Commit Graph

26336 Commits

Author SHA1 Message Date
Benjamin Kramer 951b15eb09 Revamp error checking in the ms inline asm parser.
- Actually abort when an error occurred.
- Check that the frontend lookup worked when parsing length/size/type operators.

Tested by a clang test. PR18096.

llvm-svn: 196044
2013-12-01 11:47:42 +00:00
Hal Finkel 42daeae9bd Add a scheduling model (with itinerary) for the PPC POWER7
This adds a scheduling model for the POWER7 (P7) core, and enables the
machine-instruction scheduler when targeting the P7. Scheduling for the P7,
like earlier ooo PPC cores, requires considering both dispatch group hazards,
and functional unit resources and latencies. These are both modeled in a
combined itinerary. Dispatch group formation is still handled by the post-RA
scheduler (which still needs to be updated for the P7, but nevertheless does a
pretty good job).

One interesting aspect of this change is that I've also enabled to use of AA
duing CodeGen for the P7 (just as it is for the embedded cores). The benchmark
results seem to support this decision (see below), and while this is normally
useful for in-order cores, and not for ooo cores like the P7, I think that the
dispatch slot hazards are enough like in-order resources to make the AA useful.

Test suite significant performance differences (where negative is a speedup,
and positive is a regression) vs. the current situation:

MultiSource/Benchmarks/BitBench/drop3/drop3
  with AA: N/A
  without AA: -28.7614% +/- 19.8356%
(significantly against AA)

MultiSource/Benchmarks/FreeBench/neural/neural
  with AA: -17.7406% +/- 11.2712%
  without AA: N/A
(significantly in favor of AA)

MultiSource/Benchmarks/SciMark2-C/scimark2
  with AA: -11.2079% +/- 1.80543%
  without AA: -11.3263% +/- 2.79651%

MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt
  with AA: -41.8649% +/- 17.0053%
  without AA: -34.5256% +/- 23.7072%

MultiSource/Benchmarks/mafft/pairlocalalign
  with AA: 25.3016% +/- 17.8614%
  without AA: 38.6629% +/- 14.9391%
(significantly in favor of AA)

MultiSource/Benchmarks/sim/sim
  with AA: N/A
  without AA: 13.4844% +/- 7.18195%
(significantly in favor of AA)

SingleSource/Benchmarks/BenchmarkGame/Large/fasta
  with AA: 15.0664% +/- 6.70216%
  without AA: 12.7747% +/- 8.43043%

SingleSource/Benchmarks/BenchmarkGame/puzzle
  with AA: 82.2713% +/- 26.3567%
  without AA: 75.7525% +/- 41.1842%

SingleSource/Benchmarks/Misc/flops-2
  with AA: -37.1621% +/- 20.7964%
  without AA: -35.2342% +/- 20.2999%
(significantly in favor of AA)

These are 99.5% confidence intervals from 5 runs per configuration. Regarding
the choice to turn on AA during CodeGen, of these results, four seem
significantly in favor of using AA, and one seems significantly against. I'm
not making this decision based on these numbers alone, but these results
seem consistent with results I have from other tests, and so I think that, on
balance, using AA is a win.

llvm-svn: 195981
2013-11-30 20:55:12 +00:00
Hal Finkel 46402a4211 Split some PPC itinerary classes
In preparation for adding scheduling definitions for the POWER7, split some PPC
itinerary classes so that the P7's latencies and hazards can be better
described. For the most part, this means differentiating indexed from non-index
pre-increment loads and stores. Also, differentiate single from
double-precision sqrt.

No functionality change intended (except for a more-specific latency for
single-precision sqrt on the A2).

llvm-svn: 195980
2013-11-30 20:41:13 +00:00
Zoran Jovanovic 9d86e26e62 Fixed issue with microMIPS long branch.
llvm-svn: 195975
2013-11-30 19:12:28 +00:00
Daniel Sanders 7fd68d6018 [mips][msa] MSA loads and stores have a 10-bit offset. Account for this when lowering FrameIndex.
This prevents the compiler from emitting invalid ld.[bhwd]'s and st.[bhwd]'s
when the stack frame is between 512 and 32,768 bytes in size.

llvm-svn: 195973
2013-11-30 13:47:57 +00:00
Daniel Sanders 7153414768 [mips][msa] A small refactor to reduce patch noise in my next commit
No functional change. An if-statement has been split into two nested if-statements.

llvm-svn: 195972
2013-11-30 13:15:21 +00:00
Reed Kotler ad450f239f Part 1 of 3 patches that completes very long conditional branches
in constant islands for Mips16. We introdcuce JalB16 as a synomnym
for Jal16. It makes it easier to read and is also necessary because
Jal16 is a call instruction but JalB16 is being used as a branch.
Various parts of LLVM will not work properly even in this late stage of
the backend if we use what was declared as a call instruction to function
as a branch. For one, basic block labels may not get emitted in some
situations. 

llvm-svn: 195968
2013-11-29 22:32:56 +00:00
Zoran Jovanovic 1bc3cce040 Revert revision 195965.
llvm-svn: 195967
2013-11-29 22:10:02 +00:00
Zoran Jovanovic ff2a40ce4d Fixed issue with microMIPS long branch.
llvm-svn: 195965
2013-11-29 21:41:24 +00:00
Hal Finkel 1df3205e8c Adjust PPC A2 input operand latencies
On the PPC A2, instructions are only issued after their input operands are
ready. Model this by specifying that input operands are read at dispatch (0
cycles after issue). This changes all input operand latencies from 1 to 0.

