Author: djg
Date: Tue Jul 3 09:03:57 2007
New Revision: 37851
URL: http://llvm.org/viewvc/llvm-project?rev=37851&view=rev
Log:
Fix several over-aggressive folds for undef nodes in dagcombine, to
follow the rules for undef used in instcombine.
Modified:
llvm/trunk/lib/CodeGen/SelectionDAG/DAG
Hi Dan,
> + // fold (add x, undef) -> undef
> + if (N1.getOpcode() == ISD::UNDEF)
> +return N1;
what if N0 is undef and not N1?
> - // If either operand is undef, the result is undef
> - if (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF)
> -return DAG.getNode(ISD::UNDE
On Jul 3, 2007, at 7:43 AM, Duncan Sands wrote:
> Hi Dan,
>> + // fold (add x, undef) -> undef
>> + if (N1.getOpcode() == ISD::UNDEF)
>> +return N1;
>
> what if N0 is undef and not N1?
Presumably, undef is being canonicalized to the RHS somewhere. I
don't know if this is actually true in t
On Tuesday 03 July 2007 11:23, Chris Lattner wrote:
> >> + // fold (or x, undef) -> -1
> >> + if (N1.getOpcode() == ISD::UNDEF)
> >> +return DAG.getConstant(-1, VT);
> >
> > Is this the right way to get an all-bits-one value?
>
> Yes, but this has the same problem with vectors. :)
I don't u
Author: asl
Date: Tue Jul 3 12:01:58 2007
New Revision: 37852
URL: http://llvm.org/viewvc/llvm-project?rev=37852&view=rev
Log:
Rename llvm-test => test-suite. By Dave Greene.
Modified:
llvm/trunk/autoconf/configure.ac
llvm/trunk/configure
Modified: llvm/trunk/autoconf/configure.ac
URL:
Author: johannes
Date: Tue Jul 3 12:07:33 2007
New Revision: 37853
URL: http://llvm.org/viewvc/llvm-project?rev=37853&view=rev
Log:
Some spacing fixes. Cosmetic.
Modified:
llvm/trunk/lib/Target/X86/X86FloatingPoint.cpp
llvm/trunk/lib/Target/X86/X86InstrFPStack.td
Modified: llvm/trunk/l
Author: asl
Date: Tue Jul 3 12:16:46 2007
New Revision: 37854
URL: http://llvm.org/viewvc/llvm-project?rev=37854&view=rev
Log:
Revert last change until issue reported by Owen, won't be fixed.
Modified:
llvm/trunk/autoconf/configure.ac
llvm/trunk/configure
Modified: llvm/trunk/autoconf/c
This will most likely break the nightly tester.
-Tanya
On Jul 3, 2007, at 10:01 AM, Anton Korobeynikov wrote:
> Author: asl
> Date: Tue Jul 3 12:01:58 2007
> New Revision: 37852
>
> URL: http://llvm.org/viewvc/llvm-project?rev=37852&view=rev
> Log:
> Rename llvm-test => test-suite. By Dave Gree
I received very important comments from Evan and fixed some problems. Thanks
for any suggestion.
[]'s
--
Raul Fernandes Herbster
Embedded and Pervasive Computing Laboratory - embedded.dee.ufcg.edu.br
Electrical Engineering Department - DEE - www.ee.ufcg.edu.br
Electrical Engineering and Informat
Author: tbrethou
Date: Tue Jul 3 13:11:20 2007
New Revision: 37855
URL: http://llvm.org/viewvc/llvm-project?rev=37855&view=rev
Log:
We really should not have this version number in SO many makefiles. This
should also probably change to svn instead of cvs, but leaving it for now.
Modified:
On 7/3/07, Dan Gohman <[EMAIL PROTECTED]> wrote:
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=37851&r1=37850&r2=37851&view=diff
> ==
>
Could there be an extra newline
Hello, Bill
> On 7/3/07, Dan Gohman <[EMAIL PROTECTED]> wrote:
> > URL:
> > http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=37851&r1=37850&r2=37851&view=diff
> > ==
> >
> Could
On 7/3/07, Anton Korobeynikov <[EMAIL PROTECTED]> wrote:
> Hello, Bill
>
> > On 7/3/07, Dan Gohman <[EMAIL PROTECTED]> wrote:
> > > URL:
> > > http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=37851&r1=37850&r2=37851&view=diff
> > > ===
On Jul 3, 2007, at 11:20 AM, Anton Korobeynikov wrote:
> Hello, Bill
>
>> On 7/3/07, Dan Gohman <[EMAIL PROTECTED]> wrote:
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/
>>> SelectionDAG/DAGCombiner.cpp?rev=37851&r1=37850&r2=37851&view=diff
>>> =
Author: resistor
Date: Tue Jul 3 13:37:08 2007
New Revision: 37856
URL: http://llvm.org/viewvc/llvm-project?rev=37856&view=rev
Log:
Add support for performing GVNPRE on cast instructions, and add a testcase for
this.
