oota-llvm.git
15 years agoReorder methods alphabetically. No functionality change.
Duncan Sands [Wed, 16 Jul 2008 11:41:33 +0000 (11:41 +0000)]
Reorder methods alphabetically.  No functionality change.
While this is not a wonderful organizing principle, it
does make it easy to find routines, and clear where to
insert new ones.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53672 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoTurn on LegalizeTypes by default.
Duncan Sands [Wed, 16 Jul 2008 11:36:51 +0000 (11:36 +0000)]
Turn on LegalizeTypes by default.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53671 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoDon't use ++idx_begin when I actually mean idx_begin + 1, especially since we
Matthijs Kooijman [Wed, 16 Jul 2008 10:47:35 +0000 (10:47 +0000)]
Don't use ++idx_begin when I actually mean idx_begin + 1, especially since we
also use *idx_begin in the same expression, giving unpredictable results.

This fixes this bug: http://lists.cs.uiuc.edu/pipermail/llvmdev/2008-July/015877.html

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53670 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoThe atomic.cmp.swap promotion logic is wrong: it
Duncan Sands [Wed, 16 Jul 2008 08:09:48 +0000 (08:09 +0000)]
The atomic.cmp.swap promotion logic is wrong: it
simply does the atomic.cmp.swap on the larger type,
which means it blows away whatever is sitting in
the bytes just after the memory location, i.e.
causes a buffer overflow.  This really requires
target specific code, which is why LegalizeTypes
doesn't try to handle this case generically.  The
existing (wrong) code in LegalizeDAG will go away
automatically once the type legalization code is
removed from LegalizeDAG so I'm leaving it there
for the moment.  Meanwhile, don't test for this
feature.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53669 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix PR2296. Do not transform x86_sse2_storel_dq into a full-width store.
Evan Cheng [Wed, 16 Jul 2008 07:28:14 +0000 (07:28 +0000)]
Fix PR2296. Do not transform x86_sse2_storel_dq into a full-width store.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53666 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agox86-64 PIC JIT fixes: do not generate the extra load for external GV's.
Evan Cheng [Wed, 16 Jul 2008 01:34:02 +0000 (01:34 +0000)]
x86-64 PIC JIT fixes: do not generate the extra load for external GV's.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53661 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoX86-64 PIC jump table values are different from x86-32 cases, they are dest - table...
Evan Cheng [Wed, 16 Jul 2008 01:33:08 +0000 (01:33 +0000)]
X86-64 PIC jump table values are different from x86-32 cases, they are dest - table base.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53660 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoTargetAsmInfo::SectionForGlobal showed up in a profile. Simplify it a little.
Dan Gohman [Tue, 15 Jul 2008 18:37:51 +0000 (18:37 +0000)]
TargetAsmInfo::SectionForGlobal showed up in a profile. Simplify it a little.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53639 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoSelectionDAG::AssignNodeIds is unused.
Dan Gohman [Tue, 15 Jul 2008 18:29:32 +0000 (18:29 +0000)]
SelectionDAG::AssignNodeIds is unused.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53636 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoDon't sort SDNodes by their addresses in SelectionDAG::dump. Instead,
Dan Gohman [Tue, 15 Jul 2008 18:18:54 +0000 (18:18 +0000)]
Don't sort SDNodes by their addresses in SelectionDAG::dump. Instead,
just use the AllNodes order, which is at least relatively stable
across runs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53632 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRevert this, as it seems to still be broken.
Owen Anderson [Tue, 15 Jul 2008 17:59:02 +0000 (17:59 +0000)]
Revert this, as it seems to still be broken.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53627 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoEnable local PRE by default.
Owen Anderson [Tue, 15 Jul 2008 16:28:23 +0000 (16:28 +0000)]
Enable local PRE by default.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53616 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoHave GVN do a pre-pass over the CFG that folds away unconditional branches where...
Owen Anderson [Tue, 15 Jul 2008 16:28:06 +0000 (16:28 +0000)]
Have GVN do a pre-pass over the CFG that folds away unconditional branches where possible.  This allows local PRE to be more aggressive.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53615 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoXFAIL the multdeadretval test for now, I will be fixing instcombine to make it work...
Matthijs Kooijman [Tue, 15 Jul 2008 16:05:09 +0000 (16:05 +0000)]
XFAIL the multdeadretval test for now, I will be fixing instcombine to make it work again tomorrow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53614 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoLegalizeTypes support for fabs on ppc long double.
Duncan Sands [Tue, 15 Jul 2008 15:02:44 +0000 (15:02 +0000)]
LegalizeTypes support for fabs on ppc long double.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53613 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRemove a few tests which no longer hold for deadargelim (since it is now
Matthijs Kooijman [Tue, 15 Jul 2008 14:57:01 +0000 (14:57 +0000)]
Remove a few tests which no longer hold for deadargelim (since it is now
allowed to canonicalize return values).

