Author: resistor
Date: Fri Aug 3 06:03:26 2007
New Revision: 40776
URL: http://llvm.org/viewvc/llvm-project?rev=40776view=rev
Log:
Fix a subtle iterator invalidation bug in a recursive algorithm.
Modified:
llvm/trunk/lib/Transforms/Scalar/GVN.cpp
Modified:
The attached patch implements code generation of byval arguments on
the callee. For now only the case were the struct is in the stack is
handle correctly.
With the patch, the function
---
%struct.s = type { i64, i64, i64 }
define i64 @f(%struct.s*
Author: alenhar2
Date: Fri Aug 3 12:46:27 2007
New Revision: 40786
URL: http://llvm.org/viewvc/llvm-project?rev=40786view=rev
Log:
kernel stat pass and cleanup
Modified:
poolalloc/branches/SVA/lib/DSA/DataStructure.cpp
poolalloc/branches/SVA/lib/DSA/Info.cpp
Author: johannes
Date: Fri Aug 3 15:20:50 2007
New Revision: 40792
URL: http://llvm.org/viewvc/llvm-project?rev=40792view=rev
Log:
long double patch 2 of N. Handle it in TargetData.
(I've tried to get the info right for all targets,
but I'm not expert on all of them - check yours.)
Modified:
Author: resistor
Date: Fri Aug 3 14:59:35 2007
New Revision: 40791
URL: http://llvm.org/viewvc/llvm-project?rev=40791view=rev
Log:
Fix a subtle miscompilation. This allows 197.parser to be compiled correctly.
Modified:
llvm/trunk/lib/Transforms/Scalar/GVN.cpp
Modified:
Author: johannes
Date: Fri Aug 3 15:51:37 2007
New Revision: 40793
URL: http://llvm.org/viewvc/llvm-project?rev=40793view=rev
Log:
long double patch 3 of N. Add to MVT.
Modified:
llvm/trunk/include/llvm/CodeGen/ValueTypes.h
llvm/trunk/include/llvm/CodeGen/ValueTypes.td
On Aug 3, 2007, at 9:22 AM, Rafael Espindola wrote:
The attached patch implements code generation of byval arguments on
the callee. For now only the case were the struct is in the stack is
handle correctly.
I think this patch is looking very good. As it can only affect
things marked
--- llvm/trunk/lib/Target/X86/X86TargetMachine.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetMachine.cpp Fri Aug 3 15:20:50 2007
@@ -115,8 +115,8 @@
bool is64Bit)
: Subtarget(M, FS, is64Bit),
DataLayout(Subtarget.is64Bit() ?
-
On Aug 3, 2007, at 2:06 PM, Dan Gohman wrote:
--- llvm/trunk/lib/Target/X86/X86TargetMachine.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetMachine.cpp Fri Aug 3
15:20:50 2007
@@ -115,8 +115,8 @@
bool is64Bit)
: Subtarget(M, FS, is64Bit),
Author: johannes
Date: Fri Aug 3 17:46:15 2007
New Revision: 40796
URL: http://llvm.org/viewvc/llvm-project?rev=40796view=rev
Log:
Make x86 long double alignment 32 for everything but
Darwin (which makes size within a struct==96)
Modified:
llvm/trunk/lib/Target/X86/X86TargetMachine.cpp
Author: reid
Date: Fri Aug 3 18:49:37 2007
New Revision: 40797
URL: http://llvm.org/viewvc/llvm-project?rev=40797view=rev
Log:
For PR1583:
Restore some sanity to building test programs. The output of CBE *must* be
compiled and linked with gcc not g++, even if the source program contained
On Aug 3, 2007, at 4:49 PM, Reid Spencer wrote:
Author: reid
Date: Fri Aug 3 18:49:37 2007
New Revision: 40797
@@ -387,7 +377,7 @@
# If the program requires exception handling support, enable
(potentially
# expensive) support for it.
ifdef REQUIRES_EH_SUPPORT
-JIT_OPTS +=
On Aug 3, 2007, at 3:46 PM, Dale Johannesen wrote:
Author: johannes
Date: Fri Aug 3 17:46:15 2007
New Revision: 40796
Thanks Dale,
==
--- llvm/trunk/lib/Target/X86/X86TargetMachine.cpp (original)
+++
+PPC_FP128TyID, /// 5: 128 bit floating point type (two 64-
bits)
Is there a different name for this? PowerPC isn't the only platform
that
uses (or could use) a double-double approach.
