369 Commits

Author SHA1 Message Date
Dan Gohman
26ef7c7ab7 Fix memdep's code for reasoning about dependences between two calls. A Ref
response from getModRefInfo is not useful here. Instead, check for identical
calls only in the NoModRef case.

Reapply r110270, and strengthen it to compensate for the memdep changes.
When both calls are readonly, there is no dependence between them.

llvm-svn: 110382
2010-08-05 22:09:15 +00:00
Dan Gohman
554b012f67 Revert r110270 for now. It appears to uncover a memdep bug.
llvm-svn: 110293
2010-08-05 00:43:10 +00:00
Dan Gohman
bd33dab633 The two-callsite form of AliasAnalysis::getModRefInfo is documented
to return Ref if the left callsite only reads memory read or written
by the right callsite; fix BasicAliasAnalysis to implement this.

Add AliasAnalysisEvaluator support for testing the two-callsite
form of getModRefInfo.

llvm-svn: 110270
2010-08-04 22:56:29 +00:00
Dan Gohman
5442c71f2e Thread const correctness through a bunch of AliasAnalysis interfaces and
eliminate several const_casts.

Make CallSite implicitly convertible to ImmutableCallSite.

Rename the getModRefBehavior for intrinsic IDs to
getIntrinsicModRefBehavior to avoid overload ambiguity with CallSite,
which happens to be implicitly convertible to bool.

llvm-svn: 110155
2010-08-03 21:48:53 +00:00
Dan Gohman
ad867b0aed The singular of "indices" is "index".
llvm-svn: 110135
2010-08-03 20:23:52 +00:00
Dan Gohman
852d6fc50c Delete an unused function.
llvm-svn: 110134
2010-08-03 20:20:56 +00:00
Dan Gohman
2a190081f6 Introduce a symbolic constant for ~0u for use with AliasAnalysis.
llvm-svn: 110091
2010-08-03 01:03:11 +00:00
Owen Anderson
ac4a1ede17 Add INSTANTIATE_AG_PASS, which combines RegisterPass<> with RegisterAnalysisGroup<> for pass registration.
llvm-svn: 109058
2010-07-21 23:07:00 +00:00
Owen Anderson
81781220d2 Speculatively revert r108813, in an attempt to get the self-host buildbots working again. I don't see why this patch
would cause them to fail the way they are, but none of the other intervening patches seem likely either.

llvm-svn: 108818
2010-07-20 08:26:15 +00:00
Owen Anderson
8dc129325f Reapply r108794, a fix for the failing test from last time.
llvm-svn: 108813
2010-07-20 06:52:42 +00:00
Daniel Dunbar
4a35d6f8cd Revert r108794, "Separate PassInfo into two classes: a constructor-free
superclass (StaticPassInfo) and a constructor-ful subclass (PassInfo).", it is
breaking teh everything.

llvm-svn: 108805
2010-07-20 03:06:07 +00:00
Owen Anderson
e7c5fe586a Separate PassInfo into two classes: a constructor-free superclass (StaticPassInfo) and a constructor-ful subclass (PassInfo).
llvm-svn: 108794
2010-07-20 01:19:58 +00:00
Dan Gohman
5b0a8a863f Minore code simplification.
llvm-svn: 107777
2010-07-07 14:30:04 +00:00
Dan Gohman
00ef93258a Remove interprocedural-basic-aa and associated code. The AliasAnalysis
interface needs implementations to be consistent, so any code which
wants to support different semantics must use a different interface.
It's not currently worthwhile to add a new interface for this new
concept.

Document that AliasAnalysis doesn't support cross-function queries.

llvm-svn: 107776
2010-07-07 14:27:09 +00:00
Dan Gohman
84f90a387d Remove context sensitivity concerns from interprocedural-basic-aa, and
make it more aggressive in cases where both pointers are known to live
in the same function.

llvm-svn: 107420
2010-07-01 20:08:40 +00:00
Dan Gohman
1be9e7c0b6 Fix whitespace style.
llvm-svn: 107175
2010-06-29 18:12:34 +00:00
Benjamin Kramer
80b7bc042a Use a more obvious way to avoid compiling functions which are only used when XDEBUG is enabled.
llvm-svn: 107125
2010-06-29 10:03:11 +00:00
Chandler Carruth
b1adb88d05 Jump through some silly hoops to make GCC accept that a function may not always
be called.

llvm-svn: 107124
2010-06-29 06:46:00 +00:00
Dan Gohman
0824affeff Add an Intraprocedural form of BasicAliasAnalysis, which aims to
properly handles instructions and arguments defined in different
functions, or across recursive function iterations.

