RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-26 22:42:46 +02:00
Code Issues Projects Releases Wiki Activity
66,226 Commits 4 Branches 5 Tags 704 MiB
7ee0f678cb
Commit Graph

42619 Commits

Author SHA1 Message Date
Andrew Trick
7a1dadd47d This is a prototype of an experimental register allocation 
framework. It's purpose is not to improve register allocation per se,
but to make it easier to develop powerful live range splitting. I call
it the basic allocator because it is as simple as a global allocator
can be but provides the building blocks for sophisticated register
allocation with live range splitting. 

A minimal implementation is provided that trivially spills whenever it
runs out of registers. I'm checking in now to get high-level design
and style feedback. I've only done minimal testing. The next step is
implementing a "greedy" allocation algorithm that does some register
reassignment and makes better splitting decisions.

llvm-svn: 117174
 2010-10-22 23:09:15 +00:00
Jakob Stoklund Olesen
c78f761faf Add more verification of LiveIntervals. 
llvm-svn: 117170
 2010-10-22 22:48:58 +00:00
Jakob Stoklund Olesen
68258f467a Be more strict about detecting multi-use blocks for isolation. 
When a block has exactly two uses and the register is both live-in and live-out,
don't isolate the block. We would be inserting two copies, so we haven't really
made any progress.

If the live-in and live-out values separate into disconnected components after
splitting, we would be making progress. We can't detect that for now.

llvm-svn: 117169
 2010-10-22 22:48:56 +00:00
Jim Grosbach
bba77cca02 ARM mode encoding information for CLZ, RBIT, REV*, and PKH*. 
llvm-svn: 117165
 2010-10-22 22:12:16 +00:00
Evan Cheng
2c2dcc69d8 Unbreak build. 
llvm-svn: 117155
 2010-10-22 21:49:09 +00:00
Bob Wilson
0290dbe7d4 Teach instcombine to set the alignment arguments for NEON load/store intrinsics. 
llvm-svn: 117154
 2010-10-22 21:41:48 +00:00
Evan Cheng
20b70697bb Transfer implicit ops when forming load multiple and return instructions. 
llvm-svn: 117151
 2010-10-22 21:29:58 +00:00
Jakob Stoklund Olesen
de1439ce2b Be more strict when detecting critical edges before loop splitting. 
An exit block with a critical edge must only have predecessors in the loop, or
just before the loop. This guarantees that the inserted copies in the loop
predecessors dominate the exit block.

llvm-svn: 117144
 2010-10-22 20:28:23 +00:00
Jakob Stoklund Olesen
967bcbf445 Add print methods 
llvm-svn: 117143
 2010-10-22 20:28:21 +00:00
Owen Anderson
424434414e Add correct NEON encodings for vqdmlal. 
llvm-svn: 117134
 2010-10-22 19:35:48 +00:00
Jim Grosbach
10195e9b73 Add the encoding information for the rest of the ARM mode multiply instructions. 
llvm-svn: 117133
 2010-10-22 19:15:30 +00:00
Owen Anderson
2bbdc62e17 Provide correct encodings for NEON vmlal. 
llvm-svn: 117131
 2010-10-22 19:05:25 +00:00
Evan Cheng
e5b7a3a09f Silence compiler warnings. 
llvm-svn: 117128
 2010-10-22 18:57:05 +00:00
Owen Anderson
ada2b33321 Provide correct NEON encodings for vmla. 
llvm-svn: 117126
 2010-10-22 18:54:37 +00:00
Jim Grosbach
ca2897a0ca More ARM multiply instuction binary encodings. 
llvm-svn: 117121
 2010-10-22 18:35:16 +00:00
Evan Cheng
efac5b5f8d Add fastcc cc: pass and return VFP / NEON values in registers. Controlled by -arm-fastcc for now. 
llvm-svn: 117119
 2010-10-22 18:23:05 +00:00
Jim Grosbach
eabb7c5bd1 Parameterize a bit of ARM encoding information, simplifying some instruction 
definitions.

llvm-svn: 117114
 2010-10-22 17:42:06 +00:00
Benjamin Kramer
9e988fe81e Make some symbols static, move classes into anonymous namespaces. 
llvm-svn: 117111
 2010-10-22 17:35:07 +00:00
Jim Grosbach
0ae6bf8095 More ARM multiply instruction encoding information. 
llvm-svn: 117108
 2010-10-22 17:16:17 +00:00
Wesley Peck
488027efa3 Making the e_machine configurable by the target backend in ELFObjectWriter. 
llvm-svn: 117099
 2010-10-22 15:52:49 +00:00
Andrew Trick
b9e8700beb Reverting r117031 to cleanup valgrind errors. 
It doesn't look like anything is wrong with the checkin,
but the new test cases expose a mem bug in AsmParser.

llvm-svn: 117087
 2010-10-22 03:58:29 +00:00
Eric Christopher
58a6e62c08 Add some basic ret instruction support to arm fast-isel. 
llvm-svn: 117085
 2010-10-22 01:28:00 +00:00
Sean Callanan
9a40554de4 Fixed handling of immediate operand sizes, which 
weren't properly reflecting the OperandSize attribute
of the instruction leading to improper decoding of
certain instructions with the 66H prefix.  Also added
a test case for this.

llvm-svn: 117084
 2010-10-22 01:24:11 +00:00
NAKAMURA Takumi
76a048b6e7 Win32/Signals.inc: DontRemoveFileOnSignal(): Please acquire the CriticalSection. 
It choked BugPoint on Mingw.