Add a test that checks if return value and function attributes are not removed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53612 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAdd a testcase for the canonicalizations now performed by deadargelim.
Matthijs Kooijman [Tue, 15 Jul 2008 14:42:58 +0000 (14:42 +0000)]
Add a testcase for the canonicalizations now performed by deadargelim.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53611 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAllow deadargelim to change return types even though now values were dead. This
Matthijs Kooijman [Tue, 15 Jul 2008 14:42:31 +0000 (14:42 +0000)]
Allow deadargelim to change return types even though now values were dead. This
again canonicalizes {i32} into i32 and {} into void.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53610 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRevert r53606. It turns out that explicitely tracking the liveness of the
Matthijs Kooijman [Tue, 15 Jul 2008 14:39:36 +0000 (14:39 +0000)]
Revert r53606. It turns out that explicitely tracking the liveness of the
return value as a whole in deadargelim is really not needed now that we simply
rebuild the old return value and actually prevents some canonicalization from
taking place.

This revert stops deadargelim from changing {i32} into i32 for now, but I'll
fix that next.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53609 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoMake deadargelim a bit less smart, so it doesn't choke on nested structs as
Matthijs Kooijman [Tue, 15 Jul 2008 14:03:10 +0000 (14:03 +0000)]
Make deadargelim a bit less smart, so it doesn't choke on nested structs as
return values that are still (partially) live. Instead of updating all uses of
a call instruction after removing some elements, it now just rebuilds the
original struct (With undef gaps where the unused values were) and leaves it to
instcombine to clean this up.

The added testcase still fails currently, but this is due to instcombine which
isn't good enough yet. I will fix that part next.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53608 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoDon't use isa when we can reuse a previous dyn_cast.
Matthijs Kooijman [Tue, 15 Jul 2008 13:39:08 +0000 (13:39 +0000)]
Don't use isa when we can reuse a previous dyn_cast.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53607 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoMake DeadArgElim keep liveness of the return value as a whole in addition to
Matthijs Kooijman [Tue, 15 Jul 2008 13:36:06 +0000 (13:36 +0000)]
Make DeadArgElim keep liveness of the return value as a whole in addition to
only the liveness of partial return values (for functions returning a struct).
This is more explicit to prevent unwanted changes in the return value.