llvm-svn: 107109
2010-06-29 00:50:39 +00:00
Dan Gohman
7c34ece501 Fix Value::stripPointerCasts and BasicAA to avoid trouble on
code in unreachable blocks, which have have use-def cycles.
This fixes PR7514.

llvm-svn: 107071
2010-06-28 21:16:52 +00:00
Gabor Greif
1abbde3103 use ArgOperand accessors
llvm-svn: 106697
2010-06-23 23:38:07 +00:00
Eric Christopher
7258dcd77f Revert 101465, it broke internal OpenGL testing.
Probably the best way to know that all getOperand() calls have been handled
is to replace that API instead of updating.

llvm-svn: 101579
2010-04-16 23:37:20 +00:00
Gabor Greif
f375520f7b reapply r101434
with a fix for self-hosting

rotate CallInst operands, i.e. move callee to the back
of the operand array

the motivation for this patch are laid out in my mail to llvm-commits:
more efficient access to operands and callee, faster callgraph-construction,
smaller compiler binary

llvm-svn: 101465
2010-04-16 15:33:14 +00:00
Gabor Greif
403e9694f9 back out r101423 and r101397, they break llvm-gcc self-host on darwin10
llvm-svn: 101434
2010-04-16 01:16:20 +00:00
Gabor Greif
33ae80bff7 reapply r101364, which has been backed out in r101368
with a fix

rotate CallInst operands, i.e. move callee to the back
of the operand array

the motivation for this patch are laid out in my mail to llvm-commits:
more efficient access to operands and callee, faster callgraph-construction,
smaller compiler binary

llvm-svn: 101397
2010-04-15 20:51:13 +00:00
Gabor Greif
9fd00c7d25 back out r101364, as it trips the linux nightlybot on some clang C++ tests
llvm-svn: 101368
2010-04-15 12:46:56 +00:00
Gabor Greif
aafd209632 rotate CallInst operands, i.e. move callee to the back
of the operand array

the motivation for this patch are laid out in my mail to llvm-commits:
more efficient access to operands and callee, faster callgraph-construction,
smaller compiler binary

llvm-svn: 101364
2010-04-15 10:49:53 +00:00
Dan Gohman
cb45bd9cb3 Pointers to zero-sized objects don't point to overlapping objects.
llvm-svn: 100789
2010-04-08 18:11:50 +00:00
Duncan Sands
19d0b47b1f There are two ways of checking for a given type, for example isa<PointerType>(T)
and T->isPointerTy().  Convert most instances of the first form to the second form.
Requested by Chris.

llvm-svn: 96344
2010-02-16 11:11:14 +00:00
Chris Lattner
af362f014d add some new methods to adjust this pointers. Not used yet.
llvm-svn: 94013
2010-01-20 19:26:14 +00:00
Chris Lattner
29f82a1465 move DecomposeGEPExpression out into ValueTracking.cpp
llvm-svn: 89956
2009-11-26 17:12:50 +00:00
Chris Lattner
631c5b2cb9 teach GetLinearExpression to be a bit more aggressive.
llvm-svn: 89955
2009-11-26 17:00:01 +00:00
Chris Lattner
5c1cfc252c resolve a fixme. I haven't figured out how to write a testcase
to exercise this though.

llvm-svn: 89954
2009-11-26 16:52:32 +00:00
Chris Lattner
29bc8a91d3 Teach basicaa that x|c == x+c when the c bits of x are clear. This
allows us to compile the example in readme.txt into:

LBB1_1:                                                     ## %bb
	movl	4(%rdx,%rax), %ecx
	movl	%ecx, %esi
	imull	(%rdx,%rax), %esi
	imull	%esi, %ecx
	movl	%esi, 8(%rdx,%rax)
	imull	%ecx, %esi
	movl	%ecx, 12(%rdx,%rax)
	movl	%esi, 16(%rdx,%rax)
	imull	%ecx, %esi
	movl	%esi, 20(%rdx,%rax)
	addq	$16, %rax
	cmpq	$4000, %rax
	jne	LBB1_1

instead of:

LBB1_1: 
	movl	(%rdx,%rax), %ecx
	imull	4(%rdx,%rax), %ecx
	movl	%ecx, 8(%rdx,%rax)
	imull	4(%rdx,%rax), %ecx
	movl	%ecx, 12(%rdx,%rax)
	imull	8(%rdx,%rax), %ecx
	movl	%ecx, 16(%rdx,%rax)
	imull	12(%rdx,%rax), %ecx
	movl	%ecx, 20(%rdx,%rax)
	addq	$16, %rax
	cmpq	$4000, %rax
	jne	LBB1_1

