Commit Graph

9242 Commits

Author SHA1 Message Date
Chris Lattner 1fc2773a33 roundss is an sse 4 thing, fix the test on non-sse41 builders
like llvm-gcc-x86_64-darwin10-selfhost

llvm-svn: 96417
2010-02-17 00:29:06 +00:00
Dale Johannesen cee887425e Make g5 target explicit; scheduling affects register choice.
llvm-svn: 96413
2010-02-16 23:25:23 +00:00
Chris Lattner afac7dad21 fix rdar://7653908, a crash on a case where we would fold a load
into a roundss intrinsic, producing a cyclic dag.  The root cause
of this is badness handling ComplexPattern nodes in the old dagisel
that I noticed through inspection.  Eliminate a copy of the of the
code that handled ComplexPatterns by making EmitChildMatchCode call
into EmitMatchCode.

llvm-svn: 96408
2010-02-16 22:35:06 +00:00
Dale Johannesen 0062f7bf59 Adjust register numbers in tests to compensate for the
new lack of R2.

llvm-svn: 96407
2010-02-16 22:31:31 +00:00
Chris Lattner c98beb567c filecheckize
llvm-svn: 96404
2010-02-16 22:13:43 +00:00
Devang Patel 8b9fec4428 New testcase.
llvm-svn: 96391
2010-02-16 21:16:08 +00:00
Evan Cheng 82b04130cb Look for SSE and instructions of this form: (and x, (build_vector c1,c2,c3,c4)).
If there exists a use of a build_vector that's the bitwise complement of the mask,
then transform the node to
(and (xor x, (build_vector -1,-1,-1,-1)), (build_vector ~c1,~c2,~c3,~c4)).

Since this transformation is only useful when 1) the given build_vector will
become a load from constpool, and 2) (and (xor x -1), y) matches to a single
instruction, I decided this is appropriate as a x86 specific transformation.
rdar://7323335

llvm-svn: 96389
2010-02-16 21:09:44 +00:00
David Greene 9641d06809 Add support for emitting non-temporal stores for DAGs marked
non-temporal.  Fix from r96241 for botched encoding of MOVNTDQ.

Add documentation for !nontemporal metadata.

Add a simpler movnt testcase.

llvm-svn: 96386
2010-02-16 20:50:18 +00:00
Bob Wilson 6bfacb7393 Testcase for critical edge splitting with load PRE.
llvm-svn: 96385
2010-02-16 20:48:55 +00:00
Bob Wilson 70aa8d0745 Fix pr6111: Avoid using the LR register for the target address of an indirect
branch in ARM v4 code, since it gets clobbered by the return address before
it is used.  Instead of adding a new register class containing all the GPRs
except LR, just use the existing tGPR class.

llvm-svn: 96360
2010-02-16 17:24:15 +00:00
Chris Lattner 4964ef88c2 make pcrel immediate values relative to the start of the field,
not the end of the field, fixing rdar://7651978

llvm-svn: 96330
2010-02-16 05:03:17 +00:00
Dan Gohman 521efe68ab Split the main for-each-use loop again, this time for GenerateTruncates,
as it also peeks at which registers are being used by other uses. This
makes LSR less sensitive to use-list order.

llvm-svn: 96308
2010-02-16 01:42:53 +00:00
Anton Korobeynikov ae4ccc10da Preliminary patch to improve dwarf EH generation - Hooks to return Personality / FDE / LSDA / TType encoding depending on target / options (e.g. code model / relocation model) - MCIzation of Dwarf EH printer to use encoding information - Stub generation for ELF target (needed for indirect references) - Some other small changes here and there
llvm-svn: 96285
2010-02-15 22:35:59 +00:00
Jakob Stoklund Olesen 2988d573e5 Fix PR6300.
A virtual register can be used before it is defined in the same MBB if the MBB
is part of a loop. Teach the implicit-def pass about this case.

llvm-svn: 96279
2010-02-15 22:03:29 +00:00
Bob Wilson 9be7200b08 Last week we were generating code with duplicate induction variables in this
test, but the problem seems to have gone away today.  Add a check to make sure
it doesn't come back.

llvm-svn: 96277
2010-02-15 21:56:40 +00:00
Chris Lattner 3818d9763d remove empty file.
llvm-svn: 96271
2010-02-15 21:14:50 +00:00
Chris Lattner bcbaaba532 revert r96241. It breaks two regression tests, isn't documented,
and the testcase needs improvement.

