Commit Graph

22568 Commits

Author SHA1 Message Date
Chris Lattner a39cfc5c5b silence a warning when assertions are disabled.
llvm-svn: 49283
2008-04-06 21:44:08 +00:00
Gabor Greif dd8b7a0a47 fix a warning
llvm-svn: 49282
2008-04-06 21:42:13 +00:00
Torok Edwin 613d7afe64 Prefer to expand mask for xor to -1, so we have a chance to turn it into a not.
If it cannot be expanded, it will keep the old behaviour and try to shrink the constant.
Part of enhancement for PR2191.

llvm-svn: 49280
2008-04-06 21:23:02 +00:00
Gabor Greif e9ecc68d8f API changes for class Use size reduction, wave 1.
Specifically, introduction of XXX::Create methods
for Users that have a potentially variable number of
Uses.

llvm-svn: 49277
2008-04-06 20:25:17 +00:00
Chris Lattner 8519c0ead8 disable this for now.
llvm-svn: 49248
2008-04-05 04:12:01 +00:00
Evan Cheng b5fdc923d3 1. IMPLICIT_DEF can *re-define* any register.
2. Coalescer can now create an interesting situation where a register def can
   reaches itself without being killed.

llvm-svn: 49246
2008-04-05 01:27:09 +00:00
Evan Cheng f77b5ef3d0 Favors pshufd over shufps when shuffling elements from one vector. pshufd is faster than shufps.
llvm-svn: 49244
2008-04-05 00:30:36 +00:00
Dale Johannesen 0ce4a7cc44 Make sure both PendingLoads and PendingExports are flushed
before an invoke.  Failure to do this causes references in
the landing pad to variables that were not set.  Fixes
g++.dg/eh/delayslot1.C
g++.dg/eh/fp-regs.C
g++.old-deja/g++.brendan/eh1.C

llvm-svn: 49243
2008-04-04 23:48:31 +00:00
Torok Edwin b20e659770 strdup needs <cstring>. This fixes a build error with g++-4.3.
llvm-svn: 49218
2008-04-04 16:08:00 +00:00
Chris Lattner 318d3ef88e Provide an initial cut at exposing JIT compiled symbols to performance
tools.  This is currently only enabled on the mac, but could easily be
supported by other hosts that are interested.

llvm-svn: 49207
2008-04-04 05:51:42 +00:00
Chris Lattner ee1817300a Make ExecutionEngine::updateGlobalMapping return the old mapping.
llvm-svn: 49206
2008-04-04 04:47:41 +00:00
Evan Cheng 14bee50e06 Undo PHI elimination copy placement patch. This causes coalescing (performace) issues.
llvm-svn: 49198
2008-04-04 01:20:05 +00:00
Evan Cheng 823017fdd1 This is done.
llvm-svn: 49197
2008-04-04 01:19:03 +00:00
Andrew Lenharth bfb7246fb6 if some functions don't have debug info, we were outputing the same label at the start of each of those functions. This makes assemblers unhappy
llvm-svn: 49176
2008-04-03 17:37:43 +00:00
Evan Cheng 58936a48ee - Turn copies of implicit_def into implicit_def instructions.
- Be smarter about coalescing copies from implicit_def.

llvm-svn: 49168
2008-04-03 16:41:54 +00:00
Evan Cheng 6d07b625aa Special handling of zero-sized live intervals.
llvm-svn: 49167
2008-04-03 16:40:27 +00:00
Evan Cheng 20aed56504 - Treat a live range defined by an implicit_def as a zero-sized one.
- Eliminate an implicit_def when it's being spilled.

llvm-svn: 49166
2008-04-03 16:39:43 +00:00
Evan Cheng aacf4f15b3 - PHI elimination also eliminates implicit_def that fits into a PHI node rather than copying it.
- Be (slightly) smarter about where to place the copies.

llvm-svn: 49165
2008-04-03 16:38:20 +00:00
Evan Cheng 916802a78e Start of a series of patches related to implicit_def.
There is no point in creating a long live range defined by an implicit_def. Scheduler now duplicates implicit_def instruction for each of its uses. Therefore, if an implicit_def node has multiple uses, it will become a number of very short live ranges, rather than a long one. This will make coalescer's job easier.

llvm-svn: 49164
2008-04-03 16:36:07 +00:00
Ted Kremenek 1d0436cdb0 Re-implemented Path::createDirectoryOnDisk (for Unix).
This method allows one to create a directory, and optionally create all parent
directories that do not exist.

