OSchip
/
llvm-project
mirror of https://github.com/THU-DSP-LAB/llvm-project.git
13
0
Fork 6
Code Issues Packages Projects Releases Wiki Activity
39,636 Commits 31 Branches 0 Tags 2.2 GiB
Commit Graph

4441 Commits

Author SHA1 Message Date
Evan Cheng a509537e25 The physical register + virtual register joining requirement was much too strict. 
llvm-svn: 45253
 2007-12-20 02:23:25 +00:00
Evan Cheng 61bc51ee97 Bring back a burr scheduling heuristic that's still needed. 
llvm-svn: 45252
 2007-12-20 02:22:36 +00:00
Bill Wendling 65c001e6bc Updated comments to reflect what "side effects" means in this situation. 
llvm-svn: 45245
 2007-12-20 01:08:10 +00:00
Duncan Sands e9d8861cdf Simplify LowerCallTo by using a callsite. 
llvm-svn: 45198
 2007-12-19 09:48:52 +00:00
Duncan Sands 030bce7b83 The C++ exception handling personality function wants 
to know about calls that cannot throw ('nounwind'):
if such a call does throw for some reason then the
personality will terminate the program.  The distinction
between an ordinary call and a nounwind call is that
an ordinary call gets an entry in the exception table
but a nounwind call does not.  This patch sets up the
exception table appropriately.  One oddity is that
I've chosen to bracket nounwind calls with labels (like
invokes) - the other choice would have been to bracket
ordinary calls with labels.  While bracketing
ordinary calls is more natural (because bracketing
by labels would then correspond exactly to getting an
entry in the exception table), I didn't do it because
introducing labels impedes some optimizations and I'm
guessing that ordinary calls occur more often than
nounwind calls.  This fixes the gcc filter2 eh test,
at least at -O0 (the inliner needs some tweaking at
higher optimization levels).

llvm-svn: 45197
 2007-12-19 07:36:31 +00:00
Evan Cheng 9f06e5e2df Don't leave newly created nodes around if it turns out they are not needed. 
llvm-svn: 45186
 2007-12-19 01:34:38 +00:00
Bill Wendling 166f746246 Add debugging info. Use the newly created "hasUnmodelledSideEffects" method. 
llvm-svn: 45178
 2007-12-18 21:38:04 +00:00
Anton Korobeynikov 95cc3e0e66 Support more insane CEP's in AsmPrinter (Yes, PyPy folks do really use them). 
llvm-svn: 45172
 2007-12-18 20:53:41 +00:00
Evan Cheng 483a969ece Fix PR1872: SrcValue and SrcValueOffset should not be used to compute load / store node id. 
llvm-svn: 45167
 2007-12-18 19:38:14 +00:00
Evan Cheng 78ced47a2f Also print alignment and volatileness. 
llvm-svn: 45164
 2007-12-18 19:06:30 +00:00
Evan Cheng 91e0fc9cb4 FIX for PR1799: When a load is unfolded from an instruction, check if it is a new node. If not, do not create a new SUnit. 
llvm-svn: 45157
 2007-12-18 08:42:10 +00:00
Evan Cheng e2dbba5828 SelectionDAG::dump() should print SrcValue of LoadSDNode and StoreSDNode. 
llvm-svn: 45151
 2007-12-18 07:02:08 +00:00
Duncan Sands b5a79d0eaa Make invokes of inline asm legal. Teach codegen 
how to lower them (with no attempt made to be
efficient, since they should only occur for
unoptimized code).

llvm-svn: 45108
 2007-12-17 18:08:19 +00:00
Christopher Lamb edf0788758 Change the PointerType api for creating pointer types. The old functionality of PointerType::get() has become PointerType::getUnqual(), which returns a pointer in the generic address space. The new prototype of PointerType::get() requires both a type and an address space. 
llvm-svn: 45082
 2007-12-17 01:12:55 +00:00
Owen Anderson 7b8a741189 Break local interferences in StrongPHIElimination. One step closer... 
llvm-svn: 45070
 2007-12-16 05:44:27 +00:00
Owen Anderson ccb3981256 A few more comments. 
llvm-svn: 45069
 2007-12-16 04:07:23 +00:00
Dan Gohman 8a332b235d Add explicit keywords, and fix a minor typo that they uncovered. 
llvm-svn: 45034
 2007-12-14 15:41:34 +00:00
Evan Cheng 0fcf56f8f5 Bug fix. Must also match ResNo when matching an operand with a user. 
llvm-svn: 45028
 2007-12-14 08:25:15 +00:00
Owen Anderson 53b677e4e8 Add register pairs to the list to check for local interferences. 
llvm-svn: 44987
 2007-12-13 05:53:03 +00:00
Owen Anderson 1f93edd08a Remove ugly and horrible code. It's not necessary for correctness, and can be added back later if it causes code quality issues. 
llvm-svn: 44986
 2007-12-13 05:43:37 +00:00
Evan Cheng 6e68381e02 Implicit def instructions, e.g. X86::IMPLICIT_DEF_GR32, are always re-materializable and they should not be spilled. 
llvm-svn: 44960
 2007-12-12 23:12:09 +00:00
Dan Gohman 7a7742c2fe Allow vector integer constants to be created with 
SelectionDAG::getConstant, in the same way as vector floating-point
constants. This allows the legalize expansion code for @llvm.ctpop and
friends to be usable with vector types.