llvm-svn: 96265
2010-02-15 20:53:01 +00:00
Chris Lattner 6fbfe5897c fix PR6305 by handling BlockAddress in a helper function
called by jump threading.

llvm-svn: 96263
2010-02-15 20:47:49 +00:00
David Greene 63cedef74b Add support for emitting non-temporal stores for DAGs marked
non-temporal.

llvm-svn: 96241
2010-02-15 17:02:56 +00:00
Mikhail Glushenkov 5352f07f2c Revert r96130 ("Forward parameter options as '-option=param'").
This behaviour must be configurable.

llvm-svn: 96210
2010-02-15 03:17:06 +00:00
Eric Christopher 843a4cc43c Fix a problem where we had bitcasted operands that gave us
odd offsets since the bitcasted pointer size and the offset pointer
size are going to be different types for the GEP vs base object.

llvm-svn: 96134
2010-02-13 23:38:01 +00:00
Mikhail Glushenkov 32fa169648 Forward parameter options as '-option=parameter'.
Some tools do not like the '-option parameter' form. Should this be
configurable?

llvm-svn: 96130
2010-02-13 22:37:28 +00:00
Chris Lattner f83726f6ba add encoder support and tests for rdtscp
llvm-svn: 96076
2010-02-13 03:42:24 +00:00
Jakob Stoklund Olesen b659c76c77 Fix PR6283.
When coalescing with a physreg, remember to add imp-def and imp-kill when
dealing with sub-registers.

Also fix a related bug in VirtRegRewriter where substitutePhysReg may
reallocate the operand list on an instruction and invalidate the reg_iterator.
This can happen when a register is mentioned twice on the same instruction.

llvm-svn: 96072
2010-02-13 02:06:10 +00:00
Daniel Dunbar d0c6d361fe MC/AsmParser: Attempt to constant fold expressions up-front. This ensures we avoid fixups for obvious cases like '-(16)'.
llvm-svn: 96064
2010-02-13 01:28:07 +00:00
Chris Lattner 509154e0f9 rip out the 'heinous' x86 MCCodeEmitter implementation.
We still have the templated X86 JIT emitter, *and* the
almost-copy in X86InstrInfo for getting instruction sizes.

llvm-svn: 96059
2010-02-13 00:49:29 +00:00
Chris Lattner 140caa7240 remove special cases for vmlaunch, vmresume, vmxoff, and swapgs
fix swapgs to be spelled right.

llvm-svn: 96058
2010-02-13 00:41:14 +00:00
Bob Wilson 01abf8fc2f Besides removing phi cycles that reduce to a single value, also remove dead
phi cycles.  Adjust a few tests to keep dead instructions from being optimized
away.  This (together with my previous change for phi cycles) fixes Apple
radar 7627077.

llvm-svn: 96057
2010-02-13 00:31:44 +00:00
Daniel Dunbar 224340cabe MC/X86: Push immediate operands as immediates not expressions when possible.
llvm-svn: 96055
2010-02-13 00:17:21 +00:00
Chris Lattner 34749d879d add some disassemble testcases for weird instructions
llvm-svn: 96045
2010-02-12 23:46:48 +00:00
Chris Lattner 1e827fd8ca implement the rest of correct x86-64 encoder support for
rip-relative addresses, and add a testcase.

llvm-svn: 96040
2010-02-12 23:24:09 +00:00
Dale Johannesen 26062150fa When save/restoring CR at prolog/epilog, in a large
stack frame, the prolog/epilog code was using the same
register for the copy of CR and the address of the save slot.  Oops.
This is fixed here for Darwin, sort of, by reserving R2 for this case.
A better way would be to do the store before the decrement of SP,
which is safe on Darwin due to the red zone.

SVR4 probably has the same problem, but I don't know how to fix it;
there is no red zone and R2 is already used for something else.
I'm going to leave it to someone interested in that target.

