Commit Graph

152 Commits

Author SHA1 Message Date
Daniel Dunbar
912225e185 ADT/Triple: Move a variety of clients to using isOSDarwin() and isOSWindows()
predicates.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@129816 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-19 21:14:45 +00:00
Bob Wilson
84c5eed15b This patch combines several changes from Evan Cheng for rdar://8659675.
Making use of VFP / NEON floating point multiply-accumulate / subtraction is
difficult on current ARM implementations for a few reasons.
1. Even though a single vmla has latency that is one cycle shorter than a pair
   of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause
   additional pipeline stall. So it's frequently better to single codegen
   vmul + vadd.
2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to
   stall for 4 cycles. We need to schedule them apart.
3. A vmla followed vmla is a special case. Obvious issuing back to back RAW
   vmla + vmla is very bad. But this isn't ideal either:
     vmul
     vadd
     vmla
   Instead, we want to expand the second vmla:
     vmla
     vmul
     vadd
   Even with the 4 cycle vmul stall, the second sequence is still 2 cycles
   faster.

Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough
but it isn't the optimial solution. This patch attempts to make it possible to
use vmla / vmls in cases where it is profitable.

A. Add missing isel predicates which cause vmla to be codegen'ed.
B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to
   compute a fmul and a fmla.
C. Add additional isel checks for vmla, avoid cases where vmla is feeding into
   fp instructions (except for the #3 exceptional case).
D. Add ARM hazard recognizer to model the vmla / vmls hazards.
E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the
   vmla / vmls will trigger one of the special hazards.

Enable these fp vmlx codegen changes for Cortex-A9.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@129775 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-19 18:11:57 +00:00
Jim Grosbach
f22eefba68 Tidy up.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@129034 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-06 22:35:47 +00:00
NAKAMURA Takumi
6904f05e60 Triple::MinGW64 is deprecated and removed. We can use Triple::MinGW32 generally.
No one uses *-mingw64. mingw-w64 is represented as {i686|x86_64}-w64-mingw32. In llvm side, i686 and x64 can be treated as similar way.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125747 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-17 12:24:17 +00:00
Rafael Espindola
96aa78c8c5 Add support for the --noexecstack option.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@124077 91177308-0d34-0410-b5e6-96231b3b80d8
2011-01-23 17:55:27 +00:00
Anton Korobeynikov
16c29b5f28 Rename TargetFrameInfo into TargetFrameLowering. Also, put couple of FIXMEs and fixes here and there.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@123170 91177308-0d34-0410-b5e6-96231b3b80d8
2011-01-10 12:39:04 +00:00
Evan Cheng
48575f6ea7 Making use of VFP / NEON floating point multiply-accumulate / subtraction is
difficult on current ARM implementations for a few reasons.
1. Even though a single vmla has latency that is one cycle shorter than a pair
   of vmul + vadd, a RAW hazard during the first (4? on Cortex-a8) can cause
   additional pipeline stall. So it's frequently better to single codegen
   vmul + vadd.
2. A vmla folowed by a vmul, vmadd, or vsub causes the second fp instruction to
   stall for 4 cycles. We need to schedule them apart.
3. A vmla followed vmla is a special case. Obvious issuing back to back RAW
   vmla + vmla is very bad. But this isn't ideal either:
     vmul
     vadd
     vmla
   Instead, we want to expand the second vmla:
     vmla
     vmul
     vadd
   Even with the 4 cycle vmul stall, the second sequence is still 2 cycles
   faster.

Up to now, isel simply avoid codegen'ing fp vmla / vmls. This works well enough
but it isn't the optimial solution. This patch attempts to make it possible to
use vmla / vmls in cases where it is profitable.

A. Add missing isel predicates which cause vmla to be codegen'ed.
B. Make sure the fmul in (fadd (fmul)) has a single use. We don't want to
   compute a fmul and a fmla.
C. Add additional isel checks for vmla, avoid cases where vmla is feeding into
   fp instructions (except for the #3 exceptional case).
D. Add ARM hazard recognizer to model the vmla / vmls hazards.
E. Add a special pre-regalloc case to expand vmla / vmls when it's likely the
   vmla / vmls will trigger one of the special hazards.

Work in progress, only A+B are enabled.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120960 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-05 22:04:16 +00:00
Chris Lattner
b75c651e22 tidy up
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119462 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-17 05:41:32 +00:00
Anton Korobeynikov
3346491223 First step of huge frame-related refactoring: move emit{Prologue,Epilogue} out of TargetRegisterInfo to TargetFrameInfo, which is definitely much better suitable place
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119097 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-15 00:06:54 +00:00
Eric Christopher
79ab2fe01a Revert the accidental commit I made reverting the previous commit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@118835 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-11 20:50:14 +00:00
Eric Christopher
6c50119ba3 Revert this temporarily.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@118827 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-11 19:47:02 +00:00
Rafael Espindola
0febc4657b Jim Asked us to move DataLayout on ARM back to the most specialized classes. Do
so and also change X86 for consistency.