What other target does? I've no objection to renaming if you have a
more descriptive suggestion
On Fri, 2007-08-03 at 16:58 -0700, Chris Lattner wrote:
On Aug 3, 2007, at 4:49 PM, Reid Spencer wrote:
Author: reid
Date: Fri Aug 3 18:49:37 2007
New Revision: 40797
@@ -387,7 +377,7 @@
# If the program requires exception handling support, enable
(potentially
# expensive)
Author: lattner
Date: Fri Aug 3 20:04:40 2007
New Revision: 40802
URL: http://llvm.org/viewvc/llvm-project?rev=40802view=rev
Log:
make RenamePassWorkList a local var instead of an ivar.
Modified:
llvm/trunk/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Modified:
Author: lattner
Date: Fri Aug 3 20:07:49 2007
New Revision: 40803
URL: http://llvm.org/viewvc/llvm-project?rev=40803view=rev
Log:
avoid an unneeded vector copy. This speeds up mem2reg on the testcase
in PR1432 by 6%
Modified:
llvm/trunk/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Author: lattner
Date: Fri Aug 3 20:19:38 2007
New Revision: 40804
URL: http://llvm.org/viewvc/llvm-project?rev=40804view=rev
Log:
add a typedef, no other change.
Modified:
llvm/trunk/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Modified:
Author: lattner
Date: Fri Aug 3 20:41:18 2007
New Revision: 40805
URL: http://llvm.org/viewvc/llvm-project?rev=40805view=rev
Log:
refactor some code to shrink PromoteMem2Reg::run a bit
Modified:
llvm/trunk/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Modified:
Author: lattner
Date: Fri Aug 3 20:47:41 2007
New Revision: 40806
URL: http://llvm.org/viewvc/llvm-project?rev=40806view=rev
Log:
split rewriting of single-store allocas into its own
method.
Modified:
llvm/trunk/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Modified:
Author: lattner
Date: Fri Aug 3 21:15:24 2007
New Revision: 40809
URL: http://llvm.org/viewvc/llvm-project?rev=40809view=rev
Log:
In mem2reg, when handling the single-store case, make sure to remove
a using block from the list if we handle it. Not doing this caused us
to not be able to promote
Author: lattner
Date: Fri Aug 3 21:21:22 2007
New Revision: 40810
URL: http://llvm.org/viewvc/llvm-project?rev=40810view=rev
Log:
switch from using a std::set to using a SmallPtrSet. This speeds up the
testcase in PR1432 from 6.33s to 2.90s (2.22x)
Modified:
Author: lattner
Date: Fri Aug 3 21:32:22 2007
New Revision: 40811
URL: http://llvm.org/viewvc/llvm-project?rev=40811view=rev
Log:
Three improvements:
1. Check for revisiting a block before checking domination, which is faster.
2. If the stored value isn't an instruction, we don't have to
Author: lattner
Date: Fri Aug 3 21:38:38 2007
New Revision: 40812
URL: http://llvm.org/viewvc/llvm-project?rev=40812view=rev
Log:
When we do the single-store optimization, delete both the store
and the alloca so they don't get reprocessed.
This speeds up PR1432 from 2.20s to 2.17s.
Modified:
Author: lattner
Date: Fri Aug 3 21:45:02 2007
New Revision: 40813
URL: http://llvm.org/viewvc/llvm-project?rev=40813view=rev
Log:
fix a logic bug where we wouldn't promote single store allocas if the
stored value was a non-instruction value. Doh.
This increase the # single store allocas from
25 matches
Mail list logo
