Commit Graph

1463 Commits

Author SHA1 Message Date
Chris Lattner
2636c1be17 *FINALLY* Fix a really nasty nondeterministic bug that has been haunting us
since May 1st.  In this code, the pred iterator was being invalidated sometimes
causing the wrong entries to be added to PHI nodes.

The fix for this is to defererence and safe the *PI value before we hack on
branch instructions, which changes use/def chains, which SOMETIMES invalidates
the iterator.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14278 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-21 07:19:01 +00:00
Chris Lattner
30c02f7eff Comment out the isnan stuff until we get a proper autoconf test for it
breaking the build on sparc is not acceptable.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14277 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-21 06:17:21 +00:00
Chris Lattner
beabf45a69 Make order of argument addition deterministic. In particular, the layout
of ConstantInt objects in memory used to determine which order arguments
were added in in some cases.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14276 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-21 00:07:58 +00:00
Chris Lattner
c5c5e6afe5 Make use of BinaryOperator::create* methods to shrinkify code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14262 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-20 05:04:01 +00:00
Chris Lattner
cf5933a716 Fix the inliner to be deterministic, not letting its output depend on the
relative location of Function objects in memory.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14260 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-20 04:11:48 +00:00
Chris Lattner
218a8223e6 Add some DEBUG output to the simplifycfg routines
Fix another non-deterministic behavior, this one should actually speed up the
code though as it was doing silly things.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14258 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-20 01:13:18 +00:00
Chris Lattner
56b7ee20da Now that dominator tree children are built in determinstic order, this horrible code
can go away


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14254 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 20:23:35 +00:00
Chris Lattner
aee7940d53 This will hopefully fix a heisenbug that Vladimir Merzliakov is running
into valiantly trying to compile stuff on freebsd.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14251 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 19:01:26 +00:00
Chris Lattner
2b994c7206 Fix a nasty bug, noticed by Reid
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14249 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 18:15:50 +00:00
Chris Lattner
edb8433c91 Fix one source of nondeterminism in the -licm pass: the hoist pass
was processing blocks in whatever order they happened to end up in the
dominator tree data structure.  Force an ordering.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14248 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 08:56:43 +00:00
Chris Lattner
abc35bcad3 Change to use the StableBasicBlockNumbering class
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14247 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 08:42:40 +00:00
Chris Lattner
63168d2244 Do not let the numbering of PHI nodes placed in the function depend on
non-deterministic things like the ordering of blocks in the dominance
frontier of a BB.  Unfortunately, I don't know of a better way to solve
this problem than to explicitly sort the BB's in function-order before
processing them.  This is guaranteed to slow the pass down a bit, but
is absolutely necessary to get usable diffs between two different tools
executing the mem2reg or scalarrepl pass.

Before this, bazillions of spurious diff failures occurred all over the
place due to the different order of processing PHIs:

-       %tmp.111 = getelementptr %struct.Connector_struct* %upcon.0.0, uint 0, uint 0
+       %tmp.111 = getelementptr %struct.Connector_struct* %upcon.0.1, uint 0, uint 0

Now, the diffs match.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14244 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 07:40:14 +00:00
Chris Lattner
1654cff8e8 Do not sort by the address of LLVM ConstantInt* objects. This produces
nondeterministic results that depend on where these objects land in memory.
Instead, sort by the value of the constant, which is stable.

Before this patch, the -simplifycfg pass run from two different compilers
could cause different code to be generated, though it was semantically the
same:

@@ -12258,8 +12258,8 @@
        %s_addr.1 = phi sbyte* [ %s, %entry ], [ %inc.0, %no_exit ]             ; <sbyte*> [#uses=5]
        %tmp.1 = load sbyte* %s_addr.1          ; <sbyte> [#uses=1]
        switch sbyte %tmp.1, label %no_exit [
-                sbyte 0, label %loopexit
                 sbyte 46, label %loopexit
+                sbyte 0, label %loopexit
        ]

We need to stomp all of this stuff out.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14243 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 07:02:14 +00:00
Chris Lattner
8430a4545c Do not loop over uses as we delete them. This causes iterators to be
invalidated out from under us.  This bug goes back to revision 1.1: scary.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14242 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-19 02:02:22 +00:00
Chris Lattner
5b62aa743d Implement Transforms/InstCombine/and.ll:test17, a common case that
occurs due to unordered comparison macros in math.h


