 166864
2012-10-27 04:11:32 +00:00
Nadav Rotem afae78edab Refactor the VectorTargetTransformInfo interface.
Add getCostXXX calls for different families of opcodes, such as casts, arithmetic, cmp, etc.

Port the LoopVectorizer to the new API.

The LoopVectorizer now finds instructions which will remain uniform after vectorization. It uses this information when calculating the cost of these instructions.

llvm-svn: 166836
2012-10-26 23:49:28 +00:00
Lang Hames a312bb0879 MCRegisterClass should be returned by const ref, not by value.
llvm-svn: 166822
2012-10-26 22:14:10 +00:00
Jakob Stoklund Olesen 09d69f5b0f Remove the canCombineSubRegIndices() target hook.
The new coalescer can already do all of this, so there is no need to
duplicate the efforts.

llvm-svn: 166813
2012-10-26 20:38:19 +00:00
Benjamin Kramer 6dc1e2f287 Remove LoopDependenceAnalysis.
It was unmaintained and not much more than a stub. The new DependenceAnalysis
pass is both more general and complete.

llvm-svn: 166810
2012-10-26 20:25:01 +00:00
Bill Wendling ec0fdc93b5 Remove the unneeded initializers.
llvm-svn: 166804
2012-10-26 19:52:54 +00:00
Derek Schuff 8f5eff7e76 Stop APInt::shl from generating llvm.trap
APInt::shl generated llvm.trap to guard against shifts greater than bit-width.
This was already checked with an assert, and there was a special case for
shifts equal to bit-width. Modify this check to catch shifts greater than or
equal to bit-width, so llvm.trap isn't generated.

Patch contributed by JF Bastien

llvm-svn: 166803
2012-10-26 19:52:27 +00:00
Rafael Espindola 4253bd8faf Change the internalize pass to internalize all symbols when given an empty
list of externals. This makes sense since a shared library with no symbols
can still be useful if it has static constructors.

llvm-svn: 166795
2012-10-26 18:47:48 +00:00
Chad Rosier 11c42f2d2c [ms-inline asm] Add support for the TYPE operator.
Part of rdar://12576868

llvm-svn: 166790
2012-10-26 18:04:20 +00:00
Benjamin Kramer 214935ee70 Add a basic verifier for SCEV's backedge taken counts.
Enabled with -verify-scev. This could be extended significantly but hopefully
catches the common cases now. Note that it's not enabled by default in any
configuration because the way it tries to distinguish SCEVs is still fragile and
may produce false positives. Also the test-suite isn't clean yet, one example
is that it fails if a pass drops an NSW bit but it's still present in SCEV's
cached. Cleaning up all those cases will take some time.

llvm-svn: 166786
2012-10-26 17:31:32 +00:00
Nicolas Geoffray 457b356f3a Remove GC roots that reference dead objects.
llvm-svn: 166763
2012-10-26 09:15:55 +00:00
Bill Wendling 2b4f64a30e Alphabetize the enum list.
llvm-svn: 166760
2012-10-26 07:08:58 +00:00
Hal Finkel 9dd045f178 Add VectorTargetTransform::getNumberOfParts.
As discussed on IRC, add VectorTargetTransform::getNumberOfParts
to provide a stable interface to the vector legalization splitting factor.

llvm-svn: 166751
2012-10-26 04:28:02 +00:00
Kaelyn Uhrain 367bb95c35 Fix anonymous namespace issue introduced by r166714:
include/llvm/MC/MCTargetAsmParser.h:46:8: error: 'llvm::ParseInstructionInfo' has a field 'llvm::ParseInstructionInfo::AsmRewrites' whose type uses the anonymous namespace [-Werror]

llvm-svn: 166729
2012-10-25 22:09:49 +00:00
Chad Rosier 240b7b963a [ms-inline asm] Perform field lookups with the dot operator.
llvm-svn: 166724
2012-10-25 21:51:10 +00:00
Chad Rosier f0e8720054 [ms-inline asm] Add support for creating AsmRewrites in the target specific
AsmParser logic.  To be used/tested in a subsequent commit.

