Commit Graph

300 Commits

Author SHA1 Message Date
Evan Cheng 9d768f4445 Cosmetic changes.
llvm-svn: 103155
2010-05-06 01:34:11 +00:00
Evan Cheng 718ff448df storeRegToStackSlot has forgotten about QPR_8 register class.
llvm-svn: 103154
2010-05-06 01:32:54 +00:00
Evan Cheng 250e917e9d Frame index can be negative.
llvm-svn: 102577
2010-04-29 01:13:30 +00:00
Jim Grosbach 04cbcca319 Add sizes non-floating point versions for the eh sjlj intrinsic expansions.
rdar://7895451

llvm-svn: 102526
2010-04-28 20:33:09 +00:00
Evan Cheng bcb99ecc18 Add ARM specific emitFrameIndexDebugValue.
llvm-svn: 102324
2010-04-26 07:39:25 +00:00
Dale Johannesen 60b289709e Educate GetInstrSizeInBytes implementations that
DBG_VALUE does not generate code.

llvm-svn: 100681
2010-04-07 19:51:44 +00:00
Chris Lattner 6f306d7d30 use DebugLoc default ctor instead of DebugLoc::getUnknownLoc()
llvm-svn: 100214
2010-04-02 20:16:16 +00:00
Dale Johannesen 4244d12769 Teach AnalyzeBranch, RemoveBranch and the branch
folder to be tolerant of debug info following the
branch(es) at the end of a block.

llvm-svn: 100168
2010-04-02 01:38:09 +00:00
Bob Wilson 59f75bba24 Fix VLDMQ and VSTMQ instructions to use the correct encoding and address modes.
These instructions are only needed for codegen, so I've removed all the
explicit encoding bits for now; they should be set in the same way as the for
VLDMD and VSTMD whenever we add encodings for VFP.  The use of addrmode5
requires that the instructions be custom-selected so that the number of
registers can be set in the AM5Opc value.

llvm-svn: 99309
2010-03-23 18:54:46 +00:00
Bob Wilson 9b680e21c0 Rename some instructions to match the corresponding NEON opcode.
llvm-svn: 99266
2010-03-23 06:26:18 +00:00
Bob Wilson cc0a2a75a0 Change VST1 instructions for loading Q register values to operate on pairs
of D registers.  Add a separate VST1q instruction with a Q register
source operand for use by storeRegToStackSlot.

llvm-svn: 99265
2010-03-23 06:20:33 +00:00
Bob Wilson 340861d29e Change VLD1 instructions for loading Q register values to operate on pairs
of D registers.  Add a separate VLD1q instruction with a Q register
destination operand for use by loadRegFromStackSlot.

llvm-svn: 99261
2010-03-23 05:25:43 +00:00
Bob Wilson ae08a736d6 Re-commit r98683 ("remove redundant writeback flag from ARM address mode 6")
with changes to add a separate optional register update argument.  Change all
the NEON instructions with address register writeback to use it.

llvm-svn: 99095
2010-03-20 22:13:40 +00:00
Anton Korobeynikov 422dd6608a Refactor Reg-Reg copy emission routine for ARM. This makes cross-regclass copies weirdness more straightforward. Also, add GPR <-> SPR copy support.
llvm-svn: 98887
2010-03-18 22:35:02 +00:00
Bob Wilson c7ba918b84 Revert 98683. It is breaking something in the disassembler.
llvm-svn: 98692
2010-03-16 23:01:13 +00:00
Bob Wilson c953bca10b Remove redundant writeback flag from ARM address mode 6. Also remove the
optional register update argument, which is currently unused -- when we add
support for that, it can just be a separate operand.

llvm-svn: 98683
2010-03-16 21:44:40 +00:00
Evan Cheng e9c46c25a1 - Change MachineInstr::isIdenticalTo to take a new option that determines whether it should skip checking defs or at least virtual register defs. This subsumes part of the TargetInstrInfo::isIdentical functionality.
- Eliminate TargetInstrInfo::isIdentical and replace it with produceSameValue. In the default case, produceSameValue just checks whether two machine instructions are identical (except for virtual register defs). But targets may override it to check for unusual cases (e.g. ARM pic loads from constant pools).

llvm-svn: 97628
2010-03-03 01:44:33 +00:00
Bob Wilson 37f106e18c Handle tGPR register class in a few more places. This fixes some llvm-gcc
build failures due to my fix for pr6111.

llvm-svn: 96402
2010-02-16 22:01:59 +00:00
Bob Wilson 70aa8d0745 Fix pr6111: Avoid using the LR register for the target address of an indirect
branch in ARM v4 code, since it gets clobbered by the return address before
it is used.  Instead of adding a new register class containing all the GPRs
except LR, just use the existing tGPR class.