llvm-svn: 117083
 2010-10-22 01:23:50 +00:00
Jim Grosbach
d86073aa9a ARM binary encoding for some of the multiply instructions. 
llvm-svn: 117080
 2010-10-21 22:52:30 +00:00
Jim Grosbach
7c42142b71 ARM binary encodings for MVN variants. 
llvm-svn: 117076
 2010-10-21 22:19:32 +00:00
Jim Grosbach
e6d61a87e4 Trailing whitespace. 
llvm-svn: 117073
 2010-10-21 22:04:05 +00:00
Jim Grosbach
cb41e97b5c ARM Binary encoding information for BFC/BFI instructions. 
llvm-svn: 117072
 2010-10-21 22:03:21 +00:00
Dan Gohman
fe5489a0ac Add some more documentation. 
llvm-svn: 117070
 2010-10-21 21:55:35 +00:00
Eric Christopher
59683c9d71 These don't need to be virtual. 
llvm-svn: 117068
 2010-10-21 21:47:51 +00:00
Michael J. Spencer
a43c204ef7 Fix Warnings. 
llvm-svn: 117062
 2010-10-21 20:49:38 +00:00
Michael J. Spencer
08602f9803 X86: Base _fltused on the FunctionType of the called value instead of the potentially null "CalledFunction". Thanks Duncan! 
This is needed for indirect calls.

llvm-svn: 117061
 2010-10-21 20:49:23 +00:00
Owen Anderson
f84632744e Move the encoding logic for Q registers into getMachineOpValue(). 
llvm-svn: 117060
 2010-10-21 20:49:13 +00:00
Mikhail Glushenkov
55ada6816e Trailing whitespace. 
llvm-svn: 117058
 2010-10-21 20:40:39 +00:00
Owen Anderson
ba2ac80921 ARM encodes Q registers as 2xregno (i.e. the number of the D register that corresponds to the lower 
half of the Q register), rather than with just regno.  This allows us to unify the encodings for
a lot of different NEON instrucitons that differ only in whether they have Q or D register operands.

llvm-svn: 117056
 2010-10-21 20:21:49 +00:00
Eric Christopher
7ded7e88c8 Handle storing args to the stack for calls. 
llvm-svn: 117055
 2010-10-21 20:09:54 +00:00
Wesley Peck
8a77999240 Adding initial AsmParser implementation for the MBlaze backend. It is 
mostly based on the ARM AsmParser at this time and is not particularly
functional.

Changed the MBlaze data layout from:
    "E-p:32:32-i8:8:8-i16:16:16-i64:32:32-f64:32:32-v64:32:32-v128:32:32-n32"
to:
    "E-p:32:32:32-i8:8:8-i16:16:16"
because the MicroBlaze doesn't have i64, f64, v64, or v128 data types.

Cleaned up the MBlaze source code:
    1. The floating point register class has been removed. The
       MicroBlaze does not have floating point registers. Floating
       point values are simply stored in integer registers.
    2. Renaming the CPURegs register class to GPR to reflect the
       standard naming.
    3. Removing a lot of stale code from AsmPrinter after
       the conversion to InstPrinter.
    4. Simplified sign extended loads by marking them as
       expanded in ISelLowering.

llvm-svn: 117054
 2010-10-21 19:48:38 +00:00
Dan Gohman
6fb125afc8 Explain what "constant" means here. 
llvm-svn: 117053
 2010-10-21 19:45:09 +00:00
Eric Christopher
b15029ea7e More load/store refactoring, call reg+offset simplification from within 
the emitter to handle the addresses. Only simplify the offset if we need
to - also fix bug where in addrmode 5 we weren't dividing the offset by
4, which showed up due to not always lowering.

llvm-svn: 117051
 2010-10-21 19:40:30 +00:00
Jim Grosbach
f35d3b2aa1 trailing whitespace 
llvm-svn: 117050
 2010-10-21 19:38:40 +00:00
Dan Gohman
8cb659e892 Update comments. 
llvm-svn: 117048
 2010-10-21 19:01:22 +00:00
Owen Anderson
51a6bc3b27 Add correct NEON encodings for vhadd and vrhadd. 
llvm-svn: 117047
 2010-10-21 18:55:04 +00:00
Jakob Stoklund Olesen
4794d28044 Don't include the destination interval in the union when computing 
Parent - union(Y, ...). Doh.

llvm-svn: 117042
 2010-10-21 18:47:08 +00:00
Jakob Stoklund Olesen
82da267156 Permit landing pad successor blocks when verifying basic blocks that end in an 
unconditional branch.

llvm-svn: 117041
 2010-10-21 18:47:06 +00:00
Owen Anderson
dce283c7db Add correct encodings for NEON vaddw.s* and vaddw.u*. 
llvm-svn: 117040
 2010-10-21 18:20:25 +00:00
Owen Anderson
7d90c72edf Provide correct NEON encodings for vaddl.u* and vaddl.s*. 
llvm-svn: 117039
 2010-10-21 18:09:17 +00:00
Rafael Espindola
69c973fbfd Do not recurse into symbol refs that have a variant kind. This prevents us 
from losing the variant when producing a relocation on an alias.

llvm-svn: 117037
 2010-10-21 18:00:20 +00:00
Kevin Enderby
c63105a064 More tweaks to X86 instructions to allow the 'w' suffix in places it makes 
sense, when the instruction takes the 16-bit ax register or m16 memory
location.  These changes to llvm-mc matches what the darwin assembler allows
for these instructions.  Also added the missing flex (without the wait prefix)
and ud2a as an alias to ud2 (still to add ud2b).

