Commit Graph

1239 Commits

Author SHA1 Message Date
Bill Wendling 919b7aab2e Add more plumbing. This time in the LowerArguments and "get" functions which
return partial registers. This affected the back-end lowering code some.

Also patch up some places I missed before in the "get" functions.

llvm-svn: 91880
2009-12-22 02:10:19 +00:00
Ken Dyck 798493285c In LowerEXTRACT_VECTOR_ELT, force an i32 value type for PEXTWR instead of
incrementing the simple value type of the 16-bit type, which would give the
wrong type if an intemediate MVT (such as i24) were introduced.

llvm-svn: 91602
2009-12-17 15:31:52 +00:00
Evan Cheng 1be6286028 Re-enable 91381 with fixes.
llvm-svn: 91489
2009-12-16 00:53:11 +00:00
Evan Cheng fcb5453dc7 Disable 91381 for now. It's miscompiling ARMISelDAG2DAG.cpp.
llvm-svn: 91405
2009-12-15 03:07:11 +00:00
Evan Cheng 0e8b9e32d1 Use sbb x, x to materialize carry bit in a GPR. The result is all one's or all zero's.
llvm-svn: 91381
2009-12-15 00:53:42 +00:00
Dan Gohman cecad35728 Fix integer cast code to handle vector types.
llvm-svn: 91362
2009-12-14 23:40:38 +00:00
Dan Gohman 1d459e4937 Implement vector widening, splitting, and scalarizing for SIGN_EXTEND_INREG.
llvm-svn: 91158
2009-12-11 21:31:27 +00:00
Anton Korobeynikov fc51282cbe Honour setHasCalls() set from isel.
This is used in some weird cases like general dynamic TLS model.
This fixes PR5723

llvm-svn: 91144
2009-12-11 19:39:55 +00:00
Evan Cheng 493b882f80 Optimize splat of a scalar load into a shuffle of a vector load when it's legal. e.g.
vector_shuffle (scalar_to_vector (i32 load (ptr + 4))), undef, <0, 0, 0, 0>
=>
vector_shuffle (v4i32 load ptr), undef, <1, 1, 1, 1>

iff ptr is 16-byte aligned (or can be made into 16-byte aligned).

llvm-svn: 90984
2009-12-09 21:00:30 +00:00
Evan Cheng d938faff4b Teach InferPtrAlignment to infer GV+cst alignment and use it to simplify x86 isl lowering code.
llvm-svn: 90925
2009-12-09 01:53:58 +00:00
Evan Cheng f5938d5d27 Move isConsecutiveLoad to SelectionDAG. It's not target dependent and it's primary used by selectdag passes.
llvm-svn: 90922
2009-12-09 01:36:00 +00:00
Mon P Wang 32f8bb9ed4 Added support to allow clients to custom widen. For X86, custom widen vectors for
divide/remainder since these operations can trap by unroll them and adding undefs
for the resulting vector.

llvm-svn: 90108
2009-11-30 02:42:02 +00:00
Dan Gohman de5dea869f Remove ISD::DEBUG_LOC and ISD::DBG_LABEL, which are no longer used.
Note that "hasDotLocAndDotFile"-style debug info was already broken;
people wanting this functionality should implement it in the
AsmPrinter/DwarfWriter code.

llvm-svn: 89711
2009-11-23 23:20:51 +00:00
Devang Patel ed85e12da6 We are not using DBG_STOPPOINT anymore.
llvm-svn: 89536
2009-11-21 02:46:55 +00:00
Dan Gohman 7a6611793f Target-independent support for TargetFlags on BlockAddress operands,
and support for blockaddresses in x86-32 PIC mode.

llvm-svn: 89506
2009-11-20 23:18:13 +00:00
Jeffrey Yasskin 10d3604a9e Make X86-64 in the Large model always emit 64-bit calls.
The large code model is documented at
http://www.x86-64.org/documentation/abi.pdf and says that calls should
assume their target doesn't live within the 32-bit pc-relative offset
that fits in the call instruction.

To do this, we turn off the global-address->target-global-address
conversion in X86TargetLowering::LowerCall(). The first attempt at
this broke the lazy JIT because it can separate the movabs(imm->reg)
from the actual call instruction. The lazy JIT receives the address of
the movabs as a relocation and needs to record the return address from
the call; and then when that call happens, it needs to patch the
movabs with the newly-compiled target. We could thread the call
instruction into the relocation and record the movabs<->call mapping
explicitly, but that seems to require at least as much new
complication in the code generator as this change.

To fix this, we make lazy functions _always_ go through a call
stub. You'd think we'd only have to force lazy calls through a stub on
difficult platforms, but that turns out to break indirect calls
through a function pointer. The right fix for that is to distinguish
between calls and address-of operations on uncompiled functions, but
that's complex enough to leave for someone else to do.

Another attempt at this defined a new CALL64i pseudo-instruction,
which expanded to a 2-instruction sequence in the assembly output and
was special-cased in the X86CodeEmitter's emitInstruction()
function. That broke indirect calls in the same way as above.

This patch also removes a hack forcing Darwin to the small code model.
Without far-call-stubs, the small code model requires things of the
JITMemoryManager that the DefaultJITMemoryManager can't provide.

Thanks to echristo for lots of testing!

llvm-svn: 88984
2009-11-16 22:41:33 +00:00
David Greene 1fbe054450 Add a bool flag to StackObjects telling whether they reference spill
slots.  The AsmPrinter will use this information to determine whether to
print a spill/reload comment.

Remove default argument values.  It's too easy to pass a wrong argument
value when multiple arguments have default values.  Make everything
explicit to trap bugs early.

Update all targets to adhere to the new interfaces..

llvm-svn: 87022
2009-11-12 20:49:22 +00:00
Benjamin Kramer 68e4945c03 Add compare_lower and equals_lower methods to StringRef. Switch all users of
StringsEqualNoCase (from StringExtras.h) to it.

llvm-svn: 87020
2009-11-12 20:36:59 +00:00
Nate Begeman 3a313df69b x86 vector shuffle cleanup/fixes:
1. rename the movhp patfrag to movlhps, since thats what it actually matches
2. eliminate the bogus movhps load and store patterns, they were incorrect.  The load transforms are already handled (correctly) by shufps/unpack.
3. revert a recent test change to its correct form.

llvm-svn: 86415
2009-11-07 23:17:15 +00:00
Kenneth Uildriks 07119737aa Add code to check at SelectionDAGISel::LowerArguments time to see if return values can be lowered to registers. Coming soon, code to perform sret-demotion if return values cannot be lowered to registers
llvm-svn: 86324
2009-11-07 02:11:54 +00:00
Dan Gohman f7c4299312 Initial x86 support for BlockAddresses.
llvm-svn: 85557
2009-10-30 01:28:02 +00:00
Evan Cheng 83896a59e1 Add a second ValueType argument to isFPImmLegal.
llvm-svn: 85361
2009-10-28 01:43:28 +00:00
Evan Cheng 16993aa30b Do away with addLegalFPImmediate. Add a target hook isFPImmLegal which returns true if the fp immediate can be natively codegened by target.
llvm-svn: 85281
2009-10-27 19:56:55 +00:00
Dan Gohman 4a43e3068d Make TranslateX86CC return COND_INVALID instead of aborting when it
encounters an OEQ or UNE comparison, and update its callers to check
for this return status and recover. This fixes a problem resulting from
the LowerOperation hooks being called from LegalizeVectorOps, because
LegalizeVectorOps only lowers vectors, so OEQ and UNE comparisons may
still be at large. This fixes PR5092.