The original implementation would require that *all* directories along a path
are writable by the user, including directories that already exist. For example,
suppose we wanted to create the directory "/tmp/foo/bar", and the directory
"/tmp" already exists, but not "/tmp/foo". Since "/tmp" is writable by all
users, the original implementation would work, and create "/tmp/foo", followed
by "/tmp/bar".

A problem occurred, however if one wanted to created the directory
"/Users/myuser/bar" (or equivalently "/home/myuser/bar"), and "/Users/myuser"
already existed and is writable by the current user. The directory
"/User/myuser" is writable by the user, but "/User" is not. The original
implementation of createDirectoryOnDisk would return with failure since "/User"
is not writable, even though "/User/mysuser" is writable.

The new implementation works by recursively creating parents as needed, and thus
doesn't need to check the permissions on every directory in a path.

llvm-svn: 49162
2008-04-03 16:11:31 +00:00
Evan Cheng 6c66bd368e Re-enable SSE4.
llvm-svn: 49158
2008-04-03 08:53:29 +00:00
Evan Cheng 6db4b4cc65 Fix x86-64 encoding bug. REX prefix must always follow 0x0F prefix. For example, extractps in 64bit mode: 66 REX 0F 3A 17, not 66 0F 3A REX 17.
llvm-svn: 49157
2008-04-03 08:53:17 +00:00
Evan Cheng d9129d1de3 Cosmetic
llvm-svn: 49156
2008-04-03 07:45:18 +00:00
Evan Cheng 3063c5546e Temporarily disabling SSE4 until we fix the encoding issues.
llvm-svn: 49129
2008-04-03 04:49:54 +00:00
Evan Cheng 025cea1126 Backing out 48222 temporarily.
llvm-svn: 49124
2008-04-03 03:13:16 +00:00
Dan Gohman bd72cea737 Suppress the 128-bit integer typedef on 32-bit targets, because
it causes compile errors.

llvm-svn: 49122
2008-04-02 23:52:49 +00:00
Dale Johannesen 491557712a Make EH work with unnamed functions. Reenable running
StripSymbols when EH is on.

llvm-svn: 49110
2008-04-02 20:10:52 +00:00
Dan Gohman 39d8b26322 Partial CBackend support for 128-bit integers. This is needed
now that llvm-gcc is lowering appropriately-sized struct returns
to i128 on x86-64.

llvm-svn: 49109
2008-04-02 19:40:14 +00:00
David Greene 586740f401 Iterators folloring a SmallVector erased element are invalidated so
don't access cached iterators from after the erased element.

Re-apply 49056 with SmallVector support.

llvm-svn: 49106
2008-04-02 18:24:46 +00:00
Evan Cheng d8616064d8 Now that I am told MachineRegisterInfo also tracks physical register uses / defs, I can do away with the horribleness I introduced a while back. It's impossible to detect if there is any use of a physical register below an instruction (and before any def of the register) with some cheap book keeping.
llvm-svn: 49105
2008-04-02 18:04:08 +00:00
Evan Cheng be3d44c3cb Remove #include<map> from LiveVariables.h. Not referenced.
llvm-svn: 49099
2008-04-02 17:23:50 +00:00
Dale Johannesen 8780ecbbac Cosmetic changes per EH patch review feedback.
llvm-svn: 49096
2008-04-02 17:04:45 +00:00
Anton Korobeynikov 20c9e4cbee Add new CC lowering rule: provide a list of registers, which can be 'shadowed',
when some another register is used for argument passing.
Currently is used on Win64.

