Commit Graph

136 Commits

Author SHA1 Message Date
Dan Gohman
32a81a3f6d When inserting code into a loop preheader, insert it before the
terminator, instead of after the last phi. This fixes a bug
exposed by ScalarEvolution analyzing more kinds of loops.
This fixes PR4436.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@74072 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-24 14:31:06 +00:00
Dan Gohman
51f53b7f5a Fix ScalarEvolution's backedge-taken count computations to check for
overflow when computing a integer division to round up.

Thanks to Nick Lewycky for noticing this!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73862 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-21 23:46:38 +00:00
Dan Gohman
763bad1a2b Generalize isLoopGuardedByCond's checking to consider two
SCEVUnknowns with identical Instructions to be equal. This allows
it to analze cases such as the attached testcase, where the front-end
has cloned the loop controlling expression. Along with r73805, this
lets IndVarSimplify eliminate all the sign-extend casts in the
loop in the attached testcase.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73807 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-20 00:35:32 +00:00
Dan Gohman
4e8a98519e Remove the code from IVUsers that attempted to handle
casted induction variables in cases where the cast
isn't foldable. It ended up being a pessimization in
many cases. This could be fixed, but it would require
a bunch of complicated code in IVUsers' clients. The
advantages of this approach aren't visible enough to
justify it at this time.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73706 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-18 16:54:06 +00:00
Dan Gohman
559254b697 Add -disable-output to a bunch of tests that don't care about the output.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73633 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-17 20:56:26 +00:00
Dan Gohman
61ffa8e373 Instcombine's ShrinkDemandedConstant may strip bits out of constants,
obscuring what would otherwise be a low-bits mask.  Use ComputeMaskedBits
to compute what ShrinkDemandedConstant knew about to reconstruct a
low-bits mask value.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73540 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-16 19:52:01 +00:00
Dan Gohman
bd59d7b603 Implement more aggressive folding of add operand lists when
they contain multiplications of constants with add operations.
This helps simplify several kinds of things; in particular it
helps simplify expressions like ((-1 * (%a + %b)) + %a) to %b,
as expressions like this often come up in loop trip count
computations.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73361 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-14 22:58:51 +00:00
Dan Gohman
4d8414f420 Teach SCEVExpander's visitAddRecExpr to reuse an existing canonical
induction variable when the addrec to be expanded does not require
a wider type. This eliminates the need for IndVarSimplify to
micro-manage SCEV expansions, because SCEVExpander now
automatically expands them in the form that IndVarSimplify considers
to be canonical. (LSR still micro-manages its SCEV expansions,
because it's optimizing for the target, rather than for
other optimizations.)

Also, this uses the new getAnyExtendExpr, which has more clever
expression simplification logic than the IndVarSimplify code it
replaces, and this cleans up some ugly expansions in code such as
the included masked-iv.ll testcase.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@73294 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-13 16:25:49 +00:00
Dan Gohman
ae3a0be92e Split the Add, Sub, and Mul instruction opcodes into separate
integer and floating-point opcodes, introducing
FAdd, FSub, and FMul.

For now, the AsmParser, BitcodeReader, and IRBuilder all preserve
backwards compatability, and the Core LLVM APIs preserve backwards
compatibility for IR producers. Most front-ends won't need to change
immediately.

This implements the first step of the plan outlined here:
http://nondot.org/sabre/LLVMNotes/IntegerOverflow.txt


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72897 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-04 22:49:04 +00:00
Dan Gohman
68c9344056 Don't attempt to simplify an non-affine IV expression if it can't
be simplified to a loop-invariant value. This fixes PR4315.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72798 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-03 19:11:31 +00:00
Dan Gohman
4a4f767235 Teach SCEVExpander to avoid creating over-indexed GEP indices when
possible. For example, it now emits