llvm-svn: 96360
2010-02-16 17:24:15 +00:00
Chris Lattner b06015aa69 move target-independent opcodes out of TargetInstrInfo
into TargetOpcodes.h.  #include the new TargetOpcodes.h
into MachineInstr.  Add new inline accessors (like isPHI())
to MachineInstr, and start using them throughout the 
codebase.

llvm-svn: 95687
2010-02-09 19:54:29 +00:00
Jim Grosbach a570d05228 tighten up eh.setjmp sequence a bit.
llvm-svn: 95603
2010-02-08 23:22:00 +00:00
Jim Grosbach a3575ca846 Adjust setjmp instruction sequence to not need 32-bit alignment padding
llvm-svn: 94627
2010-01-27 00:07:20 +00:00
Chris Lattner a14ac3fd80 prep work to support a future where getJumpTableInfo will return
a null pointer for functions with no jump tables.  No functionality
change.

llvm-svn: 94469
2010-01-25 23:22:00 +00:00
Jim Grosbach 04770f2aa1 For aligned load/store instructions, it's only required to know whether a
function can support dynamic stack realignment. That's a much easier question
to answer at instruction selection stage than whether the function actually
will have dynamic alignment prologue. This allows the removal of the
stack alignment heuristic pass, and improves code quality for cases where
the heuristic would result in dynamic alignment code being generated when
it was not strictly necessary.

llvm-svn: 93885
2010-01-19 18:31:11 +00:00
Jakob Stoklund Olesen 29a64c9575 Add Target hook to duplicate machine instructions.
Some instructions refer to unique labels, and so cannot be trivially cloned
with CloneMachineInstr.

llvm-svn: 92873
2010-01-06 23:47:07 +00:00
Bill Wendling a91b8207cc Remove dead variable.
llvm-svn: 92193
2009-12-28 01:57:39 +00:00
Jim Grosbach 5f9f721e95 remove out of date FIXME.
llvm-svn: 90490
2009-12-03 21:55:01 +00:00
Chris Lattner c831fac043 fix a build problem with VC++, PR5664, patch by Alp Toker!
llvm-svn: 90419
2009-12-03 06:58:32 +00:00
Jim Grosbach 36d4dec28a Thumb1 exception handling setjmp
llvm-svn: 90246
2009-12-01 18:10:36 +00:00
Bob Wilson 505ddaa4dc Remove isProfitableToDuplicateIndirectBranch target hook. It is profitable
for all the processors where I have tried it, and even when it might not help
performance, the cost is quite low.  The opportunities for duplicating
indirect branches are limited by other factors so code size does not change
much due to tail duplicating indirect branches aggressively.

llvm-svn: 90144
2009-11-30 18:35:03 +00:00
Bob Wilson d4d40670e8 Refactor target hook for tail duplication as requested by Chris.
Make tail duplication of indirect branches much more aggressive (for targets
that indicate that it is profitable), based on further experience with
this transformation.  I compiled 3 large applications with and without
this more aggressive tail duplication and measured minimal changes in code
size.  ("size" on Darwin seems to round the text size up to the nearest
page boundary, so I can only say that any code size increase was less than
one 4k page.) Radar 7421267.

llvm-svn: 89814
2009-11-24 23:35:49 +00:00
Evan Cheng 184ec26fcd Enable predication of NEON instructions in Thumb2 mode.
llvm-svn: 89748
2009-11-24 08:06:15 +00:00
Evan Cheng a33fc86be3 Add predicate operand to NEON instructions. Fix lots (but not all) 80 col violations in ARMInstrNEON.td.
llvm-svn: 89542
2009-11-21 06:21:52 +00:00
Evan Cheng bbd50b0f78 Also CSE non-pic load from constant pools.
llvm-svn: 89440
2009-11-20 02:10:27 +00:00
Bob Wilson 290e9a47a9 Add a target hook to allow changing the tail duplication limit based on the
contents of the block to be duplicated.  Use this for ARM Cortex A8/9 to
be more aggressive tail duplicating indirect branches, since it makes it
much more likely that they will be predicted in the branch target buffer.
Testcase coming soon.