Added:
llvm/trunk/test/Transforms/GVNPRE/cast.ll
Modified:
llvm/trunk/
Author: resistor
Date: Tue Jul 3 14:01:42 2007
New Revision: 37857
URL: http://llvm.org/viewvc/llvm-project?rev=37857&view=rev
Log:
Make the unary operator case a bit faster, since casts are the only kind of
unary operation.
Modified:
llvm/trunk/lib/Transforms/Scalar/GVNPRE.cpp
Modified: l
On Tuesday 03 July 2007 12:16, Anton Korobeynikov wrote:
> Author: asl
> Date: Tue Jul 3 12:16:46 2007
> New Revision: 37854
>
> URL: http://llvm.org/viewvc/llvm-project?rev=37854&view=rev
> Log:
> Revert last change until issue reported by Owen, won't be fixed.
> - test-suite) AC_CONFIG_S
On Jul 3, 2007, at 1:11 PM, David A. Greene wrote:
> On Tuesday 03 July 2007 12:16, Anton Korobeynikov wrote:
>> Author: asl
>> Date: Tue Jul 3 12:16:46 2007
>> New Revision: 37854
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=37854&view=rev
>> Log:
>> Revert last change until issue reporte
>> We overload ISD::FADD and quite a lot of others. Why not
>> ISD::ConstantFP too?
>
> Fair enough, after pondering on it, I agree with you. The proposed
> semantics are that a ConstantFP (and also a normal Constant?) produce
> the splatted immediate value?
Constant sounds good too. And U
On Tue, Jul 03, 2007 at 04:43:14PM +0200, Duncan Sands wrote:
> Hi Dan,
>
> > + // fold (add x, undef) -> undef
> > + if (N1.getOpcode() == ISD::UNDEF)
> > +return N1;
>
> what if N0 is undef and not N1?
I had thought they'd be canonicalized, but I missed that the dagcombiner
sometimes has
Author: resistor
Date: Tue Jul 3 16:36:47 2007
New Revision: 37858
URL: http://llvm.org/viewvc/llvm-project?rev=37858&view=rev
Log:
Remove BFtoLLVM. It was old, and a poor example because it didn't use best
practices for making
a LLVM frontend in C++. Maybe someday it will be rewritten..
Remo
>>> + // fold (or x, undef) -> -1
>>> + if (N1.getOpcode() == ISD::UNDEF)
>>> +return DAG.getConstant(-1, VT);
>>
>> Is this the right way to get an all-bits-one value?
>
> I guess DAG.getConstant(~0ULL, VT) is the way its written
> elsewhere. And I
> have to fix it for vectors. I'll work o
Author: asl
Date: Tue Jul 3 16:57:50 2007
New Revision: 37859
URL: http://llvm.org/viewvc/llvm-project?rev=37859&view=rev
Log:
Test modified svn mailer
Modified:
llvm/trunk/utils/NLT.schema
Modified: llvm/trunk/utils/NLT.schema
URL:
http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/NLT.
On Jul 3, 2007, at 2:13 PM, Dan Gohman wrote:
>>> We overload ISD::FADD and quite a lot of others. Why not
>>> ISD::ConstantFP too?
>>
>> Fair enough, after pondering on it, I agree with you. The proposed
>> semantics are that a ConstantFP (and also a normal Constant?) produce
>> the splatted im
Author: resistor
Date: Tue Jul 3 17:50:56 2007
New Revision: 37860
URL: http://llvm.org/viewvc/llvm-project?rev=37860&view=rev
Log:
Add functionality to value number GEP instructions. This also provides the
infrastructure that will
be used for function calls. NOTE: This does not yet do any tra
Author: void
Date: Tue Jul 3 18:04:11 2007
New Revision: 37861
URL: http://llvm.org/viewvc/llvm-project?rev=37861&view=rev
Log:
Change the place where man pages are installed to ${prefix}/share/man1. This
puts them in a more standardized place.
Modified:
llvm/trunk/Makefile.config.in
Modif
Revision: 129120
Author: bwendlin
Date: 2007-07-03 16:06:28 -0700 (Tue, 03 Jul 2007)
Log Message:
---
Change the place the man pages are installed to be /usr/local/share/man instead
of /usr/local/llvm-gcc-install/share/man.