llvm-svn: 117031
 2010-10-21 17:16:46 +00:00
Duncan Sands
669ddd943a AlignLoc is never used for anything - zap it (gcc-4.6 warning). 
llvm-svn: 117025
 2010-10-21 16:07:10 +00:00
Duncan Sands
b458f1dc85 The return value of this call is not used, so no point 
in assigning it to a variable (gcc-4.6 warning).

llvm-svn: 117024
 2010-10-21 16:06:28 +00:00
Duncan Sands
f7cfeda351 RetOp is not actually used for anything useful (though 
it looks like maybe it was supposed to be used in the
test...), so zap it (gcc-4.6 warning).

llvm-svn: 117023
 2010-10-21 16:05:44 +00:00
Duncan Sands
c40461600a The variable liTRC is not used for anything useful, zap it 
(gcc-4.6 warning).

llvm-svn: 117022
 2010-10-21 16:04:43 +00:00
Duncan Sands
0c166168a0 Remove some variables that are never really used 
(gcc-4.6 warns about these).

llvm-svn: 117021
 2010-10-21 16:03:28 +00:00
Duncan Sands
8c5f243aa0 Add parentheses to pacify gcc, which warns otherwise. 
llvm-svn: 117020
 2010-10-21 16:02:12 +00:00
Duncan Sands
acbc34528c Fix the cleanup process of exception information in JIT. Now JIT 
deregisters registered by it FDE structures allowing consecutive
JIT runs to succeed.  Patch by Yuri.  Fixes PR8285.

llvm-svn: 117004
 2010-10-21 08:57:29 +00:00
Wesley Peck
b100615cc6 Removing stale AsmPrinter directory from MicroBlaze backend. 
llvm-svn: 116998
 2010-10-21 05:05:06 +00:00
Oscar Fuentes
00ad8b9b29 Deleted lib/Target/MBlaze/AsmPrinter/CMakeLists.txt. This way the 
CMake build does not try to build that library, which collides with
MBlaze/InstPrinter.

llvm-svn: 116997
 2010-10-21 05:01:26 +00:00
Wesley Peck
d646a12f25 Recommit 116986 with capitalization typo fixed. 
llvm-svn: 116993
 2010-10-21 03:57:26 +00:00
Andrew Trick
4a3b819c1f putback r116983 and fix simple-fp-encoding.ll tests 
llvm-svn: 116992
 2010-10-21 03:40:16 +00:00
Wesley Peck
3478e641b9 Reverting the commit 116986. It was breaking the build on llvm-x86_64-linux though it 
compiles on OS X. I'll ensure that it builds on a linux machine before committing
again.

llvm-svn: 116991
 2010-10-21 03:34:22 +00:00
Owen Anderson
7da515c665 Revert r116983, which is breaking all the buildbots. 
llvm-svn: 116987
 2010-10-21 03:11:16 +00:00
Wesley Peck
c50372298d Major update of the MicroBlaze backend. The new features are: 
1. A delay slot filler that searches for valid instructions
       to fill the delay slot with. Previously NOPs would always
       be inserted into delay slots.
    2. Support for MC based instruction printer added.
    3. Support for MC based machine code generation and ELF
       file generation. ELF file generation does not yet
       completely work as much of the ELF support infrastructure
       is still x86/x86-64 specific.
    4. General clean up of the MBlaze backend code. Much of the
       tablegen code has been cleanup and simplified.

Bug Fixes:
    1. Removed duplicate periods from subtarget feature descriptions.
    2. Many of the instructions had bad machine code information
       in the tablegen files. Much of this has been fixed.

llvm-svn: 116986
 2010-10-21 03:09:55 +00:00
Michael J. Spencer
5a68d7ce94 X86: Add alloca probing to dynamic alloca on Windows. Fixes PR8424. 
llvm-svn: 116984
 2010-10-21 01:41:01 +00:00
Evan Cheng
0b9eaaf45d Add missing scheduling itineraries for transfers between core registers and VFP registers. 
llvm-svn: 116983
 2010-10-21 01:12:00 +00:00
Owen Anderson
a685f8e90a Implement correct encodings for NEON vadd, both integer and floating point. 
llvm-svn: 116981
 2010-10-21 00:48:00 +00:00
Michael J. Spencer
b9cffadc06 CodeGen-Windows: Only emit _fltused if a VarArg function is called with floating point args. 
This should be the minimum set of functions that could possibly need it.

llvm-svn: 116978
 2010-10-21 00:08:21 +00:00
Eric Christopher
7989939d87 Custom lower f64 args passed in integer registers. 
llvm-svn: 116977
 2010-10-21 00:01:47 +00:00
Michael J. Spencer
54b462089f Fix Whitespace. 
llvm-svn: 116972
 2010-10-20 23:40:27 +00:00
Bill Wendling
c7ef66fcf2 Add encoding for moving a value between two ARM core registers and a doublework 
extension register.

llvm-svn: 116970
 2010-10-20 23:37:40 +00:00
Jakob Stoklund Olesen
03fd8bf61c Remember to keep track of rematted values. 
llvm-svn: 116962
 2010-10-20 22:50:42 +00:00
Bill Wendling
0f96ff63b3 Add encodings for movement between ARM core registers and single-precision 
registers.

llvm-svn: 116961
 2010-10-20 22:44:54 +00:00
Dan Gohman
369ce9c96c Memdep says that an instruction clobbers itself 
when it means there is no specific clobber instruction.

llvm-svn: 116960
 2010-10-20 22:37:41 +00:00
Owen Anderson
9e9ae1b048 Attempt to fix valgrind complaining about (benign) leaks in pass registration by having PassRegistry 
delete PassInfo objects that were created using new-style initialization.