llvm-svn: 166714
2012-10-25 20:41:34 +00:00
Joerg Sonnenberger 635debe85b Remove exception handling usage from tblgen.
Most places can use PrintFatalError as the unwinding mechanism was not
used for anything other than printing the error. The single exception
was CodeGenDAGPatterns.cpp, where intermediate errors during type
resolution were ignored to simplify incremental platform development.
This use is replaced by an error flag in TreePattern and bailout earlier
in various places if it is set. 

llvm-svn: 166712
2012-10-25 20:33:17 +00:00
Michael Liao 8cdb7ede80 Add 'const' qualifier on member functions not changing its fields.
llvm-svn: 166708
2012-10-25 18:35:04 +00:00
Joerg Sonnenberger 356f797d66 In preparation for removing exception handling in tablegen, add
PrintFatalError, which combines PrintError with exit(1).

llvm-svn: 166690
2012-10-25 16:35:18 +00:00
Benjamin Kramer 71a3512d60 DependenceAnalysis: Push #includes down into the implementation.
llvm-svn: 166688
2012-10-25 16:15:22 +00:00
Sebastian Pop 5c87daf2c8 add TableGen support to create relationship maps between instructions
Relationship maps are represented as InstrMapping records which are parsed by
TableGen and the information is used to construct mapping tables to represent
appropriate relations between instructions. These tables are emitted into
XXXGenInstrInfo.inc file along with the functions to query them.

Patch by Jyotsna Verma <jverma@codeaurora.org>.

llvm-svn: 166685
2012-10-25 15:54:06 +00:00
Adhemerval Zanella f2aceda854 Initial TOC support for PowerPC64 object creation
This patch adds initial PPC64 TOC MC object creation using the small mcmodel
(a single 64K TOC) adding the some TOC relocations (R_PPC64_TOC,
R_PPC64_TOC16, and R_PPC64_TOC16DS).

The addition of 'undefinedExplicitRelSym' hook on 'MCELFObjectTargetWriter'
is meant to avoid the creation of an unreferenced ".TOC." symbol (used in
the .odp creation) as well to set the R_PPC64_TOC relocation target as the
temporary ".TOC." symbol. On PPC64 ABI, the R_PPC64_TOC relocation should
not point to any symbol.

llvm-svn: 166677
2012-10-25 12:27:42 +00:00
Nadav Rotem 4a87683a41 Implement a basic cost model for vector and scalar instructions.
llvm-svn: 166642
2012-10-24 23:47:38 +00:00
Nadav Rotem 30cc37ae2b Make LegalizeKind public so that we can use it outside of TargetLowering.
llvm-svn: 166623
2012-10-24 20:59:17 +00:00
Kaelyn Uhrain 79a9711918 Make ~Operator() protected so subclasses can inherit it and not have
a bunch of errors for all the Operator subclasses such as:

include/llvm/Operator.h:76:7: error: deleted function 'virtual llvm::OverflowingBinaryOperator::~OverflowingBinaryOperator()'
include/llvm/Operator.h:43:3: error: overriding non-deleted function 'virtual llvm::Operator::~Operator()'
include/llvm/Operator.h:76:7: error: 'virtual llvm::OverflowingBinaryOperator::~OverflowingBinaryOperator()' is implicitly deleted because the default definition would be ill-formed:
include/llvm/Operator.h:43:3: error: 'virtual llvm::Operator::~Operator()' is private
include/llvm/Operator.h:76:7: error: within this context