Modified Paths:
--
apple-local/branches/llv
On Jun 27, 2007, at 7:00 PM, Sheng Zhou wrote:
> Chris,
>
> Attached is the testcase, which will get:
>
> opt: /developer/home2/zsth/llvm-gcc-dev/HEAD/llvm/llvm/lib/VMCore/
> Constants.cpp:1559: static llvm::Constant*
> llvm::ConstantExpr::getZExt(llvm::Constant*, const llvm::Type*):
> Asse
On Jun 29, 2007, at 1:10 AM, Duncan Sands wrote:
> Hi,
>
> It seems like folding undef/X to undef isn't safe either though,
>
> here is my understanding of how to fold undef. I hope it clarifies
> this confusing area. Of course, I could be confused myself but I
> hope not :)
>
> (1) When is
Author: resistor
Date: Tue Jul 3 18:51:19 2007
New Revision: 37862
URL: http://llvm.org/viewvc/llvm-project?rev=37862&view=rev
Log:
Add support for performing GVNPRE on GEP instructions.
Added:
llvm/trunk/test/Transforms/GVNPRE/gep.ll
Modified:
llvm/trunk/lib/Transforms/Scalar/GVNPRE.cpp
Author: void
Date: Tue Jul 3 19:19:54 2007
New Revision: 37863
URL: http://llvm.org/viewvc/llvm-project?rev=37863&view=rev
Log:
Allow a GR64 to be moved into an MMX register via the "movd" instruction.
Still need to have JIT generate this code.
Added:
llvm/trunk/test/CodeGen/X86/2007-07-03
Author: dpatel
Date: Tue Jul 3 20:05:22 2007
New Revision: 37864
URL: http://llvm.org/viewvc/llvm-project?rev=37864&view=rev
Log:
Fix typo in assertion check.
Modified:
llvm/trunk/include/llvm/Analysis/Dominators.h
Modified: llvm/trunk/include/llvm/Analysis/Dominators.h
URL:
http://llvm.o
Author: dpatel
Date: Tue Jul 3 20:11:19 2007
New Revision: 37865
URL: http://llvm.org/viewvc/llvm-project?rev=37865&view=rev
Log:
Remove unused method - getIDomBlock().
Modified:
llvm/trunk/include/llvm/Analysis/Dominators.h
Modified: llvm/trunk/include/llvm/Analysis/Dominators.h
URL:
http
Author: void
Date: Tue Jul 3 20:29:22 2007
New Revision: 37866
URL: http://llvm.org/viewvc/llvm-project?rev=37866&view=rev
Log:
Support generation of GR64 to MMX code in the JIT.
Modified:
llvm/trunk/lib/Target/X86/X86InstrMMX.td
Modified: llvm/trunk/lib/Target/X86/X86InstrMMX.td
URL:
http
Author: reid
Date: Tue Jul 3 20:35:32 2007
New Revision: 37867
URL: http://llvm.org/viewvc/llvm-project?rev=37867&view=rev
Log:
Ignore the .svn directories so countloc.sh and llvmgrep won't traverse them.
Modified:
llvm/trunk/utils/llvmdo
Modified: llvm/trunk/utils/llvmdo
URL:
http://llvm.
Hi Dan,
This patch is breaking llvm-gcc bootstrapping on PPC.
I am not sure what exactly wrong is it. But the old code has a check
for endianness while your new code doesn't. Can you check again if
you are taking endianness into consideration?
Thanks,
Evan
On Jul 2, 2007, at 9:18 AM, Dan G
Author: resistor
Date: Tue Jul 3 23:51:16 2007
New Revision: 37868
URL: http://llvm.org/viewvc/llvm-project?rev=37868&view=rev
Log:
Fix an error in phi translation of GEPs that was causing failures.
Added:
llvm/trunk/test/Transforms/GVNPRE/2007-07-03-PhiTranslateGEP.ll
Modified:
llvm/tru
Hi Chris,
> The condition "IterationCount->getType() != LargestType" doesn't
> mean the IterationCount->getType's bitwidth < LargestType's
> so, sometimes, (like in this testcase), it need a trunc not ext.
> This patch is to fix it.
Ah, I see. Thank you for checking in the testcase. The re
On Jul 3, 2007, at 6:35 PM, Evan Cheng wrote:
> Hi Dan,
>
> This patch is breaking llvm-gcc bootstrapping on PPC.
>
> I am not sure what exactly wrong is it. But the old code has a check
> for endianness while your new code doesn't. Can you check again if
> you are taking endianness into consider
On Tue, 2007-07-03 at 18:11 +, Tanya Lattner wrote:
> Author: tbrethou
> Date: Tue Jul 3 13:11:20 2007
> New Revision: 37855
>
> URL: http://llvm.org/viewvc/llvm-project?rev=37855&view=rev
> Log:
> We really should not have this version number in SO many makefiles. This
> should also probab
On Tue, 2007-07-03 at 10:27 -0700, Tanya Lattner wrote:
> This will most likely break the nightly tester.
Yeah, it will. The nightly tester checks out "test-suite" *as* llvm-test
specifically so this configure stuff wouldn't break .. a temporary
measure till we get llvm-top all figured out.
Reid.
41 matches
Mail list logo