In particular, deadargelim now canonicalizes a function returning {i32} to
returning i32 and {} to void, if the struct returned is not used in its
entirety, but only the single element is used.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53606 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix typo.
Matthijs Kooijman [Tue, 15 Jul 2008 13:15:10 +0000 (13:15 +0000)]
Fix typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53605 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoLegalizeTypes support for promotion of bswap.
Duncan Sands [Tue, 15 Jul 2008 10:18:22 +0000 (10:18 +0000)]
LegalizeTypes support for promotion of bswap.
In LegalizeDAG the value is zero-extended to
the new type before byte swapping.  It doesn't
matter how the extension is done since the new
bits are shifted off anyway after the swap, so
extend by any old rubbish bits.  This results
in the final assembler for the testcase being
one line shorter.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53604 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoLegalizeTypes support for promotion of SIGN_EXTEND_INREG.
Duncan Sands [Tue, 15 Jul 2008 10:14:24 +0000 (10:14 +0000)]
LegalizeTypes support for promotion of SIGN_EXTEND_INREG.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53603 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoReorder the integer promotion methods alphabetically.
Duncan Sands [Tue, 15 Jul 2008 10:12:34 +0000 (10:12 +0000)]
Reorder the integer promotion methods alphabetically.
No change in functionality.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53602 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoLet DAE keep a list of live functions, instead of simply marking all arguments
Matthijs Kooijman [Tue, 15 Jul 2008 09:11:16 +0000 (09:11 +0000)]
Let DAE keep a list of live functions, instead of simply marking all arguments
and return values live for those functions. This doesn't change anything yet,
but prepares for the coming commits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53601 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoSplit DAE::MarkLive into MarkLive and PropagateLiveness.
Matthijs Kooijman [Tue, 15 Jul 2008 09:00:17 +0000 (09:00 +0000)]
Split DAE::MarkLive into MarkLive and PropagateLiveness.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53600 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoPass around const RetOrArg references instead of copying values. Also, mark
Matthijs Kooijman [Tue, 15 Jul 2008 08:56:49 +0000 (08:56 +0000)]
Pass around const RetOrArg references instead of copying values. Also, mark
RetOrArg::getDescription() as const.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53599 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoSimplify debug code by using RetOrArg::getDescription().
Matthijs Kooijman [Tue, 15 Jul 2008 08:53:36 +0000 (08:53 +0000)]
Simplify debug code by using RetOrArg::getDescription().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53598 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix indentation (intentionally left out of the previous commit).
Matthijs Kooijman [Tue, 15 Jul 2008 08:47:32 +0000 (08:47 +0000)]
Fix indentation (intentionally left out of the previous commit).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53592 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoMove the deadargelim code for intrinsically alive functions into its own
Matthijs Kooijman [Tue, 15 Jul 2008 08:45:12 +0000 (08:45 +0000)]
Move the deadargelim code for intrinsically alive functions into its own
method, to slightly simplify control flow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53591 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFixed potential bug if the source and target of a bit convert have different alignment
Mon P Wang [Tue, 15 Jul 2008 05:28:34 +0000 (05:28 +0000)]
Fixed potential bug if the source and target of a bit convert have different alignment