llvm-svn: 84640
2009-10-20 16:22:37 +00:00
Nate Begeman 18df82a20c Add support for matching shuffle patterns with palignr.
llvm-svn: 84459
2009-10-19 02:17:23 +00:00
Evan Cheng 0e9d9ca855 -Revert parts of 84326 and 84411. Distinquishing between fixed and non-fixed
stack slots and giving them different PseudoSourceValue's did not fix the
problem of post-alloc scheduling miscompiling llvm itself.
- Apply Dan's conservative workaround by assuming any non fixed stack slots can
alias other memory locations. This means a load from spill slot #1 cannot 
move above a store of spill slot #2. 
- Enable post-alloc scheduling for x86 at optimization leverl Default and above.

llvm-svn: 84424
2009-10-18 18:16:27 +00:00
Evan Cheng 0b8db2dab7 Only fixed stack objects and spill slots should be get FixedStack PseudoSourceValue.
llvm-svn: 84411
2009-10-18 06:27:36 +00:00
Evan Cheng 8759585aba Revert 84315 for now. Re-thinking the patch.
llvm-svn: 84321
2009-10-17 07:53:04 +00:00
Evan Cheng 0818d87ed1 Rename getFixedStack to getStackObject. The stack objects represented are not
necessarily fixed. Only those will negative frame indices are "fixed."

llvm-svn: 84315
2009-10-17 06:22:26 +00:00
Dan Gohman a698d7ac3c Don't forget to mark RAX as live-out of the function when arranging for
it to hold the address of an sret return value, for x86-64 ABI purposes.

Also, fix the test that was originally intended to test this to actually
test it, using FileCheck.

llvm-svn: 83853
2009-10-12 16:36:12 +00:00
Dale Johannesen e32fe29d29 Fix handling of x86 'R' constraint.
llvm-svn: 83499
2009-10-07 22:47:20 +00:00
Dan Gohman 48b185d6f7 Improve MachineMemOperand handling.
- Allocate MachineMemOperands and MachineMemOperand lists in MachineFunctions.
   This eliminates MachineInstr's std::list member and allows the data to be
   created by isel and live for the remainder of codegen, avoiding a lot of
   copying and unnecessary translation. This also shrinks MemSDNode.
 - Delete MemOperandSDNode. Introduce MachineSDNode which has dedicated
   fields for MachineMemOperands.
 - Change MemSDNode to have a MachineMemOperand member instead of its own
   fields with the same information. This introduces some redundancy, but
   it's more consistent with what MachineInstr will eventually want.
 - Ignore alignment when searching for redundant loads for CSE, but remember
   the greatest alignment.

Target-specific code which previously used MemOperandSDNodes with generic
SDNodes now use MemIntrinsicSDNodes, with opcodes in a designated range
so that the SelectionDAG framework knows that MachineMemOperand information
is available.

llvm-svn: 82794
2009-09-25 20:36:54 +00:00
Dan Gohman 08c0a95ac6 Rename several variables from EVT to more descriptive names, now that EVT
is also the name of their type, as declarations like "EVT EVT" look
really odd.

llvm-svn: 82654
2009-09-23 21:02:20 +00:00
Dan Gohman 48ade83e60 Recognize SSE min and max opportunities in even more cases.
And fix a bug with the behavior of min/max instructions formed from
fcmp uge comparisons.

Also, use FiniteOnlyFPMath() for this code instead of UnsafeFPMath,
as it is more specific.

llvm-svn: 82466
2009-09-21 18:03:22 +00:00
Daniel Dunbar 7d6781b0fe Tabs -> spaces, and remove trailing whitespace.
llvm-svn: 82355
2009-09-20 02:20:51 +00:00
Evan Cheng 9827ad39a7 Fix PR4926. When target hook EmitInstrWithCustomInserter() insert new basic blocks and update CFG, it should also inform sdisel of the changes so the phi source operands will come from the right basic blocks.
llvm-svn: 82311
2009-09-19 09:51:03 +00:00
Evan Cheng 270d0f986f Enhance EmitInstrWithCustomInserter() so target can specify CFG changes that sdisel will use to properly complete phi nodes.
Not functionality change yet.

llvm-svn: 82273
2009-09-18 21:02:19 +00:00
Chris Lattner 3a78ce3a56 Make a new X8632_MachoTargetObjectFile TLOF implementation whose
getSymbolForDwarfGlobalReference is smart enough to know that it 
needs to register the stub it references with MachineModuleInfoMachO,
so that it gets emitted at the end of the file.

Move stub emission from X86ATTAsmPrinter::doFinalization to the
new X86ATTAsmPrinter::EmitEndOfAsmFile asmprinter hook.  The important
thing here is that EmitEndOfAsmFile is called *after* the ehframes are
emitted, so we get all the stubs.

This allows us to remove a gross hack from the asmprinter where it would
"just know" that it needed to output stubs for personality functions.
Now this is all driven from a consistent interface.

The testcase change is just reordering the expected output now that the
stubs come out after the ehframe instead of before.

This also unblocks other changes that Bill wants to make.

llvm-svn: 82269
2009-09-18 20:22:52 +00:00
Dan Gohman 722b1eefdb Add support for using the FLAGS result of or, xor, and and instructions
on x86, to avoid explicit test instructions. A few existing tests changed
due to arbitrary register allocation differences.

llvm-svn: 82263
2009-09-18 19:59:53 +00:00
Chris Lattner b866602f06 Big change #1 for personality function references:
Eliminate the PersonalityPrefix/Suffix & NeedsIndirectEncoding
fields from MAI: they aren't part of the asm syntax, they are
related to the structure of the object file.

To replace their functionality, add a new 
TLOF::getSymbolForDwarfGlobalReference method which asks targets
to decide how to reference a global from EH in a pc-relative way.

The default implementation just returns the symbol.  The default
darwin implementation references the symbol through an indirect
$non_lazy_ptr stub.  The bizarro x86-64 darwin specialization
handles the weird "foo@GOTPCREL+4" hack.

DwarfException.cpp now uses this to emit the reference to the
symbol in the right way, and this also eliminates another 
horrible hack from DwarfException.cpp:

-    if (strcmp(MAI->getPersonalitySuffix(), "+4@GOTPCREL"))
-      O << "-" << MAI->getPCSymbol();

llvm-svn: 81991
2009-09-16 01:46:41 +00:00
Chris Lattner a3e68b567c add some special case handling for strangely named x86 registers.
llvm-svn: 81726
2009-09-13 22:41:48 +00:00
Duncan Sands bfb9719dad Remove a left over bit of code with no effect.
llvm-svn: 81128
2009-09-06 19:29:07 +00:00
Duncan Sands 89720bbd11 Remove some not-really-used variables, as warned
about by icc (#593, partial).  Patch by Erick Tryzelaar.

llvm-svn: 81115
2009-09-06 12:41:19 +00:00
Dan Gohman d0d5e685da Recognize more opportunities to use SSE min and max instructions,
swapping the operands if necessary.

llvm-svn: 80940
2009-09-03 20:34:31 +00:00
Mon P Wang 3e82117210 Fixed a few problems with vector shifts
- when transforming a vector shift of a non-immediate scalar shift amount, zero
    extend the i32 shift amount to i64 since the vector shift reads 64 bits
  - when transforming i16 vectors to use a vector shift, zero extend i16 shift amount
  - improve the code quality in some cases when transforming vectors to use a vector shift

llvm-svn: 80935
2009-09-03 19:56:25 +00:00
Dan Gohman 319cc69f88 Add a -disable-16bit flag and associated support for experimenting with
disabling the use of 16-bit operations on x86. This doesn't yet work for
inline asms with 16-bit constraints, vectors with 16-bit elements,
trampoline code, and perhaps other obscurities, but it's enough to try
some experiments.

