Commit Graph

1946 Commits

Author SHA1 Message Date
Chris Lattner
ef423ebdd1 gvn now hoists this load out of the hot non-call path.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@61028 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-15 06:34:48 +00:00
Chris Lattner
4807e07fff Adjust testcase to make it more stable across visitation order changes,
unbreaking it after r61024.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@61025 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-15 04:42:00 +00:00
Chris Lattner
f33131685b make GVN try to rename inputs to the resultant replaced values, which
cleans up the generated code a bit.  This should have the added benefit of
not randomly renaming functions/globals like my previous patch did. :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@61023 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-15 03:46:38 +00:00
Chris Lattner
9e59c64c14 Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which 
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:

BB1:
   X = load P
   br BB3
BB2:
   Y = load Q
   br BB3
BB3:
   R = phi [P] [Q]
   load R

turning "load R" into a phi of X/Y.  In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more 
efficient.  For example, consider:


bb1:  // has many predecessors.
   P = some_operator()
   load P

In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P.  In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something.  In many other cases though, it would scan and not
find anything useful.  MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors.  This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful.  For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!.  IMO, scalar GVN 
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.

In the future, this should be enhanced to phi translate through geps and 
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@61022 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-15 03:35:32 +00:00
Chris Lattner
5c6d91c1d1 another random testcase that shouldn't crash gvn and is
good for coverage with future changes.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@61011 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-14 21:20:46 +00:00
Chris Lattner
3f101bb1e7 RLE isn't smart enough to eliminate this safely yet.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60994 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-13 21:04:20 +00:00
Chris Lattner
4f118f0a20 rename some tests to be more uniform in naming convention.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60988 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-13 18:47:40 +00:00
Chris Lattner
c2f33f24af gvn should never crash on this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60987 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-13 18:39:44 +00:00
Bill Wendling
8295e49e76 Temporarily revert r60973. It's inexplicably causing a failure when self-hosting LLVM:
llvm[2]: Linking Release executable opt (without symbols)
...
Undefined symbols:
  "llvm::APFloat::IEEEsingle", referenced from:
      __ZN4llvm7APFloat10IEEEsingleE$non_lazy_ptr in libLLVMCore.a(Constants.o)
      __ZN4llvm7APFloat10IEEEsingleE$non_lazy_ptr in libLLVMCore.a(AsmWriter.o)
      __ZN4llvm7APFloat10IEEEsingleE$non_lazy_ptr in libLLVMCore.a(ConstantFold.o)
  "llvm::APFloat::IEEEdouble", referenced from:
      __ZN4llvm7APFloat10IEEEdoubleE$non_lazy_ptr in libLLVMCore.a(Constants.o)
      __ZN4llvm7APFloat10IEEEdoubleE$non_lazy_ptr in libLLVMCore.a(AsmWriter.o)
      __ZN4llvm7APFloat10IEEEdoubleE$non_lazy_ptr in libLLVMCore.a(ConstantFold.o)
ld: symbol(s) not found

This is in release mode. To replicate, compile llvm and llvm-gcc in optimized
mode. Then build llvm, in optimized mode, with the newly created compiler.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60977 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-13 09:28:44 +00:00
Chris Lattner
879922932f make RLE preserve the name of the load that it replaces. This is just
a pretification of the IR.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60973 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-13 07:22:47 +00:00
Chris Lattner
5a45bf1b48 loosen up an assertion that isn't valid when called from
invalidateCachedPointerInfo.  Thanks to Bill for sending me
a testcase.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60805 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-09 22:45:32 +00:00
Chris Lattner
20d6f0982a Teach BasicAA::getModRefInfo(CallSite, CallSite) some
tricks based on readnone/readonly functions.

Teach memdep to look past readonly calls when analyzing
deps for a readonly call.  This allows elimination of a
few more calls from 403.gcc:

before:
     63 gvn    - Number of instructions PRE'd
 153986 gvn    - Number of instructions deleted
  50069 gvn    - Number of loads deleted

after:
     63 gvn    - Number of instructions PRE'd
 153991 gvn    - Number of instructions deleted
  50069 gvn    - Number of loads deleted

5 calls isn't much, but this adds plumbing for the next change.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60794 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-09 21:19:42 +00:00
Devang Patel
51d8b669f1 Actually test something. Use PR3170 test case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60727 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-08 23:44:46 +00:00
Devang Patel
6d9896f153 Undo previous patch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60701 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-08 17:02:37 +00:00
Chris Lattner
a16ead8b10 fix a bug I introduced in simplifycfg handling single entry phi
nodes. FoldSingleEntryPHINodes deletes the PHI, so there is no
need to delete it afterward.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60653 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-07 07:22:45 +00:00
Chris Lattner
5e600e67b1 Reimplement the inner loop of DSE. It now uniformly uses getDependence(),
doesn't do its own local caching, and is slightly more aggressive about
free/store dse (see testcase).  This eliminates the last external client 
of MemDep::getDependenceFrom().