llvm-svn: 44954
 2007-12-12 22:21:26 +00:00
Owen Anderson 499e5bffcf Forgot to remove a register from the PHI-union after I'd determined that it 
interfered with other registers.  Seems like that might be a good thing to do. :-)

llvm-svn: 44902
 2007-12-12 01:25:08 +00:00
Evan Cheng 6766d2fa4f If deleting a reload instruction due to reuse (value is available in register R and reload is targeting R), make sure to invalidate the kill information of the last kill. 
llvm-svn: 44894
 2007-12-11 23:36:57 +00:00
Bill Wendling 38236ef6cb Need to grow the indexed map. Added debug statements. 
llvm-svn: 44892
 2007-12-11 23:27:51 +00:00
Bill Wendling 642e15a7cb Simplify slightly. 
llvm-svn: 44881
 2007-12-11 22:22:22 +00:00
Owen Anderson f24dd1c1eb More progress on StrongPHIElimination. Now we actually USE the DomForest! 
llvm-svn: 44877
 2007-12-11 20:12:11 +00:00
Bill Wendling b678ae7c38 Blark! How in the world did this work without this?! 
llvm-svn: 44874
 2007-12-11 19:40:06 +00:00
Bill Wendling 7717a8a37d - Update the virtual reg to machine instruction map when hoisting. 
- Fix subtle bug when creating initially creating this map.

llvm-svn: 44873
 2007-12-11 19:17:04 +00:00
Bill Wendling 5143d898c8 Checking for "zero operands" during the "CanHoistInst()" method isn't necessary 
because those with side effects will be caught by other checks in here.

Also, simplify the check for a BB in a sub loop.

llvm-svn: 44871
 2007-12-11 18:45:11 +00:00
Evan Cheng 303417d242 Switch over to MachineLoopInfo. 
llvm-svn: 44838
 2007-12-11 02:09:15 +00:00
Evan Cheng f54030231e Pretty print shuffle mask operand. 
llvm-svn: 44837
 2007-12-11 02:08:35 +00:00
Gordon Henriksen 7843c16f31 CollectorMetadata and Collector are rejiggered to get along with 
per-function collector model. Collector is now the factory for
CollectorMetadata, so the latter may be subclassed.

llvm-svn: 44827
 2007-12-11 00:30:17 +00:00
Owen Anderson ba61806ef1 A little more progress on StrongPHIElimination, now that I have a better sense of 
how the CodeGen machinery works.

llvm-svn: 44786
 2007-12-10 08:07:09 +00:00
Christopher Lamb d202e03fe5 Improve branch folding by recgonizing that explict successor relationships impact the value of fall-through choices. 
llvm-svn: 44785
 2007-12-10 07:24:06 +00:00
Chris Lattner 64443973c0 Duncan points out that the subtraction is unneeded since hte code 
knows the vector is not pow2

llvm-svn: 44740
 2007-12-09 17:56:34 +00:00
Chris Lattner 69d3298777 Add support for splitting the operand of a return instruction. 
llvm-svn: 44728
 2007-12-09 00:06:19 +00:00
Bill Wendling 3f19dfe794 Reverting 44702. It wasn't correct to rename them. 
llvm-svn: 44727
 2007-12-08 23:58:46 +00:00
Chris Lattner e48fc80446 add many new cases to SplitResult. SplitResult now handles all the cases that LegalizeDAG does. 
llvm-svn: 44726
 2007-12-08 23:58:27 +00:00
Chris Lattner de9046af54 Implement splitting support for store, allowing us to compile: 
%f8 = type <8 x float>

define void @test_f8(%f8* %P, %f8* %Q, %f8* %S) {
	%p = load %f8* %P		; <%f8> [#uses=1]
	%q = load %f8* %Q		; <%f8> [#uses=1]
	%R = add %f8 %p, %q		; <%f8> [#uses=1]
	store %f8 %R, %f8* %S
	ret void
}

into:

_test_f8:
	movaps	16(%rdi), %xmm0
	addps	16(%rsi), %xmm0
	movaps	(%rdi), %xmm1
	addps	(%rsi), %xmm1
	movaps	%xmm0, 16(%rdx)
	movaps	%xmm1, (%rdx)
	ret

llvm-svn: 44725
 2007-12-08 23:24:26 +00:00
Chris Lattner de87224cd9 implement vector splitting of load, undef, and binops. 
llvm-svn: 44724
 2007-12-08 23:08:49 +00:00
Chris Lattner 1ef437d4e1 implement some methods. 
llvm-svn: 44723
 2007-12-08 22:40:18 +00:00
Chris Lattner a5e7db115e add scaffolding for splitting of vectors. 
llvm-svn: 44722
 2007-12-08 22:37:41 +00:00
Chris Lattner 8c8eaf6b92 reorganize header to separate into functional blocks. 
llvm-svn: 44719
 2007-12-08 21:59:32 +00:00
Chris Lattner 4063bd6eae split scalarization out to its own file. 
llvm-svn: 44718
 2007-12-08 20:30:28 +00:00
Chris Lattner 5c7c46baaf Split expansion out into its own file. 
llvm-svn: 44717
 2007-12-08 20:27:32 +00:00
Chris Lattner 029c816460 Split promotion support out to its own file. 
llvm-svn: 44716
 2007-12-08 20:24:38 +00:00
Chris Lattner 757d4beba9 Rename LegalizeDAGTypes.cpp -> LegalizeTypes.cpp 
llvm-svn: 44715
 2007-12-08 20:17:13 +00:00
Chris Lattner 92288147b6 Split the class definition of DAGTypeLegalizer out into a header. 
Leave it visibility hidden, but not in an anon namespace.

llvm-svn: 44714
 2007-12-08 20:16:06 +00:00
Bill Wendling 2b07d8c5a0 Renaming: 
isTriviallyReMaterializable -> hasNoSideEffects
  isReallyTriviallyReMaterializable -> isTriviallyReMaterializable

llvm-svn: 44702
 2007-12-08 07:17:56 +00:00
Bill Wendling 4375173ba0 Incorporated comments from Evan and Chris: 
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071203/056043.html
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20071203/056048.html

llvm-svn: 44696
 2007-12-08 01:47:01 +00:00
Bill Wendling fb706bc52b Initial commit of the machine code LICM pass. It successfully hoists this: 
_foo:
        li r2, 0
LBB1_1: ; bb
        li r5, 0
        stw r5, 0(r3)
        addi r2, r2, 1
        addi r3, r3, 4
        cmplw cr0, r2, r4
        bne cr0, LBB1_1 ; bb
LBB1_2: ; return
        blr 

to:

_foo:
        li r2, 0
        li r5, 0
LBB1_1: ; bb
        stw r5, 0(r3)
        addi r2, r2, 1
        addi r3, r3, 4
        cmplw cr0, r2, r4
        bne cr0, LBB1_1 ; bb
LBB1_2: ; return
        blr

ZOMG!! :-)

Moar to come...

llvm-svn: 44687
 2007-12-07 21:42:31 +00:00
Evan Cheng 85cdba29b0 Add an option to control this heuristic tweak so I can test it. 
llvm-svn: 44671
 2007-12-07 00:28:32 +00:00
Dale Johannesen 5eff4de9c8 Redo previous patch so optimization only done for i1. 
Simpler and safer.

llvm-svn: 44663
 2007-12-06 17:53:31 +00:00
Evan Cheng 8393dc7378 Turning simple splitting on. Start testing new coalescer heuristics as new llcbeta. 
llvm-svn: 44660
 2007-12-06 08:54:31 +00:00
Chris Lattner eedaf92fcf third time around: instead of disabling this completely, 
only disable it if we don't know it will be obviously profitable.
Still fixme, but less so. :)

llvm-svn: 44658
 2007-12-06 07:47:55 +00:00
Chris Lattner b5fdfb9612 Actually, disable this code for now. More analysis and improvements to 
the X86 backend are needed before this should be enabled by default.

llvm-svn: 44657
 2007-12-06 07:44:31 +00:00
Chris Lattner 7c709a5d08 implement a readme entry, compiling the code into: 
_foo:
	movl	$12, %eax
	andl	4(%esp), %eax
	movl	_array(%eax), %eax
	ret

instead of:

_foo:
	movl	4(%esp), %eax
	shrl	$2, %eax
	andl	$3, %eax
	movl	_array(,%eax,4), %eax
	ret

As it turns out, this triggers all the time, in a wide variety of
situations, for example, I see diffs like this in various programs:

-       movl    8(%eax), %eax
-       shll    $2, %eax
-       andl    $1020, %eax
-       movl    (%esi,%eax), %eax
+       movzbl  8(%eax), %eax
+       movl    (%esi,%eax,4), %eax


-       shll    $2, %edx
-       andl    $1020, %edx
-       movl    (%edi,%edx), %edx
+       andl    $255, %edx
+       movl    (%edi,%edx,4), %edx

Unfortunately, I also see stuff like this, which can be fixed in the
X86 backend:

-       andl    $85, %ebx
-       addl    _bit_count(,%ebx,4), %ebp
+       shll    $2, %ebx
+       andl    $340, %ebx
+       addl    _bit_count(%ebx), %ebp

llvm-svn: 44656
 2007-12-06 07:33:36 +00:00
Chris Lattner 42558bf664 implement the rest of the functionality from SelectionDAGLegalize::ScalarizeVectorOp 
llvm-svn: 44654
 2007-12-06 05:53:43 +00:00
Dale Johannesen 05bbbda78a Fix PR1842. 
llvm-svn: 44649
 2007-12-06 01:43:46 +00:00
Evan Cheng 7fc1d98353 Fix for PR1831: if all defs of an interval are re-materializable, then it's a preferred spill candiate. 
llvm-svn: 44644
 2007-12-06 00:01:56 +00:00
Evan Cheng 678b86d6ce MachineInstr can change. Store indexes instead. 
llvm-svn: 44612
 2007-12-05 10:24:35 +00:00
Evan Cheng 06353b48b5 If a split live interval is spilled again, remove the kill marker on its last use. 
llvm-svn: 44611
 2007-12-05 09:51:10 +00:00
Evan Cheng 64b3baaaea Clobber more bugs. 
llvm-svn: 44610
 2007-12-05 09:05:34 +00:00
Evan Cheng d7de56ac93 Fix kill info for split intervals. 
llvm-svn: 44609
 2007-12-05 08:16:32 +00:00
Chris Lattner c9693c60a5 more scalarization 
llvm-svn: 44608
 2007-12-05 07:45:02 +00:00
Chris Lattner 1a0d49a63c scalarize vector binops 
llvm-svn: 44607
 2007-12-05 07:36:58 +00:00
Evan Cheng 269dbd31d0 - Mark last use of a split interval as kill instead of letting spiller track it. 
This allows an important optimization to be re-enabled.
- If all uses / defs of a split interval can be folded, give the interval a
  low spill weight so it would not be picked in case spilling is needed (avoid
  pushing other intervals in the same BB to be spilled).

llvm-svn: 44601
 2007-12-05 03:22:34 +00:00
Evan Cheng bb26301864 Add a argument to storeRegToStackSlot and storeRegToAddr to specify whether 
the stored register is killed.

llvm-svn: 44600
 2007-12-05 03:14:33 +00:00
Evan Cheng e412a4427b Remove a unsafe optimization. This fixes 401.bzip2. 
llvm-svn: 44587
 2007-12-04 23:57:55 +00:00
Evan Cheng cd8a89b3cd Spiller unfold optimization bug: do not clobber a reusable stack slot value unless it can be modified. 
llvm-svn: 44575
 2007-12-04 19:19:45 +00:00
Chris Lattner b892225fb9 Implement framework for scalarizing node results. This is sufficient 
to codegen this:

define float @test_extract_elt(<1 x float> * %P) {
	%p = load <1 x float>* %P
	%R = extractelement <1 x float> %p, i32 0
	ret float %R
}

llvm-svn: 44570
 2007-12-04 07:48:46 +00:00
Chris Lattner 681c9d6697 start providing framework for scalarizing vectors. 
llvm-svn: 44569
 2007-12-04 07:29:51 +00:00
Evan Cheng d1badb960e Discard split intervals made empty due to folding. 
llvm-svn: 44565
 2007-12-04 00:32:23 +00:00
Evan Cheng 40965448ff Bug fixes. 
llvm-svn: 44549
 2007-12-03 21:31:55 +00:00
Duncan Sands 38ef3a8ec7 Rather than having special rules like "intrinsics cannot 
throw exceptions", just mark intrinsics with the nounwind
attribute.  Likewise, mark intrinsics as readnone/readonly
and get rid of special aliasing logic (which didn't use
anything more than this anyway).