llvm-svn: 80930
2009-09-03 17:18:51 +00:00
Sandeep Patel 68c5f477fa Retype from unsigned to CallingConv::ID accordingly. Approved by Bob Wilson.
llvm-svn: 80773
2009-09-02 08:44:58 +00:00
Chris Lattner d5f4fcceae refactor select 'sched insertion' out to its own method.
llvm-svn: 80764
2009-09-02 05:57:00 +00:00
Duncan Sands 9cf8bcb69d Revert commit 80428. It completely broke exception
handling on x86-32 linux.

llvm-svn: 80592
2009-08-31 16:45:16 +00:00
Bill Wendling 39bb29f7fe - Add target lowering methods to get the preferred format for the FDE and LSDA
encodings.
- Make some of the values emitted by the FDEs dependent upon the pointer
  size. This is in line with how GCC does things. And it has the benefit of
  working for Darwin in 64-bit mode now.

llvm-svn: 80428
2009-08-29 12:20:54 +00:00
Eric Christopher 69baa5139b Add FIXME for when we support more specific XMM registers.
llvm-svn: 80269
2009-08-27 18:08:16 +00:00
Eric Christopher 799c92bcfb Nuke trailing whitespace.
llvm-svn: 80268
2009-08-27 18:07:15 +00:00
Dan Gohman f1abb5511b Expand i8 selects into control flow instead of 16-bit conditional
moves. This avoids the need to promote the operands (or implicitly
extend them, a partial register update condition), and can reduce
i8 register pressure. This substantially speeds up code such as
write_hex in lib/Support/raw_ostream.cpp.

subclass-coalesce.ll is too trivial and no longer tests what it was
originally intended to test.

llvm-svn: 80184
2009-08-27 00:14:12 +00:00
Devang Patel 0939595711 Record variable debug info at ISel time directly.
llvm-svn: 79742
2009-08-22 17:12:53 +00:00
Eric Christopher 9fe912de5f Implement sse4.2 string/text processing instructions:
Add patterns and instruction encoding information.
Add custom lowering to deal with hardwired return register of
uncertain type (xmm0).

llvm-svn: 79377
2009-08-18 22:50:32 +00:00
Dan Gohman 47a31a29c5 Fix function alignment at -Os on x86 to be 1, not 2. getFunctionAlignment
returns a log2 value.

llvm-svn: 79293
2009-08-18 00:20:06 +00:00
Dan Gohman f0b984021c Avoid emitting XMM save code in soft-float or no-implicit-float mode
or some other situation where no xmm registers need to be saved.

llvm-svn: 79207
2009-08-16 21:24:25 +00:00
Bill Wendling bae6b2cca3 Reapply r79127. It was fixed by d0k.
llvm-svn: 79136
2009-08-15 21:21:19 +00:00
Bill Wendling d3fade656f Revert r79127. It was causing compilation errors.
llvm-svn: 79135
2009-08-15 21:14:01 +00:00
Evan Cheng 52d4e64711 Change allowsUnalignedMemoryAccesses to take type argument since some targets
support unaligned mem access only for certain types. (Should it be size
instead?)

ARM v7 supports unaligned access for i16 and i32, some v6 variants support it
as well.

llvm-svn: 79127
2009-08-15 19:23:44 +00:00
Dan Gohman 0700a56860 On x86-64, for a varargs function, don't store the xmm registers to
the register save area if %al is 0. This avoids touching xmm
regsiters when they aren't actually used.

llvm-svn: 79061
2009-08-15 01:38:56 +00:00
Anton Korobeynikov 75821f7c69 Properly handle indirect win64 args when they're passed in memory
llvm-svn: 79009
2009-08-14 18:19:10 +00:00
Owen Anderson 55f1c09e31 Push LLVMContexts through the IntegerType APIs.
llvm-svn: 78948
2009-08-13 21:58:54 +00:00
Owen Anderson c6daf8f17c Fix warnings.
llvm-svn: 78725
2009-08-11 21:59:30 +00:00
Owen Anderson 9f94459d24 Split EVT into MVT and EVT, the former representing _just_ a primitive type, while
the latter is capable of representing either a primitive or an extended type.

llvm-svn: 78713
2009-08-11 20:47:22 +00:00
Owen Anderson 53aa7a960c Rename MVT to EVT, in preparation for splitting SimpleValueType out into its own struct type.
llvm-svn: 78610
2009-08-10 22:56:29 +00:00
Owen Anderson 3e77df2bcd SimpleValueType-ify a few more methods on TargetLowering.
llvm-svn: 78595
2009-08-10 20:46:15 +00:00
Owen Anderson 246617857f Continue the SimpleValueType-ification.
llvm-svn: 78593
2009-08-10 20:18:46 +00:00
Owen Anderson c30530d105 Start moving TargetLowering away from using full MVTs and towards SimpleValueType, which will simplify the privatization of IntegerType in the future.
llvm-svn: 78584
2009-08-10 18:56:59 +00:00
Anton Korobeynikov 741ea0d7fd Better handle kernel code model. Also, generalize the things and fix one
subtle bug with small code model.

llvm-svn: 78255
2009-08-05 23:01:26 +00:00
Dan Gohman f9bbcd1afd Major calling convention code refactoring.
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.

This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.

This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.

llvm-svn: 78142
2009-08-05 01:29:28 +00:00
Anton Korobeynikov 7d80ab1593 Perform bitconvert to proper type
llvm-svn: 77965
2009-08-03 08:14:14 +00:00
Anton Korobeynikov 442beabbf7 Add 'Indirect' LocInfo class and use to pass __m128 on win64. Also minore fixes here and there (mostly __m64).
llvm-svn: 77964
2009-08-03 08:13:56 +00:00
Anton Korobeynikov 72bc3846bc Cleanup Darwin MMX calling conv stuff - make the stuff more generic. This also fixes a subtle bug, when 6th v1i64 argument passed wrongly.
llvm-svn: 77963
2009-08-03 08:13:24 +00:00
Anton Korobeynikov 71386e08fe Unbreak Win64 CC. Step one: honour register save area, fix some alignment and provide a different set of call-clobberred registers.
llvm-svn: 77962
2009-08-03 08:12:53 +00:00
Rafael Espindola 854d34a9fb Remove a bitcast that was a no-op.
Thanks to Eli Friedman for noticing it.

llvm-svn: 77942
2009-08-03 03:00:05 +00:00
Rafael Espindola 18ba271a79 Use movq to move 64 bits in and out of mmx registers.
Fixes PR4669

llvm-svn: 77940
2009-08-03 02:45:34 +00:00
Dan Gohman c278621694 Minor code cleanups.
llvm-svn: 77795
2009-08-01 19:14:37 +00:00
Chris Lattner 51d5b43cda refactor section construction in TLOF to be through an explicit
initialize method, which can be called when an MCContext is available.

llvm-svn: 77687
2009-07-31 17:42:42 +00:00
Dan Gohman 013f007762 Rename GRAD to GR32_AD, to follow the naming convention of other
classes. And define its SubRegClassList.

llvm-svn: 77601
2009-07-30 17:02:08 +00:00
Evan Cheng e62288fdd4 Optimize some common usage patterns of atomic built-ins __sync_add_and_fetch() and __sync_sub_and_fetch.
When the return value is not used (i.e. only care about the value in the memory), x86 does not have to use add to implement these. Instead, it can use add, sub, inc, dec instructions with the "lock" prefix.

This is currently implemented using a bit of instruction selection trick. The issue is the target independent pattern produces one output and a chain and we want to map it into one that just output a chain. The current trick is to select it into a merge_values with the first definition being an implicit_def. The proper solution is to add new ISD opcodes for the no-output variant. DAG combiner can then transform the node before it gets to target node selection.