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53590 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoCorrect this inversion!
Nick Lewycky [Tue, 15 Jul 2008 03:47:44 +0000 (03:47 +0000)]
Correct this inversion!
I swear that didn't show up in svn diff...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53587 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix up comments.
Nick Lewycky [Tue, 15 Jul 2008 03:40:27 +0000 (03:40 +0000)]
Fix up comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53586 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFixed call stack alignment. Improved AsmPrinter alignment issues.
Bruno Cardoso Lopes [Tue, 15 Jul 2008 02:03:36 +0000 (02:03 +0000)]
Fixed call stack alignment. Improved AsmPrinter alignment issues.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53585 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoLinkOnce definitions have default scope, like weak definitions. Otherwise, the linker...
Devang Patel [Tue, 15 Jul 2008 00:00:11 +0000 (00:00 +0000)]
LinkOnce definitions have default scope, like weak definitions. Otherwise, the linker may not be able to match LinkOnce definition from one module with an exteranl reference from other module.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53580 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoGoodbye tail duplication (for good this time).
Evan Cheng [Mon, 14 Jul 2008 22:29:15 +0000 (22:29 +0000)]
Goodbye tail duplication (for good this time).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53574 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoThis replaces all $(SolutionDir) macros with $(ProjectDir)..\
Chris Lattner [Mon, 14 Jul 2008 18:46:16 +0000 (18:46 +0000)]
This replaces all $(SolutionDir) macros with $(ProjectDir)..\
Patch by Nicolas Capens!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53571 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoReapply 53476 and 53480, with a fix so that it properly updates
Dan Gohman [Mon, 14 Jul 2008 18:19:29 +0000 (18:19 +0000)]
Reapply 53476 and 53480, with a fix so that it properly updates
the BB member to the current basic block after emitting
instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53567 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix uninitialized use of the Changed variable.
Dan Gohman [Mon, 14 Jul 2008 17:55:01 +0000 (17:55 +0000)]
Fix uninitialized use of the Changed variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53564 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoImprove debug output for MemOperandSDNode. PseudoSourceValue nodes
Dan Gohman [Mon, 14 Jul 2008 17:51:24 +0000 (17:51 +0000)]
Improve debug output for MemOperandSDNode. PseudoSourceValue nodes
don't have value names, so use print instead of getName() to get a
useful string.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53563 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix edito in the PseudoSourceValue name list.
Dan Gohman [Mon, 14 Jul 2008 17:45:47 +0000 (17:45 +0000)]
Fix edito in the PseudoSourceValue name list.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53562 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoReformat this message to fit in 80 cols.
Dan Gohman [Mon, 14 Jul 2008 17:43:27 +0000 (17:43 +0000)]
Reformat this message to fit in 80 cols.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53561 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoI don't think BUILD_PAIR can have a vector result.
Duncan Sands [Mon, 14 Jul 2008 17:34:19 +0000 (17:34 +0000)]
I don't think BUILD_PAIR can have a vector result.
Remove support for this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53559 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoTighten up some checks. Fix FPOWI splitting for
Duncan Sands [Mon, 14 Jul 2008 17:33:37 +0000 (17:33 +0000)]
Tighten up some checks.  Fix FPOWI splitting for
non-power-of-two vectors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53558 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoReapply r53540, now with the matching header!
Chris Lattner [Mon, 14 Jul 2008 17:32:59 +0000 (17:32 +0000)]
Reapply r53540, now with the matching header!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53557 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAn INSERT_VECTOR_ELT can insert a larger value
Duncan Sands [Mon, 14 Jul 2008 17:32:02 +0000 (17:32 +0000)]
An INSERT_VECTOR_ELT can insert a larger value
than the vector element type.  Don't forget to
handle this when the insertion index is not a
constant.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53556 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAccording to the docs, it is possible to have an
Duncan Sands [Mon, 14 Jul 2008 17:27:46 +0000 (17:27 +0000)]
According to the docs, it is possible to have an
extending load of a vector.  Handle this case when
splitting vector loads.  I'm not completely sure
what is supposed to happen, but I think it means
hi should be set to undef.  LegalizeDAG does not
consider this case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53555 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoThere should be no extending loads or truncating
Duncan Sands [Mon, 14 Jul 2008 17:22:31 +0000 (17:22 +0000)]
There should be no extending loads or truncating
stores of one-element vectors.  Also, neaten the
handling of INSERT_VECTOR_ELT when the inserted
type is larger than the vector element type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53554 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoIgnore TargetConstant with an illegal type. These
Duncan Sands [Mon, 14 Jul 2008 17:15:45 +0000 (17:15 +0000)]
Ignore TargetConstant with an illegal type.  These
are used for passing huge immediates in inline ASM
from the front-end straight down to the ASM writer.
Of course this is a hack, but it is simple, limited
in scope, works in practice, and is what LegalizeDAG
does.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53553 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAdded Subtarget support into RegisterInfo
Bruno Cardoso Lopes [Mon, 14 Jul 2008 14:42:54 +0000 (14:42 +0000)]
Added Subtarget support into RegisterInfo
Added HasABICall and HasAbsoluteCall (equivalent to gcc -mabicall and
-mno-shared). HasAbsoluteCall is not implemented but HasABICall is the
default for o32 ABI. Now, both should help into a more accurate
relocation types implementation.
Added IsLinux is needed to choose between asm directives.
Instruction name strings cleanup.
AsmPrinter improved.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53551 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRevert r53540 - it does not compile.
Duncan Sands [Mon, 14 Jul 2008 07:59:28 +0000 (07:59 +0000)]
Revert r53540 - it does not compile.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53549 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoReimplement LinkFunctionProtos in terms of GetLinkageResult. This fixes
Chris Lattner [Mon, 14 Jul 2008 07:23:24 +0000 (07:23 +0000)]
Reimplement LinkFunctionProtos in terms of GetLinkageResult.  This fixes
the second half of link-global-to-func.ll and causes some minor changes in
messages.

There are two TODOs here.  First, this causes a regression in
2008-07-06-AliasWeakDest.ll, which is now failing (so I xfailed it).  Anton,
I would really appreciate it if you could take a look at this.  It should be
a matter of adding proper alias support to GetLinkageResult, and was probably
already a latent bug that would manifest with globals.

