Evan Cheng [Tue, 11 Aug 2009 08:52:18 +0000 (08:52 +0000)]
Fix Thumb2 load / store addressing mode matching code. Do not use so_reg form to
match base only address, i.e. [r] since Thumb2 requires a offset register field.
For those, use [r + imm12] where the immediate is zero.
Note the generated assembly code does not look any different after the patch.
But the bug would have broken the JIT (if there is Thumb2 support) and it can
break later passes which expect the address mode to be well-formed.
Rebuild RegScavenger::DistanceMap each time it is needed.
The register scavenger maintains a DistanceMap that maps MI pointers to their
distance from the top of the current MBB. The DistanceMap is built
incrementally in forward() and in bulk in findFirstUse(). It is used by
scavengeRegister() to determine which candidate register has the longest
unused interval.
Unfortunately the DistanceMap contents can become outdated. The first time
scavengeRegister() is called, the DistanceMap is filled to cover the MBB. If
then instructions are inserted in the MBB (as they always are following
scavengeRegister()), the recorded distances are too short. This causes bad
behaviour in the included test case where a register use /after/ the current
position is ignored because findFirstUse() thinks is is /before/ the current
position. A "using an undefined register" assertion follows promptly.
The fix is to build a fresh DistanceMap at the top of scavengeRegister(), and
discard it after use. This means that DistanceMap is no longer needed as a
RegScavenger member variable, and forward() doesn't need to update it.
The fix then discloses issue number two in the same test case: The candidate
search in scavengeRegister() finds a CSR that has been saved in the prologue,
but is currently unused. It would be both inefficient and wrong to spill such
a register in the emergency spill slot. In the present case, the emergency
slot restore is placed immediately before the normal epilogue restore, leading
to a "Redefining a live register" assertion.
Fix number two: When scavengerRegister() stumbles upon an unused register that
is overwritten later in the MBB, return that register early. It is important
to verify that the register is defined later in the MBB, otherwise it might be
an unspilled CSR.
Bob Wilson [Tue, 11 Aug 2009 05:39:44 +0000 (05:39 +0000)]
Use vAny type to get rid of Neon intrinsics that differed only in whether
the overloaded vector types allowed floating-point or integer vector elements.
Most of these operations actually depend on the element type, so bitcasting
was not an option.
If you include the vpadd intrinsics that I updated earlier, this gets rid
of 20 intrinsics.
Daniel Dunbar [Tue, 11 Aug 2009 03:42:33 +0000 (03:42 +0000)]
llvm-mc: Fix darwin .section parsing. It was skipping the section name and a ','
(and outputting a diagnostic pointing at the wrong place), all of which lead to
much confusion.
Daniel Dunbar [Tue, 11 Aug 2009 02:59:53 +0000 (02:59 +0000)]
llvm-mc/AsmParser: Implement automatic classification of RegisterClass operands.
- This drops us to 123 ambiguous instructions (previously ~500) on X86.
Sean Callanan [Tue, 11 Aug 2009 01:09:06 +0000 (01:09 +0000)]
Added the x86 INT instructions; both the special-case INT 3 and the general-case
INT i8. These instructions are only for interpretation by disassemblers, not
for emission, so they do not as yet have patterns.
Gabor Greif [Tue, 11 Aug 2009 00:59:39 +0000 (00:59 +0000)]
Lay the groundwork for my upcoming ilist sentinel shrinking patch
by defining a LLVM_COMPACT_SENTINELS symbol to 0 or 1 in config.h.
I'm asking for 3 favors:
- may an autofoo expert look at this and suggest improvements?
- may a cmake expert suggest analogous functionality for config.h?
- may somebody with the right autofoo mix regenerate configure? (mine is too new)
Thanks!
--This line, and those below, will be ignored--
Jim Grosbach [Tue, 11 Aug 2009 00:09:57 +0000 (00:09 +0000)]
SjLj based exception handling unwinding support. This patch is nasty, brutish
and short. Well, it's kinda short. Definitely nasty and brutish.
The front-end generates the register/unregister calls into the SjLj runtime,
call-site indices and landing pad dispatch. The back end fills in the LSDA
with the call-site information provided by the front end. Catch blocks are
not yet implemented.
Built on Darwin and verified no llvm-core "make check" regressions.
Lang Hames [Mon, 10 Aug 2009 23:43:28 +0000 (23:43 +0000)]
Modified VNInfo. The "copy" member is now a union which holds the copy for a register interval, or the defining register for a stack interval. Access is via getCopy/setCopy and getReg/setReg.
Dan Gohman [Mon, 10 Aug 2009 23:43:19 +0000 (23:43 +0000)]
Fix a bug in the DAGCombiner's handling of multiple linked
MERGE_VALUES nodes. Replacing the result values with the
operands in one MERGE_VALUES node may cause another
MERGE_VALUES node be CSE'd with the first one, and bring
its uses along, so that the first one isn't dead, as this
code expects. Fix this by iterating until the node is
really dead. This fixes PR4699.
Daniel Dunbar [Mon, 10 Aug 2009 21:00:45 +0000 (21:00 +0000)]
llvm-mc/AsmParser: Allow .td users to redefine the names of the methods to call
on target specific operands for testing class membership and converting to
MCInst operands.
Chris Lattner [Mon, 10 Aug 2009 17:58:51 +0000 (17:58 +0000)]
Fix a weird ppc64-specific link error during an llvm-gcc build:
ld: bad offset (0x00000091) for lo14 instruction pic-base fix-up in ___popcountdi2 from libgcc/./_popcountsi2_s.o
The problem is that the non lazy symbol pointers need to be 8 byte aligned
on ppc64 and .section doesn't have an implicit alignment like ".non_lazy_symbol_pointer"
does.
David Goodwin [Mon, 10 Aug 2009 15:55:25 +0000 (15:55 +0000)]
Post RA scheduler changes. Introduce a hazard recognizer that uses the target schedule information to accurately model the pipeline. Update the scheduler to correctly handle multi-issue targets.
Chris Lattner [Mon, 10 Aug 2009 01:39:42 +0000 (01:39 +0000)]
Make the big switch: Change MCSectionMachO to represent a section *semantically*
instead of syntactically as a string. This means that it keeps track of the
segment, section, flags, etc directly and asmprints them in the right format.
This also includes parsing and validation support for llvm-mc and
"attribute(section)", so we should now start getting errors about invalid
section attributes from the compiler instead of the assembler on darwin.
Still todo:
1) Uniquing of darwin mcsections
2) Move all the Darwin stuff out to MCSectionMachO.[cpp|h]
3) there are a few FIXMEs, for example what is the syntax to get the
S_GB_ZEROFILL segment type?
Daniel Dunbar [Sun, 9 Aug 2009 08:19:00 +0000 (08:19 +0000)]
llvm-mc/AsmParser: Add hack to ignore Int_* and *_Int instructions for now, to
make it easier to see interesting ambiguities.
- Also, check that user doesn't try to redefine the super class. This is a wart
in the current design, in that assembler match classes aren't explicitly
declared somewhere (so there isn't a unique place to declare the super
class). This should probably be fixed.
Daniel Dunbar [Sun, 9 Aug 2009 06:05:33 +0000 (06:05 +0000)]
llvm-mc/AsmParser: Separate instruction ordering for ambiguity detection.
- We want the ordering operation to be simple, since we run it on every
match. The old ordering is also not a strict weak ordering when there are
ambiguities, which makes MSVC unhappy.
- While we are at it, detect all ambiguities instead of just the adjacent
ones. There are actually 655, for X86.
projects / llvm / log