llvm-svn: 89187
2009-11-18 03:34:27 +00:00
Jim Grosbach 01c1cae34d Detect need for autoalignment of the stack earlier to catch spills more
conservatively. eliminateFrameIndex() machinery adjust to handle addr mode
6 (vld1/vst1) used for spills. Fix tests to expect aligned Q-reg spilling

llvm-svn: 88874
2009-11-15 21:45:34 +00:00
Jim Grosbach 74ae3e5b0e set the def of the VLD1q64 properly
llvm-svn: 88873
2009-11-15 21:05:07 +00:00
Evan Cheng 6ad7da96fe - Change TargetInstrInfo::reMaterialize to pass in TargetRegisterInfo.
- If destination is a physical register and it has a subreg index, use the
  sub-register instead.
This fixes PR5423.

llvm-svn: 88745
2009-11-14 02:55:43 +00:00
Jim Grosbach d7cf55cd0e Use Unified Assembly Syntax for the ARM backend.
llvm-svn: 86494
2009-11-09 00:11:35 +00:00
Jim Grosbach a15c3b7124 Use aligned load/store instructions for spilling Q registers when we know the stack slot is 128 bit aligned
llvm-svn: 86425
2009-11-08 00:27:19 +00:00
Evan Cheng fe864425cb Refactor code.
llvm-svn: 86423
2009-11-08 00:15:23 +00:00
Jim Grosbach 4e9f379554 80-column cleanup of file header comments
llvm-svn: 86408
2009-11-07 22:00:39 +00:00
Evan Cheng a8e8a7c976 Refactor code. Fix a potential missing check. Teach isIdentical() about tLDRpci_pic.
llvm-svn: 86330
2009-11-07 04:04:34 +00:00
Evan Cheng b376ce0169 Fix t2Int_eh_sjlj_setjmp. Immediate form of orr is a 32-bit instruction. So it should be 22 bytes instead of 20 bytes long.
llvm-svn: 85965
2009-11-03 23:13:34 +00:00
Evan Cheng 31c2f4701b Trim unnecessary include.
llvm-svn: 85878
2009-11-03 07:08:08 +00:00
Evan Cheng 23c009f125 Clean up copyRegToReg.
llvm-svn: 85870
2009-11-03 05:51:39 +00:00
Anton Korobeynikov d195f9e5c3 Turn neon reg-reg moves fixup code into separate pass. This should reduce the compile time.
llvm-svn: 85850
2009-11-03 01:04:26 +00:00
Evan Cheng 1708b06c0e Unbreak ARMBaseRegisterInfo::copyRegToReg.
llvm-svn: 85787
2009-11-02 04:44:55 +00:00
Anton Korobeynikov 14635da94b Use NEON reg-reg moves, where profitable. This reduces "domain-cross" stalls, when we used to mix vfp and neon code (the former were used for reg-reg moves)
llvm-svn: 85764
2009-11-02 00:10:38 +00:00
Bob Wilson 73789b848d Add a Thumb BRIND pattern. Change the ARM BRIND assembly to separate the
opcode and operand with a tab.  Check for these instructions in the usual
places.

llvm-svn: 85411
2009-10-28 18:26:41 +00:00
Evan Cheng 5d1b849658 Don't forget subreg indices when folding load / store.
llvm-svn: 85048
2009-10-25 07:52:27 +00:00
Evan Cheng 46ed1f8341 80 col violation.
llvm-svn: 84986
2009-10-24 02:07:42 +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 4729191bb2 Distinquish stack slots from other stack objects. They (and fixed objects) get FixedStack PseudoSourceValues.
llvm-svn: 84326
2009-10-17 09:20:14 +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
Anton Korobeynikov 75b59fb055 Add PseudoSourceValues for constpool stuff on ELF (Darwin should use something similar)
and register spills.

llvm-svn: 83435
2009-10-07 00:06:35 +00:00
Jakob Stoklund Olesen dc9efe8078 Introduce the TargetInstrInfo::KILL machine instruction and get rid of the
unused DECLARE instruction.

KILL is not yet used anywhere, it will replace TargetInstrInfo::IMPLICIT_DEF
in the places where IMPLICIT_DEF is just used to alter liveness of physical
registers.

llvm-svn: 83006
2009-09-28 20:32:26 +00:00
Evan Cheng 83e0d481ae Make ARM and Thumb2 32-bit immediate materialization into a single 32-bit pseudo
instruction. This makes it re-materializable.