The second todo is to reimplement LinkAlias in the same pattern as
function and global linking.  This should be pretty straight-forward for
someone who knows aliases, but isn't a requirement for correctness.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53548 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agodon't do any linkage, not even type resolution, of symbols that have
Chris Lattner [Mon, 14 Jul 2008 06:52:19 +0000 (06:52 +0000)]
don't do any linkage, not even type resolution, of symbols that have
internal linkage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53547 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoimplement linking of globals to functions, in one direction
Chris Lattner [Mon, 14 Jul 2008 06:49:45 +0000 (06:49 +0000)]
implement linking of globals to functions, in one direction
(replacing a function with a global).  This is needed when building
llvm itself with LTO on darwin, because of the EXPLICIT_SYMBOL hack
in lib/system/DynamicLibrary.cpp.

Implementation of linking the other way will need to wait for a
cleanup of LinkFunctionProtos.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53546 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agowrap long lines, remove some code from a non-assert build.
Chris Lattner [Mon, 14 Jul 2008 05:52:33 +0000 (05:52 +0000)]
wrap long lines, remove some code from a non-assert build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53545 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix a bunch of bugs handling vector compare constant expressions, fixing
Chris Lattner [Mon, 14 Jul 2008 05:17:31 +0000 (05:17 +0000)]
Fix a bunch of bugs handling vector compare constant expressions, fixing
PR2317.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53544 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoDocument and fix Constant::getVectorElements to return an empty vector
Chris Lattner [Mon, 14 Jul 2008 05:10:41 +0000 (05:10 +0000)]
Document and fix Constant::getVectorElements to return an empty vector
when presented with a constant expr.

If ConstantExpr::getV[IF]Cmp to work when ConstantFoldCompareInstruction
returns an undef or constant expr.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53541 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoIf a function calls setjmp, never inline it into other functions. This is
Chris Lattner [Mon, 14 Jul 2008 00:46:56 +0000 (00:46 +0000)]
If a function calls setjmp, never inline it into other functions.  This is
a hack around the fact that we don't represent the CFG correctly for sj/lj.
It fixes PR2486.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53540 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agosimplify some code, shuffle and insertelt always return a vector.
Chris Lattner [Mon, 14 Jul 2008 00:32:20 +0000 (00:32 +0000)]
simplify some code, shuffle and insertelt always return a vector.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53538 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agowhitespace fix.
Chris Lattner [Mon, 14 Jul 2008 00:28:45 +0000 (00:28 +0000)]
whitespace fix.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53537 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agodoxygenate comments and wrap to 80 cols.
Chris Lattner [Mon, 14 Jul 2008 00:27:31 +0000 (00:27 +0000)]
doxygenate comments and wrap to 80 cols.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53536 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAdd a note.
Chris Lattner [Mon, 14 Jul 2008 00:19:59 +0000 (00:19 +0000)]
Add a note.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53535 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix PR2506 by being a bit more careful about reverse fact propagation when
Chris Lattner [Mon, 14 Jul 2008 00:15:52 +0000 (00:15 +0000)]
Fix PR2506 by being a bit more careful about reverse fact propagation when
disproving a condition.  This actually compiles the existing testcase
(udiv_select_to_select_shift) to:

define i64 @test(i64 %X, i1 %Cond) {
entry:
%divisor1.t = lshr i64 %X, 3 ; <i64> [#uses=1]
%quotient2 = lshr i64 %X, 3 ; <i64> [#uses=1]
%sum = add i64 %divisor1.t, %quotient2 ; <i64> [#uses=1]
ret i64 %sum
}

instead of:

define i64 @test(i64 %X, i1 %Cond) {
entry:
%quotient1.v = select i1 %Cond, i64 3, i64 4 ; <i64> [#uses=1]
%quotient1 = lshr i64 %X, %quotient1.v ; <i64> [#uses=1]
%quotient2 = lshr i64 %X, 3 ; <i64> [#uses=1]
%sum = add i64 %quotient1, %quotient2 ; <i64> [#uses=1]
ret i64 %sum
}