llvm-svn: 166611
2012-10-24 19:17:42 +00:00
Pete Cooper f4d5a7908d Improve DenseMap checks for power of 2 growth. Thanks for the tip Jakob
llvm-svn: 166609
2012-10-24 18:50:44 +00:00
Micah Villmow bf3eeb2dfc Add some cleanup to the DataLayout changes requested by Chandler.
llvm-svn: 166607
2012-10-24 18:36:13 +00:00
Micah Villmow 51e7246cb4 Back out r166591, not sure why this made it through since I cancelled the command. Bleh, sorry about this!
llvm-svn: 166596
2012-10-24 17:25:11 +00:00
Nadav Rotem 2289f2c932 Implement a basic VectorTargetTransformInfo interface to be used by the loop and bb vectorizers for modeling the cost of instructions.
llvm-svn: 166593
2012-10-24 17:22:41 +00:00
Micah Villmow 6a8f3f9e20 Delete a directory that wasn't supposed to be checked in yet.
llvm-svn: 166591
2012-10-24 17:20:04 +00:00
Micah Villmow 12d9127833 Add in support for getIntPtrType to get the pointer type based on the address space.
This checkin also adds in some tests that utilize these paths and updates some of the
clients.

llvm-svn: 166578
2012-10-24 15:52:52 +00:00
Elena Demikhovsky d6afb03bc9 Special calling conventions for Intel OpenCL built-in library.
llvm-svn: 166566
2012-10-24 14:46:16 +00:00
Richard Smith 1f6f455f7c Fix ODR violations: a virtual function must be defined, even if it's never
called. Provide an (asserting) definition of Operator's private destructor.
Remove destructors from all classes derived from Operator. We don't need them
for safety, because their implicit definitions would be ill-formed (they'd call
Operator's private destructor), and we don't need them to avoid emitting
vtables, because we don't do anything with Operator subclasses which would
trigger vtable instantiation.

The Operator hierarchy is still a complete disaster with regard to undefined
behavior, but this at least allows LLVM to link when using Clang's
-fcatch-undefined-behavior with a new vptr-based type checking mechanism.

llvm-svn: 166530
2012-10-24 00:30:41 +00:00
Michael J. Spencer 8e95113597 [Support/StringSet] Fix memory leak when inserted key already exists.
llvm-svn: 166517
2012-10-23 22:55:54 +00:00
Kevin Enderby dccdac6a06 Make branch heavy code for generating marked up disassembly simpler
and easier to read by adding a couple helper functions.  Suggestion by
Chandler Carruth and seconded by Meador Inge!

llvm-svn: 166515
2012-10-23 22:52:52 +00:00
Pete Cooper 41ef09a4b5 Change DenseMap to use a power of 2 growth if one is given instead of the next power of 2. This was causing DenseMaps to grow 4x instead of 2x. I'll keep an eye on the buildbots as this could impact performance
llvm-svn: 166493
2012-10-23 19:34:36 +00:00
Pete Cooper 8ba353da39 Fixed bug in SmallDenseMap where it wouldn't leave enough space for an empty bucket if the number of values was exactly equal to the small capacity. This led to an infinite loop when finding a non-existent element
llvm-svn: 166492
2012-10-23 18:47:35 +00:00
Chad Rosier 37e755cee2 [ms-inline asm] Add an implementation of the offset operator. This is a follow
on patch to r166433.
rdar://12470317

llvm-svn: 166488
2012-10-23 17:43:43 +00:00
Lang Hames dd8574e8fb Use ilist rather than std::list for Node and Edge lists in the PBQP graph. This
should fix an issue (described at http://stackoverflow.com/questions/10065384/instantiation-of-a-list-with-an-incomplete-type-in-a-typedef)
that was preventing LLVMCodeGen from building with libc++ in C++11 mode.

llvm-svn: 166484
2012-10-23 17:10:51 +00:00
Eli Friedman 0f4871d487 [ms-inline-asm] Implement _emit directive (which is roughly equivalent to .byte).
<rdar://problem/12470345>.

llvm-svn: 166451
2012-10-22 23:58:19 +00:00
Kevin Enderby 62183c4e18 Add support for annotated disassembly output for X86 and arm.
Per the October 12, 2012 Proposal for annotated disassembly output sent out by
Jim Grosbach this set of changes implements this for X86 and arm.  The llvm-mc
tool now has a -mdis option to produced the marked up disassembly and a couple
of small example test cases have been added.