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14221 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-18 06:07:51 +00:00
Chris Lattner
e5ad50b2e8 Do not function resolve intrinsics. This prevents warnings and possible bad
things from happening due to

declare bool %llvm.isunordered(double, double)
declare bool %llvm.isunordered(float, float)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14219 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-18 05:50:48 +00:00
Brian Gaeke
afe16e1742 I love the smell of a freshly broken PowerPC build in the morning.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14206 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-17 22:27:04 +00:00
Chris Lattner
5253f05709 Fix compilation problem on freebsd. Problem noted by Vladimir Merzliakov in
PR371


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14203 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-17 21:20:52 +00:00
Chris Lattner
f70c22b019 Rename Type::PrimitiveID to TypeId and ::getPrimitiveID() to ::getTypeID()
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14201 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-17 18:19:28 +00:00
Chris Lattner
5dd04027c7 Rename Type::PrimitiveID to TypeId and ::getPrimitiveID() to ::getTypeID()
Delete two functions that are now methods on the Type class


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14200 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-17 18:16:02 +00:00
Brian Gaeke
802daf8b7d Fix typo in DEBUG printout.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14196 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-17 07:26:52 +00:00
Brian Gaeke
83a7049270 Um, did someone make a typo or something?
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14192 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-15 23:09:50 +00:00
Chris Lattner
c45ef6d395 Remove support for the isnan intrinsic
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14186 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-15 21:37:54 +00:00
Brian Gaeke
fda2cce3bf Quick hack to get this file compiling again on Mac OS X. The right thing to do
is write an autoconf macro that checks whether __isnan or isnan actually works
**using the C++ compiler after #include <cmath>**, instead of doing it the easy
way with AC_CHECK_FUNCS().


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14171 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-14 06:33:19 +00:00
Alkis Evlogimenos
df497317f1 Add constant folding capabilities to the isunordered intrinsic.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14168 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-13 01:23:56 +00:00
Chris Lattner
cf11035a6f Constant fold the isnan intrinsic
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14150 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-11 06:16:23 +00:00
Chris Lattner
3285a6f426 Fix a bug in my checkin from last night that caused miscompilations of
186.crafty, fhourstones and 132.ijpeg.

Bugpoint makes really nasty miscompilations embarassingly easy to find.  It
narrowed it down to the instcombiner and this testcase (from fhourstones):

bool %l7153_l4706_htstat_loopentry_2E_4_no_exit_2E_4(int* %i, [32 x int]* %works, int* %tmp.98.out) {
newFuncRoot:
        %tmp.96 = load int* %i          ; <int> [#uses=1]
        %tmp.97 = getelementptr [32 x int]* %works, long 0, int %tmp.96         ; <int*> [#uses=1]
        %tmp.98 = load int* %tmp.97             ; <int> [#uses=2]
        %tmp.99 = load int* %i          ; <int> [#uses=1]
        %tmp.100 = and int %tmp.99, 7           ; <int> [#uses=1]
        %tmp.101 = seteq int %tmp.100, 7                ; <bool> [#uses=2]
        %tmp.102 = cast bool %tmp.101 to int            ; <int> [#uses=0]
        br bool %tmp.101, label %codeRepl4.exitStub, label %codeRepl3.exitStub

codeRepl4.exitStub:             ; preds = %newFuncRoot
        store int %tmp.98, int* %tmp.98.out
        ret bool true

codeRepl3.exitStub:             ; preds = %newFuncRoot
        store int %tmp.98, int* %tmp.98.out
        ret bool false
}

... which only has one combination performed on it:

$ llvm-as < t.ll | opt -instcombine -debug | llvm-dis
IC: Old =       %tmp.101 = seteq int %tmp.100, 7                ; <bool> [#uses=1]
    New =       setne int %tmp.100, 0           ; <bool>:<badref> [#uses=0]
IC: MOD =       br bool %tmp.101, label %codeRepl3.exitStub, label %codeRepl4.exitStub
IC: MOD =       %tmp.97 = getelementptr [32 x int]* %works, uint 0, int %tmp.96         ; <int*> [#uses=1]