llvm-svn: 116959
 2010-10-20 22:22:30 +00:00
Dan Gohman
3376c4f08d Factor out the main aliasing check into a separate function. 
llvm-svn: 116958
 2010-10-20 22:11:14 +00:00
Evan Cheng
9a06e4c7c8 More accurate estimate / tracking of register pressure. 
- Initial register pressure in the loop should be all the live defs into the
  loop. Not just those from loop preheader which is often empty.
- When an instruction is hoisted, update register pressure from loop preheader
  to the original BB.
- Treat only use of a virtual register as kill since the code is still SSA.

llvm-svn: 116956
 2010-10-20 22:03:58 +00:00
Dale Johannesen
a324c8c6bd Fix crash introduced in 116852. 8573915. 
llvm-svn: 116955
 2010-10-20 22:03:37 +00:00
Dan Gohman
599a664758 Fix comments; the type graph is currently a tree, not a DAG. 
llvm-svn: 116954
 2010-10-20 22:02:58 +00:00
Jakob Stoklund Olesen
9121d7b293 Move some of the InlineSpiller rematerialization code into LiveRangeEdit. 
llvm-svn: 116951
 2010-10-20 22:00:51 +00:00
Dale Johannesen
58fe3193a6 Remove Synthesizable from the Type system; as MMX vector 
types are no longer Legal on X86, we don't need it.
No functional change.  8499854.

llvm-svn: 116947
 2010-10-20 21:32:10 +00:00
Jakob Stoklund Olesen
c04876b607 When SimpleRegisterCoalescing is trimming kill flags on a physical register 
operand, also check if subregisters are killed.

Add <imp-def> operands for subregisters that remain alive after a super register
is killed.

I don't have a testcase for this that reproduces on trunk. <rdar://problem/8441758>

llvm-svn: 116940
 2010-10-20 18:45:55 +00:00
Rafael Espindola
0a50bbaee1 Handle _GLOBAL_OFFSET_TABLE_ correctly. 
llvm-svn: 116932
 2010-10-20 16:46:08 +00:00
Michael J. Spencer
fd40803107 Use C++03... 
llvm-svn: 116927
 2010-10-20 16:00:45 +00:00
Michael J. Spencer
2da558d210 System-Win32/Path: Fix incorrect assumption in isValid. 
A recent commit to clang exposed a bug in the Win32 Path code. This is a
minimal fix for it.

llvm-svn: 116925
 2010-10-20 15:23:58 +00:00
Chandler Carruth
55850cad48 Remove remaining uses of ATTRIBUTE_UNUSED on variables, and delete three 
#includes in the process.

llvm-svn: 116919
 2010-10-20 08:27:02 +00:00
Eric Christopher
78c7e84ce5 Fix a TODO by removing some unnecesary copies. 
llvm-svn: 116915
 2010-10-20 08:02:24 +00:00
NAKAMURA Takumi
0f54a81d63 Add ATTRIBUTE_UNUSED for -Asserts. 
llvm-svn: 116909
 2010-10-20 04:05:29 +00:00
Tobias Grosser
88ce93b0eb Add RegionPass support. 
A RegionPass is executed like a LoopPass but on the regions detected by the
RegionInfo pass instead of the loops detected by the LoopInfo pass.

llvm-svn: 116905
 2010-10-20 01:54:44 +00:00
Douglas Gregor
1e64599cb3 Fix CMake build 
llvm-svn: 116903
 2010-10-20 01:36:56 +00:00
Jim Grosbach
ebb4bc2c05 Fix backwards conditional. 
llvm-svn: 116897
 2010-10-20 01:10:01 +00:00
Dan Gohman
c781a28a1d Make CodeGen TBAA-aware. 
llvm-svn: 116890
 2010-10-20 00:31:05 +00:00
Jim Grosbach
c77e19ac99 Add dynamic realignment when rematerializing the base register. 
llvm-svn: 116886
 2010-10-20 00:02:50 +00:00
Jim Grosbach
5f33ecfc98 Nuke a commented out bit that got missed a while back. 
llvm-svn: 116883
 2010-10-19 23:48:47 +00:00
Jim Grosbach
a8c0be5343 Add a pre-dispatch SjLj EH hook on the unwind edge for targets to do any 
setup they require. Use this for ARM/Darwin to rematerialize the base
pointer from the frame pointer when required. rdar://8564268

llvm-svn: 116879
 2010-10-19 23:27:08 +00:00
Dan Gohman
399ef0b44a Move NoAA out of BasicAliasAnalysis.cpp into its own file, now that 
it doesn't have a special relationship with BasicAliasAnalysis
anymore.

llvm-svn: 116876
 2010-10-19 23:09:08 +00:00
Dan Gohman
befe1767ff Reapply r116831 and r116839, converting AliasAnalysis to use 
uint64_t, plus fixes for places I missed before.

llvm-svn: 116875
 2010-10-19 22:54:46 +00:00
Douglas Gregor
a598c0fbce Extend StringRef's edit-distance algorithm to permit an upper bound on the allowed edit distance 
llvm-svn: 116867
 2010-10-19 22:13:48 +00:00
Jim Grosbach
d7916d7468 Update comments to remove obsolete references. 
llvm-svn: 116863
 2010-10-19 21:34:47 +00:00
Dan Gohman
de3c49438f Revert r116831 and r116839, which are breaking selfhost builds. 
llvm-svn: 116858
 2010-10-19 21:06:16 +00:00
Jakob Stoklund Olesen
02d7f65c49 Shrink MachineOperand from 40 to 32 bytes on 64-bit hosts. 
Pull an unsigned out of the Contents union such that it has the same size as two
pointers and no padding.

Arrange members such that the Contents union and all pointers can be 8-byte
aligned without padding.