Problem #2 is we are adding a whole bunch of x86 atomic instructions when in fact these instructions are identical to the non-lock versions. We need a way to add target specific information to target nodes and have this information carried over to machine instructions. Asm printer (or JIT) can use this information to add the "lock" prefix.

llvm-svn: 77582
2009-07-30 08:33:02 +00:00
Eric Christopher 77268a56ff Add llvm_unreachable for ... unreachable code!
llvm-svn: 77480
2009-07-29 18:14:04 +00:00
Chris Lattner 6b6dbb3bd8 whitespace cleanup.
llvm-svn: 77438
2009-07-29 05:48:09 +00:00
Eric Christopher 99f5534296 Fix comment.
llvm-svn: 77415
2009-07-29 01:01:19 +00:00
Eric Christopher f7802a33ce Add support for gcc __builtin_ia32_ptest{z,c,nzc} intrinsics. Lower
to ptest instruction plus setcc. Revamp ptest instruction. Add test.

llvm-svn: 77407
2009-07-29 00:28:05 +00:00
Owen Anderson 4aa3295a65 Return ConstantVector to 2.5 API.
llvm-svn: 77366
2009-07-28 21:19:26 +00:00
Chris Lattner a3242e93b7 the apple "ld_classic" linker doesn't support .literal16 in 32-bit
mode, and "ld64" (the default linker) falls back to it in -static
mode.

llvm-svn: 77334
2009-07-28 17:50:28 +00:00
Chris Lattner 5e693ed07b Rip all of the global variable lowering logic out of TargetAsmInfo. Since
it is highly specific to the object file that will be generated in the end,
this introduces a new TargetLoweringObjectFile interface that is implemented
for each of ELF/MachO/COFF/Alpha/PIC16 and XCore.

Though still is still a brutal and ugly refactoring, this is a major step
towards goodness.

This patch also:
1. fixes a bunch of dangling pointer problems in the PIC16 backend.
2. disables the TargetLowering copy ctor which PIC16 was accidentally using.
3. gets us closer to xcore having its own crazy target section flags and
   pic16 not having to shadow sections with its own objects.
4. fixes wierdness where ELF targets would set CStringSection but not
   CStringSection_.  Factor the code better.
5. fixes some bugs in string lowering on ELF targets.

llvm-svn: 77294
2009-07-28 03:13:23 +00:00
Owen Anderson 69c464dec4 Move ConstantFP construction back to the 2.5-ish API.
llvm-svn: 77247
2009-07-27 20:59:43 +00:00
Owen Anderson edb4a70325 Revert the ConstantInt constructors back to their 2.5 forms where possible, thanks to contexts-on-types. More to come.
llvm-svn: 77011
2009-07-24 23:12:02 +00:00
Eric Christopher f37ea3ad75 Update insertps handling based on feedback. Move to a v4f32 style
to support vector arguments and scalar arguments correctly. Update
lowering and fix comment to refer to pinsr* instead of insertps.

llvm-svn: 76921
2009-07-24 00:33:09 +00:00
Eli Friedman caccc0081a Add support for MMX VSETCC.
llvm-svn: 76713
2009-07-22 01:06:52 +00:00
Owen Anderson 47db941fd3 Get rid of the Pass+Context magic.
llvm-svn: 76702
2009-07-22 00:24:57 +00:00
Eli Friedman da9eda8ef6 Remove shift amount flavor. It isn't actually complete enough to
be useful, and it's currently unused.  (Some issues: it isn't actually 
rich enough to capture the semantics on many architectures, and
semantics can vary depending on the type being shifted.)

llvm-svn: 76633
2009-07-21 20:12:16 +00:00
Dale Johannesen 68cb8d0310 revert 76503 while I figure out what's going on
llvm-svn: 76517
2009-07-21 00:12:29 +00:00
Dale Johannesen a2d2adfa4e Make sure a global matching asm 'i' constraint gets its
flags set properly.  (hasMemory is clearly irrelevant
when matching 'i', I don't understand what this was
supposed to be doing.)
gcc.apple/asm-block-25.c (test passed before by
accident, but generated code was wrong)

llvm-svn: 76503
2009-07-20 23:39:13 +00:00
Chris Lattner 5849d22bd1 Copy ExpandInlineAsm to TargetLowering from TargetAsmInfo.
llvm-svn: 76441
2009-07-20 17:51:36 +00:00
Evan Cheng 18fe458103 Fix x86 inline ams 'q' constraint support. In 32-bit mode, it's just like 'Q', i.e. EAX, EDX, ECX, EBX. In 64-bit mode, it just means all the i64r registers. Yeah, that makes sense.
llvm-svn: 76248
2009-07-17 22:13:25 +00:00
Owen Anderson f945a9ed07 Move a few more convenience factory functions from Constant to LLVMContext.
llvm-svn: 75840
2009-07-15 21:51:10 +00:00
Torok Edwin fbcc663cbf llvm_unreachable->llvm_unreachable(0), LLVM_UNREACHABLE->llvm_unreachable.
This adds location info for all llvm_unreachable calls (which is a macro now) in
!NDEBUG builds.
In NDEBUG builds location info and the message is off (it only prints
"UREACHABLE executed").

llvm-svn: 75640
2009-07-14 16:55:14 +00:00
Chris Lattner e91900097e Fix PR4533, which is about buggy codegen in x86-64 -static mode.
Basically, using:
  lea symbol(%rip), %rax

is not valid in -static mode, because the current RIP may not be
within 32-bits of "symbol" when an app is built partially pic and
partially static.  The fix for this is to compile it to:

  lea symbol, %rax

It would be better to codegen this as:

  movq $symbol, %rax

but that will come next.


The hard part of fixing this bug was fixing abi-isel, which was actively
testing for the wrong behavior.  Also, the RUN lines are completely impossible
to understand what they are testing.  To help with this, convert the -static 
x86-64 codegen tests to use filecheck.  This is much more stable and makes it
more clear what the codegen is expected to be.

llvm-svn: 75382
2009-07-11 20:29:19 +00:00
Torok Edwin 56d0659726 assert(0) -> LLVM_UNREACHABLE.
Make llvm_unreachable take an optional string, thus moving the cerr<< out of
line.
LLVM_UNREACHABLE is now a simple wrapper that makes the message go away for
NDEBUG builds.

llvm-svn: 75379
2009-07-11 20:10:48 +00:00
Chris Lattner 21c2940553 remove the now-dead TM argument to these methods.
llvm-svn: 75276
2009-07-10 21:00:45 +00:00
Chris Lattner e2f524f176 add a couple of predicates to test for "stub style pic in PIC mode" and "stub style pic in dynamic-no-pic" mode.
llvm-svn: 75273
2009-07-10 20:47:30 +00:00
Chris Lattner 93f0f79178 eliminate GVRequiresRegister, replacing it with predicates we
need for other purposes.

llvm-svn: 75243
2009-07-10 07:38:24 +00:00
Chris Lattner 2ff35f528c change a bunch of logic in LowerGlobalAddress to leverage the work
done in ClassifyGlobalReference instead of reconstructing the info
awkwardly.

llvm-svn: 75240
2009-07-10 07:34:39 +00:00
Chris Lattner dc842c06c2 move some classification logic around. Now GVRequiresExtraLoad
is just a trivial wrapper around "ClassifyGlobalReference", which
stole a ton of logic from LowerGlobalAddress.

llvm-svn: 75237
2009-07-10 07:20:05 +00:00
Chris Lattner ca9d784bf1 change isGlobalStubReference to take target flags instead of a MachineOperand.
llvm-svn: 75236
2009-07-10 06:29:59 +00:00
Chris Lattner b9af63a4d2 GVRequiresExtraLoad is now never used for calls, simplify it based on this.
llvm-svn: 75232
2009-07-10 05:52:02 +00:00
Chris Lattner ace6ec26d9 actually, just eliminate PCRelGVRequiresExtraLoad. It makes the code
more complex and slow than just directly testing what we care about.