rdar://11764962

llvm-svn: 166445
2012-10-22 22:31:46 +00:00
Chad Rosier 5bca3f9b8e [ms-inline asm] Add the isOffsetOf() function.
Part of rdar://12470317

llvm-svn: 166436
2012-10-22 19:50:35 +00:00
Nadav Rotem dbf4783634 Add the "ForceSizeOpt" attribute.
Patch by Quentin Colombet <qcolombet@apple.com>

Original description:
"""
The attached patch is the first step to have a better control on Oz related optimizations.
The Oz optimization level focuses on code size, thus I propose to add an attribute called ForceSizeOpt.
"""

llvm-svn: 166422
2012-10-22 17:33:31 +00:00
Benjamin Kramer a74129adad Symbol hygiene: Make sure declarations and definitions match, make helper functions static.
llvm-svn: 166376
2012-10-20 12:53:26 +00:00
Shuxin Yang 1479fcdef1 1. Remove noreturn attribute from __builtin_debugtrap().
(The change at Clang side was committed in r166345)

2. Cosmetic change in order to conform to coding standards. 

llvm-svn: 166350
2012-10-19 23:00:20 +00:00
Eric Christopher 5d7e15edc5 Grammar-o.
llvm-svn: 166343
2012-10-19 22:10:54 +00:00
Nadav Rotem 4dc976fbcb revert r166264 because the LTO build is still failing
llvm-svn: 166340
2012-10-19 21:28:43 +00:00
Chad Rosier 0f48c55e70 [ms-inline asm] Have the TargetParser callback to Sema to determine the size of
a memory operand.  Retain this information and then add the sizing directives
to the IR.  This allows the backend to do proper instruction selection.

llvm-svn: 166316
2012-10-19 20:57:14 +00:00
Michael Liao 14c43496c4 Add 'IntrNoReturn' for longjmp intrinsics
llvm-svn: 166314
2012-10-19 20:43:54 +00:00
Chad Rosier 7a05864b48 [ms-inline asm] Add a MCAsmParserSemaCallback to the TargetAsmParser.
llvm-svn: 166308
2012-10-19 20:35:42 +00:00
Shuxin Yang cdde059a34 This patch is to fix radar://8426430. It is about llvm support of __builtin_debugtrap()
which is supposed to consistently raise SIGTRAP across all systems. In contrast,
__builtin_trap() behave differently on different systems. e.g. it raises SIGTRAP on ARM, and
SIGILL on X86. The purpose of __builtin_debugtrap() is to consistently provide "trap"
functionality, in the mean time preserve the compatibility with on gcc on __builtin_trap().

  The X86 backend is already able to handle debugtrap(). This patch is to:
  1) make front-end recognize "__builtin_debugtrap()" (emboddied in the one-line change to Clang).
  2) In DAG legalization phase, by default, "debugtrap" will be replaced with "trap", which
     make the __builtin_debugtrap() "available" to all existing ports without the hassle of
     changing their code.
  3) If trap-function is specified (via -trap-func=xyz to llc), both __builtin_debugtrap() and
     __builtin_trap() will be expanded into the function call of the specified trap function.
    This behavior may need change in the future.

  The provided testing-case is to make sure 2) and 3) are working for ARM port, and we
already have a testing case for x86. 

llvm-svn: 166300
2012-10-19 20:11:16 +00:00
Chad Rosier 75f0b2f2bd [ms-inline asm] Add the isParsingInlineAsm() function to the MCAsmTargetParser.
llvm-svn: 166292
2012-10-19 17:57:49 +00:00
Evgeniy Stepanov 8eb77d847e Move SplitBlockAndInsertIfThen to BasicBlockUtils.
llvm-svn: 166278
2012-10-19 10:48:31 +00:00
Nick Lewycky ac612277cb Pacify -Wnon-virtual-dtor.
llvm-svn: 166270
2012-10-19 07:00:09 +00:00
Nadav Rotem 4985ddc5e0 recommit the patch that makes LSR and LowerInvoke use the TargetTransform interface.
llvm-svn: 166264
2012-10-19 04:27:49 +00:00
Nadav Rotem 5dc203e8f4 Reapply the TargerTransformInfo changes, minus the changes to LSR and Lowerinvoke.
llvm-svn: 166248
2012-10-18 23:22:48 +00:00
Jordan Rose ed8560b09c Fix case for include of Compiler.h.
llvm-svn: 166243
2012-10-18 22:36:01 +00:00
Jordan Rose ec43d52336 Add move constructors for OwningPtr and OwningArrayPtr.
While LLVM itself is still C++03, there's no reason why tools built on
top of it can't use C++11 features.