llvm-svn: 49079
2008-04-02 05:23:57 +00:00
Owen Anderson 2412158111 In some situations, we need to check for local interferences between the PHI
node and its inputs.

llvm-svn: 49070
2008-04-02 03:00:13 +00:00
Owen Anderson edfc2eb558 Correctly mark a valno that was previous defined by a PHI node as having an
unknown defining inst after PHI elimination.

llvm-svn: 49069
2008-04-02 02:12:45 +00:00
Dale Johannesen fd967cf3fa Recommitting EH patch; this should answer most of the
review feedback.
-enable-eh is still accepted but doesn't do anything.
EH intrinsics use Dwarf EH if the target supports that,
and are handled by LowerInvoke otherwise.
The separation of the EH table and frame move data is,
I think, logically figured out, but either one still
causes full EH info to be generated (not sure how to
split the metadata correctly).
MachineModuleInfo::needsFrameInfo is no longer used and
is removed.

llvm-svn: 49064
2008-04-02 00:25:04 +00:00
Evan Cheng ac38d444e2 1. Drop default inline threshold back down to 200.
2. Do not use # of basic blocks as part of the cost computation since it doesn't really figure into function size.
3. More aggressively inline function with vector code.

llvm-svn: 49061
2008-04-01 23:59:29 +00:00
Tanya Lattner 052838c55d Reverting 49056 due to the build being broken.
llvm-svn: 49060
2008-04-01 23:41:44 +00:00
Evan Cheng b86595fb0a ReMat of load from stub in pic mode extends the life of pic base. Currently spiller doesn't do a good job of estimating the impact. Disable for now.
llvm-svn: 49059
2008-04-01 23:26:12 +00:00
David Greene 7f7edc3824 Iterators folloring a SmallVector erased element are invalidated so
don't access cached iterators from after the erased element.

llvm-svn: 49056
2008-04-01 22:14:23 +00:00
Bill Wendling fc9f25dab0 Fix compilation errors on MSVC. Patch by Argiris Kirtzidis!
llvm-svn: 49055
2008-04-01 22:09:20 +00:00
Evan Cheng 19a6dd9f2a Remove unnecessary and non-deterministic checking code. Re-enable remat of load from gv stub.
llvm-svn: 49054
2008-04-01 21:38:20 +00:00
Evan Cheng 985a0b51d7 Re-materialization is for uses only.
llvm-svn: 49053
2008-04-01 21:37:32 +00:00
Dan Gohman cb9f8f6e4e Don't use __bzero for memset if the second argument isn't zero.
llvm-svn: 49050
2008-04-01 20:56:18 +00:00
Dan Gohman 980d7200c1 Speculatively micro-optimize memory-zeroing calls on Darwin 10.
llvm-svn: 49048
2008-04-01 20:38:36 +00:00
Dale Johannesen 5e4e051c2a Revert 49006 for the moment.
llvm-svn: 49046
2008-04-01 20:00:57 +00:00
Owen Anderson 49dd9f16a9 Don't dereference MBB->end().
llvm-svn: 49043
2008-04-01 18:05:08 +00:00
Chris Lattner 6f2ffdb73f Change the MemoryBuffer::getFile* methods to take just a pointer to the
start of a filename, not a filename+length.  All clients can produce a
null terminated name, and the system api's require null terminated 
strings anyway.

llvm-svn: 49041
2008-04-01 18:04:03 +00:00
Evan Cheng 306e3dcff4 Disabling remat of load from gv stub (temporarily) again to fix llvmgcc bootstrap miscompare.
llvm-svn: 49037
2008-04-01 07:33:13 +00:00
Chris Lattner ba98fca5e5 add some #includes.
llvm-svn: 49036
2008-04-01 06:25:23 +00:00
Chris Lattner 18356d8857 MappedFile is dead, remove it.
llvm-svn: 49035
2008-04-01 06:20:44 +00:00
Chris Lattner 4c5e15f7da Implement Path::MapInFilePages/UnMapFilePages on unix, which
provides fast MappedFile::getFile for large files.