This speeds up code generation by 0.8% on a 64-bit host. 32-bit hosts should be
unaffected.

llvm-svn: 116857
 2010-10-19 20:56:32 +00:00
Owen Anderson
deee1e18f1 Passes do not need to recursively initialize passes that they preserve, if 
they do not also require them.  This allows us to reduce inter-pass linkage
dependencies.

llvm-svn: 116854
 2010-10-19 20:08:44 +00:00
Dale Johannesen
ee87cbe4e9 Enable using vdup for vector constants which are splat of 
integers by default, and remove the controlling flag, now
that LICM will hoist such vdup's.  8003375.

llvm-svn: 116852
 2010-10-19 20:00:17 +00:00
Rafael Espindola
5f43aeb079 Small cleanups and fixes in preparation for fixing _GLOBAL_OFFSET_TABLE_. 
llvm-svn: 116848
 2010-10-19 19:31:37 +00:00
Evan Cheng
1c8dafd12a Re-enable register pressure aware machine licm with fixes. Hoist() may have 
erased the instruction during LICM so UpdateRegPressureAfter() should not
reference it afterwards.

llvm-svn: 116845
 2010-10-19 18:58:51 +00:00
Dan Gohman
37a9b813f5 Oops, check in all the files for converting AliasAnalysis to 
use uint64_t.

llvm-svn: 116839
 2010-10-19 18:08:27 +00:00
Dan Gohman
03f86df2e5 Change AliasAnalysis and its clients to use uint64_t instead of unsigned 
for representing object sizes, for consistency with other parts of LLVM.

llvm-svn: 116831
 2010-10-19 18:00:02 +00:00
Owen Anderson
46990c17f7 Get rid of static constructors for pass registration. Instead, every pass exposes an initializeMyPassFunction(), which 
must be called in the pass's constructor.  This function uses static dependency declarations to recursively initialize
the pass's dependencies.

Clients that only create passes through the createFooPass() APIs will require no changes.  Clients that want to use the
CommandLine options for passes will need to manually call the appropriate initialization functions in PassInitialization.h
before parsing commandline arguments.

I have tested this with all standard configurations of clang and llvm-gcc on Darwin.  It is possible that there are problems
with the static dependencies that will only be visible with non-standard options.  If you encounter any crash in pass
registration/creation, please send the testcase to me directly.

llvm-svn: 116820
 2010-10-19 17:21:58 +00:00
Daniel Dunbar
6ff550c84d Revert r116781 "- Add a hook for target to determine whether an instruction def 
is", which breaks some nightly tests.

llvm-svn: 116816
 2010-10-19 17:14:24 +00:00
Dan Gohman
6e3755905e Consistently use AliasAnalysis::UnknownSize instead of hardcoding ~0u. 
llvm-svn: 116815
 2010-10-19 17:06:23 +00:00
Mikhail Glushenkov
0c09a4b97f GlobalOpt: EvaluateFunction() must not evaluate stores to weak_odr globals. 
Fixes PR8389.

llvm-svn: 116812
 2010-10-19 16:47:23 +00:00
Che-Liang Chiou
bc4f00f42d Add lower argument and return of device function 
llvm-svn: 116805
 2010-10-19 13:14:40 +00:00
Bill Wendling
f937f0ed59 Fix spelling. 
llvm-svn: 116804
 2010-10-19 10:18:23 +00:00
Michael J. Spencer
e528f2588f X86: Add MS-CRT libcalls. 
llvm-svn: 116801
 2010-10-19 07:32:52 +00:00
Michael J. Spencer
cd6be63d05 Fix Whitespace. 
llvm-svn: 116800
 2010-10-19 07:32:42 +00:00
NAKAMURA Takumi
4a85646d6c lib/CodeGen/TargetLoweringObjectFileImpl.cpp: Tweak to emit ".{section}${name}" instead of ".{section}$linkonce_{name}" for linkonce sections. 
It seems GNU ld/PECOFF relies on section names, linking with g++'s libstdc++.a would fail.

llvm-svn: 116791
 2010-10-19 03:24:42 +00:00
Andrew Trick
3df2830391 Fix for machine licm assert: RCCost <= RegPressure[RCId] 
in MultiSource/Benchmarks/VersaBench/beamformer/beamformer.
SmallSet.insert returns true if the element is inserted.

llvm-svn: 116790
 2010-10-19 02:50:50 +00:00
Rafael Espindola
2ba55832f3 Fix PR8300 by remembering to keep the bitcast in all cases. 
llvm-svn: 116788
 2010-10-19 02:02:57 +00:00
NAKAMURA Takumi
1c62027520 lib/System/Win32/ThreadLocal.inc: Suppress "unused" warning on -Asserts. 
llvm-svn: 116785
 2010-10-19 01:22:01 +00:00
NAKAMURA Takumi
8c67d3f400 lib/Support/raw_ostream.cpp: Fix Cygwin's build. 
setmode is provided by io.h on Cygwin.

llvm-svn: 116784
 2010-10-19 01:21:55 +00:00
Evan Cheng
9c3f6f486e - Add a hook for target to determine whether an instruction def is 
"long latency" enough to hoist even if it may increase spilling. Reloading
  a value from spill slot is often cheaper than performing an expensive
  computation in the loop. For X86, that means machine LICM will hoist
  SQRT, DIV, etc. ARM will be somewhat aggressive with VFP and NEON
  instructions.
- Enable register pressure aware machine LICM by default.