Investigating if this can be improved a bit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@115469 91177308-0d34-0410-b5e6-96231b3b80d8
2010-10-03 18:59:45 +00:00
Jason W Kim
d4d4f4f488 I added a new file ARMAsmBackend which stubs out in similar ways to
the eqv X86 class.
For now, I split the ELFARMAsmBackend from the DarwinARMAsmBackend
(also mimicking X86)

Tested against -r115126



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@115129 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-30 02:17:26 +00:00
Nick Lewycky
a6136b7156 Resolve this GCC warning:
ARMTargetMachine.cpp:53: error: control reaches end of non-void function


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@114992 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-28 21:40:26 +00:00
Rafael Espindola
fd9493d74e Odd additional stub framework for the ARM MC ELF emission.
llc now recognizes the "intent" to support MC/obj emission for ARM, but
given that they are all stubs, it asserts on --filetype=obj --march=arm

Patch by Jason Kim.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@114856 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-27 18:31:37 +00:00
Bob Wilson
bd916c54b7 Convert some VTBL and VTBX instructions to use pseudo instructions prior to
register allocation.  Remove the NEONPreAllocPass, which is no longer needed.
Yeah!!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@113818 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-13 23:55:10 +00:00
Evan Cheng
7b4d31176e Report error if codegen tries to instantiate a ARM target when the cpu does support it. e.g. cortex-m* processors.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@110798 91177308-0d34-0410-b5e6-96231b3b80d8
2010-08-11 07:17:46 +00:00
Evan Cheng
e44be63816 Change -prefer-32bit-thumb to attribute -mattr=+32bit instead to disable more 32-bit to 16-bit optimizations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@110584 91177308-0d34-0410-b5e6-96231b3b80d8
2010-08-09 18:35:19 +00:00
Evan Cheng
e8846feaa1 Add an option to disable 32 -> 16-bit Thumb2 size reduction pass for experimentation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@110579 91177308-0d34-0410-b5e6-96231b3b80d8
2010-08-09 17:16:10 +00:00
Anton Korobeynikov
cec36f4c11 Hook in GlobalMerge pass
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@109359 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-24 21:52:08 +00:00
Evan Cheng
dca653951c Remove early IT block formation. It's not used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@107513 91177308-0d34-0410-b5e6-96231b3b80d8
2010-07-02 21:07:09 +00:00
Bob Wilson
ade57fa619 Add missing ARM and Thumb data layout info for vector types.
Radar 8128745.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106820 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-25 04:41:08 +00:00
Evan Cheng
8acf67672b Oops. IT block formation pass needs to be run at any optimization level.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106775 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-24 19:10:14 +00:00
Evan Cheng
96c3da6436 Move ARM if-conversion before post-ra scheduling.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106355 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-18 23:32:07 +00:00
Evan Cheng
86050dc8cc Allow ARM if-converter to be run after post allocation scheduling.
- This fixed a number of bugs in if-converter, tail merging, and post-allocation
  scheduler. If-converter now runs branch folding / tail merging first to
  maximize if-conversion opportunities.
- Also changed the t2IT instruction slightly. It now defines the ITSTATE
  register which is read by instructions in the IT block.
- Added Thumb2 specific hazard recognizer to ensure the scheduler doesn't
  change the instruction ordering in the IT block (since IT mask has been
  finalized). It also ensures no other instructions can be scheduled between
  instructions in the IT block.

This is not yet enabled.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106344 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-18 23:09:54 +00:00
Evan Cheng
46df4eb46e Make post-ra scheduling, anti-dep breaking, and register scavenger (conservatively) aware of predicated instructions. This enables ARM to move if-conversion before post-ra scheduler.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@106091 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-16 07:35:02 +00:00
Evan Cheng
34aa423814 Typo.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105677 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-09 03:49:12 +00:00
Evan Cheng
d847124211 Thumb2 IT blocks are fairly expensive. When there are multiple selects using
the same condition, it's important to make sure they are scheduled together
to avoid forming multiple IT blocks. I'm adding a pre-regalloc pass that forms
IT blocks early (by re-scheduling instructions and split basic blocks) to
attempt to fix this. This is not turned on by default since I am not sure this
is the right fix.