llvm-svn: 44544
 2007-12-03 20:06:50 +00:00
Evan Cheng 196faa9dc5 Typo 
llvm-svn: 44532
 2007-12-03 10:00:00 +00:00
Evan Cheng 85ef9834a6 Update kill info for uses of split intervals. 
llvm-svn: 44531
 2007-12-03 09:58:48 +00:00
Evan Cheng f45a1d623c Remove redundant foldMemoryOperand variants and other code clean up. 
llvm-svn: 44517
 2007-12-02 08:30:39 +00:00
Evan Cheng 388f6f51a0 Fix a bug where splitting cause some unnecessary spilling. 
llvm-svn: 44482
 2007-12-01 04:42:39 +00:00
Evan Cheng 69fda0a716 Allow some reloads to be folded in multi-use cases. Specifically testl r, r -> cmpl [mem], 0. 
llvm-svn: 44479
 2007-12-01 02:07:52 +00:00
Evan Cheng b10dc27b20 Do not fold reload into an instruction with multiple uses. It issues one extra load. 
llvm-svn: 44467
 2007-11-30 21:23:43 +00:00
Devang Patel cc45c338d1 Provide a way to update DescGlobals cache directly. 
llvm-svn: 44446
 2007-11-30 00:51:33 +00:00
Evan Cheng d35b5acae4 Do not lose rematerialization info when spilling already split live intervals. 
llvm-svn: 44443
 2007-11-29 23:02:50 +00:00
Evan Cheng 8494ee175c Fix a major performance issue with splitting. If there is a def (not def/use) 
in the middle of a split basic block, create a new live interval starting at
the def. This avoid artifically extending the live interval over a number of
cycles where it is dead. e.g.

bb1:
       = vr1204   (use / kill) <= new interval starts and ends here.
...
...
vr1204 =          (new def)   <= start a new interval here.
       = vr1204   (use)

llvm-svn: 44436
 2007-11-29 10:12:14 +00:00
Evan Cheng f85c063ec0 Replace the odd kill# hack with something less fragile. 
llvm-svn: 44434
 2007-11-29 09:49:23 +00:00
Evan Cheng be255b0650 Fixed various live interval splitting bugs / compile time issues. 
llvm-svn: 44428
 2007-11-29 01:06:25 +00:00
Evan Cheng 147f7799c5 Kill info update bug. 
llvm-svn: 44427
 2007-11-29 01:05:47 +00:00
Duncan Sands 5208d1ab4a Add some convenience methods for querying attributes, and 
use them.

llvm-svn: 44403
 2007-11-28 17:07:01 +00:00
Duncan Sands 45a0c3265f Add missing newlines at EOF. 
llvm-svn: 44399
 2007-11-28 10:13:38 +00:00
Evan Cheng c1648b6a0d Recover compile time regression. 
llvm-svn: 44386
 2007-11-28 01:28:46 +00:00
Owen Anderson 30767b15e9 Add MachineLoopInfo. This is not yet tested. 
llvm-svn: 44384
 2007-11-27 22:47:08 +00:00
Nate Begeman 6f026a654c Support returning non-power-of-2 vectors to unblock some work 
llvm-svn: 44371
 2007-11-27 19:28:48 +00:00
Duncan Sands ad0ea2d430 Fix PR1146: parameter attributes are longer part of 
the function type, instead they belong to functions
and function calls.  This is an updated and slightly
corrected version of Reid Spencer's original patch.
The only known problem is that auto-upgrading of
bitcode files doesn't seem to work properly (see
test/Bitcode/AutoUpgradeIntrinsics.ll).  Hopefully
a bitcode guru (who might that be? :) ) will fix it.

llvm-svn: 44359
 2007-11-27 13:23:08 +00:00
Chris Lattner 698b1cb28d err, no really. 
llvm-svn: 44352
 2007-11-27 06:14:32 +00:00
Chris Lattner 28caf2717a don't depend on ADL. 
llvm-svn: 44351
 2007-11-27 06:14:12 +00:00
Dan Gohman 9a69341725 Don't lower srem/urem X%C to X-X/C*C unless the division is actually 
optimized. This avoids creating illegal divisions when the combiner is
running after legalize; this fixes PR1815. Also, it produces better
code in the included testcase by avoiding the subtract and multiply
when the division isn't optimized.

llvm-svn: 44341
 2007-11-26 23:46:11 +00:00
Chris Lattner cab915f9cf Implement expand support for MERGE_VALUEs that only produces one result. 
llvm-svn: 44304
 2007-11-24 19:12:15 +00:00
Chris Lattner 6e3641897b Implement support for custom legalization in DAGTypeLegalizer::ExpandOperand. 
Improve a comment.
Unbreak Duncan's carefully written path compression where I didn't realize
what was happening!