  %p.2.ip.1 = getelementptr [3 x [3 x double]]* %p, i64 2, i64 %tmp, i64 1

instead of the equivalent but less obvious

  %p.2.ip.1 = getelementptr [3 x [3 x double]]* %p, i64 0, i64 %tmp, i64 19


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72452 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-27 02:00:53 +00:00
Dan Gohman
f876ad0a70 In cases where a pointer value is an operand of a multiplication or
division operation, don't attempt to use the operation's value as
the base of a getelementptr. This fixes PR4271.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72422 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-26 17:41:16 +00:00
Dan Gohman
95bdbfa066 When rewriting the loop exit test with the canonical induction variable,
leave the original comparison in place if it has other uses, since the
other uses won't be dominated by the new comparison instruction.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72369 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-24 19:11:38 +00:00
Dan Gohman
453aa4fbf1 Generalize SCEVExpander::visitAddRecExpr's GEP persuit, and avoid
sending SCEVUnknowns to expandAddToGEP. This avoids the need for
expandAddToGEP to bend the rules and peek into SCEVUnknown
expressions.

Factor out the code for testing whether a SCEV can be factored by
a constant for use in a GEP index. This allows it to handle
SCEVAddRecExprs, by recursing.

As a result, SCEVExpander can now put more things in GEP indices,
so it emits fewer explicit mul instructions.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72366 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-24 18:06:31 +00:00
Torok Edwin
b679de2a21 The rewriter may hold references to instructions that are deleted because they are trivially dead.
Fix by clearing the rewriter cache before deleting the trivially dead
instructions.
Also make InsertedExpressions use an AssertingVH to catch these
bugs easier.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72364 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-24 14:23:16 +00:00
Dan Gohman
d6d0294e35 Teach IndVarSimplify's FixUsesBeforeDefs to handle InvokeInsts by
assuming that the use of the value is in a block dominated by the
"normal" destination. LangRef.html and other documentation sources
don't explicitly guarantee this, but it seems to be assumed in
other places in LLVM at least.

This fixes an assertion failure on the included testcase, which
is derived from the Ada testsuite.

FixUsesBeforeDefs is a temporary measure which I'm looking to
replace with a more capable solution.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72266 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-22 16:47:11 +00:00
Dan Gohman
0d56b06f86 Fix a thinko in the code that adapted SCEVMulExpr operands for
use in expanding SCEVAddExprs with GEPs. The operands of a
SCEVMulExpr need to be multiplied together, not added.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72250 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-22 07:14:20 +00:00
Dan Gohman
5be18e8476 Teach SCEVExpander to expand arithmetic involving pointers into GEP
instructions. It attempts to create high-level multi-operand GEPs,
though in cases where this isn't possible it falls back to casting
the pointer to i8* and emitting a GEP with that. Using GEP instructions
instead of ptrtoint+arithmetic+inttoptr helps pointer analyses that
don't use ScalarEvolution, such as BasicAliasAnalysis.

Also, make the AddrModeMatcher more aggressive in handling GEPs.
Previously it assumed that operand 0 of a GEP would require a register
in almost all cases. It now does extra checking and can do more
matching if operand 0 of the GEP is foldable. This fixes a problem
that was exposed by SCEVExpander using GEPs.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72093 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-19 02:15:55 +00:00
Dan Gohman
04623272d2 Add nounwind to a few tests.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72002 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-18 15:16:49 +00:00
Dan Gohman
81db61a2e6 Factor the code for collecting IV users out of LSR into an IVUsers class,
and generalize it so that it can be used by IndVarSimplify. Implement the
base IndVarSimplify transformation code using IVUsers. This removes
TestOrigIVForWrap and associated code, as ScalarEvolution now has enough
builtin overflow detection and folding logic to handle all the same cases,
and more. Run "opt -iv-users -analyze -disable-output" on your favorite
loop for an example of what IVUsers does.

This lets IndVarSimplify eliminate IV casts and compute trip counts in
more cases. Also, this happens to finally fix the remaining testcases
in PR1301.