llvm-svn: 166242
2012-10-18 22:22:58 +00:00
Jordan Rose 43805d73c7 Add a T&& constructor to llvm::Optional.
This allows llvm::Optional to be used with movable-but-not-copyable types.
While LLVM itself is still C++03, there's no reason why tools built on
top of it can't use C++11 features.

llvm-svn: 166241
2012-10-18 22:22:55 +00:00
Chad Rosier d48d078487 [ms-inline asm] Add a size argument to the LookupInlineAsmIdentifier() callback,
which will be used by the asm matcher in the near future.

llvm-svn: 166222
2012-10-18 20:27:15 +00:00
Sebastian Pop fdd94d4955 Change MachineFrameInfo::StackObject::Alloca from Value* to AllocaInst*
This more accurately reflects what is actually being stored in the
field.

No functionality change intended.

Author:    Matthew Curtis <mcurtis@codeaurora.org>
llvm-svn: 166215
2012-10-18 19:53:45 +00:00
Chad Rosier f641baa030 [ms-inline asm] Have the LookupInlineAsmIdentifier() callback function return a
*NamedDecl.  In turn, build the expressions after we're finished parsing the
asm.  This avoids a crasher if the lookup fails.

llvm-svn: 166212
2012-10-18 19:39:30 +00:00
Chad Rosier 8bce664144 [ms-inline asm] Move most of the AsmParsing logic in clang back into the MC
layer.  Add the ParseMSInlineAsm() function, which is the new interface to 
clang.  Also expose the new MCAsmParserSemaCallback interface, which is used
by the back-end to do name lookup in Sema.  Finally, remove the now defunct
APIs introduced in r165946.

llvm-svn: 166183
2012-10-18 15:49:34 +00:00
Chandler Carruth e8479e15f5 Introduce a BarrierNoop pass, a hack designed to allow *some* control
over the implicitly-formed-and-nesting CGSCC pass manager and function
pass managers, especially when using them on the opt commandline or
using extension points in the module builder. The '-barrier' opt flag
(or the pass itself) will create a no-op module pass in the pipeline,
resetting the pass manager stack, and allowing the creation of a new
pipeline of function passes or CGSCC passes to be created that is
independent from any previous pipelines.

For example, this can be used to test running two CGSCC passes in
independent CGSCC pass managers as opposed to in the same CGSCC pass
manager. It also allows us to introduce a further hack into the
PassManagerBuilder to separate the O0 pipeline extension passes from the
always-inliner's CGSCC pass manager, which they likely do not want to
participate in... At the very least none of the Sanitizer passes want
this behavior.

This fixes a bug with ASan at O0 currently, and I'll commit the ASan
test which covers this pass. I'm happy to add a test case that this pass
exists and works, but not sure how much time folks would like me to
spend adding test cases for the details of its behavior of partition
pass managers.... The whole thing is just vile, and mostly intended to
unblock ASan, so I'm hoping to rip this all out in a brave new pass
manager world.

llvm-svn: 166172
2012-10-18 08:05:46 +00:00
Bob Wilson d6d9ccca38 Temporarily revert the TargetTransform changes.
The TargetTransform changes are breaking LTO bootstraps of clang.  I am
working with Nadav to figure out the problem, but I am reverting it for now
to get our buildbots working.