llvm-svn: 44301
 2007-11-24 18:11:42 +00:00
Chris Lattner f81d5886c6 Several changes: 
1) Change the interface to TargetLowering::ExpandOperationResult to 
   take and return entire NODES that need a result expanded, not just
   the value.  This allows us to handle things like READCYCLECOUNTER,
   which returns two values.
2) Implement (extremely limited) support in LegalizeDAG::ExpandOp for MERGE_VALUES.
3) Reimplement custom lowering in LegalizeDAGTypes in terms of the new
   ExpandOperationResult.  This makes the result simpler and fully 
   general.
4) Implement (fully general) expand support for MERGE_VALUES in LegalizeDAGTypes.
5) Implement ExpandOperationResult support for ARM f64->i64 bitconvert and ARM
   i64 shifts, allowing them to work with LegalizeDAGTypes.
6) Implement ExpandOperationResult support for X86 READCYCLECOUNTER and FP_TO_SINT,
   allowing them to work with LegalizeDAGTypes.

LegalizeDAGTypes now passes several more X86 codegen tests when enabled and when
type legalization in LegalizeDAG is ifdef'd out.

llvm-svn: 44300
 2007-11-24 07:07:01 +00:00
Duncan Sands b87dde7e8e Fix a bug in which node A is replaced by node B, but later 
node A gets back into the DAG again because it was hiding in
one of the node maps: make sure that node replacement happens
in those maps too.

llvm-svn: 44263
 2007-11-21 16:43:19 +00:00
Dale Johannesen 763e110a9f Fix .eh table linkage issues on Darwin. Some EH support 
for Darwin PPC, but it's not fully working yet.

llvm-svn: 44258
 2007-11-20 23:24:42 +00:00
Chris Lattner 09c0393d5e ExpandUnalignedLoad doesn't handle vectors right at all apparently. 
Fix a couple of problems:
1. Don't assume the VT-1 is a VT that is half the size.
2. Treat vectors of FP in the vector path, not the FP path.

This has a couple of remaining problems before it will work with
the code in PR1811: the code below this change assumes that it can
use extload/shift/or to construct the result, which isn't right for
vectors.

This also doesn't handle vectors of 1 or vectors that aren't pow-2.

llvm-svn: 44243
 2007-11-19 21:38:03 +00:00
Chris Lattner 6fa95ec19d Implement vector expand support for shuffle_vector. This fixes PR1811. 
llvm-svn: 44242
 2007-11-19 21:16:54 +00:00
Chris Lattner 67d77945e7 Implement splitting of UNDEF nodes. This is the first step towards fixing PR1811 
llvm-svn: 44239
 2007-11-19 20:21:32 +00:00
Dan Gohman 36347a26f9 Add support in SplitVectorOp for remainder operators. 
llvm-svn: 44233
 2007-11-19 15:15:03 +00:00
Nate Begeman d4d45c268c Add support for vectors to int <-> float casts. 
llvm-svn: 44204
 2007-11-17 03:58:34 +00:00
Evan Cheng 8e22379303 Live interval splitting: 
When a live interval is being spilled, rather than creating short, non-spillable
intervals for every def / use, split the interval at BB boundaries. That is, for
every BB where the live interval is defined or used, create a new interval that
covers all the defs and uses in the BB.

This is designed to eliminate one common problem: multiple reloads of the same
value in a single basic block. Note, it does *not* decrease the number of spills
since no copies are inserted so the split intervals are *connected* through
spill and reloads (or rematerialization). The newly created intervals can be
spilled again, in that case, since it does not span multiple basic blocks, it's
spilled in the usual manner. However, it can reuse the same stack slot as the
previously split interval.

This is currently controlled by -split-intervals-at-bb.

llvm-svn: 44198
 2007-11-17 00:40:40 +00:00
Anton Korobeynikov 66b91e66ec Implement necessary bits for flt_rounds gcc builtin. 
Codegen bits and llvm-gcc support will follow.