Thumb2 will split it back out into two instructions so IT pass will generate the
right mask. Also, this expose opportunies to optimize the movw to a 16-bit move.

llvm-svn: 82982
2009-09-28 09:14:39 +00:00
Anton Korobeynikov 8d0fbebb9f Add QPR_VFP2 regclass and add copy_to_regclass nodes, where needed to
constraint the register usage.

llvm-svn: 81635
2009-09-12 22:21:08 +00:00
Anton Korobeynikov 59e2b8e894 Add NEON 'laned' operations. This fixes another bunch of gcc testsuite fails and
makes the code faster.

llvm-svn: 81220
2009-09-08 15:22:32 +00:00
Evan Cheng 7a37b1a2ca Fix PR4789. Teach eliminateFrameIndex how to handle VLDRQ and VSTRQ which cannot fold any immediate offset.
llvm-svn: 80191
2009-08-27 01:23:50 +00:00
Chris Lattner e9a75a6654 rename TAI -> MAI, being careful not to make MAILJMP instructions :)
llvm-svn: 79777
2009-08-22 21:43:10 +00:00
Chris Lattner 7b26fce23e Rename TargetAsmInfo (and its subclasses) to MCAsmInfo.
llvm-svn: 79763
2009-08-22 20:48:53 +00:00
Devang Patel 0939595711 Record variable debug info at ISel time directly.
llvm-svn: 79742
2009-08-22 17:12:53 +00:00
Jim Grosbach 841850ed26 Add Thumb2 eh_sjlj_setjmp implementation
llvm-svn: 78701
2009-08-11 19:42:21 +00:00
Jim Grosbach 1d5350c08f fix GetInstSizeInBytes for eh_sjlj_setjmp
llvm-svn: 78683
2009-08-11 17:08:15 +00:00
Jim Grosbach f24f9d9cb6 Whitespace cleanup. Remove trailing whitespace.
llvm-svn: 78666
2009-08-11 15:33:49 +00:00
Evan Cheng 092b701a2c Add support for folding loads / stores into 16-bit moves used by Thumb2.
llvm-svn: 78558
2009-08-10 06:32:05 +00:00
Evan Cheng 55c014a9f3 80 col violation.
llvm-svn: 78557
2009-08-10 05:51:48 +00:00
Anton Korobeynikov 887d05ce9b Use VLDM / VSTM to spill/reload 128-bit Neon registers
llvm-svn: 78468
2009-08-08 13:35:48 +00:00
Evan Cheng 2aa91cc2be Code refactoring. No functionality change.
llvm-svn: 78455
2009-08-08 03:20:32 +00:00
Evan Cheng 4c3b1ca5a0 Fix support to use NEON for single precision fp math.
llvm-svn: 78397
2009-08-07 19:30:41 +00:00
Evan Cheng b972e5633f It turns out most of the thumb2 instructions are not allowed to touch SP. The semantics of such instructions are unpredictable. We have just been lucky that tests have been passing.
This patch takes pain to ensure all the PEI lowering code does the right thing when lowering frame indices, insert code to manipulate stack pointers, etc. It's also custom lowering dynamic stack alloc into pseudo instructions so we can insert the right instructions at scheduling time.

This fixes PR4659 and PR4682.

llvm-svn: 78361
2009-08-07 00:34:42 +00:00
David Goodwin e5b5d8fbb3 When using NEON for single-precision FP, the NEON result must be placed in D0-D15 as these are the only D registers with S subregs. Introduce a new regclass to represent D0-D15 and use it in the NEON single-precision FP patterns.
llvm-svn: 78244
2009-08-05 21:02:22 +00:00
Chris Lattner e98a3c3ca3 Move the getInlineAsmLength virtual method from TAI to TII, where
the only real caller (GetFunctionSizeInBytes) uses it.

The custom ARM implementation of this is basically reimplementing
an assembler poorly for negligible gain.  It should be removed 
IMNSHO, but I'll leave that to ARMish folks to decide.

llvm-svn: 77877
2009-08-02 05:20:37 +00:00
Evan Cheng e64f48ba8b Workaround a couple of Darwin assembler bugs.
llvm-svn: 77781
2009-08-01 06:13:52 +00:00
Evan Cheng 95d6325859 t2BR_JT is mov pc, it's 2 byte long, not 4.
llvm-svn: 77744
2009-07-31 22:22:22 +00:00
Evan Cheng f6d0fa3d33 - Teach TBB / TBH offset limits are 510 and 131070 respectively since the offset
is scaled by two.
- Teach GetInstSizeInBytes about TBB and TBH.

llvm-svn: 77701
2009-07-31 18:28:05 +00:00
Evan Cheng 780748d565 - More refactoring. This gets rid of all of the getOpcode calls.
- This change also makes it possible to switch between ARM / Thumb on a
  per-function basis.
- Fixed thumb2 routine which expand reg + arbitrary immediate. It was using
  using ARM so_imm logic.
- Use movw and movt to do reg + imm when profitable.
- Other code clean ups and minor optimizations.