This reverts svn commits: 165665 165669 165670 165786 165787 165997
and I have also reverted clang svn 165741

llvm-svn: 166168
2012-10-18 05:43:52 +00:00
Jakob Stoklund Olesen 7a9f0c09de Switch MRI::UsedPhysRegs to a register unit bit vector.
This is a more compact, less redundant representation, and it avoids
scanning long lists of aliases for ARM D-registers, for example.

llvm-svn: 166124
2012-10-17 20:26:33 +00:00
Evan Cheng 839fb650b2 Add a really faster pre-RA scheduler (-pre-RA-sched=linearize). It doesn't use
any scheduling heuristics nor does it build up any scheduling data structure
that other heuristics use. It essentially linearize by doing a DFA walk but
it does handle glues correctly.

IMPORTANT: it probably can't handle all the physical register dependencies so
it's not suitable for x86. It also doesn't deal with dbg_value nodes right now
so it's definitely is still WIP.

rdar://12474515

llvm-svn: 166122
2012-10-17 19:39:36 +00:00
Jakob Stoklund Olesen 0736442683 Merge MRI::isPhysRegOrOverlapUsed() into isPhysRegUsed().
All callers of these functions really want the isPhysRegOrOverlapUsed()
functionality which also checks aliases. For historical reasons, targets
without register aliases were calling isPhysRegUsed() instead.

Change isPhysRegUsed() to also check aliases, and switch all
isPhysRegOrOverlapUsed() callers to isPhysRegUsed().

llvm-svn: 166117
2012-10-17 18:44:18 +00:00
Nadav Rotem 6b94c2a09b Add a loop vectorizer.
llvm-svn: 166112
2012-10-17 18:25:06 +00:00
Andrew Trick 0b1d8d04b9 misched: Better handling of invalid latencies in the machine model
llvm-svn: 166107
2012-10-17 17:27:10 +00:00
Craig Topper 1958f04bda Remove LLVM_DELETED_FUNCTION from destructors that override non-deleted base class destructors. This isn't legal by the C++11 standard and clang now checks for it. Curiously gcc didn't catch this, possibly because of the template usage.
llvm-svn: 166089
2012-10-17 05:15:58 +00:00
Jakob Stoklund Olesen a2136be107 Use a SparseSet instead of a BitVector for UsedInInstr in RAFast.
This is just as fast, and it makes it possible to avoid leaking the
UsedPhysRegs BitVector implementation through
MachineRegisterInfo::addPhysRegsUsed().

llvm-svn: 166083
2012-10-17 01:37:59 +00:00
Eric Christopher 587e153197 Spacing.
llvm-svn: 166074
2012-10-16 23:46:19 +00:00
Chad Rosier e4ad2a0b96 [ms-inline asm] Add the helper function, isParseringInlineAsm(). To be used in a future commit.
llvm-svn: 166054
2012-10-16 20:16:20 +00:00
Bill Wendling 7b9eb5c64a And now we can call the other 'get' method from this one and not duplicate the code.
llvm-svn: 166037
2012-10-16 18:20:09 +00:00
Bill Wendling 53a6f63c26 Use the appropriate Attributes::get method to create an Attributes object.
llvm-svn: 166035
2012-10-16 18:06:06 +00:00
Owen Anderson 544284eb31 Speculative fix the mask constants to be of type uintptr_t. I don't know of any case where the old form was incorrect, but I'm more confident that such cases don't exist in this version.
llvm-svn: 166031
2012-10-16 17:10:33 +00:00
Dmitri Gribenko 610a86e6bf Fix function parameter spelling in comments. Caught by -Wdocumentation.
llvm-svn: 166024
2012-10-16 15:37:50 +00:00
Stepan Dyatkovskiy e59a920b0c Issue:
Stack is formed improperly for long structures passed as byval arguments for
EABI mode.

If we took AAPCS reference, we can found the next statements:

A: "If the argument requires double-word alignment (8-byte), the NCRN (Next
Core Register Number) is rounded up to the next even register number." (5.5
Parameter Passing, Stage C, C.3).

B: "The alignment of an aggregate shall be the alignment of its most-aligned
component." (4.3 Composite Types, 4.3.1 Aggregates).