llvm-svn: 49034
2008-04-01 06:16:24 +00:00
Chris Lattner a542518315 Reimplement MemoryBuffer::getFile with three enhancements:
1) stop using MappedFile.
2) if profitable use the sys::path::MapInFilePages api to
   read the file.
3) otherwise fallback to read.

When sys::path::MapInFilePages is implemented, this provides
several benefits:

: this avoids fragmenting memory for small files.
: this avoids extraneous stat calls when the file size is known.
: this only keeps the file descriptor open while reading the 
    file, not for the duration of the lifetime of the memory 
    buffer.  This fixes a serious clang FD 'leak' problem.

I believe that this will work on a win32 machine, but I don't have
one to test on.  I'd appreciate it if someone could check.

llvm-svn: 49031
2008-04-01 06:05:21 +00:00
Chris Lattner 3089e1d82e Stub out some sys::Path::MapInFilePages/UnMapFilePages methods.
llvm-svn: 49030
2008-04-01 06:00:12 +00:00
Chris Lattner d4310a27aa change the archive stuff to use MemoryBuffer instead of mappedfile.
MemoryBuffer is higher level and more closely matches the model
needed.

llvm-svn: 49029
2008-04-01 04:26:46 +00:00
Chris Lattner 4ac7bb0294 rewrite SourceFile to be in terms of MemoryBuffer, not MappedFile.
llvm-svn: 49027
2008-04-01 03:59:34 +00:00
Chris Lattner 64959dcf03 Make MappedFile::map return a const correct pointer, don't leak address space on Unix platforms.
llvm-svn: 49026
2008-04-01 03:49:38 +00:00
Chris Lattner bcfa564759 Remove the MappedFile::charBase member, rename base -> getBase() and
make getBase() return a const-correct pointer.

llvm-svn: 49025
2008-04-01 03:40:53 +00:00
Chris Lattner 314a1414d0 Change DiffFilesWithTolerance to be written in terms of MemoryBuffer,
not an mmapped file.  This more closely matches its requirements and
provides an implicitly null terminated buffer, something this
routine had to emulate itself before.

llvm-svn: 49024
2008-04-01 03:39:49 +00:00
Chris Lattner 62385ba3a0 Remove MappedFile support for mapping files for write and exec
and shared.  This complicates the design, is not used, and probably
doesn't even work.

llvm-svn: 49022
2008-04-01 03:10:22 +00:00
Chris Lattner 3d7127b3e3 add missing #includes
llvm-svn: 49020
2008-04-01 02:58:05 +00:00
Evan Cheng 0bd72c5ccd More soft fp fixes.
llvm-svn: 49016
2008-04-01 02:18:22 +00:00
Evan Cheng 4cabe4b452 Pasto.
llvm-svn: 49014
2008-04-01 02:00:09 +00:00
Evan Cheng 611abc03ed Add comment.
llvm-svn: 49013
2008-04-01 01:51:26 +00:00
Evan Cheng 86e476b7cb Unbreak ARM / Thumb soft FP support.
llvm-svn: 49012
2008-04-01 01:50:16 +00:00
Dale Johannesen efa81a6979 Accept 'y' constraint (MMX) in inline asm.
llvm-svn: 49011
2008-04-01 00:57:48 +00:00
Chris Lattner efecfbc113 cleanup the MappedFile API and comments. This removes and updates
tons of out of date comments (really nothing throws here!) and fixes
some other fairly glaring issues: "size" used to return the size of 
the file *and* change it, depending on how you called it.