llvm-svn: 44182
 2007-11-15 23:25:33 +00:00
Nate Begeman bd117f06ba Basic non-power-of-2 vector support 
llvm-svn: 44181
 2007-11-15 21:15:26 +00:00
Duncan Sands d4494352f8 This assertion was bogus. 
llvm-svn: 44167
 2007-11-15 09:54:37 +00:00
Evan Cheng 2c1a50455c Fix a thinko in post-allocation coalescer. 
llvm-svn: 44166
 2007-11-15 08:13:29 +00:00
Bill Wendling b3712f8146 Adding debug output during coalescing. 
llvm-svn: 44154
 2007-11-15 02:06:30 +00:00
Bill Wendling 8269925b1e Need to increment the iterator. 
llvm-svn: 44153
 2007-11-15 00:40:48 +00:00
Anton Korobeynikov 2c6387803e Fix PIC jump table codegen on x86-32/linux. In fact, such thing should be applied 
to all targets uses GOT-relative offsets for PIC (Alpha?)

llvm-svn: 44108
 2007-11-14 09:18:41 +00:00
Evan Cheng 7f02cfa599 Clean up sub-register implementation by moving subReg information back to 
MachineOperand auxInfo. Previous clunky implementation uses an external map
to track sub-register uses. That works because register allocator uses
a new virtual register for each spilled use. With interval splitting (coming
soon), we may have multiple uses of the same register some of which are
of using different sub-registers from others. It's too fragile to constantly
update the information.

llvm-svn: 44104
 2007-11-14 07:59:08 +00:00
Owen Anderson d8167ab332 Run computeDomForest() on the set of registers that need to be tested for 
interference.

llvm-svn: 44064
 2007-11-13 20:13:24 +00:00
Owen Anderson 569ef71e44 Preserve LiveVariables when doing critical edge splitting. 
llvm-svn: 44063
 2007-11-13 20:04:45 +00:00
Dale Johannesen 7a7085f6d3 Add parameter to getDwarfRegNum to permit targets 
to use different mappings for EH and debug info;
no functional change yet.
Fix warning in X86CodeEmitter.

llvm-svn: 44056
 2007-11-13 19:13:01 +00:00
Bill Wendling f359fed9f9 Unify CALLSEQ_{START,END}. They take 4 parameters: the chain, two stack 
adjustment fields, and an optional flag. If there is a "dynamic_stackalloc" in
the code, make sure that it's bracketed by CALLSEQ_START and CALLSEQ_END. If
not, then there is the potential for the stack to be changed while the stack's
being used by another instruction (like a call).

This can only result in tears...

llvm-svn: 44037
 2007-11-13 00:44:25 +00:00
Owen Anderson c520c4b325 Break critical edges coming into blocks with PHI nodes. 
llvm-svn: 44019
 2007-11-12 17:27:27 +00:00
Evan Cheng be51f28e2b Refactor some code. 
llvm-svn: 44010
 2007-11-12 06:35:08 +00:00
Owen Anderson a1cd45213d As Chris and Evan pointed out, BreakCriticalMachineEdges doesn't really need 
to be a pass of its own.  Instead, move it out into a helper method.

llvm-svn: 44002
 2007-11-12 01:05:09 +00:00
Hartmut Kaiser 67297144ab Fixed a strange construct. Please review. 
llvm-svn: 43960
 2007-11-09 19:59:00 +00:00
Duncan Sands e795efea5b Move MinAlign to MathExtras.h. 
llvm-svn: 43944
 2007-11-09 13:41:39 +00:00
Duncan Sands e7a9ac929f Fix some load/store logic that would be wrong for 
apints on big-endian machines if the bitwidth is
not a multiple of 8.  Introduce a new helper,
MVT::getStoreSizeInBits, and use it.

llvm-svn: 43934
 2007-11-09 08:57:19 +00:00
Duncan Sands bab9dc9433 Add terminating newline. 
llvm-svn: 43933
 2007-11-09 08:30:21 +00:00
Evan Cheng 797d56ff17 Much improved pic jumptable codegen: 
Then:
        call    "L1$pb"
"L1$pb":
        popl    %eax
		...
LBB1_1: # entry
        imull   $4, %ecx, %ecx
        leal    LJTI1_0-"L1$pb"(%eax), %edx
        addl    LJTI1_0-"L1$pb"(%ecx,%eax), %edx
        jmpl    *%edx

        .align  2
        .set L1_0_set_3,LBB1_3-LJTI1_0
        .set L1_0_set_2,LBB1_2-LJTI1_0
        .set L1_0_set_5,LBB1_5-LJTI1_0
        .set L1_0_set_4,LBB1_4-LJTI1_0
LJTI1_0:
        .long    L1_0_set_3
        .long    L1_0_set_2

Now:
        call    "L1$pb"
"L1$pb":
        popl    %eax
		...
LBB1_1: # entry
        addl    LJTI1_0-"L1$pb"(%eax,%ecx,4), %eax
        jmpl    *%eax