Now that IndVarSimplify is being more aggressive, it occasionally runs
into the problem where ScalarEvolutionExpander's code for avoiding
duplicate expansions makes it difficult to ensure that all expanded
instructions dominate all the instructions that will use them. As a
temporary measure, IndVarSimplify now uses a FixUsesBeforeDefs function
to fix up instructions inserted by SCEVExpander. Fortunately, this code
is contained, and can be easily removed once a more comprehensive
solution is available.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71535 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-12 02:17:14 +00:00
Dan Gohman
efb9fbfdab When forgetting SCEVs for loop PHIs, don't forget SCEVUnknown values.
These values aren't analyzable, so they don't care if more information
about the loop trip count can be had. Also, SCEVUnknown is used for
a PHI while the PHI itself is being analyzed, so it needs to be left
in the Scalars map. This fixes a variety of subtle issues.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@71533 91177308-0d34-0410-b5e6-96231b3b80d8
2009-05-12 01:27:58 +00:00
Dan Gohman
cab9cf01ce Make this testcase slightly less trivial, so that it doesn't fail
if indvars happens to optimize away the unused primary induction
variable.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@70333 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-28 22:03:26 +00:00
Dale Johannesen
617d108a63 Fix PR 4086, a bug in FP IV elimination.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@70247 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-27 21:03:15 +00:00
Dan Gohman
d6c329532b Handle ands with ~0 correctly too. This fixes PR4052.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@70176 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-27 01:41:10 +00:00
Dan Gohman
2c73d5fb9e Handle ands with 0 and shifts by 0 correctly. These aren't
common, but indvars shouldn't crash on them. This fixes PR4054.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@70051 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-25 17:05:40 +00:00
Dan Gohman
e22d503ee5 Fix an error in this test.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69893 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-23 15:22:28 +00:00
Dan Gohman
752ec7da50 Change SCEVExpander's expandCodeFor to provide more flexibility
with the persistent insertion point, and change IndVars to make
use of it. This fixes a bug where IndVars was holding on to a
stale insertion point and forcing the SCEVExpander to continue to
use it.

This fixes PR4038.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69892 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-23 15:16:49 +00:00
Evan Cheng
9c15949967 Make sure both operands have binary instructions have the same type.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69844 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-22 23:39:28 +00:00
Evan Cheng
1abe64663e Avoid deferencing use_begin() if value does not have a use.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69836 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-22 22:45:37 +00:00
Dan Gohman
aabb04f527 SCEVExpander's InsertCastOfTo knows how to move existing cast
instructions in order to avoid inserting new ones. However, if
the cast instruction is the SCEVExpander's InsertPt, this
causes subsequently emitted instructions to be inserted near
the cast, and not at the location of the original insert point.
Fix this by adjusting the insert point in such cases.
This fixes PR4009.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69808 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-22 16:11:16 +00:00
Dale Johannesen
764eccf025 Another testcase for IV shortening.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69247 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-16 00:45:21 +00:00
Dale Johannesen
dd1f9e4bf6 Enhance induction variable code to remove the
sext around sext(shorter IV + constant), using a
longer IV instead, when it can figure out the
add can't overflow.  This comes up a lot in
subscripting; mainly affects 64 bit.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@69123 91177308-0d34-0410-b5e6-96231b3b80d8
2009-04-15 01:10:12 +00:00
Dan Gohman
15cab2817b Back out the change in 64918 that used sign-extensions when promoting
trip counts that use signed comparisons. It's not obviously the best
approach for preserving trip count information, and at any rate there
isn't anything in the tree right now that makes use of that, so for
now always using zero-extensions is preferable.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@65347 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-23 23:20:35 +00:00
Dan Gohman
f5a309e989 Use a sign-extend instead of a zero-extend when promoting a
trip count value when the original loop iteration condition is
signed and the canonical induction variable won't undergo signed
overflow. This isn't required for correctness; it just preserves
more information about original loop iteration values.

Add a getTruncateOrSignExtend method to ScalarEvolution,
following getTruncateOrZeroExtend.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@64918 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-18 17:22:41 +00:00
Dan Gohman
d2067fd730 Fix a corner case in the new indvars promotion logic: if there
are multiple IV's in a loop, some of them may under go signed
or unsigned wrapping even if the IV that's used in the loop
exit condition doesn't. Restrict sign-extension-elimination
and zero-extension-elimination to only those that operate on
the original loop-controlling IV.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@64866 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-18 00:52:00 +00:00
Dan Gohman
74786c0f43 Rename IndVarsSimplify to IndVarSimplify, to be consistent with
the name used in the code that these tests are for.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@64624 91177308-0d34-0410-b5e6-96231b3b80d8
2009-02-16 00:56:15 +00:00