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60619 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-06 00:53:22 +00:00
Chris Lattner
7f7c736682 Fix test/Transforms/GVN/pre-load.ll
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60594 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-05 17:04:12 +00:00
Devang Patel
3831005eb1 Rewrite code that 1) filters loops and 2) calculates new loop bounds.
This fixes many bugs. I will add more test cases in a separate check-in.

Some day, the code that manipulates CFG and updates dom. info could use refactoring help.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60554 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-04 21:38:42 +00:00
Chris Lattner
2a5dc33c57 testcase for br undef folding.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60471 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-03 07:48:27 +00:00
Chris Lattner
421fa9e32e Teach jump threading some more simple tricks:
1) have it fold "br undef", which does occur with
   surprising frequency as jump threading iterates.
2) teach j-t to delete dead blocks.  This removes the successor
   edges, reducing the in-edges of other blocks, allowing 
   recursive simplification.
3) Fold things like:
     br COND, BBX, BBY
  BBX:
     br COND, BBZ, BBW

   which also happens because jump threading iterates.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60470 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-03 07:48:08 +00:00
Chris Lattner
1ea7771e09 don't spew tons of stuff to the output. This testcase is *not* for
loop deletion (it is for a ton of passes), which is very bad.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60465 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-03 06:41:50 +00:00
Chris Lattner
c89c6a964c Implement PRE of loads in the GVN pass with a pretty cheap and
straight-forward implementation.  This does not require any extra
alias analysis queries beyond what we already do for non-local loads.

Some programs really really like load PRE.  For example, SPASS triggers
this ~1000 times, ~300 times in 255.vortex, and ~1500 times on 403.gcc.

The biggest limitation to the implementation is that it does not split
critical edges.  This is a huge killer on many programs and should be
addressed after the initial patch is enabled by default.

The implementation of this should incidentally speed up rejection of 
non-local loads because it avoids creating the repl densemap in cases 
when it won't be used for fully redundant loads.

This is currently disabled by default.
Before I turn this on, I need to fix a couple of miscompilations in
the testsuite, look at compile time performance numbers, and look at
perf impact.  This is pretty close to ready though.




git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60408 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-02 08:16:11 +00:00
Owen Anderson
e209d9a3e6 Add a test for my previous PRE fix.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60394 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-02 04:25:42 +00:00
Bill Wendling
03aae5f7e1 Use m_Specific() instead of double matching.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60341 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-01 08:09:47 +00:00
Chris Lattner
d8aafcb200 simplify these patterns using m_Specific. No need to grep for
xor in testcase (or is a substring).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60328 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-01 05:16:26 +00:00
Chris Lattner
05f18920e1 Teach inst combine to merge GEPs through PHIs. This is really
important because it is sinking the loads using the GEPs, but
not the GEPs themselves.  This triggers 647 times on 403.gcc
and makes the .s file much much nicer.  For example before:

        je      LBB1_87 ## bb78
LBB1_62:        ## bb77
        leal    84(%esi), %eax
LBB1_63:        ## bb79
        movl    (%eax), %eax
...
LBB1_87:        ## bb78
        movl    $0, 4(%esp)
        movl    %esi, (%esp)
        call    L_make_decl_rtl$stub
        jmp     LBB1_62 ## bb77


after:

        jne     LBB1_63 ## bb79
LBB1_62:        ## bb78
        movl    $0, 4(%esp)
        movl    %esi, (%esp)
        call    L_make_decl_rtl$stub
LBB1_63:        ## bb79
        movl    84(%esi), %eax

The input code was (and the GEPs are merged and
the PHI is now eliminated by instcombine):

        br i1 %tmp233, label %bb78, label %bb77
bb77:           
        %tmp234 = getelementptr %struct.tree_node* %t_addr.3, i32 0, i32 0, i32 22              
        br label %bb79
bb78:           
        call void @make_decl_rtl(%struct.tree_node* %t_addr.3, i8* null) nounwind
        %tmp235 = getelementptr %struct.tree_node* %t_addr.3, i32 0, i32 0, i32 22              
        br label %bb79
bb79:           
        %iftmp.12.0.in = phi %struct.rtx_def** [ %tmp235, %bb78 ], [ %tmp234, %bb77 ]           
        %iftmp.12.0 = load %struct.rtx_def** %iftmp.12.0.in             



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60322 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-01 02:34:36 +00:00
Chris Lattner
67afda6c1e testcase for my previous commit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60315 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-01 01:42:03 +00:00
Bill Wendling
b3833d1eb9 Implement ((A|B)&1)|(B&-2) -> (A&1) | B transformation. This also takes care of
permutations of this pattern.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60312 91177308-0d34-0410-b5e6-96231b3b80d8
2008-12-01 01:07:11 +00:00
Bill Wendling
b01865c210 Add instruction combining for ((A&~B)|(~A&B)) -> A^B and all permutations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60291 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-30 13:52:49 +00:00
Bill Wendling
7f0ef6b325 Implement (A&((~A)|B)) -> A&B transformation in the instruction combiner. This
takes care of all permutations of this pattern.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60290 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-30 13:08:13 +00:00
Bill Wendling
6e1783fb4a getSExtValue() doesn't work for ConstantInts with bitwidth > 64 bits. Use all
APInt calls instead.