llvm-svn: 116781
 2010-10-19 00:55:07 +00:00
Chandler Carruth
755dd8000d Move the definition of this to the source file to anchor the vtable. 
llvm-svn: 116779
 2010-10-19 00:37:30 +00:00
Bob Wilson
c3fb4427f4 Support alignment for NEON vld-lane and vst-lane instructions. 
llvm-svn: 116776
 2010-10-19 00:16:32 +00:00
Kevin Enderby
8a3030c6af Added a few tweaks to the Intel Descriptor-table support instructions to allow 
word forms and suffixed versions to match the darwin assembler in 32-bit and
64-bit modes.  This is again for use just with assembly source for llvm-mc .

llvm-svn: 116773
 2010-10-19 00:01:44 +00:00
Jim Grosbach
10fd33ecfc ARM encoding information for [SU]SAT* instructions. 
llvm-svn: 116768
 2010-10-18 23:35:38 +00:00
Dan Gohman
8b332ffb8c Make the representation of AliasSets explicitly differentiate 
between "not known yet" and "known no tbaa info" so that it
can merge them properly.

llvm-svn: 116767
 2010-10-18 23:31:47 +00:00
Eric Christopher
5850afe5f2 Revert r116220 - thus turning arm fast isel back on by default. 
llvm-svn: 116762
 2010-10-18 22:53:53 +00:00
Daniel Dunbar
f5c3546f2b CrashRecoveryContext: Add missing return, so that the signal fires after we our 
routine is off the stack. Otherwise we show up rather confusingly in the stack
trace.

llvm-svn: 116755
 2010-10-18 21:55:18 +00:00
Dan Gohman
6efd04961b Don't pass the raw invalid pointer used to represent conflicting 
TBAA information to AliasAnalysis.

llvm-svn: 116751
 2010-10-18 21:28:00 +00:00
Bill Wendling
3b3d9aaa86 Don't recompute MachineRegisterInfo in the Optimize* method. 
llvm-svn: 116750
 2010-10-18 21:22:31 +00:00
Mikhail Glushenkov
48553997d2 Trailing whitespace. 
llvm-svn: 116749
 2010-10-18 21:16:00 +00:00
Rafael Espindola
9a98116bb7 Implement R_386_GOT32. 
llvm-svn: 116744
 2010-10-18 20:47:21 +00:00
Dan Gohman
64affb7be2 Make AliasSetTracker TBAA-aware, enabling TBAA-enabled LICM. 
llvm-svn: 116743
 2010-10-18 20:44:50 +00:00
Rafael Espindola
c75defe576 Relocate with .bss instead of using the symbol. Matches gas behavior. 
llvm-svn: 116741
 2010-10-18 20:25:33 +00:00
Rafael Espindola
6504490709 Reenable assert. 
llvm-svn: 116738
 2010-10-18 19:33:01 +00:00
Devang Patel
dbe2dbee5b Transfer debug loc to lowered call. 
Patch by Alexander Herz!

llvm-svn: 116733
 2010-10-18 18:53:44 +00:00
Dan Gohman
df5c887e86 Add TypeBasedAliasAnalysis to the standard pass lists. Note that it 
is currently inert by default.

llvm-svn: 116732
 2010-10-18 18:50:27 +00:00
Dan Gohman
35def103f7 Fix BasicAA to pass TBAAInfo through to the chained analysis. 
llvm-svn: 116730
 2010-10-18 18:45:11 +00:00
Jim Grosbach
ffb06eda57 For Thumb2, try to use frame pointer references for stack slots even when a 
base register is available. rdar://8525298

llvm-svn: 116729
 2010-10-18 18:39:46 +00:00
Rafael Espindola
7cc236c87f Produce ELF::R_386_GOTPC relocations. 
llvm-svn: 116728
 2010-10-18 18:36:12 +00:00
Dan Gohman
7820328076 Make TypeBasedAliasAnalysis default to doing nothing, with a command-line 
option to enable it.

llvm-svn: 116722
 2010-10-18 18:17:47 +00:00
Dan Gohman
97f3b0cac1 Use chaining in TypeBasedAliasAnalysis::pointsToConstantMemory. 
llvm-svn: 116721
 2010-10-18 18:10:31 +00:00
Dan Gohman
6aff5b94ff Make BasicAliasAnalysis a normal AliasAnalysis implementation which 
does normal initialization and normal chaining. Change the default
AliasAnalysis implementation to NoAlias.

Update StandardCompileOpts.h and friends to explicitly request
BasicAliasAnalysis.

Update tests to explicitly request -basicaa.

llvm-svn: 116720
 2010-10-18 18:04:47 +00:00
Rafael Espindola
d74116b1e8 Make the bots happy. 
llvm-svn: 116719
 2010-10-18 18:03:28 +00:00
Kevin Enderby
cb18d38b76 Added a handful of x86-32 instructions that were missing so that llvm-mc would 
be more complete.  These are only expected to be used by llvm-mc with assembly
source so there is no pattern, [], in the .td files.  Most are being added to
X86InstrInfo.td as Chris suggested and only comments about register uses are
added.  Suggestions welcome on the .td changes as I'm not sure on every detail
of the x86 records.  More missing instructions will be coming.

llvm-svn: 116716
 2010-10-18 17:04:36 +00:00
Rafael Espindola
bf9107e924 Produce a R_386_PLT32 when needed. Moved the default cases of switches to the 
start for consistency.