llvm-svn: 75231
2009-07-10 05:48:03 +00:00
Chris Lattner 7277a807f0 There is only one case where GVRequiresExtraLoad returns true for calls:
split its handling out to PCRelGVRequiresExtraLoad, and simplify code
based on this.

llvm-svn: 75230
2009-07-10 05:45:15 +00:00
Chris Lattner 1cc7ae7c3b the "isDirectCall" operand of GVRequiresRegister is always false, eliminate it.
llvm-svn: 75229
2009-07-10 05:37:11 +00:00
Owen Anderson 0504e0a222 Thread LLVMContext through MVT and related parts of SDISel.
llvm-svn: 75153
2009-07-09 17:57:24 +00:00
Chris Lattner 212c0506cd simplify this logic a bit more.
llvm-svn: 75118
2009-07-09 07:02:30 +00:00
Chris Lattner 72e3deca47 move reasoning about darwin $non_lazy_ptr stubs from asmprinter into
isel.

llvm-svn: 75117
2009-07-09 06:59:17 +00:00
Chris Lattner e0dae0e716 make isel use MO_PIC_BASE_OFFSET when lowering globalvalues on darwin in pic
mode, instead of having asmprinter just "know" to print them.

llvm-svn: 75109
2009-07-09 05:47:33 +00:00
Chris Lattner d047d06358 make isel decide whether to emit $stub's on darwin instead of asmprinter.
llvm-svn: 75107
2009-07-09 05:27:35 +00:00
Chris Lattner d5d80ab6b6 Make isel determine where to emit PLT-relative calls instead of having
asmprinter do it.

llvm-svn: 75104
2009-07-09 05:02:21 +00:00
Chris Lattner fef11d6e77 simplify some code based on the fact that picstyles != none are only valid
in pic or dynamic-no-pic mode. Also, x86-64 never used picstylegot.

llvm-svn: 75101
2009-07-09 04:39:06 +00:00
Chris Lattner 62f568c037 all this logic always returns true because GOT mode is never active in x86-64 mode.
Simplify it away, someone should evaluate this.

llvm-svn: 75100
2009-07-09 04:27:47 +00:00
Chris Lattner 3945dd0c44 isPICStyleRIPRel() and friends are never true in -static mode.
Simplify code based on this.

llvm-svn: 75099
2009-07-09 04:24:46 +00:00
Chris Lattner 1c5bf9d26d When in -static mode, force the PIC style to none. Doing this requires fixing
code which conflated RIPRel PIC with x86-64.  Fix these to just check for X86-64
directly.

llvm-svn: 75092
2009-07-09 03:15:51 +00:00
Chris Lattner 384a738b4a merge two identical functions and simplify things that are GOT specific
llvm-svn: 75091
2009-07-09 02:55:47 +00:00
Chris Lattner 4d1e9bf717 hoist check for IsTailCall to callers. Eliminate redundant check for
x86-64: GOT-style PIC is never used on x86-64.

llvm-svn: 75090
2009-07-09 02:46:53 +00:00
Chris Lattner 88765d4831 change a few methods to be static functions.
llvm-svn: 75089
2009-07-09 02:44:11 +00:00
Chris Lattner 47f64ea174 move handling of dllimport linkage in isel, not in asmprinter.
llvm-svn: 75086
2009-07-09 00:58:53 +00:00
Torok Edwin fb8d6d5b58 Implement changes from Chris's feedback.
Finish converting lib/Target.

llvm-svn: 75043
2009-07-08 20:53:28 +00:00
Torok Edwin fa04002254 Convert more abort() calls to llvm_report_error().
Also remove trailing semicolon.

llvm-svn: 75027
2009-07-08 19:04:27 +00:00
Torok Edwin 6dd2730024 Start converting to new error handling API.
cerr+abort -> llvm_report_error
assert(0)+abort -> LLVM_UNREACHABLE (assert(0)+llvm_unreachable-> abort() included)

llvm-svn: 75018
2009-07-08 18:01:40 +00:00
Dale Johannesen 56a53d02e1 Don't accept globals as matching 'i' constraint
in PIC modes (in accordance with existing comment).
gcc.apple/asm-block-25.c

llvm-svn: 74886
2009-07-07 00:18:49 +00:00
Tilmann Scheller aea6059ed4 Add NumFixedArgs attribute to CallSDNode which indicates the number of fixed arguments in a vararg call.
With the SVR4 ABI on PowerPC, vector arguments for vararg calls are passed differently depending on whether they are a fixed or a variable argument. Variable vector arguments always go into memory, fixed vector arguments are put 
into vector registers. If there are no free vector registers available, fixed vector arguments are put on the stack.

The NumFixedArgs attribute allows to decide for an argument in a vararg call whether it belongs to the fixed or variable portion of the parameter list.

llvm-svn: 74764
2009-07-03 06:44:53 +00:00
Bill Wendling 512ff7353e Update comments to make it clear that the function alignment is the Log2 of the
bytes and not bytes.

llvm-svn: 74624
2009-07-01 18:50:55 +00:00
Bill Wendling 31ceb1bcba Add an "alignment" field to the MachineFunction object. It makes more sense to
have the alignment be calculated up front, and have the back-ends obey whatever
alignment is decided upon.

This allows for future work that would allow for precise no-op placement and the
like.

llvm-svn: 74564
2009-06-30 22:38:32 +00:00
David Greene 8adf1fdc80 Add a 256-bit register class and YMM registers.
llvm-svn: 74469
2009-06-29 22:50:51 +00:00
Owen Anderson 45c299ef65 Add a target-specific DAG combine on X86 to fold the common pattern of
fence-atomic-fence down to just the atomic op.  This is possible thanks to
X86's relatively strong memory model, which guarantees that locked instructions
(which are used to implement atomics) are implicit fences.

llvm-svn: 74435
2009-06-29 18:04:45 +00:00
David Greene f92ba97cda Add more vector ValueTypes for AVX and other extended vector instruction
sets.

llvm-svn: 74427
2009-06-29 16:47:10 +00:00
Chris Lattner ae0acfcef0 pull @GOT, @GOTOFF, @GOTPCREL handling into isel from the asmprinter.
llvm-svn: 74378
2009-06-27 05:39:56 +00:00
Chris Lattner fea81da433 Reimplement rip-relative addressing in the X86-64 backend. The new
implementation primarily differs from the former in that the asmprinter
doesn't make a zillion decisions about whether or not something will be
RIP relative or not.  Instead, those decisions are made by isel lowering
and propagated through to the asm printer.  To achieve this, we:

1. Represent RIP relative addresses by setting the base of the X86 addr
   mode to X86::RIP.
2. When ISel Lowering decides that it is safe to use RIP, it lowers to
   X86ISD::WrapperRIP.  When it is unsafe to use RIP, it lowers to
   X86ISD::Wrapper as before.
3. This removes isRIPRel from X86ISelAddressMode, representing it with
   a basereg of RIP instead.
4. The addressing mode matching logic in isel is greatly simplified.
5. The asmprinter is greatly simplified, notably the "NotRIPRel" predicate
   passed through various printoperand routines is gone now.
6. The various symbol printing routines in asmprinter now no longer infer
   when to emit (%rip), they just print the symbol.