llvm-svn: 49009
2008-04-01 00:53:25 +00:00
Chris Lattner 1fa4dca692 remove DEFINING_FILE_FOR for MappedFile.h
llvm-svn: 49008
2008-04-01 00:35:55 +00:00
Dale Johannesen 7d02cf3c9c Emit exception handling info for functions which are
not marked nounwind, or for all functions when -enable-eh
is set, provided the target supports Dwarf EH.

llvm-gcc generates nounwind in the right places; other FEs
will need to do so also.  Given such a FE, -enable-eh should
no longer be needed.

llvm-svn: 49006
2008-03-31 23:40:23 +00:00
Evan Cheng e4f77c69ac It's not safe to fold a load from GV stub or constantpool into a two-address use.
llvm-svn: 49002
2008-03-31 23:19:51 +00:00
Nate Begeman f8be3836e3 Actually disable crash reporting on Mac OS X, returning bugpoint to speedy
crash miscompilations.

llvm-svn: 49000
2008-03-31 22:19:25 +00:00
Dan Gohman 3717cdaf22 Set blockBegin to point to the beginning of the block,
not the end.

llvm-svn: 48999
2008-03-31 22:08:00 +00:00
Evan Cheng ed6e34fe41 Move reMaterialize() from TargetRegisterInfo to TargetInstrInfo.
llvm-svn: 48995
2008-03-31 20:40:39 +00:00
Dan Gohman f549b26254 Fix a DAGCombiner optimization to respect volatile qualification.
llvm-svn: 48994
2008-03-31 20:32:52 +00:00
Erick Tryzelaar 4a0da98825 Expose Function::viewCFG and Function::viewCFGOnly to bindings.
llvm-svn: 48982
2008-03-31 16:22:09 +00:00
Evan Cheng 1973a46cd3 Re-apply 48911.
llvm-svn: 48977
2008-03-31 07:54:19 +00:00
Evan Cheng 73d7c3bfba The support for remat of instructions with a register operand is hackish, to say the least. Since the register operand guaranteed to be PIC base and that it is already live at all uses, we are making sure it will not be spilled after its uses are rematerialized for both performance and correctness reasons.
llvm-svn: 48976
2008-03-31 07:53:30 +00:00
Owen Anderson f28fc71c93 Fix a major bug in the DFS calculation. Thanks for Christopher Lamb for pointing this out.
llvm-svn: 48973
2008-03-31 01:39:20 +00:00
Nate Begeman f2b0b0eb17 Don't eliminate bitcast instructions that change the type of a pointer
llvm-svn: 48971
2008-03-31 00:22:16 +00:00
Duncan Sands 7d6c8ae488 Fix comment typo.
llvm-svn: 48967
2008-03-30 19:38:55 +00:00
Nick Lewycky 9fb8908457 Moved from PR1570.
llvm-svn: 48965
2008-03-30 19:07:11 +00:00
Chris Lattner 0f760dfe09 Fix "Control reaches the end of non-void function" warnings,
patch by David Chisnall.

llvm-svn: 48963
2008-03-30 18:22:13 +00:00
Chris Lattner 519a51a76f minor code cleanups, allow constant folding sinf/cosf.
llvm-svn: 48961
2008-03-30 18:02:00 +00:00
Evan Cheng 16d72072df Cosmetic changes.
llvm-svn: 48947
2008-03-29 18:34:22 +00:00
Chris Lattner 4311ad2dae change iterator invalidation avoidance to just move the iterator backward
when something changes, instead of moving forward.  This allows us to 
simplify memset lowering, inserting the memset at the end of the range of 
stuff we're touching instead of at the start.

