Changes in directory llvm/lib/CodeGen/SelectionDAG:
SelectionDAGISel.cpp updated: 1.454 - 1.455
---
Log message:
Since TypeInfos are passed as i8 pointers, a NULL TypeInfo should be passed
as a null i8 pointer not as a 0 i32.
---
Diffs of the changes: (+2 -4)
SelectionDAGISel.cpp |6
Changes in directory llvm/include/llvm/Target:
TargetLowering.h updated: 1.126 - 1.127
---
Log message:
Target specific ifcvt code duplication limit.
---
Diffs of the changes: (+21 -3)
TargetLowering.h | 24 +---
1 files changed, 21 insertions(+), 3 deletions(-)
Changes in directory llvm/lib/CodeGen:
IfConversion.cpp updated: 1.19 - 1.20
---
Log message:
Ifcvt triangle: don't ifcvt 'true' BB if it has other predecessors; don't merge
'false' BB if it has other predecessors.
---
Diffs of the changes: (+18 -3)
IfConversion.cpp | 21
Changes in directory llvm/lib/Target/ARM:
ARMISelLowering.cpp updated: 1.55 - 1.56
---
Log message:
Set ARM ifcvt duplication limit to 3 for now.
---
Diffs of the changes: (+1 -0)
ARMISelLowering.cpp |1 +
1 files changed, 1 insertion(+)
Index: llvm/lib/Target/ARM/ARMISelLowering.cpp
Hi Chris, this breaks the Ada build since the TYPE_SIZE may not be
a constant:
+ if (TYPE_SIZE(type) == 0 ||
+ PadStartBits = int(TREE_INT_CST_LOW(TYPE_SIZE(type))) ||
I've attached a fix, which includes some other tweaks for arrays
(not very important) plus whitespace
Hi Dale,
Arrange for only 1 of multiple branches to landing pad to be kept.
you seem to assume that there is at most one landing pad block for
a given MBB, i.e. that all invokes in the MBB unwind to the same
landing pad. Are you sure that's true?
Ciao,
Duncan.
Changes in directory llvm-www:
OldNews.html updated: 1.5 - 1.6
---
Log message:
various updates
---
Diffs of the changes: (+5 -3)
OldNews.html |8 +---
1 files changed, 5 insertions(+), 3 deletions(-)
Index: llvm-www/OldNews.html
diff -u llvm-www/OldNews.html:1.5
Changes in directory llvm-www:
www-index.html updated: 1.142 - 1.143
---
Log message:
actually link to the dev mtg slides :)
---
Diffs of the changes: (+4 -4)
www-index.html |8
1 files changed, 4 insertions(+), 4 deletions(-)
Index: llvm-www/www-index.html
diff -u
Changes in directory llvm-www:
header.incl updated: 1.59 - 1.60
---
Log message:
add dev mtg to sidebar
---
Diffs of the changes: (+9 -0)
header.incl |9 +
1 files changed, 9 insertions(+)
Index: llvm-www/header.incl
diff -u llvm-www/header.incl:1.59
Changes in directory llvm/lib/Transforms/Scalar:
GVNPRE.cpp updated: 1.8 - 1.9
---
Log message:
Fix Expression comparison, which in turn fixes a value numbering error.
---
Diffs of the changes: (+11 -12)
GVNPRE.cpp | 23 +++
1 files changed, 11 insertions(+), 12
Does this actually cause a problem? This function is only called if
the outer struct has constant size. TreeToLLVM::EmitAggregateCopy
checks that TREE_CODE(TYPE_SIZE(type)) == INTEGER_CST is true before
calling into this code.
Yes, it crashes the compiler. Remember that the function is
Hi Dale,
Arrange for only 1 of multiple branches to landing pad to be kept.
it struck me that this patch seems to assume that if a successor
is a landing pad then it is necessarily a landing pad for the
current MBB. But couldn't it be the landing pad for some other
MBB, and simply a normal
Hi Chris,
Okay, do you have a testcase that demonstrates this?
I will synthesize one for you.
Also, the code for handling large arrays isn't needed. This code
is only called for fixed size structures that are small.
Hey, you are the one who used getInt64 to get the array length! :)
On Jun 1, 2007, at 12:56 PM, Duncan Sands wrote:
Hi Dale,
Arrange for only 1 of multiple branches to landing pad to be kept.
That should be successors, not branches, but you've gotten past
that
it struck me that this patch seems to assume that if a successor
is a landing pad then it
Hi Dale,
it struck me that this patch seems to assume that if a successor
is a landing pad then it is necessarily a landing pad for the
current MBB.
I wasn't intending to assume that; how so?