I think this is a big improvement over the previous situation.  It does have
two small caveats though: 1. I implemented a horrible "no-rip" modifier for
the inline asm "P" constraint modifier.  This is a short term hack, there is
a much better, but more involved, solution.  2. I had to xfail an 
-aggressive-remat testcase because it isn't handling the use of RIP in the
constant-pool reading instruction.  This specific test is easy to fix without
-aggressive-remat, which I intend to do next.

llvm-svn: 74372
2009-06-27 04:16:01 +00:00
Chris Lattner 49ed726e46 Move all the TLS processing logic into isel, don't do it in asmprinter at all.
llvm-svn: 74327
2009-06-26 21:20:29 +00:00
Chris Lattner 2ed6a9d7bd move magic for PIC constantpool references from asmprinter to isel.
llvm-svn: 74313
2009-06-26 19:22:52 +00:00
Chris Lattner 2aaad91bbe start adding logic in isel to determine asm printer semantics, step N of M.
llvm-svn: 74246
2009-06-26 00:43:52 +00:00
Chris Lattner a3da048c82 indentation fix
llvm-svn: 73840
2009-06-21 02:22:34 +00:00
Eli Friedman 48021d15d6 Misc accumulated tweaks to legalization logic for various targets.
llvm-svn: 73476
2009-06-16 06:40:59 +00:00
Chris Lattner c68a564cdd I got J and K backward, many thanks to Eli for spotting this!
llvm-svn: 73372
2009-06-15 04:39:05 +00:00
Chris Lattner ea3621a6b1 implement support for the 'K' asm constraint, PR4347
llvm-svn: 73366
2009-06-15 04:01:39 +00:00
Arnold Schwaighofer e3a018d707 Fix Bug 4278: X86-64 with -tailcallopt calling convention
out of sync with regular cc.

The only difference between the tail call cc and the normal
cc was that one parameter register - R9 - was reserved for
calling functions through a function pointer. After time the
tail call cc has gotten out of sync with the regular cc. 

We can use R11 which is also caller saved but not used as
parameter register for potential function pointers and
remove the special tail call cc on x86-64.

llvm-svn: 73233
2009-06-12 16:26:57 +00:00
Anton Korobeynikov 06039d1190 Silence a warning
llvm-svn: 73152
2009-06-09 23:00:39 +00:00
Eli Friedman 0d4234416f Get rid of some unnecessary code.
llvm-svn: 73017
2009-06-07 07:28:45 +00:00
Eli Friedman 3234587213 Slightly generalize the code that handles shuffles of consecutive loads
on x86 to handle more cases.  Fix a bug in said code that would cause it 
to read past the end of an object.  Rewrite the code in 
SelectionDAGLegalize::ExpandBUILD_VECTOR to be a bit more general. 
Remove PerformBuildVectorCombine, which is no longer necessary with 
these changes.  In addition to simplifying the code, with this change, 
we can now catch a few more cases of consecutive loads.

llvm-svn: 73012
2009-06-07 06:52:44 +00:00
Eli Friedman 75c496f920 Avoid crashing on a variable-index insertelement with element type i16.
llvm-svn: 72991
2009-06-06 06:32:50 +00:00
Eli Friedman 1b1844ad1f Get rid of some bogus patterns for X86vzmovl. Don't create VZEXT_MOVL
nodes for vectors with an i16 element type.  Add an optimization for 
building a vector which is all zeros/undef except for the bottom 
element, where the bottom element is an i8 or i16.

llvm-svn: 72988
2009-06-06 06:05:10 +00:00
Eli Friedman b45e8ce69a PR2598: make sure to expand illegal forms of integer/floating-point
conversions for x86, like <2 x i32> -> <2 x float> and <4 x i16> -> 
<4 x float>.

llvm-svn: 72983
2009-06-06 03:57:58 +00:00
Devang Patel d1c7d34924 Add new function attribute - noimplicitfloat
Update code generator to use this attribute and remove NoImplicitFloat target option.
Update llc to set this attribute when -no-implicit-float command line option is used.

llvm-svn: 72959
2009-06-05 21:57:13 +00:00
Nate Begeman 624690c6b2 Adapt the x86 build_vector dagcombine to the current state of the legalizer.
build vectors with i64 elements will only appear on 32b x86 before legalize.
Since vector widening occurs during legalize, and produces i64 build_vector 
elements, the dag combiner is never run on these before legalize splits them
into 32b elements.

Teach the build_vector dag combine in x86 back end to recognize consecutive 
loads producing the low part of the vector.

Convert the two uses of TLI's consecutive load recognizer to pass LoadSDNodes
since that was required implicitly.

Add a testcase for the transform.

Old:
	subl	$28, %esp
	movl	32(%esp), %eax
	movl	4(%eax), %ecx
	movl	%ecx, 4(%esp)
	movl	(%eax), %eax
	movl	%eax, (%esp)
	movaps	(%esp), %xmm0
	pmovzxwd	%xmm0, %xmm0
	movl	36(%esp), %eax
	movaps	%xmm0, (%eax)
	addl	$28, %esp
	ret

New:
	movl	4(%esp), %eax
	pmovzxwd	(%eax), %xmm0
	movl	8(%esp), %eax
	movaps	%xmm0, (%eax)
	ret

llvm-svn: 72957
2009-06-05 21:37:30 +00:00
Devang Patel 54707b420a Evan thinks NoImplicitFloat check is not required here.
llvm-svn: 72954
2009-06-05 18:48:29 +00:00
Dan Gohman 11231d0c75 Remove unnecessary #includes.
llvm-svn: 72782
2009-06-03 16:47:12 +00:00
Dale Johannesen 5234d3795f Revert 72707 and 72709, for the moment.
llvm-svn: 72712
2009-06-02 03:12:52 +00:00
Dale Johannesen 0b8ca79253 Make the implicit inputs and outputs of target-independent
ADDC/ADDE use MVT::i1 (later, whatever it gets legalized to)
instead of MVT::Flag.  Remove CARRY_FALSE in favor of 0; adjust
all target-independent code to use this format.

Most targets will still produce a Flag-setting target-dependent
version when selection is done.  X86 is converted to use i32
instead, which means TableGen needs to produce different code
in xxxGenDAGISel.inc.  This keys off the new supportsHasI1 bit
in xxxInstrInfo, currently set only for X86; in principle this
is temporary and should go away when all other targets have
been converted.  All relevant X86 instruction patterns are
modified to represent setting and using EFLAGS explicitly.  The
same can be done on other targets.

The immediate behavior change is that an ADC/ADD pair are no
longer tightly coupled in the X86 scheduler; they can be
separated by instructions that don't clobber the flags (MOV).
I will soon add some peephole optimizations based on using
other instructions that set the flags to feed into ADC.

llvm-svn: 72707
2009-06-01 23:27:20 +00:00
Bill Wendling 09f17a8479 Untabification.
llvm-svn: 72604
2009-05-30 01:09:53 +00:00
Evan Cheng a9cda8abf2 Added optimization that narrow load / op / store and the 'op' is a bit twiddling instruction and its second operand is an immediate. If bits that are touched by 'op' can be done with a narrower instruction, reduce the width of the load and store as well. This happens a lot with bitfield manipulation code.
e.g.
orl     $65536, 8(%rax)
=>
orb     $1, 10(%rax)

Since narrowing is not always a win, e.g. i32 -> i16 is a loss on x86, dag combiner consults with the target before performing the optimization.

llvm-svn: 72507
2009-05-28 00:35:15 +00:00
Eli Friedman a56159b7e9 Ger rid of some dead code.
llvm-svn: 72494
2009-05-27 20:39:00 +00:00
Eli Friedman acb851a8c0 Don't abuse the quirky behavior of LegalizeDAG for XINT_TO_FP and
FP_TO_XINT.  Necessary for some cleanups I'm working on.  Updated 
from the previous version (r72431) to fix a bug and make some things a 
bit clearer.

llvm-svn: 72445
2009-05-27 00:47:34 +00:00
Daniel Dunbar d96b117872 Back out r72431, it is causing a number of compilation crashes with clang.
llvm-svn: 72436
2009-05-26 21:27:02 +00:00
Eli Friedman 8c7bff96ed Don't abuse the quirky behavior of LegalizeDAG for XINT_TO_FP and
FP_TO_XINT.  Necessary for some cleanups I'm working on. 

llvm-svn: 72431
2009-05-26 19:18:56 +00:00
Eli Friedman 2199ed399f Make the X86 backend mark EXTRACT_SUBVECTOR as Expand, at least for the
moment.

llvm-svn: 72350
2009-05-23 22:44:52 +00:00
Eli Friedman dfe4f25355 Make the x86 backend custom-lower UINT_TO_FP and FP_TO_UINT on 32-bit
systems instead of attempting to promote them to a 64-bit SINT_TO_FP or 
FP_TO_SINT.  This is in preparation for removing the type legalization 
code from LegalizeDAG: once type legalization is gone from LegalizeDAG, 
it won't be able to handle the i64 operand/result correctly.