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53534 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix mishandling of the infinite loop case when merging two blocks. This
Chris Lattner [Sun, 13 Jul 2008 22:23:11 +0000 (22:23 +0000)]
Fix mishandling of the infinite loop case when merging two blocks.  This
fixes PR2540.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53533 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agomore refactoring. Use early exits instead of really complex logic.
Chris Lattner [Sun, 13 Jul 2008 22:04:41 +0000 (22:04 +0000)]
more refactoring.  Use early exits instead of really complex logic.
No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53532 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoimprove comments.
Chris Lattner [Sun, 13 Jul 2008 21:55:46 +0000 (21:55 +0000)]
improve comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53531 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agofactor another large hunk of code out into its own function.
Chris Lattner [Sun, 13 Jul 2008 21:53:26 +0000 (21:53 +0000)]
factor another large hunk of code out into its own function.
No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53530 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoadd a description.
Chris Lattner [Sun, 13 Jul 2008 21:27:36 +0000 (21:27 +0000)]
add a description.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53529 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFinal bit of simplification for FoldBranchToCommonDest.
Chris Lattner [Sun, 13 Jul 2008 21:20:19 +0000 (21:20 +0000)]
Final bit of simplification for FoldBranchToCommonDest.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53528 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agosimplify logic a bit
Chris Lattner [Sun, 13 Jul 2008 21:15:11 +0000 (21:15 +0000)]
simplify logic a bit

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53527 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRefactor some code out into its own helper function, getting rid of crazy multiline
Chris Lattner [Sun, 13 Jul 2008 21:12:01 +0000 (21:12 +0000)]
Refactor some code out into its own helper function, getting rid of crazy multiline
conditionals and commenting the code better.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53526 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix comment.
Duncan Sands [Sun, 13 Jul 2008 20:14:38 +0000 (20:14 +0000)]
Fix comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53525 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix PR2231 - opt -internalize -std-compile-opts should run internalize first
Chris Lattner [Sun, 13 Jul 2008 19:35:21 +0000 (19:35 +0000)]
Fix PR2231 - opt -internalize -std-compile-opts should run internalize first

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53523 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRemove deleted files from VC++ project.
Cedric Venet [Sun, 13 Jul 2008 11:12:36 +0000 (11:12 +0000)]
Remove deleted files from VC++ project.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53521 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoInline typedef for alist_iterator::pointer to work with MSVC++.
Ted Kremenek [Sat, 12 Jul 2008 18:28:46 +0000 (18:28 +0000)]
Inline typedef for alist_iterator::pointer to work with MSVC++.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53517 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoStop creating extraneous smax/umax in SCEV. This removes a regression where we
Nick Lewycky [Sat, 12 Jul 2008 07:41:32 +0000 (07:41 +0000)]
Stop creating extraneous smax/umax in SCEV. This removes a regression where we
started complicating many loops ('for' loops, in fact).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53508 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agooperator[] is not defined for list::iterator. Overload it in ilist::iterator
Nick Lewycky [Sat, 12 Jul 2008 07:00:52 +0000 (07:00 +0000)]
operator[] is not defined for list::iterator. Overload it in ilist::iterator
to prevent silly things from happening accidentally. PR2171

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53507 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoEnhance analysis of srem.
Nick Lewycky [Sat, 12 Jul 2008 05:04:38 +0000 (05:04 +0000)]
Enhance analysis of srem.

Remove dead code analyzing urem. 'urem' of power-of-2 is canonicalized to an
'and' instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53506 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoImplement llvm.atomic.cmp.swap.i32 on PPC. Patch by Gary Benson!
Evan Cheng [Sat, 12 Jul 2008 02:23:19 +0000 (02:23 +0000)]
Implement llvm.atomic.cmp.swap.i32 on PPC. Patch by Gary Benson!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53505 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoTypos.
Evan Cheng [Sat, 12 Jul 2008 02:22:07 +0000 (02:22 +0000)]
Typos.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53504 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix PR2536: a nasty spiller bug. If a two-address instruction uses a register but...
Evan Cheng [Sat, 12 Jul 2008 01:56:02 +0000 (01:56 +0000)]
Fix PR2536: a nasty spiller bug. If a two-address instruction uses a register but the use portion of its live range is not part of its liveinterval, it must be defined by an implicit_def. In that case, do not spill the use. e.g.
8   %reg1024<def> = IMPLICIT_DEF
12  %reg1024<def> = INSERT_SUBREG %reg1024<kill>, %reg1025, 2