		.align  2
		.set L1_0_set_3,LBB1_3-"L1$pb"
		.set L1_0_set_2,LBB1_2-"L1$pb"
		.set L1_0_set_5,LBB1_5-"L1$pb"
		.set L1_0_set_4,LBB1_4-"L1$pb"
LJTI1_0:
        .long    L1_0_set_3
        .long    L1_0_set_2

llvm-svn: 43924
 2007-11-09 01:32:10 +00:00
Evan Cheng f14006f4d6 Didn't mean to check these in. 
llvm-svn: 43923
 2007-11-09 01:28:33 +00:00
Evan Cheng 1bf166312b Bug fix. Passive nodes are not in SUnitMap. 
llvm-svn: 43922
 2007-11-09 01:27:11 +00:00
Owen Anderson 65d2fcdd2a This preserves critical edge breaking. 
llvm-svn: 43911
 2007-11-08 22:23:57 +00:00
Owen Anderson 3bc8124a66 Make BreakCriticalMachineEdges available as a pass that can be depended on. 
llvm-svn: 43910
 2007-11-08 22:20:23 +00:00
Evan Cheng ece4c68b82 If both parts of smul_lohi, etc. are used, don't simplify. If only one part is used, try simplify it. 
llvm-svn: 43888
 2007-11-08 09:25:29 +00:00
Owen Anderson 0be8c1dafe Add the majority of machine-level critical edge breaking pass. Most of this was written by Fernando, cleanup and updating to TOT by me. 
This still needs a bit of work, particularly to handle jump tables properly.

llvm-svn: 43885
 2007-11-08 07:55:43 +00:00
Owen Anderson bfbc12973d Take another stab at getting isLiveIn() and isLiveOut() right. 
llvm-svn: 43869
 2007-11-08 01:32:45 +00:00
Owen Anderson 9d86ef12c8 Bring UsedBlocks back. StrongPHIElimination needs this information. 
llvm-svn: 43866
 2007-11-08 01:20:48 +00:00
Evan Cheng e742ee1dbe Simplify my (il)logic. 
llvm-svn: 43819
 2007-11-07 08:08:25 +00:00
Owen Anderson c6a5387d09 Add some more of StrongPHIElim. 
llvm-svn: 43805
 2007-11-07 05:17:15 +00:00
Dan Gohman ccfc028283 Remainder operations must be either integer or floating-point. 
llvm-svn: 43781
 2007-11-06 22:11:54 +00:00
Evan Cheng dd71a5c37b When the allocator rewrite a spill register with new virtual register, it replaces other operands of the same register. Watch out for situations where 
only some of the operands are sub-register uses.

llvm-svn: 43776
 2007-11-06 21:12:10 +00:00
Evan Cheng d5d59ad634 First step towards moving the coalescer to priority_queue based machinery. 
llvm-svn: 43764
 2007-11-06 08:52:21 +00:00
Evan Cheng 92d23e5204 Fix a bug where a def use operand isn't being detected as a sub-register use. 
llvm-svn: 43763
 2007-11-06 08:50:44 +00:00
Evan Cheng 2dbffa4e76 Add pseudo dependency to force two-address instruction to be scheduled after 
other uses. There was a overly restricted check that prevented some obvious
cases.

llvm-svn: 43762
 2007-11-06 08:44:59 +00:00
Owen Anderson d378cea030 Add a few comments. 
llvm-svn: 43755
 2007-11-06 05:26:02 +00:00
Owen Anderson eb964eb2c8 DomForest is a forest of registers, not instructions. 
llvm-svn: 43754
 2007-11-06 05:22:43 +00:00
Owen Anderson a9057f0b97 StrongPHIElimination requires LiveVariables. 
llvm-svn: 43751
 2007-11-06 04:49:43 +00:00
Dan Gohman 08143e397d Add support for vector remainder operations. 
llvm-svn: 43744
 2007-11-05 23:35:22 +00:00
Rafael Espindola fa0df55bdd Move the LowerMEMCPY and LowerMEMCPYCall to a common place. 
Thanks for the suggestions Bill :-)

llvm-svn: 43742
 2007-11-05 23:12:20 +00:00
Dale Johannesen 4646aa3e33 Make labels work in asm blocks; allow labels as 
parameters.  Rename ValueRefList to ParamList
in AsmParser, since its only use is for parameters.

llvm-svn: 43734
 2007-11-05 21:20:28 +00:00
Duncan Sands f7ae8bd090 Don't output ABI size padding twice. By using the store 
size for the field we get ABI padding automatically, so
no need to put it in again when we emit the field.

llvm-svn: 43720
 2007-11-05 18:03:02 +00:00