So if we have structure with doubles (9 double fields) and 3 Core unused
registers (r1, r2, r3): caller should use r2 and r3 registers only.
Currently r1,r2,r3 set is used, but it is invalid.

Callee VA routine should also use r2 and r3 regs only. All is ok here. This
behaviour is guessed by rounding up SP address with ADD+BFC operations.

Fix:
Main fix is in ARMTargetLowering::HandleByVal. If we detected AAPCS mode and
8 byte alignment, we waste odd registers then.

P.S.:
I also improved LDRB_POST_IMM regression test. Since ldrb instruction will
not generated by current regression test after this patch. 

llvm-svn: 166018
2012-10-16 07:16:47 +00:00
Bill Wendling 118a78b977 Cleanup whitespace.
llvm-svn: 166016
2012-10-16 06:10:45 +00:00
Owen Anderson 04b8daa970 Fix a bug in the set(I,E)/reset(I,E) methods that I recently added. The boundary condition for checking if I and E were in the same word were incorrect, and, beyond that, the mask computation was not using a wide enough constant.
llvm-svn: 166015
2012-10-16 06:04:27 +00:00
Bill Wendling 3ffbac4432 Have AttrBuilder defriend the Attributes class.
llvm-svn: 166011
2012-10-16 05:55:09 +00:00
Bill Wendling c6a15cf519 Use the Attributes::get method which takes an AttrVal value directly to simplify the code a bit. No functionality change.
llvm-svn: 166009
2012-10-16 05:23:31 +00:00
Bill Wendling a517c30ed3 Put simple c'tors inline.
llvm-svn: 166008
2012-10-16 05:22:28 +00:00
Andrew Trick d9d4be0d57 misched: Added handleMove support for updating all kill flags, not just for allocatable regs.
This is a medium term workaround until we have a more robust solution
in the form of a register liveness utility for postRA passes.

llvm-svn: 166001
2012-10-16 00:22:51 +00:00
Jakob Stoklund Olesen f67bf3e0ea Remove RegisterClassInfo::isReserved() and isAllocatable().
Clients can use the equivalent functions in MRI.

llvm-svn: 165990
2012-10-15 22:41:03 +00:00
Jakob Stoklund Olesen cea596acf7 Remove LIS::isAllocatable() and isReserved() helpers.
All callers can simply use the corresponding MRI functions.

llvm-svn: 165985
2012-10-15 22:14:34 +00:00
Owen Anderson 6b7bdf88c9 Add range-based set()/reset() to BitVector. These allow fast setting/resetting of ranges of bits, particularly useful when dealing with very large BitVector's.
llvm-svn: 165984
2012-10-15 22:05:27 +00:00
Jakob Stoklund Olesen c30a9af2d7 Switch most getReservedRegs() clients to the MRI equivalent.
Using the cached bit vector in MRI avoids comstantly allocating and
recomputing the reserved register bit vector.

llvm-svn: 165983
2012-10-15 21:57:41 +00:00
Jakob Stoklund Olesen 57e310613c Freeze the reserved registers as soon as isel is complete.
Also provide an MRI::getReservedRegs() function to access the frozen
register set, and isReserved() and isAllocatable() methods to test
individual registers.

The various implementations of TRI::getReservedRegs() are quite
complicated, and many passes need to look at the reserved register set.
This patch makes it possible for these passes to use the cached copy in
MRI, avoiding a lot of malloc traffic and repeated calculations.

llvm-svn: 165982
2012-10-15 21:33:06 +00:00
Bill Wendling 50d27849f6 Move the Attributes::Builder outside of the Attributes class and into its own class named AttrBuilder. No functionality change.
llvm-svn: 165960
2012-10-15 20:35:56 +00:00
Bill Wendling 6fe20085ac Add comments.
llvm-svn: 165958
2012-10-15 19:58:25 +00:00
Andrew Trick 90f711da9a misched: ILP scheduler for experimental heuristics.
llvm-svn: 165950
2012-10-15 18:02:27 +00:00
Chad Rosier 0d6f149e62 [ms-inline asm] Add a few new APIs to the AsmParser class in support of MS-Style
inline assembly.  For the time being, these will be called directly by clang.
However, in the near future I expect these to be sunk back into the MC layer
and more basic APIs (e.g., getClobbers(), getConstraints(), etc.) will be called
by clang.