This isn't quite ideal, but I don't think any other operation for any 
target ends up in this situation, so treating this case specially seems 
reasonable.

llvm-svn: 72324
2009-05-23 09:59:16 +00:00
Evan Cheng ab0d23396a Run code placement optimization for targets that want it (arm and x86 for now).
llvm-svn: 71726
2009-05-13 21:42:09 +00:00
Chris Lattner f1d9b91434 Fix PR4152: asm constraint validation happens before dag combine, so we
need to work a bit to combine things like (x+c1+c2) into x+c3.

llvm-svn: 71232
2009-05-08 18:23:14 +00:00
Nate Begeman 7e6e352735 Fix infinite recursion in the C++ code which handles movddup by making it unnecessary.
llvm-svn: 70425
2009-04-29 22:47:44 +00:00
Nate Begeman 5f829d896d Implement review feedback for vector shuffle work.
llvm-svn: 70372
2009-04-29 05:20:52 +00:00
Nate Begeman 8d6d4b9289 2nd attempt, fixing SSE4.1 issues and implementing feedback from duncan.
PR2957

ISD::VECTOR_SHUFFLE now stores an array of integers representing the shuffle
mask internal to the node, rather than taking a BUILD_VECTOR of ConstantSDNodes
as the shuffle mask.  A value of -1 represents UNDEF.

In addition to eliminating the creation of illegal BUILD_VECTORS just to 
represent shuffle masks, we are better about canonicalizing the shuffle mask,
resulting in substantially better code for some classes of shuffles.

llvm-svn: 70225
2009-04-27 18:41:29 +00:00
Rafael Espindola c1396a2313 Fix PR 4004 by including the call to __tls_get_addr in X86tlsaddr. This is not
very elegant, but neither is the tls specification :-(

llvm-svn: 69968
2009-04-24 12:59:40 +00:00
Rafael Espindola b93db668b3 Revert 69952. Causes testsuite failures on linux x86-64.
llvm-svn: 69967
2009-04-24 12:40:33 +00:00
Nate Begeman bb881d66f4 PR2957
ISD::VECTOR_SHUFFLE now stores an array of integers representing the shuffle
mask internal to the node, rather than taking a BUILD_VECTOR of ConstantSDNodes
as the shuffle mask.  A value of -1 represents UNDEF.

In addition to eliminating the creation of illegal BUILD_VECTORS just to 
represent shuffle masks, we are better about canonicalizing the shuffle mask,
resulting in substantially better code for some classes of shuffles.

A clean up of x86 shuffle code, and some canonicalizing in DAGCombiner is next.

llvm-svn: 69952
2009-04-24 03:42:54 +00:00
Duncan Sands 7ce5cc6bd1 Get rid of what looks like a copy-and-pasted typo.
Spotted by gcc-4.5.

llvm-svn: 69673
2009-04-21 09:44:39 +00:00
Bob Wilson f8b85477ae Move duplicated AddLiveIn function from X86 and ARM backends to be a method
in the MachineFunction class, renaming it to addLiveIn for consistency with
the same method in MachineBasicBlock.  Thanks for Anton for suggesting this.

llvm-svn: 69615
2009-04-20 18:36:57 +00:00
Rafael Espindola 355fe12c82 For general dynamic TLS access we must use
leaq	foo@TLSGD(%rip), %rdi

as part of the instruction sequence. Using a register other than %rdi and then
copying it to %rdi is not valid.

llvm-svn: 69350
2009-04-17 14:35:58 +00:00
Rafael Espindola 6d6c6043ea X86-64 TLS support for local exec and initial exec.
llvm-svn: 68947
2009-04-13 13:02:49 +00:00
Dan Gohman de912e2475 Remove the obsolete SelectionDAG::getNodeValueTypes and simplify
code that uses it by using SelectionDAG::getVTList instead.

llvm-svn: 68744
2009-04-09 23:54:40 +00:00
Dan Gohman f15454866c Fix grammaros in comments.
llvm-svn: 68666
2009-04-09 02:06:09 +00:00
Rafael Espindola 3b2df10c9e Re-apply 68552.
Tested by bootstrapping llvm-gcc and using that to build llvm.

llvm-svn: 68645
2009-04-08 21:14:34 +00:00
Rafael Espindola d173f4237d Avoid a hard coded constant.
llvm-svn: 68603
2009-04-08 08:09:33 +00:00
Dan Gohman ad3e549a53 Implement support for using modeling implicit-zero-extension on x86-64
with SUBREG_TO_REG, teach SimpleRegisterCoalescing to coalesce
SUBREG_TO_REG instructions (which are similar to INSERT_SUBREG
instructions), and teach the DAGCombiner to take advantage of this on
targets which support it. This eliminates many redundant
zero-extension operations on x86-64.

This adds a new TargetLowering hook, isZExtFree. It's similar to
isTruncateFree, except it only applies to actual definitions, and not
no-op truncates which may not zero the high bits.

Also, this adds a new optimization to SimplifyDemandedBits: transform
operations like x+y into (zext (add (trunc x), (trunc y))) on targets
where all the casts are no-ops. In contexts where the high part of the
add is explicitly masked off, this allows the mask operation to be
eliminated. Fix the DAGCombiner to avoid undoing these transformations
to eliminate casts on targets where the casts are no-ops.

Also, this adds a new two-address lowering heuristic. Since
two-address lowering runs before coalescing, it helps to be able to
look through copies when deciding whether commuting and/or
three-address conversion are profitable.

Also, fix a bug in LiveInterval::MergeInClobberRanges. It didn't handle
the case that a clobber range extended both before and beyond an
existing live range. In that case, multiple live ranges need to be
added. This was exposed by the new subreg coalescing code.

Remove 2008-05-06-SpillerBug.ll. It was bugpoint-reduced, and the
spiller behavior it was looking for no longer occurrs with the new
instruction selection.

llvm-svn: 68576
2009-04-08 00:15:30 +00:00
Bill Wendling 4aa25b79f9 Temporarily revert r68552. This was causing a failure in the self-hosting LLVM
builds.

--- Reverse-merging (from foreign repository) r68552 into '.':
U    test/CodeGen/X86/tls8.ll
U    test/CodeGen/X86/tls10.ll
U    test/CodeGen/X86/tls2.ll
U    test/CodeGen/X86/tls6.ll
U    lib/Target/X86/X86Instr64bit.td
U    lib/Target/X86/X86InstrSSE.td
U    lib/Target/X86/X86InstrInfo.td
U    lib/Target/X86/X86RegisterInfo.cpp
U    lib/Target/X86/X86ISelLowering.cpp
U    lib/Target/X86/X86CodeEmitter.cpp
U    lib/Target/X86/X86FastISel.cpp
U    lib/Target/X86/X86InstrInfo.h
U    lib/Target/X86/X86ISelDAGToDAG.cpp
U    lib/Target/X86/AsmPrinter/X86ATTAsmPrinter.cpp
U    lib/Target/X86/AsmPrinter/X86IntelAsmPrinter.cpp
U    lib/Target/X86/AsmPrinter/X86ATTAsmPrinter.h
U    lib/Target/X86/AsmPrinter/X86IntelAsmPrinter.h
U    lib/Target/X86/X86ISelLowering.h
U    lib/Target/X86/X86InstrInfo.cpp
U    lib/Target/X86/X86InstrBuilder.h
U    lib/Target/X86/X86RegisterInfo.td

llvm-svn: 68560
2009-04-07 22:35:25 +00:00
Rafael Espindola 1edda06792 Reduce code duplication on the TLS implementation.
This introduces a small regression on the generated code
quality in the case we are just computing addresses, not
loading values.

