Commit Graph

37352 Commits

Author SHA1 Message Date
Jakob Stoklund Olesen 710c6892be Fix typo. Thank you, valgrind.
llvm-svn: 99974
2010-03-31 00:40:08 +00:00
Chris Lattner c263b42d0d Fix a major source of compile-time slowness at -O0 -g by optimizing
the storage of !dbg metadata kinds in the instruction themselves.
The on-the-side hash table works great for metadata that not-all
instructions get, or for metadata that only exists when optimizing.
But when compile-time is everything, it isn't great.

I'm not super thrilled with the fact that this plops a TrackingVH in
Instruction, because it grows it by 3 words.  I'm investigating 
alternatives, but this should be a step in the right direction in any
case.

llvm-svn: 99957
2010-03-30 23:03:27 +00:00
Jakob Stoklund Olesen 6f6ebb663c Enable -sse-domain-fix by default. Now with tests!
llvm-svn: 99954
2010-03-30 22:47:00 +00:00
Jakob Stoklund Olesen 3493398f13 V_SETALLONES is an integer instruction.
Since it is just a pxor in disguise, we should probably expand it to a full
polymorphic triple.

llvm-svn: 99953
2010-03-30 22:46:55 +00:00
Jakob Stoklund Olesen dbff4e8103 Renumber SSE execution domains for better code size.
SSEDomainFix will collapse to the domain with the lower number when it has a
choice. The SSEPackedSingle domain often has smaller instructions, so prefer
that.

llvm-svn: 99952
2010-03-30 22:46:53 +00:00
Bob Wilson 6f7fd28824 Revert Mon Ping's change 99928, since it broke all the llvm-gcc buildbots.
llvm-svn: 99948
2010-03-30 22:27:04 +00:00
Devang Patel 57c644f926 Ignore invalid metadata.
llvm-svn: 99938
2010-03-30 22:09:52 +00:00
Jakob Stoklund Olesen cf35648ebe Revert "Enable -sse-domain-fix by default. What could possibly go wrong?"
Not running 'make check-all' before committing is a bad idea.

llvm-svn: 99933
2010-03-30 21:36:32 +00:00
Jakob Stoklund Olesen a654df84e6 Enable -sse-domain-fix by default. What could possibly go wrong?
llvm-svn: 99931
2010-03-30 21:09:31 +00:00
Mon P Wang 7460571381 Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset,
e.g., llvm.memcpy.i32(i8*, i8*, i32, i32) -> llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)
A update of langref will occur in a subsequent checkin.

llvm-svn: 99928
2010-03-30 20:55:56 +00:00
Chris Lattner a3b94ba2c5 move some method definitions to files that make sense.
llvm-svn: 99927
2010-03-30 20:48:48 +00:00
Benjamin Kramer 04c713dd45 Introduce SpecificBumpPtrAllocator, a wrapper for BumpPtrAllocator which allows
only a single type of object to be allocated. Use it to make VNInfo destruction
typesafe.

llvm-svn: 99919
2010-03-30 20:16:45 +00:00
Torok Edwin 921e80b1f5 Typo noticed by Duncan.
llvm-svn: 99918
2010-03-30 20:15:13 +00:00
Dan Gohman 39027c403c Fix a grammaro.
llvm-svn: 99917
2010-03-30 20:04:57 +00:00
Jakob Stoklund Olesen 3b9af40938 Add cross-block inference to SSEDomainFix.
llvm-svn: 99916
2010-03-30 20:04:01 +00:00
Gabor Greif b469818279 fix two cases where the arguments were extracted from the wrong range out of the InvokeInst
spotted by baldrick -- thanks\!

llvm-svn: 99914
2010-03-30 19:20:53 +00:00
Eric Christopher 6ad8167714 Remove the pmulld intrinsic and autoupdate it as a vector multiply.
Rewrite the pmulld patterns, and make sure that they fold in loads of
arguments into the instruction.

llvm-svn: 99910
2010-03-30 18:49:01 +00:00
Chris Lattner 9897043928 Rip out the 'is temporary' nonsense from the MCContext interface to
create symbols.  It is extremely error prone and a source of a lot
of the remaining integrated assembler bugs on x86-64.

This fixes rdar://7807601.

llvm-svn: 99902
2010-03-30 18:10:53 +00:00
Evan Cheng 85eea4e031 Funky indentation.
llvm-svn: 99901
2010-03-30 18:08:53 +00:00
Devang Patel 67d94ab0f8 There is no need to fall through after processing DBG_VALUE machine instruction.
llvm-svn: 99899
2010-03-30 18:07:00 +00:00
Daniel Dunbar 692d165c3b Fix -Asserts warning.
llvm-svn: 99895
2010-03-30 17:57:42 +00:00
Douglas Gregor 83fd0156a9 Introduce namespace-scope functions to enable LLVM statistics without
passing the command-line parameter "-stats" and to print the resulting
statistics without calling llvm_shutdown().

llvm-svn: 99893
2010-03-30 17:32:08 +00:00
Benjamin Kramer 0c1dcb083e XFAIL some PIC16 tests when running under valgrind-leaks. I don't expect these
to be fixed any time soon.

llvm-svn: 99888
2010-03-30 14:34:13 +00:00
Benjamin Kramer 7cc264bf19 PIC16: Plug a leak in PIC16Section by allocating name & address strings in the
MCContext. There is still one leak left in PIC16Section (the Items vector).

llvm-svn: 99887
2010-03-30 13:28:42 +00:00
Torok Edwin cc1db61180 Don't overwrite previous value, if it succeeded.
llvm-svn: 99886
2010-03-30 12:52:03 +00:00
Torok Edwin 61b798a5ac Honour addGlobalMapping() in the interpreter, if it was used to add mappings for
external Functions (the JIT does honour this).

