Commit Graph

54 Commits

Author SHA1 Message Date
Owen Anderson 6c18d1aac0 Get rid of static constructors for pass registration. Instead, every pass exposes an initializeMyPassFunction(), which
must be called in the pass's constructor.  This function uses static dependency declarations to recursively initialize
the pass's dependencies.

Clients that only create passes through the createFooPass() APIs will require no changes.  Clients that want to use the
CommandLine options for passes will need to manually call the appropriate initialization functions in PassInitialization.h
before parsing commandline arguments.

I have tested this with all standard configurations of clang and llvm-gcc on Darwin.  It is possible that there are problems
with the static dependencies that will only be visible with non-standard options.  If you encounter any crash in pass
registration/creation, please send the testcase to me directly.

llvm-svn: 116820
2010-10-19 17:21:58 +00:00
Owen Anderson 8ac477ffb5 Begin adding static dependence information to passes, which will allow us to
perform initialization without static constructors AND without explicit initialization
by the client.  For the moment, passes are required to initialize both their
(potential) dependencies and any passes they preserve.  I hope to be able to relax
the latter requirement in the future.

llvm-svn: 116334
2010-10-12 19:48:12 +00:00
Owen Anderson df7a4f2515 Now with fewer extraneous semicolons!
llvm-svn: 115996
2010-10-07 22:25:06 +00:00
Chris Lattner 19199cce55 stop forcing a noop AssemblyAnnotationWriter to silence #uses
comments, these don't happen anymore.

llvm-svn: 112901
2010-09-02 23:03:10 +00:00
Dan Gohman 110ed64fbb Revert 112442 and 112440 until the compile time problems introduced
by 112440 are resolved.

llvm-svn: 112692
2010-09-01 01:45:53 +00:00
Dan Gohman 3a08ed7904 Make IVUsers iterative instead of recursive.
This has the side effect of reversing the order of most of
IVUser's results.

llvm-svn: 112442
2010-08-29 16:40:03 +00:00
Dan Gohman ed2b005842 Tweak IVUsers' concept of "interesting" to exclude add recurrences
where the step value is an induction variable from an outer loop, to
avoid trouble trying to re-expand such expressions. This effectively
hides such expressions from indvars and lsr, which prevents them
from getting into trouble.

llvm-svn: 111317
2010-08-17 22:50:37 +00:00
Owen Anderson a7aed18624 Reapply r110396, with fixes to appease the Linux buildbot gods.
llvm-svn: 110460
2010-08-06 18:33:48 +00:00
Owen Anderson bda59bd247 Revert r110396 to fix buildbots.
llvm-svn: 110410
2010-08-06 00:23:35 +00:00
Owen Anderson 755aceb5d0 Don't use PassInfo* as a type identifier for passes. Instead, use the address of the static
ID member as the sole unique type identifier.  Clean up APIs related to this change.

llvm-svn: 110396
2010-08-05 23:42:04 +00:00
Owen Anderson a57b97e7e7 Fix batch of converting RegisterPass<> to INTIALIZE_PASS().
llvm-svn: 109045
2010-07-21 22:09:45 +00:00
Dan Gohman e637ff5e9a Remove the Expr member from IVUsers. Instead of remembering the expression,
just ask ScalarEvolution for it on demand. This helps IVUsers be more robust
in the case of expressions changing underneath it. This fixes PR6862.

llvm-svn: 101819
2010-04-19 21:48:58 +00:00
Dan Gohman 008a38b1d6 Add a cast to void to show that the return value is being
intentionally ignored.

llvm-svn: 100984
2010-04-11 19:30:19 +00:00
Dan Gohman 7841a6ecd2 Delete a dead check.
llvm-svn: 100983
2010-04-11 19:29:41 +00:00
Dan Gohman 2532856704 Delete dead code.
llvm-svn: 100981
2010-04-11 19:28:47 +00:00
Dan Gohman ee6451dca1 Fix a bug in IVUsers which was permitting non-affine addrecs to
be sent to LSR, which it isn't prepared to handle.