This, in turn, allows us to make use of the addressing instructions already
used in the function instead of inserting our own.  For example, we now
codegen:

	%tmp41 = getelementptr [8 x i8]* %ref_idx, i32 0, i32 0		; <i8*> [#uses=2]
	call void @llvm.memset.i64( i8* %tmp41, i8 -1, i64 8, i32 1 )

instead of:

	%tmp20 = getelementptr [8 x i8]* %ref_idx, i32 0, i32 7		; <i8*> [#uses=1]
	%ptroffset = getelementptr i8* %tmp20, i64 -7		; <i8*> [#uses=1]
	call void @llvm.memset.i64( i8* %ptroffset, i8 -1, i64 8, i32 1 )

llvm-svn: 48940
2008-03-29 05:15:47 +00:00
Chris Lattner ac95515741 make the common case of a single store (which clearly shouldn't be turned
into a memset!) faster by avoiding an allocation of an std::list node.

llvm-svn: 48939
2008-03-29 04:52:12 +00:00
Chris Lattner d528b21a65 give form-memset a significantly more sane heuristic, enable it by default.
llvm-svn: 48937
2008-03-29 04:36:18 +00:00
Owen Anderson 8b22873bdd Remove some unneeded code for LiveInterval joining, and fix a bug in the Phi elimination algorithm where we were accidentally reasoning about
the source rather than the destination.

llvm-svn: 48936
2008-03-29 01:58:47 +00:00
Dan Gohman fd2eb00cc2 Fix a tokenfactor node to use the load chain rather than the
load value. This fixes PR2177.

llvm-svn: 48932
2008-03-28 23:45:16 +00:00
Evan Cheng b8654202dd Backing out 48911 for now. It's breaking stuff.
llvm-svn: 48922
2008-03-28 17:49:06 +00:00
Chris Lattner a148acdc82 ifdef out a dead function. Should this be removed?
llvm-svn: 48916
2008-03-28 15:36:27 +00:00
Duncan Sands 35c7cdac07 Rename getAnyLoad to getLoad is suggested by Evan.
llvm-svn: 48914
2008-03-28 09:45:24 +00:00
Evan Cheng 81e0c9a32c New entry.
llvm-svn: 48912
2008-03-28 07:07:06 +00:00
Evan Cheng 9ae4d7b719 Load from stub is already re-materializable.
llvm-svn: 48911
2008-03-28 06:49:25 +00:00
Nick Lewycky f88c84f690 Update example to new syntax.
llvm-svn: 48910
2008-03-28 06:46:51 +00:00
Chris Lattner d62964a7d8 make memset inference significantly more powerful: it can now handle
memsets that initialize "structs of arrays" and other store sequences
that are not sequential.  This is still only enabled if you pass 
-form-memset-from-stores.  The flag is not heavily tested and I haven't
analyzed the perf regressions when -form-memset-from-stores is passed
either, but this causes no make check regressions.

llvm-svn: 48909
2008-03-28 06:45:13 +00:00
Evan Cheng 87bac50d7b New entry.
llvm-svn: 48908
2008-03-28 06:34:23 +00:00
Duncan Sands f740509e58 Implement LegalizeTypes support for softfloat LOAD.
In order to handle indexed nodes I had to introduce
a new constructor, and since I was there I factorized
the code in the various load constructors.

llvm-svn: 48894
2008-03-27 20:23:40 +00:00
Dan Gohman cad51cb671 Avoid creating chain dependencies from CopyToReg nodes to load and store
nodes. This doesn't currently have much impact the generated code, but it
does produce simpler-looking SelectionDAGs, and consequently
simpler-looking ScheduleDAGs, because there are fewer spurious
dependencies.

In particular, CopyValueToVirtualRegister now uses the entry node as the
input chain dependency for new CopyToReg nodes instead of calling getRoot
and depending on the most recent memory reference.

Also, rename UnorderedChains to PendingExports and pull it up from being
a local variable in SelectionDAGISel::BuildSelectionDAG to being a
member variable of SelectionDAGISel, so that it doesn't have to be
passed around to all the places that need it.

llvm-svn: 48893
2008-03-27 19:56:19 +00:00
Devang Patel eb1e3fcbe0 PHI->removeIncomingValue may remove PHInode.
Increment iterator in advance.

llvm-svn: 48890
2008-03-27 17:32:46 +00:00