Commit Graph

136 Commits

Author SHA1 Message Date
Dan Gohman
6ebfd72f37 Fix SCEVExpander::visitAddRecExpr so that it remembers the induction variable
it inserted rather than using LoopInfo::getCanonicalInductionVariable to
rediscover it, since that doesn't work on non-canonical loops. This fixes
infinite recurrsion on such loops; PR7562.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@109419 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-26 18:28:14 +00:00
Dan Gohman
133e295b36 Simplify this code; LoopInfo::getCanonicalInductionVariable will only
find integer induction variables.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@108853 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-20 16:46:58 +00:00
Dan Gohman
7c58dbd88c Make getOrInsertCanonicalInductionVariable guarantee that its
result is a PHINode*.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@108852 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-20 16:44:52 +00:00
Dan Gohman
bb5d92741b Fix the order that SCEVExpander considers add operands in so that
it doesn't miss an opportunity to form a GEP, regardless of the
relative loop depths of the operands. This fixes rdar://8197217.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@108475 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-15 23:38:13 +00:00
Gabor Greif
19e5ada58a remove useless cast and fix typos in comment
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@107989 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-09 16:42:04 +00:00
Gabor Greif
f64f9cf7ec cache result of operator*
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@107988 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-09 16:39:02 +00:00
Gabor Greif
7656018c22 cache result of operator*
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@107976 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-09 15:40:10 +00:00
Dan Gohman
204b7cad7a Simplify.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@107248 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-30 06:09:46 +00:00
Dan Gohman
403a8cdda5 Use A.append(...) instead of A.insert(A.end(), ...) when A is a
SmallVector, and other SmallVector simplifications.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106452 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-21 19:47:52 +00:00
Dan Gohman
6f5fed253f Restore a call to rememberInstruction which was accidentally dropped
in refactoring.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106398 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-19 22:50:35 +00:00
Dan Gohman
485c43fc47 Factor out duplicated code for reusing and inserting casts into
a helper function.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106388 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-19 13:25:23 +00:00
Jim Grosbach
08f55d02da A few more places where SCEVExpander bits need to skip over debug intrinsics
when iterating through instructions. Yet more work for rdar://7797940

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106149 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-16 21:13:38 +00:00
Dan Gohman
25fcaff409 LSR needs to remember inserted instructions even in postinc mode, because
there could be multiple subexpressions within a single expansion which
require insert point adjustment. This fixes PR7306.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105510 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-05 00:33:07 +00:00
Dan Gohman
deff621abd Use getConstant instead of getIntegerSCEV. The two are basically the
same, now that getConstant has overloads consistent with ConstantInt::get.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102965 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-03 22:09:21 +00:00
Dan Gohman
948c8a3e3d When checking whether the special handling for an addrec increment which
doesn't dominate the header is needed, don't check whether the increment
expression has computable loop evolution. While the operands of an
addrec are required to be loop-invariant, they're not required to 
dominate any part of the loop. This fixes PR6914.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102389 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-26 21:46:36 +00:00
Dan Gohman
b35798347e Fix a bunch of namespace polution.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101376 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 17:08:50 +00:00
Dan Gohman
f8d0578e4c When emitting code for an add, don't force a SCEVUnknown wrapper around
a hoisted intermediate result if the intermediate result isn't an
Instruction.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100884 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-09 19:14:31 +00:00
Dan Gohman
53b73a283e Add a comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100874 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-09 18:20:03 +00:00
Dan Gohman
448db1cdef Generalize IVUsers to track arbitrary expressions rather than expressions
explicitly split into stride-and-offset pairs. Also, add the
ability to track multiple post-increment loops on the same expression.

This refines the concept of "normalizing" SCEV expressions used for
to post-increment uses, and introduces a dedicated utility routine for
normalizing and denormalizing expressions.