The live range [12, 14) are not part of the r1024 live interval since it's defined by an implicit def. It will not conflicts with live interval of r1025. Now suppose both registers are spilled, you can easily see a situation where both registers are reloaded before the INSERT_SUBREG and both target registers that would overlap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53503 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoBack out 53476 and 53480 for now. Somehow they cause llc to miscompile 179.art.
Evan Cheng [Sat, 12 Jul 2008 01:38:51 +0000 (01:38 +0000)]
Back out 53476 and 53480 for now. Somehow they cause llc to miscompile 179.art.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53502 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAdd a utility function to MachineInstr for testing whether an instruction
Dan Gohman [Sat, 12 Jul 2008 00:10:52 +0000 (00:10 +0000)]
Add a utility function to MachineInstr for testing whether an instruction
has exactly one MachineMemOperand, and change some X86 lowering code to
make use of it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53498 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix spelling of "hierarchy" in comments.
Dan Gohman [Fri, 11 Jul 2008 22:51:32 +0000 (22:51 +0000)]
Fix spelling of "hierarchy" in comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53489 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix typos in comments.
Dan Gohman [Fri, 11 Jul 2008 22:48:20 +0000 (22:48 +0000)]
Fix typos in comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53487 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoInclude a frame index in the "fixed stack" pseudo source value
Dan Gohman [Fri, 11 Jul 2008 22:44:52 +0000 (22:44 +0000)]
Include a frame index in the "fixed stack" pseudo source value
instead of using the frame index for the SVOffset, which was
inconsistent.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53486 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoMinor tweaks to the ImmutableList iterator interface.
Ted Kremenek [Fri, 11 Jul 2008 22:43:07 +0000 (22:43 +0000)]
Minor tweaks to the ImmutableList iterator interface.
Added partial specialization of DenseMapInfo<T> for ImmutableList.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53485 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoPorting r53478 into mainline:
Bill Wendling [Fri, 11 Jul 2008 22:42:10 +0000 (22:42 +0000)]
Porting r53478 into mainline:

Update to build_llvm. Don't output a floating point number for the version.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53484 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFix an obsolete top-level comment.
Dan Gohman [Fri, 11 Jul 2008 22:39:58 +0000 (22:39 +0000)]
Fix an obsolete top-level comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53481 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoFactor out debugging code into the common base class.
Dan Gohman [Fri, 11 Jul 2008 22:36:22 +0000 (22:36 +0000)]
Factor out debugging code into the common base class.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53480 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAdd support for putting NamedRegionTimers in TimerGroups, and
Dan Gohman [Fri, 11 Jul 2008 21:54:34 +0000 (21:54 +0000)]
Add support for putting NamedRegionTimers in TimerGroups, and
use a timer group for the timers in SelectionDAGISel. Also,
Split scheduling out from emitting, to give each their own
timer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53476 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoUse find instead of lower_bound.
Dan Gohman [Fri, 11 Jul 2008 20:58:19 +0000 (20:58 +0000)]
Use find instead of lower_bound.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53474 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoadd support for returning i128, PR2532.
Chris Lattner [Fri, 11 Jul 2008 20:53:00 +0000 (20:53 +0000)]
add support for returning i128, PR2532.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53472 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoTrim unnecessary #includes.
Dan Gohman [Fri, 11 Jul 2008 20:38:31 +0000 (20:38 +0000)]
Trim unnecessary #includes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53471 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoDon't call lookupNumber more than we have to.
Owen Anderson [Fri, 11 Jul 2008 20:05:13 +0000 (20:05 +0000)]
Don't call lookupNumber more than we have to.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53470 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoRemove an apparently useless routine: there should
Duncan Sands [Fri, 11 Jul 2008 17:02:09 +0000 (17:02 +0000)]
Remove an apparently useless routine: there should
be no need to split the result of a vector RET node,
since they are always already legal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53462 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoIt is pointless to turn a UINT_TO_FP into an
Duncan Sands [Fri, 11 Jul 2008 17:00:14 +0000 (17:00 +0000)]
It is pointless to turn a UINT_TO_FP into an
SINT_TO_FP libcall plus additional operations:
it might as well be a direct UINT_TO_FP libcall.
So only turn it into an SINT_TO_FP if the target
has special handling for SINT_TO_FP.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53461 91177308-0d34-0410-b5e6-96231b3b80d8

15 years agoAdd two missing SINT_TO_FP libcalls.
Duncan Sands [Fri, 11 Jul 2008 16:57:02 +0000 (16:57 +0000)]
Add two missing SINT_TO_FP libcalls.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53460 91177308-0d34-0410-b5e6-96231b3b80d8