llvm-svn: 100839
2010-04-09 01:22:56 +00:00
Dan Gohman d006ab90dd 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.

llvm-svn: 100699
2010-04-07 22:27:08 +00:00
Dan Gohman 91ce8e9a5c Add a const qualifier.
llvm-svn: 100515
2010-04-06 01:31:12 +00:00
Dan Gohman 8b0a419eb1 Spelling fixes.
llvm-svn: 97453
2010-03-01 17:49:51 +00:00
Dan Gohman 8c16b38262 Remove unused variables and parameters.
llvm-svn: 96780
2010-02-22 04:11:59 +00:00
Dan Gohman f446713fd0 Simplify this code; no need for a custom subclass if it doesn't need
to override anything from the parent class.

llvm-svn: 96150
2010-02-14 02:48:58 +00:00
Dan Gohman 45774ce0ad 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.

llvm-svn: 95975
2010-02-12 10:34:29 +00:00
Dan Gohman c42c5243a1 Use an AssemblyAnnotatorWriter to clean up IVUsers' debug output.
The "uses=" comments are just clutter in this context.

llvm-svn: 95799
2010-02-10 20:42:37 +00:00
Bill Wendling c5829c4a50 Add "dump" method to IVUsersOneStride.
llvm-svn: 95022
2010-02-01 22:51:23 +00:00
Dan Gohman 51ad99d2c5 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.

llvm-svn: 94061
2010-01-21 02:09:26 +00:00
Dan Gohman d693472821 Add a new helper function to IVUsers for returning the "canonical"
form of an expression. This is the expression without the
post-increment adjustment made, which is useful in determining
which registers will be used by the expansion.

llvm-svn: 93921
2010-01-19 21:55:32 +00:00
Dan Gohman bc694918cc Use WriteAsOperand instead of getName() to print loop header names,
so that unnamed blocks are handled.

llvm-svn: 93059
2010-01-09 18:17:45 +00:00
David Greene 069857ea31 Convert debug messages to use dbgs(). Generally this means
s/errs/dbgs/g except for certain special cases.

llvm-svn: 92037
2009-12-23 20:20:46 +00:00
Dan Gohman 18fa5686f6 Add Loop contains utility methods for testing whether a loop
contains another loop, or an instruction. The loop form is
substantially more efficient on large loops than the typical
code it replaces.

llvm-svn: 91654
2009-12-18 01:24:09 +00:00
Dan Gohman 92c3696524 Reapply LoopStrengthReduce and IVUsers cleanups, excluding the part
of 91296 that caused trouble -- the Processed list needs to be
preserved for the livetime of the pass, as AddUsersIfInteresting
is called from other passes.

llvm-svn: 91641
2009-12-18 00:06:20 +00:00
Evan Cheng 090ac0865a Revert 91280-91283, 91286-91289, 91291, 91293, 91295-91296. It apparently introduced a non-deterministic behavior in the optimizer somewhere.
llvm-svn: 91598
2009-12-17 09:39:49 +00:00
Dan Gohman 2a07fd94f1 Clear the Processed set when it is no longer used, and clear the
IVUses list in releaseMemory().

llvm-svn: 91296
2009-12-14 17:35:17 +00:00
Dan Gohman fbeec7270c Fix a thinko; isNotAlreadyContainedIn had a built-in negative, so the
condition was inverted when the code was converted to contains().

llvm-svn: 91295
2009-12-14 17:31:01 +00:00
Dan Gohman 57eb6cda7a Drop Loop::isNotAlreadyContainedIn in favor of Loop::contains. The
former was just exposing a LoopInfoBase implementation detail.

llvm-svn: 91286
2009-12-14 17:06:50 +00:00
Jim Grosbach bf2956a88b enable iv-users simplification by default
llvm-svn: 89713
2009-11-23 23:25:54 +00:00
Jim Grosbach 50d67e791b Teach IVUsers to keep things simpler and track loop-invariant strides only
for uses inside the loop. This works better with LSR. Disabled behind
-simplify-iv-users while benchmarking.