It doesn't get much better than this.  :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14109 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-10 02:33:20 +00:00
Chris Lattner
448c323cba More minor cleanups
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14108 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-10 02:12:35 +00:00
Chris Lattner
48595f18cf Eliminate many occurrances of Instruction::
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14107 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-10 02:07:29 +00:00
Chris Lattner
457dd829ad Implement InstCombine/select.ll:test15*
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14095 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-09 07:59:58 +00:00
Chris Lattner
f523d06816 Be more careful about the order we put stuff onto the worklist. This allow us to
collapse this:
bool %le(int %A, int %B) {
        %c1 = setgt int %A, %B
        %tmp = select bool %c1, int 1, int 0
        %c2 = setlt int %A, %B
        %result = select bool %c2, int -1, int %tmp
        %c3 = setle int %result, 0
        ret bool %c3
}

into:

bool %le(int %A, int %B) {
        %c3 = setle int %A, %B          ; <bool> [#uses=1]
        ret bool %c3
}

which is handy, because the Java FE makes these sequences all over the place.

This is tested as: test/Regression/Transforms/InstCombine/JavaCompare.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14086 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-09 05:08:07 +00:00
Chris Lattner
2be51ae22d Implement select.ll:test14*
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14083 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-09 04:24:29 +00:00
Brian Gaeke
f36b5f0fb6 Expand head-of-file comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13982 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-03 05:03:02 +00:00
Brian Gaeke
dfb2e7d7c9 Use new form of unconditional branch constructor.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13930 91177308-0d34-0410-b5e6-96231b3b80d8
2004-06-01 20:06:10 +00:00
Chris Lattner
11b9be5a1a Fix one of the major things that is causing the C Backend to infinite loop
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13872 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-28 05:02:13 +00:00
John Criswell
e7b9a1a5f2 Fix a bug in the -deadtypeelim pass. The SymbolTable re-write changed it
to eliminate the wrong type.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13855 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-27 21:16:46 +00:00
Chris Lattner
e1368aef23 Fix InstCombine/load.ll & PR347.
This code hadn't been updated after the "structs with more than 256 elements"
related changes to the GEP instruction.  Also it was not handling the
ConstantAggregateZero class.

Now it does!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13834 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-27 17:30:27 +00:00
Chris Lattner
8dfe5705b1 Implement constant folding of fmod, which is used a lot in povray
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13823 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-27 07:25:00 +00:00
Chris Lattner
b18b9d7374 Restructure call constant folding code a bit to make it simpler
Add support for acos/asin/atan.  188.ammp contains three calls to acos with
constant arguments.  Constant folding it allows elimination of those 3 calls
and three FP divisions of the results.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13821 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-27 06:26:28 +00:00
Alkis Evlogimenos
fe3a093bc6 Do not pass a null pointer if this instruction is not prepended or
appended anywhere.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13798 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-26 22:50:28 +00:00
Alkis Evlogimenos
99c58f4910 Use one destination constructor for the unconditional branch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13792 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-26 21:38:14 +00:00
Reid Spencer
9231ac8b6f Convert to SymbolTable's new iteration interface.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13754 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-25 08:53:40 +00:00
Reid Spencer
567bc2cc1e Convert to SymbolTable's new lookup and iteration interfaces.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13751 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-25 08:52:20 +00:00
Reid Spencer
c8a1fcdb48 Remove unused header file.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13750 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-25 08:51:36 +00:00
Reid Spencer
3f0635e012 Make this pass simply invoke SymbolTable::strip().
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13749 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-25 08:51:25 +00:00
Chris Lattner
3c6a0d4ae2 Implement InstCombine:shift.ll:test16, which turns (X >> C1) & C2 != C3
into (X & (C2 << C1)) != (C3 << C1), where the shift may be either left or
right and the compare may be any one.

This triggers 1546 times in 176.gcc alone, as it is a common pattern that
occurs for bitfield accesses.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13740 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-25 06:32:08 +00:00
Chris Lattner
a710ddc9c4 Implement instcombine/cast.ll:test16:
Canonicalize cast X to bool into a setne instruction


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13736 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-25 04:29:21 +00:00
Chris Lattner
b11a99bd39 Fix a bug in my previous checkin
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13717 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-24 06:24:46 +00:00
Chris Lattner
f52988a92d Spelling people's names right is kinda important
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13702 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-23 21:27:29 +00:00
Chris Lattner
befa499d45 Fix cases where we missed inlining some more obvious candidates because the
caller was in an SCC.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@13693 91177308-0d34-0410-b5e6-96231b3b80d8
2004-05-23 21:22:17 +00:00