Another issue is llvm selects are modeled as two-address conditional moves.
This can be very bad when the copies before the conditional moves are not
coalesced away. Teach IT formation pass to move the copies above the IT block
(when legal) to avoid breaking the IT block.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105669 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-09 01:46:50 +00:00
Dan Gohman
ff7a562751 Implement a bunch more TargetSelectionDAGInfo infrastructure.
Move EmitTargetCodeForMemcpy, EmitTargetCodeForMemset, and
EmitTargetCodeForMemmove out of TargetLowering and into
SelectionDAGInfo to exercise this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@103481 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-11 17:31:57 +00:00
Anton Korobeynikov
4b38debf59 Remove late ARM codegen optimization pass committed by accident.
It is not ready for public yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100673 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-07 18:23:27 +00:00
Anton Korobeynikov
ebd4cb43c2 Move NEON-VFP domain fixer upper, so post-RA scheduler would benefit from it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100668 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-07 18:21:46 +00:00
Anton Korobeynikov
1e7b324fe5 Some initial version of global merger
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@100641 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-07 18:19:07 +00:00
Daniel Dunbar
5d067fe158 TargetRegistry: Fix create{AsmInfo,MCDisassembler} to return non-const objects.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99097 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-20 22:36:22 +00:00
Chris Lattner
e0faa54705 remove dead code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@95134 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-02 21:38:59 +00:00
Chris Lattner
f1d6b107d2 eliminate all the dead addSimpleCodeEmitter implementations.
eliminate random "code emitter" stuff in Alpha, except for
the JIT path.  Next up, remove the template cruft.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@95131 91177308-0d34-0410-b5e6-96231b3b80d8
2010-02-02 21:31:47 +00:00
Jim Grosbach
e45ab8a0a9 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.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@93885 91177308-0d34-0410-b5e6-96231b3b80d8
2010-01-19 18:31:11 +00:00
Jim Grosbach
e27d205d5d Factor the stack alignment calculations out into a target independent pass.
No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90336 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-02 19:30:24 +00:00
Jim Grosbach
a44321776e 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

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@88874 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-15 21:45:34 +00:00
Chris Lattner
59a9178fbe indicate what the native integer types for the target are.
Please verify.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86397 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-07 19:07:32 +00:00
Evan Cheng
b9803a8fa6 - Add pseudo instructions tLDRpci_pic and t2LDRpci_pic which does a pc-relative
load of a GV from constantpool and then add pc. It allows the code sequence to
  be rematerializable so it would be hoisted by machine licm.
- Add a late pass to break these pseudo instructions into a number of real
  instructions. Also move the code in Thumb2 IT pass that breaks up t2MOVi32imm
  to this pass. This is done before post regalloc scheduling to allow the
  scheduler to proper schedule these instructions. It also allow them to be
  if-converted and shrunk by later passes.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86304 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-06 23:52:48 +00:00
Daniel Dunbar
2928c83b01 Pass StringRef by value.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86251 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-06 10:58:06 +00:00
Anton Korobeynikov
747409a290 Move subtarget check upper for NEON reg-reg fixup pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85914 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-03 18:46:11 +00:00
Anton Korobeynikov
7aaf94bb0d Turn neon reg-reg moves fixup code into separate pass. This should reduce the compile time.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85850 91177308-0d34-0410-b5e6-96231b3b80d8
2009-11-03 01:04:26 +00:00
Bob Wilson
a597103c32 Revert r85346 change to control tail merging by CodeGenOpt::Level.
I'm going to redo this using the OptimizeForSize function attribute.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85426 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 20:46:46 +00:00
Bob Wilson
cd4f04d6bc Record CodeGen optimization level in the BranchFolding pass so that we can
use it to control tail merging when there is a tradeoff between performance
and code size.  When there is only 1 instruction in the common tail, we have
been merging.  That can be good for code size but is a definite loss for
performance.  Now we will avoid tail merging in that case when the
optimization level is "Aggressive", i.e., "-O3".  Radar 7338114.

Since the IfConversion pass invokes BranchFolding, it too needs to know
the optimization level.  Note that I removed the RegisterPass instantiation
for IfConversion because it required a default constructor.  If someone
wants to keep that for some reason, we can add a default constructor with
a hard-wired optimization level.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85346 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-27 23:49:38 +00:00
Bob Wilson
bac6ed4ba4 Revert 84843. Evan, this was breaking some of the if-conversion tests.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84868 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-22 16:52:21 +00:00
Evan Cheng
87689d3b70 Move if-conversion before post-regalloc scheduling so the predicated instruction get scheduled properly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84843 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-22 06:48:32 +00:00
Evan Cheng
268c79350e Trim include.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84831 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-22 05:08:49 +00:00
Evan Cheng
62a1b5db44 Move load / store multiple before post-alloc scheduling.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83236 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-02 04:57:15 +00:00
Evan Cheng
792e1f6df9 Add a option which would move ld/st multiple pass before post-alloc scheduling.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@83145 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-30 08:53:01 +00:00