Will work on it and on X86-64 support.

llvm-svn: 68552
2009-04-07 21:37:46 +00:00
Mon P Wang 9c186c5d27 Added a x86 dag combine to increase the chances to use a
movq for v2i64 on x86-32.

llvm-svn: 68368
2009-04-03 02:43:30 +00:00
Chris Lattner d2eb0a63a1 silence warning in release-asserts build.
llvm-svn: 68253
2009-04-01 22:14:45 +00:00
Evan Cheng d9d6e427d6 i128 shift libcalls are not available on x86.
llvm-svn: 68133
2009-03-31 19:38:51 +00:00
Evan Cheng a84a318873 When optimzing a mul by immediate into two, the resulting mul's should get a x86 specific node to avoid dag combiner from hacking on them further.
llvm-svn: 68066
2009-03-30 21:36:47 +00:00
Rafael Espindola 6ff3dabbb4 Have only one definition of X86AddrNumOperands.
llvm-svn: 67949
2009-03-28 18:55:31 +00:00
Evan Cheng fd81c73cde Optimize some 64-bit multiplication by constants into two lea's or one lea + shl since imulq is slow (latency 5). e.g.
x * 40
=>
shlq    $3, %rdi
leaq    (%rdi,%rdi,4), %rax

This has the added benefit of allowing more multiply to be folded into addressing mode. e.g.
a * 24 + b
=>
leaq    (%rdi,%rdi,2), %rax
leaq    (%rsi,%rax,8), %rax

llvm-svn: 67917
2009-03-28 05:57:29 +00:00
Rafael Espindola e728019392 I am trying to add a segment to the X86 addresses matching to
improve TLS support (see http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20090309/075220.html), but that code is VERY brittle.

This patch just makes it a bit more resistant.

llvm-svn: 67843
2009-03-27 15:26:30 +00:00
Evan Cheng d88ebc352c -no-implicit-float means explicit fp operations are legal.
llvm-svn: 67784
2009-03-26 23:06:32 +00:00
Bill Wendling aa28be652c Pull transform from target-dependent code into target-independent code.
llvm-svn: 67742
2009-03-26 06:14:09 +00:00
Bill Wendling 94f299f2c5 Match this pattern so that we can generate simpler code:
%a = ...
  %b = and i32 %a, 2
  %c = srl i32 %b, 1
  %d = br i32 %c, 

into

  %a = ...
  %b = and %a, 2
  %c = X86ISD::CMP %b, 0
  %d = X86ISD::BRCOND %c ...

This applies only when the AND constant value has one bit set and the SRL
constant is equal to the log2 of the AND constant. The back-end is smart enough
to convert the result into a TEST/JMP sequence.

llvm-svn: 67728
2009-03-26 01:47:50 +00:00
Bill Wendling 798fd56d0f These instructions have special lowering that may lower them to SSE
instructions. Prevent that if we don't want implicit uses of SSE.

llvm-svn: 66877
2009-03-13 08:41:47 +00:00
Evan Cheng 1fb8aedd1e Fix some significant problems with constant pools that resulted in unnecessary paddings between constant pool entries, larger than necessary alignments (e.g. 8 byte alignment for .literal4 sections), and potentially other issues.
1. ConstantPoolSDNode alignment field is log2 value of the alignment requirement. This is not consistent with other SDNode variants.
2. MachineConstantPool alignment field is also a log2 value.
3. However, some places are creating ConstantPoolSDNode with alignment value rather than log2 values. This creates entries with artificially large alignments, e.g. 256 for SSE vector values.
4. Constant pool entry offsets are computed when they are created. However, asm printer group them by sections. That means the offsets are no longer valid. However, asm printer uses them to determine size of padding between entries.
5. Asm printer uses expensive data structure multimap to track constant pool entries by sections.
6. Asm printer iterate over SmallPtrSet when it's emitting constant pool entries. This is non-deterministic.


Solutions:
1. ConstantPoolSDNode alignment field is changed to keep non-log2 value.
2. MachineConstantPool alignment field is also changed to keep non-log2 value.
3. Functions that create ConstantPool nodes are passing in non-log2 alignments.
4. MachineConstantPoolEntry no longer keeps an offset field. It's replaced with an alignment field. Offsets are not computed when constant pool entries are created. They are computed on the fly in asm printer and JIT.
5. Asm printer uses cheaper data structure to group constant pool entries.
6. Asm printer compute entry offsets after grouping is done.
7. Change JIT code to compute entry offsets on the fly.

llvm-svn: 66875
2009-03-13 07:51:59 +00:00
Chris Lattner 99cc133710 generalize the previous code to use the full generality of LEA
for i32/i64 expressions (we could also do i16 on cpus where
i16 lea is fast, but I didn't add this).  On the example, we now
generate:

_test:
	movl	4(%esp), %eax
	cmpl	$42, (%eax)
	setl	%al
	movzbl	%al, %eax
	leal	4(%eax,%eax,8), %eax
	ret

instead of:

_test:
	movl	4(%esp), %eax
	cmpl	$41, (%eax)
	movl	$4, %ecx
	movl	$13, %eax
	cmovg	%ecx, %eax
	ret

llvm-svn: 66869
2009-03-13 05:53:31 +00:00
Chris Lattner 4be6df5d86 optimize the case of cond ? 42 : 41 and friends. This compiles the
example to:

_test:
	movl	4(%esp), %eax
	cmpl	$41, (%eax)
	setg	%al
	movzbl	%al, %eax
	orl	$4294967294, %eax
	ret

instead of:

        movl    4(%esp), %eax
        cmpl    $41, (%eax)
	movl	$4294967294, %ecx
	movl	$4294967295, %eax
	cmova	%ecx, %eax
	ret

which is smaller in code size and faster. rdar://6668608

llvm-svn: 66868
2009-03-13 05:22:11 +00:00
Chris Lattner 4147f08e44 Move 3 "(add (select cc, 0, c), x) -> (select cc, x, (add, x, c))"
related transformations out of target-specific dag combine into the
ARM backend.  These were added by Evan in r37685 with no testcases
and only seems to help ARM (e.g. test/CodeGen/ARM/select_xform.ll).

Add some simple X86-specific (for now) DAG combines that turn things
like cond ? 8 : 0  -> (zext(cond) << 3).  This happens frequently
with the recently added cp constant select optimization, but is a
very general xform.  For example, we now compile the second example
in const-select.ll to:

_test:
        movsd   LCPI2_0, %xmm0
        ucomisd 8(%esp), %xmm0
        seta    %al
        movzbl  %al, %eax
        movl    4(%esp), %ecx
        movsbl  (%ecx,%eax,4), %eax
        ret

instead of:

_test:
        movl    4(%esp), %eax
        leal    4(%eax), %ecx
        movsd   LCPI2_0, %xmm0
        ucomisd 8(%esp), %xmm0
        cmovbe  %eax, %ecx
        movsbl  (%ecx), %eax
        ret

This passes multisource and dejagnu.

llvm-svn: 66779
2009-03-12 06:52:53 +00:00