Commit Graph

4411 Commits

Author SHA1 Message Date
Owen Anderson
516eb1c139 Put a heuristic in place to prevent GVN from falling into bad cases with massively complicated CFGs.
This speeds up a particular testcase from 12+ hours to 5 seconds with little perceptible loss of quality.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55391 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-26 22:07:42 +00:00
Devang Patel
a0b3909d43 If IV is used in a int-to-float cast inside the loop then try to eliminate the cast operation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55374 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-26 17:57:54 +00:00
Chris Lattner
7eb01bfc16 improve encapsulation of the BBExecutable set.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55271 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-23 23:39:31 +00:00
Chris Lattner
cf712dee93 Switch an assortment of maps, sets and vectors to more efficient versions,
patch contributed by m-s!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55270 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-23 23:36:38 +00:00
Chris Lattner
944fac71e0 Switch the asmprinter (.ll) and all the stuff it requires over to
use raw_ostream instead of std::ostream.  Among other goodness,
this speeds up llvm-dis of kc++ with a release build from 0.85s
to 0.49s (88% faster).

Other interesting changes:
 1) This makes Value::print be non-virtual.
 2) AP[S]Int and ConstantRange can no longer print to ostream directly, 
    use raw_ostream instead.
 3) This fixes a bug in raw_os_ostream where it didn't flush itself 
    when destroyed.
 4) This adds a new SDNode::print method, instead of only allowing "dump".


A lot of APIs have both std::ostream and raw_ostream versions, it would
be useful to go through and systematically anihilate the std::ostream 
versions.

This passes dejagnu, but there may be minor fallout, plz let me know if
so and I'll fix it.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55263 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-23 22:23:09 +00:00
Chris Lattner
3793325cb3 Add a new trivial -inst-namer pass which makes it possible to diff the
before/after effects of a pass, crazy!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55230 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-23 06:07:02 +00:00
Chris Lattner
88e6dc8bf1 Fix PR2423 by checking all indices for out of range access, not only
indices that start with an array subscript.  x->field[10000] is just 
as bad as (*X)[14][10000].


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55226 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-23 05:21:06 +00:00
Chris Lattner
2dba6a1bdb consolidate DenseMapInfo implementations, and add one for std::pair.
Patch contributed by m-s.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55167 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-22 05:08:25 +00:00
Nick Lewycky
5d52c4501a Revert r54876 r54877 r54906 and r54907. Evan found that these caused a 20%
slowdown in bzip2.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55113 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-21 05:56:10 +00:00
Evan Cheng
ab5d5e360e Silence a compiler warning.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55087 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-20 23:36:48 +00:00
Mon P Wang
4f5ca2cf80 Fixed shuffle optimizations to handle non power of 2 vectors
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@55035 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-20 02:23:25 +00:00
Chris Lattner
7de3b5db26 don't use the result of WriteAsOperand
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54979 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-19 04:45:19 +00:00
Nick Lewycky
e86389dd2a Make this comment clearer. Instead of using an ambiguous ~ (not) on an icmp
predicate, swap the order of the operands.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54907 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 20:02:02 +00:00
Nick Lewycky
91a0f78265 Consider the case where xor by -1 and xor by 128 have been combined already to
produce an xor by 127.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54906 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 19:58:24 +00:00
Gordon Henriksen
5eca075b74 Rename some GC classes so that their roll will hopefully be clearer.
In particular, Collector was confusing to implementors. Several
thought that this compile-time class was the place to implement
their runtime GC heap. Of course, it doesn't even exist at runtime.
Specifically, the renames are:

  Collector               -> GCStrategy
  CollectorMetadata       -> GCFunctionInfo
  CollectorModuleMetadata -> GCModuleInfo
  CollectorRegistry       -> GCRegistry
  Function::getCollector  -> getGC (setGC, hasGC, clearGC)

Several accessors and nested types have also been renamed to be
consistent. These changes should be obvious.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54899 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 18:44:35 +00:00
Evan Cheng
ee2aa85f44 Revert 54821. It's miscompiling 252.eon and 447.dealII
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54878 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 08:07:31 +00:00
Nick Lewycky
cbe23e33ba I found a better place for this optz'n.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54877 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 07:54:14 +00:00
Nick Lewycky
1b344bc986 Xor'ing both sides of icmp by sign-bit is equivalent to swapping signedness of
the predicate.

Also, make this optz'n apply in more cases where it's safe to do so.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54876 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 07:34:14 +00:00
Chris Lattner
fad86b003a Rework the routines that convert AP[S]Int into a string. Now, instead of
returning an std::string by value, it fills in a SmallString/SmallVector
passed in.  This significantly reduces string thrashing in some cases.

More specifically, this:
 - Adds an operator<< and a print method for APInt that allows you to 
   directly send them to an ostream.
 - Reimplements APInt::toString to be much simpler and more efficient
   algorithmically in addition to not thrashing strings quite as much.