llvm-svn: 99885
2010-03-30 12:31:58 +00:00
Torok Edwin 8061bb141e Reapply r99881 with some fixes: only call destructor in releaseMemory!
llvm-svn: 99883
2010-03-30 11:17:48 +00:00
Torok Edwin a2d1dc42e1 Revert 99881, it brooke smooshlab's llvm-gcc-i386-darwin9.
llvm-svn: 99882
2010-03-30 10:25:08 +00:00
Torok Edwin ba6d13c0ea Introduce another Reset() method in BumpPtrAllocator that calls a destructor
on all objects it has allocated, if they are all of the same size and alignment.
Use this to destruct all VNInfos allocated in LiveIntervalAnalysis (PR6653).

valnos is not reliable for this purpose, as seen in r99400
(which still leaked, and sometimes caused double frees).

llvm-svn: 99881
2010-03-30 10:08:26 +00:00
Evan Cheng cbcccce420 Avoid being influenced by the presence of dbg_value instructions.
llvm-svn: 99879
2010-03-30 05:49:07 +00:00
Chris Lattner 39e56a05f7 stringref'ize Timer apis
llvm-svn: 99877
2010-03-30 05:34:02 +00:00
Chris Lattner 90fe73dc92 finally, maintain a global list of timer groups, allowing us to
implement TimerGroup::printAll, which prints and resets all active
timers.

llvm-svn: 99876
2010-03-30 05:27:58 +00:00
Chris Lattner 4cd6571fef add a new TimerGroup::print method, and refactor away the bogus
TimerGroup copy ctor and assignment operator.

llvm-svn: 99875
2010-03-30 05:20:02 +00:00
Chris Lattner 6cbd8d118e rename GetLibSupportInfoOutputFile -> CreateInfoOutputFile and
have it always return a new stream to simplify clients.

llvm-svn: 99874
2010-03-30 05:01:08 +00:00
Chris Lattner dcd68b7a65 if a timergroup is destroyed before its timers, print times.
llvm-svn: 99873
2010-03-30 04:58:26 +00:00
Chris Lattner 9a608d248e change TimerGroup to keep a linked list of active timers
instead of just a count of them, and refactor the guts of
report printing out of removeTimer into its own method.
Refactor addTimerToPrint away.

llvm-svn: 99872
2010-03-30 04:40:01 +00:00
Chris Lattner 707431cf26 reapply my timer rewrite with a change for PassManager to store
timers by pointer instead of by-value.

llvm-svn: 99871
2010-03-30 04:03:22 +00:00
Chris Lattner ec8ef9b643 revert r99862 which is causing FNT failures.
llvm-svn: 99870
2010-03-30 03:57:00 +00:00
Chris Lattner 57a0542397 fairly major rewrite of various timing related stuff.
llvm-svn: 99862
2010-03-30 02:38:19 +00:00
Eric Christopher c1ddaaf5b1 Add FIXME for operand promotion.
llvm-svn: 99859
2010-03-30 01:04:59 +00:00
Jakob Stoklund Olesen 486aa2eadc Be gentle to MSVC. C++ is hard, after all.
llvm-svn: 99855
2010-03-30 00:09:32 +00:00
Daniel Dunbar c95156262d MC/Mach-O/x86_64: Support @GOTPCREL on symbols, even for non-PCrel relocations!
llvm-svn: 99853
2010-03-29 23:56:40 +00:00
Bill Wendling b2799edaf6 Re-add back in the slow way of determining of a clean-up should become a
catch-all. The "dominates" way won't catch all of the selectors which must be
changed.

llvm-svn: 99850
2010-03-29 23:37:07 +00:00
Jakob Stoklund Olesen b551aa4da5 Basic implementation of SSEDomainFix pass.
Cross-block inference is primitive and wrong, but the pass is working otherwise.

llvm-svn: 99848
2010-03-29 23:24:21 +00:00
Bill Wendling decbb74d68 A more general (and simpler!) implementation of r99671. It performs a similar
transform. I.e., if a clean-up eh.selector call dominates the invoke of an
_Unwind_Resume_or_Rethrow, then we convert the eh.selector into a
catch-all. This patch, however, uses the DominatorTree information, and doesn't
go through the whole rigmarole of starting at the eh.exception call, finding the
corresponding URoR and eh.selector calls, and trying to trace through any number
of instruction types to get to them.

llvm-svn: 99846
2010-03-29 23:02:46 +00:00
Devang Patel 23b2ae621c Encode start location of debug value, communicated through DBG_VALUE machine instruction, in a variable's DIE.
llvm-svn: 99845
2010-03-29 22:59:58 +00:00
Chris Lattner dcd7f927bd move a function into a more logical place in the file
llvm-svn: 99842
2010-03-29 21:34:06 +00:00
Chris Lattner 09a4074869 remove support for per-time peak memory tracking, this
isn't used by anyone and is better exposed as a non-per-timer 
thing.  Also, stop including System/Mutex.h in Timer.h

llvm-svn: 99841
2010-03-29 21:28:41 +00:00
Evan Cheng 742db6874a Fix PR4975. Avoid referencing empty vector.
llvm-svn: 99840
2010-03-29 21:27:30 +00:00
Chris Lattner b355db7877 various timer fixes: move operator= out of line,
eliminate the per-timer lock (timers should be 
externally locked if needed), the info-output-stream
can never be dbgs(), so drop the check.  Make some
stuff private.

llvm-svn: 99839
2010-03-29 21:24:52 +00:00