But couldn't it be the landing pad for some other
MBB, and simply a normal successor for
On Jun 1, 2007, at 1:24 PM, Duncan Sands wrote:
Suppose a successor S1 has isLandingPad true because it is
a landing pad for some other MBB, and some later successor (S2)
has isLandingPad true because it is the landing pad for this
MBB. What will happen? When we get to S1 foundPad is set to
Okay, do you have a testcase that demonstrates this?
I will synthesize one for you.
Actually a testcase would be pointless, since the compiler crashes
during the (Ada) compiler build. It wouldn't be pointless if I
could make a C or C++ testcase, but I don't think it can be done
except in
Revision: 127931
Author: clattner
Date: 2007-06-01 13:42:56 -0700 (Fri, 01 Jun 2007)
Log Message:
---
Fix union comparison code when handling hard ada cases :)
Patch by Duncan Sands. Thanks for beating this into me Duncan.
Modified Paths:
--
Changes in directory llvm/lib/Target/ARM:
ARMInstrInfo.td updated: 1.108 - 1.109
---
Log message:
Opcode modifier s comes after condition code. e.g. addlts, not addslt.
---
Diffs of the changes: (+20 -20)
ARMInstrInfo.td | 40
1 files changed, 20
Changes in directory llvm/lib/Transforms/Scalar:
GVNPRE.cpp updated: 1.9 - 1.10
---
Log message:
clean() needs to process things in topological order.
---
Diffs of the changes: (+21 -28)
GVNPRE.cpp | 49 +
1 files changed, 21
Changes in directory llvm/lib/Transforms/Scalar:
LICM.cpp updated: 1.101 - 1.102
---
Log message:
Insert new instructions in AliasSet.
---
Diffs of the changes: (+5 -1)
LICM.cpp |6 +-
1 files changed, 5 insertions(+), 1 deletion(-)
Index: llvm/lib/Transforms/Scalar/LICM.cpp
Changes in directory llvm/lib/ExecutionEngine/Interpreter:
Execution.cpp updated: 1.184 - 1.185
Interpreter.cpp updated: 1.39 - 1.40
---
Log message:
For PR1486: http://llvm.org/PR1486 :
Avoid overwriting the APInt instance with 0 bytes which causes the bitwidth
to be set to 0 (illegal)
Changes in directory llvm/test/CodeGen/X86:
test-pic-jtbl.ll updated: 1.4 - 1.5
---
Log message:
Implement smarter algorithm for choosing which blocks to tail-merge.
See test/CodeGen/X86/test-pic-jtbl.ll for a case where it works well;
shaves another 10K off our favorite benchmark. I was
Changes in directory llvm/lib/CodeGen:
BranchFolding.cpp updated: 1.61 - 1.62
---
Log message:
Implement smarter algorithm for choosing which blocks to tail-merge.
See test/CodeGen/X86/test-pic-jtbl.ll for a case where it works well;
shaves another 10K off our favorite benchmark. I was
Changes in directory llvm/lib/CodeGen:
README.txt updated: 1.6 - 1.7
---
Log message:
Fancier algorithm in tail-merge comment implemented, so remove comment.
---
Diffs of the changes: (+0 -18)
README.txt | 18 --
1 files changed, 18 deletions(-)
Index:
Changes in directory llvm/lib/CodeGen:
BranchFolding.cpp updated: 1.62 - 1.63
---
Log message:
Fix CorrectExtraCFGEdges to allow for multiple LandingPad targets.
---
Diffs of the changes: (+6 -9)
BranchFolding.cpp | 15 ++-
1 files changed, 6 insertions(+), 9 deletions(-)
On Jun 1, 2007, at 1:28 PM, Dale Johannesen wrote:
On Jun 1, 2007, at 1:24 PM, Duncan Sands wrote:
Suppose a successor S1 has isLandingPad true because it is
a landing pad for some other MBB, and some later successor (S2)
has isLandingPad true because it is the landing pad for this
MBB.
Changes in directory llvm-test/SingleSource/UnitTests/Integer:
reduce_xor.reference_output updated: 1.1 - 1.2
---
Log message:
Add the exit code to the reference output.
---
Diffs of the changes: (+1 -0)
reduce_xor.reference_output |1 +
1 files changed, 1 insertion(+)
Index:
Changes in directory llvm-test/SingleSource/UnitTests/Integer:
bits.h updated: 1.11 - 1.12
---
Log message:
Add int8
---
Diffs of the changes: (+1 -0)
bits.h |1 +
1 files changed, 1 insertion(+)
Index: llvm-test/SingleSource/UnitTests/Integer/bits.h
diff -u
Changes in directory llvm/lib/CodeGen:
IntrinsicLowering.cpp updated: 1.85 - 1.86
---
Log message:
Make LowerCTPOP() support arbitrary bitwidth integer type.
---
Diffs of the changes: (+22 -9)
IntrinsicLowering.cpp | 31 ++-
1 files changed, 22 insertions(+),
30 matches
Mail list logo