Better still would be to rewrite the CR-saving code completely;
spilling each CR subregister individually is horrible code.

llvm-svn: 96015
2010-02-12 21:35:34 +00:00
Chris Lattner 392be58cad Add support for a union type in LLVM IR. Patch by Talin!
llvm-svn: 96011
2010-02-12 20:49:41 +00:00
Chris Lattner 75879be9d8 1. modernize the constantmerge pass, using densemap/smallvector.
2. don't bother trying to merge globals in non-default sections,
   doing so is quite dubious at best anyway.
3. fix a bug reported by Arnaud de Grandmaison where we'd try to
   merge two globals in different address spaces.

llvm-svn: 95995
2010-02-12 18:17:23 +00:00
Chris Lattner 554003f481 rename test
llvm-svn: 95993
2010-02-12 18:05:00 +00:00
Anton Korobeynikov b9ce3cc458 Testcases for recent stdcall / fastcall mangling improvements
llvm-svn: 95982
2010-02-12 15:29:13 +00:00
Anton Korobeynikov c9276dfe04 Cleanup stdcall / fastcall name mangling.
This should fix alot of problems we saw so far, e.g. PRs 5851 & 2936

llvm-svn: 95980
2010-02-12 15:28:40 +00:00
Dan Gohman 45774ce0ad Reapply the new LoopStrengthReduction code, with compile time and
bug fixes, and with improved heuristics for analyzing foreign-loop
addrecs.

This change also flattens IVUsers, eliminating the stride-oriented
groupings, which makes it easier to work with.

llvm-svn: 95975
2010-02-12 10:34:29 +00:00
Evan Cheng 0e4df63bd5 Update test to match 95961.
llvm-svn: 95971
2010-02-12 07:48:46 +00:00
Evan Cheng 993bd1b7da Test for 95961.
llvm-svn: 95962
2010-02-12 02:35:03 +00:00
Evan Cheng 67e45e1670 Test case for 95958.
llvm-svn: 95959
2010-02-12 02:02:23 +00:00
Bob Wilson 0827e040e0 Add a new pass on machine instructions to optimize away PHI cycles that
reduce down to a single value.  InstCombine already does this transformation
but DAG legalization may introduce new opportunities.  This has turned out to
be important for ARM where 64-bit values are split up during type legalization:
InstCombine is not able to remove the PHI cycles on the 64-bit values but
the separate 32-bit values can be optimized.  I measured the compile time 
impact of this (running llc on 176.gcc) and it was not significant.

llvm-svn: 95951
2010-02-12 01:30:21 +00:00
Chris Lattner 1572e760bc fix the encodings of monitor and mwait, which were completely
busted in both encoders.  I'm not bothering to fix it in the
old one at this point.

llvm-svn: 95947
2010-02-12 01:06:22 +00:00
Charles Davis be5557e86b Add a new function attribute, 'alignstack'. It will indicate (when the backends
implement support for it) that the stack should be forcibly realigned in the
prologue (and the process reversed in the epilogue).

llvm-svn: 95945
2010-02-12 00:31:15 +00:00
Jakob Stoklund Olesen 93c92225af Reapply coalescer fix for better cross-class coalescing.
This time with fixed test cases.

llvm-svn: 95938
2010-02-11 23:55:29 +00:00
Eric Christopher cccdc13662 Make sure that ConstantExpr offsets also aren't off of extern
symbols.

Thanks to Duncan Sands for the testcase!

llvm-svn: 95877
2010-02-11 17:44:04 +00:00
Chris Lattner 4e8137d678 Rename ValueRequiresCast to ShouldOptimizeCast, to better reflect
what it does.  Enhance it to return false to optimizing vector
sign extensions from vector comparisions, which is the idiom used
to get a splatted vector for a vector comparison.

Doing this breaks vector-casts.ll, add some compensating 
transformations to handle the important case they cover without
depending on this canonicalization.

This fixes rdar://7434900 a serious pessimization of vector compares.

llvm-svn: 95855
2010-02-11 06:26:33 +00:00
Chris Lattner 1d4eb8fac4 convert to filecheck.
llvm-svn: 95854
2010-02-11 06:24:37 +00:00
Chris Lattner c053cbbc4d Make DSE only scan blocks that are reachable from the entry
block.  Other blocks may have pointer cycles that will crash
basicaa and other alias analyses.  In any case, there is no
point wasting cycles optimizing dead blocks.  This fixes 
rdar://7635088

llvm-svn: 95852
2010-02-11 05:11:54 +00:00
Chris Lattner f492ece81e a testcase that doesn't crash GVN but could someday.
llvm-svn: 95851
2010-02-11 05:08:05 +00:00