llvm-svn: 165946
2012-10-15 17:19:13 +00:00
Micah Villmow 4bb926d91d Resubmit the changes to llvm core to update the functions to support different pointer sizes on a per address space basis.
llvm-svn: 165941
2012-10-15 16:24:29 +00:00
Adhemerval Zanella ef206f19a4 PowerPC: add EmitTCEntry class for TOC creation
This patch replaces the EmitRawText by a EmitTCEntry class (specialized for
each Streamer) in PowerPC64 TOC entry creation.

llvm-svn: 165940
2012-10-15 15:43:14 +00:00
Kostya Serebryany b0e2506d97 [asan] make AddressSanitizer to be a FunctionPass instead of ModulePass. This will simplify chaining other FunctionPasses with asan. Also some minor cleanup
llvm-svn: 165936
2012-10-15 14:20:06 +00:00
Bill Wendling fbd38fe2e3 Add an enum for the return and function indexes into the AttrListPtr object. This gets rid of some magic numbers.
llvm-svn: 165924
2012-10-15 07:29:08 +00:00
Bill Wendling 79d45dbbf9 Use a ::get method to create the attribute from Attributes::AttrVals instead of a constructor.
llvm-svn: 165923
2012-10-15 06:53:28 +00:00
Bill Wendling 4ec54eeed6 Supply a default 'operator=' method.
llvm-svn: 165922
2012-10-15 06:34:18 +00:00
Bill Wendling 8c3e65db52 Move the AttributesImpl header file into the VMCore directory so that it can be opaque.
llvm-svn: 165920
2012-10-15 05:40:12 +00:00
Bill Wendling d079a446d7 Attributes Rewrite
Convert the internal representation of the Attributes class into a pointer to an
opaque object that's uniqued by and stored in the LLVMContext object. The
Attributes class then becomes a thin wrapper around this opaque
object. Eventually, the internal representation will be expanded to include
attributes that represent code generation options, etc.

llvm-svn: 165917
2012-10-15 04:46:55 +00:00
Benjamin Kramer 43f2507fce Fix a typo that made ImmutableMap::getMaxElement() useless.
Add a basic unit test for ImmutableMap. Found by inspection.

llvm-svn: 165907
2012-10-14 15:56:39 +00:00
Bill Wendling 5ae76a379c Remove dead methods.
llvm-svn: 165902
2012-10-14 09:21:44 +00:00
Bill Wendling 76d2cd2f60 Remove operator cast method in favor of querying with the correct method.
llvm-svn: 165899
2012-10-14 08:54:26 +00:00
Bill Wendling 2a3c1cca7d Remove the bitwise AND operators from the Attributes class. Replace it with the equivalent from the builder class.
llvm-svn: 165896
2012-10-14 07:52:48 +00:00
Bill Wendling 722b26c0f2 Remove the bitwise assignment OR operator from the Attributes class. Replace it with the equivalent from the builder class.
llvm-svn: 165895
2012-10-14 07:35:59 +00:00
Bill Wendling 5c407ed3ab Remove the bitwise OR operator from the Attributes class. Replace it with the equivalent from the builder class.
llvm-svn: 165894
2012-10-14 07:17:34 +00:00
Bill Wendling a05b043c4a Remove the bitwise XOR operator from the Attributes class. Replace it with the equivalent from the builder class.
llvm-svn: 165893
2012-10-14 06:56:13 +00:00
Bill Wendling 85a64c217f Remove the bitwise NOT operator from the Attributes class. Replace it with the equivalent from the builder class.
llvm-svn: 165892
2012-10-14 06:39:53 +00:00
Bill Wendling 1fcc82225a Decode the LLVM attributes from bitcode using the attributes builder.
llvm-svn: 165891
2012-10-14 04:10:01 +00:00