llvm-svn: 116715
 2010-10-18 16:58:03 +00:00
Jim Grosbach
3e41568604 ARM addrmode4 instructions (ldm, stm and friends) can't encode an immediate 
offset for stack references. Make sure we take that into account when
deciding whether to reserver an emergency spill slot for the register
scavenger. rdar://8559625

llvm-svn: 116714
 2010-10-18 16:48:59 +00:00
Jim Grosbach
f93e92a4c1 Grammar tweak. 
llvm-svn: 116712
 2010-10-18 16:38:50 +00:00
Rafael Espindola
4a7459403a Handle GOTOFF correctly on i386. 
llvm-svn: 116711
 2010-10-18 16:38:04 +00:00
Jim Grosbach
315bb0cf2b Trivial grammar tweak. 
llvm-svn: 116710
 2010-10-18 16:29:26 +00:00
Kalle Raiskila
3cdfdd9383 Improve lowering of sext to i128 on SPU. 
The old algorithm inserted a 'rotqmbyi' instruction which was
both redundant and wrong - it made shufb select bytes from the
wrong end of the input quad.

llvm-svn: 116701
 2010-10-18 09:34:19 +00:00
Eric Christopher
e453bd6697 Remove the check for invalid calling conventions. Testing shows that they're 
working just fine.

llvm-svn: 116698
 2010-10-18 06:49:12 +00:00
Eric Christopher
6287738105 Lift arg promotion from the X86 backend. This should be unified at some point. 
llvm-svn: 116694
 2010-10-18 02:17:53 +00:00
Eric Christopher
79c2206640 Now that we handle all allocas via a non-SP reg offset remove all of the 
special case handling for ARM::SP.

llvm-svn: 116688
 2010-10-17 11:08:44 +00:00
Benjamin Kramer
9248e59b35 std::string cleanup. 
llvm-svn: 116686
 2010-10-17 07:39:34 +00:00
Benjamin Kramer
e8c2e82a84 Fix a typo and silence unused variable warnings in -Asserts build. 
llvm-svn: 116685
 2010-10-17 07:38:40 +00:00
Eric Christopher
d99f79f92b Allow more load types to be materialized through the allocas. 
llvm-svn: 116683
 2010-10-17 06:07:26 +00:00
Oscar Fuentes
a6ffccab95 Build with RTTI and exceptions disabled. Only in GCC for now. 
llvm-svn: 116682
 2010-10-17 02:26:16 +00:00
Eric Christopher
5fcaab5d0e Optimize GEP off of intermediate allocas. 
llvm-svn: 116681
 2010-10-17 01:51:42 +00:00
Eric Christopher
20bbcbeceb Fix comment. 
llvm-svn: 116680
 2010-10-17 01:42:53 +00:00
Eric Christopher
d9b4b47b9a Turn on AddOperator folding in GEP. 
llvm-svn: 116679
 2010-10-17 01:41:46 +00:00
Eric Christopher
490f82bc32 Use the i12 immediate versions of the load instructions - they're handled 
more in the post-passes.

llvm-svn: 116678
 2010-10-17 01:40:27 +00:00
Rafael Espindola
be5c52d2dc Add a MCObjectFormat class so that code common to all targets that use a 
single object format can be shared.

This also adds support for

mov zed+(bar-foo), %eax

on ELF and COFF targets.

llvm-svn: 116675
 2010-10-16 18:23:53 +00:00
Benjamin Kramer
a0e156fe6d Eliminate some calls to Value::getNameStr. 
llvm-svn: 116670
 2010-10-16 11:28:23 +00:00
Michael J. Spencer
f9a7c39ecc MC-COFF: Add support for default-null weak externals. 
llvm-svn: 116666
 2010-10-16 08:25:57 +00:00
Michael J. Spencer
e57b670425 X86-Windows: Emit an undefined global __fltused symbol when targeting Windows 
if any floating point arguments are passed to an external function.

llvm-svn: 116665
 2010-10-16 08:25:41 +00:00
Michael J. Spencer
16ad2c129c Whitespace! 
llvm-svn: 116664
 2010-10-16 08:25:21 +00:00
Daniel Dunbar
fedc37ce15 MC/AsmParser: Report .stabs directive as unsupported. 
llvm-svn: 116659
 2010-10-16 04:56:42 +00:00
Evan Cheng
5bd7971922 More machine LICM work. It now tracks register pressure for path from preheader to current BB and use the information determine whether hoisting is worthwhile. 
llvm-svn: 116654
 2010-10-16 02:20:26 +00:00
Eric Christopher
98c6ce755a Fix some funky formatting that got through. 
llvm-svn: 116653
 2010-10-16 01:10:35 +00:00
Bill Wendling
8e4ceb45f2 ARMCodeEmitter::emitMiscInstruction is dead. Long live 
ARMCodeEmitter::emitMiscInstruction!

llvm-svn: 116644
 2010-10-15 23:35:12 +00:00
Eric Christopher
69eaf30a42 Make sure offset is 0 for load/store register to the stack call. 
llvm-svn: 116640
 2010-10-15 23:07:10 +00:00
Owen Anderson
4373b4516b Generalize MemCpyOpt's handling of call slot forwarding to function properly when the call slot 
forwarding is implemented with a load/store pair rather than a memcpy.