This fixes PR3144.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60288 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-30 12:38:24 +00:00
Eli Friedman
d83ae7d698 Optimize memmove and memset into the LLVM builtins. Note that these
only show up in code from front-ends besides llvm-gcc, like clang.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60287 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-30 08:32:11 +00:00
Bill Wendling
13441c57ec Strengthen check for div inst-combining.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60276 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-30 04:33:53 +00:00
Bill Wendling
e1196d6d3e Instcombine was illegally transforming -X/C into X/-C when either X or C
overflowed on negation. This commit checks to make sure that neithe C nor X
overflows. This requires that the RHS of X (a subtract instruction) be a
constant integer.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60275 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-30 03:42:12 +00:00
Chris Lattner
cae30afc34 don't require GVN to work on dead values, just make the
test return the loaded value.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60252 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-29 21:21:48 +00:00
Chris Lattner
3a76be584b Fix a thinko that manifested as a crash on clamav last night.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60251 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-29 20:29:04 +00:00
Chris Lattner
d3d12ecadd Fix PR3141 by ensuring that MemoryDependenceAnalysis::removeInstruction
properly updates the reverse dependency map when it installs updated 
dependencies for instructions that depend on the removed instruction.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60222 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-28 22:51:08 +00:00
Chris Lattner
f5102a0f08 don't call MergeBasicBlockIntoOnlyPred on a block whose only
predecessor is itself.  This doesn't make sense, and this is
a dead infinite loop anyway.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60210 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-28 19:54:49 +00:00
Nick Lewycky
8ca5248522 Chris prefers icmp/select over udiv!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60187 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-27 22:41:10 +00:00
Nick Lewycky
895f085385 Add a couple of missed optimizations on integer vectors. Multiply and divide
by 1, as well as multiply by -1.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60182 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-27 20:21:08 +00:00
Chris Lattner
3d86d242c6 Fix PR3138: if we merge the entry block into another block, make sure to
move the other block back up into the entry position!



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60179 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-27 19:25:19 +00:00
Chris Lattner
69e067fdd8 Make jump threading substantially more powerful, in the following ways:
1. Make it fold blocks separated by an unconditional branch.  This enables
   jump threading to see a broader scope.
2. Make jump threading able to eliminate locally redundant loads when they
   feed the branch condition of a block.  This frequently occurs due to
   reg2mem running.
3. Make jump threading able to eliminate *partially redundant* loads when
   they feed the branch condition of a block.  This is common in code with
   lots of loads and stores like C++ code and 255.vortex.

This implements thread-loads.ll and rdar://6402033.

Per the fixme's, several pieces of this should be moved into Transforms/Utils.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60148 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-27 05:07:53 +00:00
Evan Cheng
88f2d6c167 convertToSignExtendedInteger should return opInvalidOp instead of asserting if sematics of float does not allow arithmetics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60042 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-25 19:00:29 +00:00
Chris Lattner
8d0d603a0b reenable test
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59986 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-24 21:27:20 +00:00
Bill Wendling
612e0b701f Temporarily XFAIL this test. r59976 and r59972 broke it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59981 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-24 20:43:33 +00:00
Chris Lattner
47f5751c80 Fix 3113: If we have a dead cyclic PHI, replace the whole thing
with an undef.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59972 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-24 19:25:36 +00:00
Nick Lewycky
0c73079855 Optimize (x/y)*y into x-(x%y) in general. Div and rem are about the same, and
a subtract is cheaper than a multiply. This generalizes an existing transform.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59800 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-21 07:33:58 +00:00
Devang Patel
13877bf6c2 Give SIToFPInst preference over UIToFPInst because it is faster on platforms that are widely used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59476 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-18 00:40:02 +00:00
Devang Patel
cd40233429 While handling floating point IVs lift restrictions on initial value and increment value.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59471 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-17 23:27:13 +00:00
Chris Lattner
a74cefa6d3 Handle the case where there is no "not". It is possible it got
folded into the select.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@59389 91177308-0d34-0410-b5e6-96231b3b80d8
2008-11-16 04:25:26 +00:00