GCC (4.2) doesn't seem to be able to eliminate the loads in this 
testcase either, it generates:

L2:
	movl	(%rdx), %eax
	imull	4(%rdx), %eax
	movl	%eax, 8(%rdx)
	imull	4(%rdx), %eax
	movl	%eax, 12(%rdx)
	imull	8(%rdx), %eax
	movl	%eax, 16(%rdx)
	imull	12(%rdx), %eax
	movl	%eax, 20(%rdx)
	addl	$4, %ecx
	addq	$16, %rdx
	cmpl	$1002, %ecx
	jne	L2

llvm-svn: 89952
2009-11-26 16:26:43 +00:00
Chris Lattner
12dacdd359 teach basicaa that A[i] != A[i+1].
llvm-svn: 89951
2009-11-26 16:18:10 +00:00
Chris Lattner
7a5b56aca9 Change the other half of aliasGEP (which handles GEP differencing) to use DecomposeGEPExpression. This dramatically simplifies and shrinks the code by eliminating the horrible CheckGEPInstructions method, fixes a miscompilation (@test3) and makes the code more aggressive. In particular, we now handle the @test4 case, which is reduced from the SmallPtrSet constructor. Missing this caused us to emit a variable length memset instead of a fixed size one.
llvm-svn: 89922
2009-11-26 02:17:34 +00:00
Chris Lattner
5341c9679d Generalize DecomposeGEPExpression to exactly handle what Value::getUnderlyingObject does (when TD is around). This allows us to avoid calling DecomposeGEPExpression unless the ultimate alias check we care about passes, speedup up BasicAA a bit.
llvm-svn: 89920
2009-11-26 02:14:59 +00:00
Chris Lattner
8928899413 Implement a new DecomposeGEPExpression method, which decomposes a GEP into a list of scaled offsets. Use this to eliminate some previous ad-hoc code which was subtly broken (it assumed all Constant*'s were non-zero, but strange constant express could be zero).
llvm-svn: 89915
2009-11-26 02:13:03 +00:00
Chris Lattner
a99edbedd1 Use GEPOperator more pervasively to simplify code.
llvm-svn: 89914
2009-11-26 02:11:08 +00:00
Chris Lattner
214ef9a4ef fix comment, thanks all :)
llvm-svn: 89666
2009-11-23 17:07:35 +00:00
Chris Lattner
1e7b37ebba use the new isNoAlias method to simplify some code, only do an escaping check if
we have a non-constant pointer.  Constant pointers can't be local.

llvm-svn: 89665
2009-11-23 16:46:41 +00:00
Chris Lattner
98e25326a0 whitespace cleanup, tidying
llvm-svn: 89664
2009-11-23 16:45:27 +00:00
Chris Lattner
84ed59abcf speed up BasicAA a bit by implementing a long-standing TODO.
llvm-svn: 89663
2009-11-23 16:44:43 +00:00
Chris Lattner
a13c9d1a5e add fixme for dubious code. Duncan, what do you think?
llvm-svn: 89602
2009-11-22 16:16:48 +00:00
Chris Lattner
db1e9f1290 remove a silly condition that doesn't make a lot of sense anymore.
llvm-svn: 89601
2009-11-22 16:15:59 +00:00
Chris Lattner
d6a49ad25a reduce indentation, no functionality change.
llvm-svn: 89600
2009-11-22 16:05:05 +00:00
Chris Lattner
88874ec60a Remove the AliasAnalysis::getMustAliases method, which is dead.
The hasNoModRefInfoForCalls isn't worth it as a filter because 
basicaa provides m/r info and everything chains to it, so remove
it.

llvm-svn: 89599
2009-11-22 16:01:44 +00:00
Dan Gohman
94e617627d Extend CaptureTracking to indicate when a value is never stored, even
if it is not ultimately captured. Teach BasicAliasAnalysis that a 
local object address which does not escape and is never stored does
not alias with a value resulting from a load.

llvm-svn: 89398
2009-11-19 21:57:48 +00:00
Nick Lewycky
c53e2ecf02 Teach BasicAA that a constant expression can't alias memory provably not
allocated until runtime (such as an alloca). Patch by Hans Wennborg!

llvm-svn: 88760
2009-11-14 06:15:14 +00:00
Dan Gohman
ccb4584edd Default-addressspace null pointers don't alias anything. This allows
GVN to be more aggressive. Patch by Hans Wennborg! (with a comment added by me)

llvm-svn: 86582
2009-11-09 19:29:11 +00:00