This speeds up llvm-dis on kc++ by 7%, and may also slightly speed up the
asmprinter.  This also fixes a bug I introduced into the asmwriter in a
previous patch w.r.t. alias printing.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54873 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-17 07:19:36 +00:00
Owen Anderson
3688f268cb Remove GCSE, ValueNumbering, and LoadValueNumbering. These have been deprecated for almost a year; it's finally time for them to go away.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54822 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-15 21:31:02 +00:00
Devang Patel
35115f92e4 Reapply 54786. Add overflow and number of mantissa bits checks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54821 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-15 21:21:34 +00:00
Evan Cheng
21066541ae Revert 54786. It's not checking for overflows, etc.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54813 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-15 08:12:11 +00:00
Chris Lattner
2806dffc73 use smallvector instead of vector for a couple worklists. This speeds up instcombine
by ~10% on some testcases.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54811 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-15 04:03:01 +00:00
Bill Wendling
7a7cf6b984 Temporarily revert r54792. It's causing an ICE during bootstrapping.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54804 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-14 23:05:24 +00:00
Devang Patel
1df183f554 Use DenseMap. Patch by Pratik Solanki.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54792 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-14 21:31:10 +00:00
Devang Patel
79ceb4463e If IV is used in a int-to-float cast inside the loop then try to eliminate the cast opeation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54786 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-14 20:58:31 +00:00
Dan Gohman
a8c763b307 Use empty() instead of begin() == end().
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54780 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-14 18:13:49 +00:00
Matthijs Kooijman
c1f1d46f00 Replace two for loops with while(!X->use_empty()) loops. This prevents
invalidating the iterator by deleting the current use. This fixes a segfault on
64 bit linux reported in PR2675.

Also remove an unneeded if.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54778 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-14 15:03:05 +00:00
Dan Gohman
a60832b018 Fix a bogus srem rule - a negative value srem'd by a power-of-2
can have a non-negative result; for example, -16%16 is 0. Also,
clarify the related comments. This fixes PR2670.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54767 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-13 23:12:35 +00:00
Dan Gohman
2edeb63ece Fix SCCP's handling of struct value loads and stores. SCCP doesn't
track individual leaf values in such cases, so it needs to treat
struct values as normal values in this case.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54760 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-13 21:22:48 +00:00
Devang Patel
c677de2713 Rename. s/FindIVForUser/FindIVUserForCond/g
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54754 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-13 20:31:11 +00:00
Devang Patel
d16aba22c9 Check sign to detect overflow before changing compare stride.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54710 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-13 02:05:14 +00:00
Bill Wendling
f9e67aca52 Remove tabs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54707 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-12 23:15:44 +00:00
Chris Lattner
b8cd4d3d49 Implement support for simplifying vector comparisons by 0.0 and 1.0 like we
do for scalars.  Patch contributed by Nicolas Capens

This also generalizes the previous xforms to work on long double, now that 
isExactlyValue works for long double.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54653 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-11 22:06:05 +00:00
Eric Christopher
7a61d701c0 Have IRBuilder take a template argument on whether or not to preserve
names. This can save a lot of allocations if you aren't going to be
looking at the output.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54546 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-08 19:39:37 +00:00
Matthijs Kooijman
7d942001fa Let SRETPromotion properly preserve the function name instead of (implicitly)
postfixing it with a number.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54468 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-07 16:01:23 +00:00
Matthijs Kooijman
a7bab2c81c Fix SRETPromotion, it was generating functions without returns statements since
r53941 (but this was not noticed due to the lack of a basic test for
SRETPromotion). 


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54467 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-07 15:58:09 +00:00
Matthijs Kooijman
81ec2486b3 Add some debug output to SRETPromotion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54464 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-07 15:14:04 +00:00
Dan Gohman
4ce9627e97 Fix a shufflevector instcombine that was emitting invalid masks indices
when it meant to be emitting undef indices.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54417 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-06 18:17:32 +00:00
Evan Cheng
f5e25f32c7 Fix PR2355: bug in ChangeCompareStride. When the loop termination compare is the only use of its iv stride, the stride can be eliminated by moving it to another stride. If the scale is negative, swap the predicate instead of using a inverse predicate.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54415 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-06 18:04:43 +00:00
Chris Lattner
46bbad217b optimize a common idiom generated by clang for bitfield access, PR2638.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54408 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-06 07:35:52 +00:00
Chris Lattner
5af5f463e1 Zap sitofp/fptoui pairs. In all cases when the sign difference
matters, the result is undefined anyway.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54396 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-06 05:13:06 +00:00
Nick Lewycky
b30591ec64 Reinstate this optimization, but without the miscompile. Thanks to Bill for
tracking down that this was breaking llvm-gcc bootstrap on Linux.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54394 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-06 04:54:03 +00:00
Dan Gohman
5a6c448a7e Pass the computed iteration count value to RewriteLoopExitValues
instead of having it call getIterationCount again.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54380 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-05 22:34:21 +00:00
Bill Wendling
021507be82 Revert r53282. This was causing a miscompile on Linux. Also, the transformation
looks bogus. Please see PR2629 for details on why this is breaking things.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54372 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-05 21:23:45 +00:00
Dan Gohman
8f613f30a7 Trim #includes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54350 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-05 15:32:23 +00:00
Duncan Sands
58921feaa9 Fix comment typos.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54266 91177308-0d34-0410-b5e6-96231b3b80d8
2008-08-01 12:23:49 +00:00
Nate Begeman
5bc1ea0736 Add vector shifts to the IR, patch by Eli Friedman.
CodeGen & Clang work coming next.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54161 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-29 15:49:41 +00:00
Matthijs Kooijman
75cf9cc527 Add -unroll-allow-partial command line option that enabled the loop unroller to
partially unroll a loop when fully unrolling would not fit under the threshold.

Patch by Mikael Lepistö.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54160 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-29 13:21:23 +00:00
Matthijs Kooijman
477f5a2f11 Restructure ArgumentPromotion a bit. Instead of just having a single boolean
that says "unconditional loads from this argument are safe", we now keep track
of the safety per set of indices from which loads happen. This prevents
ArgPromotion from promoting loads that aren't really valid. As an added effect,
this will now disregard the the type of the indices passed to a GEP, so
"load GEP %A, i32 1" and "load GEP %A, i64 1" will result in a single argument,
not two.

This fixes PR2598, for which a testcase has been added as well.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@54159 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-29 10:00:13 +00:00