This fixes the expansion of expressions which are post-increment users
of more than one loop at a time. More broadly, this takes LSR another
step closer to being able to reason about more than one loop at a time.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100699 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-07 22:27:08 +00:00
Dan Gohman
e059ee832c Don't back past debug info intrinsics; SCEVExpander's strategy
for ignoring debug info intrinsics everywhere else is to advance
past them, and it needs to be consistent.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99332 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-23 21:53:22 +00:00
Dan Gohman
6c7ed6b549 Fix more places to more thoroughly ignore debug intrinsics. This fixes
use-before-def errors in SCEVExpander-produced code in sqlite3 when debug
info with optimization is enabled, though the testcases for this are
dependent on use-list order.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99001 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-19 21:51:03 +00:00
Dan Gohman
f9e64729af Reapply r98755 with a thinko which miscompiled gengtype fixed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98793 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-18 01:17:13 +00:00
Dan Gohman
ebf78f18df Revert 98755, which may be causing trouble.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98762 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-17 19:54:53 +00:00
Dan Gohman
0afc29c3e6 Change SCEVNAryExpr's operand array from a SmallVector to a plain
pointer and length, and allocate the arrays in ScalarEvolution's
BumpPtrAllocator, so that they get released when their owning
SCEV gets released. SCEVs are immutable, so they don't need to worry
about operand array resizing. This fixes a memory leak reported
in PR6637.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98755 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-17 18:51:01 +00:00
Dale Johannesen
768069ee7a Another place where debug info affected codegen.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98026 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-09 01:08:11 +00:00
Dale Johannesen
621e06f9df Fix another case where LSR was affected by debug info.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97865 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-06 02:45:26 +00:00
Dale Johannesen
8d50ea7603 Fix a case where LSR is sensitive to debug info.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97830 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-05 21:12:40 +00:00
Dan Gohman
087bd1e3a1 Make SCEVExpander and LSR more aggressive about hoisting expressions out
of loops.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97642 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-03 05:29:13 +00:00
Dan Gohman
ed78dbafd0 Revert r97580; that's not the right way to fix this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97639 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-03 04:36:42 +00:00
Dan Gohman
c4f7ec85ec When expanding an expression such as (A + B + C + D), sort the operands
by loop depth and emit loop-invariant subexpressions outside of loops.
This speeds up MultiSource/Applications/viterbi and others.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97580 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-02 19:32:21 +00:00
Dan Gohman
069d6f3396 Non-affine post-inc SCEV expansions have more code which must be
emitted after the increment. Make sure the insert position
reflects this. This fixes PR6453.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97537 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-02 01:59:21 +00:00
Dan Gohman
3f46a3abee Spelling fixes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@97453 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-01 17:49:51 +00:00
Dan Gohman
9feae9f0de Fix SCEVExpander's existing PHI reuse checking to recognize the
case where there are loop-invariant instructions somehow left
inside the loop, and in a position where they won't dominate
the IV increment position.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96448 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-17 02:39:31 +00:00
Duncan Sands
1df9859c40 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.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96344 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-16 11:11:14 +00:00
Dan Gohman
22e621908f When reusing an existing PHI node in a loop, be even more
strict about the requirements.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96301 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-16 00:20:08 +00:00
Duncan Sands
b0bc6c361d Uniformize the names of type predicates: rather than having isFloatTy and
isInteger, we now have isFloatTy and isIntegerTy.  Requested by Chris!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96223 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-15 16:12:20 +00:00
Dan Gohman
4559855013 When restoring a saved insert location, check to see if the saved
insert location has become an "inserted" instruction since the time
it was saved. If so, advance to the first non-"inserted" instruction.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96203 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-15 00:21:43 +00:00
Dan Gohman
1d826a76f5 In rememberInstruction, if the value being remembered is the
current insertion point, advance the current insertion point.
This avoids a use-before-def situation in a testcase extracted
from clang which is difficult to reduce to a reasonable-sized
regression test.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96151 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-14 03:12:47 +00:00
Dan Gohman
0a799ab158 Fix a case of mismatched types in an Add that turned up in 447.dealII.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96007 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-12 20:39:25 +00:00
Dan Gohman
572645cf84 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.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@95975 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-12 10:34:29 +00:00
Dan Gohman
4f8eea82d8 Generalize target-independent folding rules for sizeof to handle more
cases, and implement target-independent folding rules for alignof and
offsetof. Also, reassociate reassociative operators when it leads to
more folding.

Generalize ScalarEvolution's isOffsetOf to recognize offsetof on
arrays. Rename getAllocSizeExpr to getSizeOfExpr, and getFieldOffsetExpr
to getOffsetOfExpr, for consistency with analagous ConstantExpr routines.

Make the target-dependent folder promote GEP array indices to
pointer-sized integers, to make implicit casting explicit and exposed
to subsequent folding.

And add a bunch of testcases for this new functionality, and a bunch
of related existing functionality.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94987 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-01 18:27:38 +00:00
Dan Gohman
150dfa8711 Check Type::isSized before calling ScalarEvolution::getAllocSizeExpr,
rather than after.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94742 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-28 06:32:46 +00:00
Dan Gohman
0f5efe5625 Remove SCEVAllocSizeExpr and SCEVFieldOffsetExpr, and in their place
use plain SCEVUnknowns with ConstantExpr::getSizeOf and
ConstantExpr::getOffsetOf constants. This eliminates a bunch of
special-case code.

Also add code for pattern-matching these expressions, for clients that
want to recognize them.

Move ScalarEvolution's logic for expanding array and vector sizeof
expressions into an element count times the element size, to expose
the multiplication to subsequent folding, into the regular constant
folder.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94737 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-28 02:15:55 +00:00
Dan Gohman
c37e3d5d7c When re-using an existing cast for a user, it's still necessary to call
rememberInstruction so that future users of that user will be inserted
in the correct position. This fixes the Darwin selfhost.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94070 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-21 10:08:42 +00:00
Dan Gohman
a10756ee65 Re-implement the main strength-reduction portion of LoopStrengthReduction.
This new version is much more aggressive about doing "full" reduction in
cases where it reduces register pressure, and also more aggressive about
rewriting induction variables to count down (or up) to zero when doing so
reduces register pressure.

It currently uses fairly simplistic algorithms for finding reuse
opportunities, but it introduces a new framework allows it to combine
multiple strategies at once to form hybrid solutions, instead of doing
all full-reduction or all base+index.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94061 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-21 02:09:26 +00:00
Dan Gohman
3abf905b50 Fix a typo and an 80-column violation in comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@93931 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-19 22:26:02 +00:00
Bob Wilson
eb35699e64 Fix a comment typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90511 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-04 01:33:04 +00:00
Chris Lattner
7896c9f436 improve portability to avoid conflicting with std::next in c++'0x.
Patch by Howard Hinnant!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90365 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-03 00:50:42 +00:00
Duncan Sands
ac53a0b272 Introduce and use convenience methods for getting pointer types
where the element is of a basic builtin type.  For example, to get
an i8* use getInt8PtrTy.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83379 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-06 15:40:36 +00:00
Dan Gohman
f3f1be6f06 When extending the operands of an addrec, iterate through all
the operands, rather than trying to partition them into a start
and a step. This handles non-affine add recurrences correctly.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83011 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-28 21:01:47 +00:00