llvm-svn: 77300
2009-07-28 05:48:47 +00:00
Evan Cheng 0e075e2429 convertToThreeAddress can't handle Thumb2 instructions (which don't have same address mode as ARM instructions).
llvm-svn: 77230
2009-07-27 18:44:00 +00:00
Evan Cheng 8f2ed1bc5a Clean up.
llvm-svn: 77221
2009-07-27 18:25:24 +00:00
Evan Cheng 056c669e93 Get rid of some more getOpcode calls.
This also fixes potential problems in ARMBaseInstrInfo routines not recognizing thumb1 instructions when 32-bit and 16-bit instructions mix.

llvm-svn: 77218
2009-07-27 18:20:05 +00:00
Evan Cheng 371ec9e810 If CPSR is modified but the def is dead, then it's ok to fold the load / store.
llvm-svn: 77182
2009-07-27 04:18:04 +00:00
Evan Cheng c47e109335 Use t2LDRi12 and t2STRi12 to load / store to / from stack frames. Eliminate more getOpcode calls.
llvm-svn: 77181
2009-07-27 03:14:20 +00:00
Evan Cheng 186332f898 Use the right instructions to copy between GPR and the more strictive tGPR classes. t2MOV does not match the RC requirements.
llvm-svn: 77175
2009-07-27 00:33:08 +00:00
Evan Cheng 0e5b149930 Merge isLoadFromStackSlot into one since it behaves the same regardless of sub-target.
llvm-svn: 77174
2009-07-27 00:24:36 +00:00
Evan Cheng 26b51b15ed Just use a single isMoveInstr to catch all the cases.
llvm-svn: 77173
2009-07-27 00:05:15 +00:00
Evan Cheng f3a1fce8ae Change Thumb2 jumptable codegen to one that uses two level jumps:
Before:
      adr r12, #LJTI3_0_0
      ldr pc, [r12, +r0, lsl #2]
LJTI3_0_0:
      .long    LBB3_24
      .long    LBB3_30
      .long    LBB3_31
      .long    LBB3_32

After:
      adr r12, #LJTI3_0_0
      add pc, r12, +r0, lsl #2
LJTI3_0_0:
      b.w    LBB3_24
      b.w    LBB3_30
      b.w    LBB3_31
      b.w    LBB3_32

This has several advantages.
1. This will make it easier to optimize this to a TBB / TBH instruction +
   (smaller) table.
2. This eliminate the need for ugly asm printer hack to force the address
   into thumb addresses (bit 0 is one).
3. Same codegen for pic and non-pic.
4. This eliminate the need to align the table so constantpool island pass
   won't have to over-estimate the size.

Based on my calculation, the later is probably slightly faster as well since
ldr pc with shifter address is very slow. That is, it should be a win as long
as the HW implementation can do a reasonable job of branch predict the second
branch.

llvm-svn: 77024
2009-07-25 00:33:29 +00:00
Evan Cheng 666c912ce3 Make sure thumb2 jumptable entries are aligned.
llvm-svn: 76986
2009-07-24 18:20:44 +00:00
Eli Friedman 95fc6ee51a Remove unused member functions.
llvm-svn: 76960
2009-07-24 07:43:59 +00:00
Evan Cheng 6cfbe61361 FLDD, FLDS, FCPYD, FCPYS, FSTD, FSTS, VMOVD, VMOVQ maps to the same instructions on all sub-targets.
llvm-svn: 76925
2009-07-24 00:53:56 +00:00
David Goodwin cdd405d804 Correctly handle the Thumb-2 imm8 addrmode. Specialize frame index elimination more exactly for Thumb-2 to get better code gen.
llvm-svn: 76919
2009-07-24 00:16:18 +00:00
Anton Korobeynikov c5df7e2dc1 Emit cross regclass register moves for thumb2.
Minor code duplication cleanup.

llvm-svn: 76124
2009-07-16 23:26:06 +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
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
Evan Cheng bf041366c4 80 col violation.
llvm-svn: 75358
2009-07-11 06:37:27 +00:00
Evan Cheng 3b88dd6900 Move isPredicated from .cpp to .h
llvm-svn: 75217
2009-07-10 01:38:27 +00:00
Evan Cheng e3a53c448b Change how so_imm and t2_so_imm are handled. At instruction selection time, the immediates are no longer encoded in the imm8 + rot format, that are left as it is. The encoding is now done in ams printing and code emission time instead.
llvm-svn: 75048
2009-07-08 21:03:57 +00:00
David Goodwin af7451b674 Checkpoint Thumb2 Instr info work. Generalized base code so that it can be shared between ARM and Thumb2. Not yet activated because register information must be generalized first.
llvm-svn: 75010
2009-07-08 16:09:28 +00:00