llvm-svn: 116637
 2010-10-15 22:52:12 +00:00
Eric Christopher
c6e4aaec0a Formatting. 
llvm-svn: 116635
 2010-10-15 22:49:28 +00:00
Eric Christopher
7f8c3dff07 Fix else if -> if in store machinery. 
llvm-svn: 116628
 2010-10-15 22:32:37 +00:00
Bill Wendling
64d2bf006c Reformatting. No functionalogicality changes. 
llvm-svn: 116625
 2010-10-15 21:50:45 +00:00
Eric Christopher
a875589580 Refactor ARM fast-isel reg + offset to be a base + offset. 
llvm-svn: 116622
 2010-10-15 21:32:12 +00:00
Jim Grosbach
5d7a5f8d33 Encoding information for the various ARM saturating add/sub instructions. 
llvm-svn: 116612
 2010-10-15 19:49:46 +00:00
Jim Grosbach
785858b2ba ARM binary encoding information for RSB and RSC instructions. 
llvm-svn: 116604
 2010-10-15 18:42:41 +00:00
Jim Grosbach
440b0e6b34 Don't mark argument value stores as immutable, as otherwise the post-RA 
scheduler may reorder loads from them before the stores and other such
badness. PR8347. Patch by David Meyer

llvm-svn: 116602
 2010-10-15 18:34:47 +00:00
Bob Wilson
3b63270ed3 Use simple RegState::Define flag instead of getDefRegState(true). 
llvm-svn: 116601
 2010-10-15 18:25:59 +00:00
Rafael Espindola
125360f835 Refactor alias handling to AliasedSymbol. 
llvm-svn: 116600
 2010-10-15 18:25:33 +00:00
Eric Christopher
b287ab82d7 Expand GEP handling for constant offsets. 
llvm-svn: 116594
 2010-10-15 18:02:07 +00:00
Jim Grosbach
cb52aee60e When expanding the MOVsr[la]_flag pseudos, the CPSR implicit def becomes 
an explicit def. Make sure to capture that properly. rdar://8556556

llvm-svn: 116591
 2010-10-15 17:35:17 +00:00
Jim Grosbach
67f94c42d8 ARM mode encoding information for UBFX and SBFX instructions. 
llvm-svn: 116588
 2010-10-15 17:15:16 +00:00
Jakob Stoklund Olesen
05c010aa88 Remove unused accessor. 
llvm-svn: 116580
 2010-10-15 16:06:40 +00:00
Rafael Espindola
9b114d966a Refactor code a bit and avoid creating unnecessary entries in the string 
map.

llvm-svn: 116579
 2010-10-15 15:39:06 +00:00
Bob Wilson
6b6b53ad6f Remove unused ARMISD::AND selection DAG node. 
llvm-svn: 116566
 2010-10-15 04:34:40 +00:00
Bob Wilson
fcc42f2f3a ARM instructions that are both predicated and set the condition codes 
have been printed with the "S" modifier after the predicate.  With ARM's
unified syntax, they are supposed to go in the other order.  We fixed this
for Thumb when we switched to unified syntax but missed changing it for
ARM.  Apparently we don't generate these instructions often because no one
noticed until now.  Thanks to Bill Wendling for the testcase!

llvm-svn: 116563
 2010-10-15 03:23:44 +00:00
Jim Grosbach
b01bcbd047 Encoding info for extension instructions. 
llvm-svn: 116560
 2010-10-15 02:29:58 +00:00
Jakob Stoklund Olesen
b569fb0904 Eliminate curli from SplitEditor. Use the LiveRangeEdit reference instead. 
llvm-svn: 116547
 2010-10-15 00:34:01 +00:00
Jakob Stoklund Olesen
5d5bcb8ee4 Move stack slot assignments into LiveRangeEdit. 
All registers created during splitting or spilling are assigned to the same
stack slot as the parent register.

When splitting or rematting, we may not spill at all. In that case the stack
slot is still assigned, but it will be dead.

llvm-svn: 116546
 2010-10-15 00:16:55 +00:00
Jakob Stoklund Olesen
acca65b973 Create a new LiveRangeEdit class to keep track of the new registers created when 
splitting or spillling, and to help with rematerialization.

Use LiveRangeEdit in InlineSpiller and SplitKit. This will eventually make it
possible to share remat code between InlineSpiller and SplitKit.

llvm-svn: 116543
 2010-10-14 23:49:52 +00:00
Jim Grosbach
6da1b52d7b Add missing Rd encoding for MOVs instruction. 
llvm-svn: 116537
 2010-10-14 23:28:31 +00:00
Jim Grosbach
804505c7d4 Refactor the MOVsr[al]_flag and RRX pseudo-instructions to really be pseudos 
and let the ARMExpandPseudoInsts pass fix them up into the real (MOVs)
instruction form.

llvm-svn: 116534
 2010-10-14 22:57:13 +00:00
Dan Gohman
d93278f6b6 Tolerate a null parent pointer. 
llvm-svn: 116533
 2010-10-14 22:55:57 +00:00
Jim Grosbach
29dc23398f Tweak the ARM backend to use the RRX mnemonic instead of the 'mov a, b, rrx' 
pseudonym.

llvm-svn: 116512
 2010-10-14 20:43:44 +00:00
Francois Pichet
ad0548d742 Always use binary mode for output stream. This is important to prevent unwanted end of line conversion on Windows. Should not affect Unix where O_BINARY is not defined. This fix /clang/test/lexer/preamble.c XFAIL on WIN32. 
llvm-svn: 116509
 2010-10-14 20:30:58 +00:00
Jim Grosbach
73c78f8790 MOVi16 and MOVT ARM mode encodings. 
llvm-svn: 116498
 2010-10-14 18:54:27 +00:00
Jakob Stoklund Olesen
e0b9dbf91b Only split around a loop if the live range has uses outside the loop periphery. 
Before we would also split around a loop if any peripheral block had multiple
uses. This could cause repeated splitting when splitting a different live range
would insert uses into the periphery.

Now -spiller=inline passes the nightly test suite again.

llvm-svn: 116494
 2010-10-14 18:26:45 +00:00
Jim Grosbach
6d5cb39757 Simplify encoding information and add 'dst' operand info for TAILJMP. 
llvm-svn: 116488
 2010-10-14 17:24:28 +00:00
Rafael Espindola
2cdc3d6235 Remove some code duplication. 
llvm-svn: 116484
 2010-10-14 16:34:44 +00:00