llvm-svn: 89299
2009-11-19 02:05:44 +00:00
Evan Cheng 85a9f430e9 - Teach LSR to avoid changing cmp iv stride if it will create an immediate that
cannot be folded into target cmp instruction.
- Avoid a phase ordering issue where early cmp optimization would prevent the
  later count-to-zero optimization.
- Add missing checks which could cause LSR to reuse stride that does not have
  users.
- Fix a bug in count-to-zero optimization code which failed to find the pre-inc
  iv's phi node.
- Remove, tighten, loosen some incorrect checks disable valid transformations.
- Quite a bit of code clean up.

llvm-svn: 86969
2009-11-12 07:35:05 +00:00
Dan Gohman 2267f8aa72 Fix IVUsers to avoid assuming that the loop has a unique backedge.
llvm-svn: 86161
2009-11-05 19:41:37 +00:00
Dan Gohman 6daa565cf9 Instead of testing whether an instruction dominates the loop preheader,
test whether it properly dominates the loop header. This is equivalent
when the loop has a preheader, and has the advantage of working when
the loop doesn't have a preheader. Since IVUsers doesn't Require
LoopSimplify, the loop isn't guaranteed to have a preheader.

llvm-svn: 82899
2009-09-27 15:30:00 +00:00
Dan Gohman 0523507a2f Remove a redundant #include.
llvm-svn: 82897
2009-09-27 15:21:52 +00:00
Chris Lattner 13d62b60bd remove a few dead insertion methods.
llvm-svn: 79882
2009-08-24 02:39:26 +00:00
Chris Lattner 4dc3edde9f remove a few DOUTs here and there.
llvm-svn: 79832
2009-08-23 06:35:02 +00:00
Chris Lattner b25de3ff60 eliminate the "Value" printing methods that print to a std::ostream.
This required converting a bunch of stuff off DOUT and other cleanups.

llvm-svn: 79819
2009-08-23 04:37:46 +00:00
Dan Gohman 74cb168a9e Fix more missing newlines.
llvm-svn: 79244
2009-08-17 15:29:31 +00:00
Daniel Dunbar 0dd5e1ed39 More migration to raw_ostream, the water has dried up around the iostream hole.
- Some clients which used DOUT have moved to DEBUG. We are deprecating the
   "magic" DOUT behavior which avoided calling printing functions when the
   statement was disabled. In addition to being unnecessary magic, it had the
   downside of leaving code in -Asserts builds, and of hiding potentially
   unnecessary computations.

llvm-svn: 77019
2009-07-25 00:23:56 +00:00
Dan Gohman f2c9922f50 Add a newline, now that Value's operator<< doesn't print one.
llvm-svn: 75568
2009-07-14 00:32:49 +00:00
Dan Gohman cc85ae132c Make Loop and MachineLoop be subclasses of LoopBase, rather than typedefs,
using the Curiously Recurring Template Pattern with LoopBase.

This will help further refactoring, and future functionality for
Loop. Also, Headers can now foward-declare Loop, instead of pulling
in LoopInfo.h or doing tricks.

llvm-svn: 75519
2009-07-13 21:51:15 +00:00
Dan Gohman af75234955 Change all SCEV* to SCEV *.
llvm-svn: 74918
2009-07-07 17:06:11 +00:00
Owen Anderson 65b6056e37 SCEVHandle is no more!
llvm-svn: 73906
2009-06-22 21:39:50 +00:00
Dan Gohman 2199dbc5ff Don't (unconditionally) use getSCEVAtScope to simplify the step
expression in IVUsers, because in the case of a use of a non-linear
addrec outside of a loop, this causes the addrec to be evaluated as
a linear addrec.

llvm-svn: 73774
2009-06-19 17:33:15 +00:00