Significant test-suite performance changes (these are 99.5% confidence
intervals on 6 runs for both before and after):

speedups:
MultiSource/Benchmarks/sim/sim
	-1.21915% +/- 0.175063%
MultiSource/Benchmarks/TSVC/LinearDependence-flt/LinearDependence-flt
	-1.23946% +/- 1.05133%
SingleSource/Benchmarks/Misc/flops-2
	-1.24237% +/- 0.681362%
MultiSource/Applications/JM/lencod/lencod
	-1.33992% +/- 0.757498%
MultiSource/Benchmarks/TSVC/InductionVariable-flt/InductionVariable-flt
	-1.51802% +/- 1.21468%
MultiSource/Benchmarks/TSVC/GlobalDataFlow-flt/GlobalDataFlow-flt
	-2.18818% +/- 1.28605%
MultiSource/Benchmarks/TSVC/Packing-flt/Packing-flt
	-2.21977% +/- 1.19499%
SingleSource/Benchmarks/BenchmarkGame/spectral-norm
	-2.29822% +/- 0.671871%
MultiSource/Benchmarks/TSVC/Packing-dbl/Packing-dbl
	-2.40975% +/- 0.355931%
SingleSource/Benchmarks/Misc/fp-convert
	-2.41899% +/- 1.04751%
MultiSource/Benchmarks/TSVC/Searching-dbl/Searching-dbl
	-2.50349% +/- 0.126765%
SingleSource/Benchmarks/Misc/flops-3
	-3.00214% +/- 0.700795%
MultiSource/Benchmarks/TSVC/LoopRestructuring-flt/LoopRestructuring-flt
	-3.56995% +/- 3.2929%
MultiSource/Applications/sgefa/sgefa
	-4.24908% +/- 2.00413%
MultiSource/Benchmarks/ASC_Sequoia/IRSmk/IRSmk
	-18.1294% +/- 3.96489%

regressions:
MultiSource/Benchmarks/TSVC/Reductions-dbl/Reductions-dbl
	1.03249% +/- 0.178547%
MultiSource/Applications/hexxagon/hexxagon
	1.16597% +/- 0.285235%
MultiSource/Benchmarks/TSVC/IndirectAddressing-flt/IndirectAddressing-flt
	1.39576% +/- 1.07855%
SingleSource/Benchmarks/Misc-C++/stepanov_v1p2
	1.71539% +/- 0.173182%
MultiSource/Benchmarks/Fhourstones-3.1/fhourstones3.1
	1.90013% +/- 0.866472%
MultiSource/Benchmarks/TSVC/Recurrences-dbl/Recurrences-dbl
	2.39854% +/- 1.05914%
MultiSource/Benchmarks/TSVC/ControlFlow-dbl/ControlFlow-dbl
	2.4402% +/- 0.817904%
MultiSource/Benchmarks/TSVC/LoopRestructuring-dbl/LoopRestructuring-dbl
	5.87997% +/- 3.3172%
MultiSource/Benchmarks/Trimaran/netbench-crc/netbench-crc
	9.02643% +/- 5.79591%
MultiSource/Benchmarks/VersaBench/bmm/bmm
	10.3517% +/- 1.227%

Obviously, there are data points on both sides of this; but I think, overall,
this supports making the change.

llvm-svn: 195951
2013-11-29 07:04:59 +00:00
Hal Finkel 5a7162f36b Create a PPC440 SchedMachineModel
Some of the older PPC processor definitions don't have associated
SchedMachineModels; correct this for the PPC440.

llvm-svn: 195949
2013-11-29 06:32:17 +00:00
Hal Finkel 4035e8d86a Fixup PPC440 load/store operand latencies
The operand latencies for loads and stores in the PPC440 itinerary were wrong
(the store operands are all inputs, and the "with update" (pre-increment)
instructions need a latency for the additional output).

llvm-svn: 195948
2013-11-29 06:19:43 +00:00
Hal Finkel a10bd1d23a Adjust PPC440 operand latencies
The operand latencies for the PPC440 should be specified relative to dispatch,
not relative to the initial fetch-and-decode stages. Because most instructions
(ignoring bypass) wait in dispatch until their operands are ready, this is
modeled as reading input operands "at dispatch" (0 cycles after issue), and so
every input and output operand has 4 cycles subtracted from it.

This could alter scheduling slightly, but I don't expect a large effect.

llvm-svn: 195947
2013-11-29 05:59:00 +00:00
Hal Finkel dd06369913 Don't model the fetch and decode units for the PPC440
Modeling the fetch and decode units in the PPC440 itinerary does not add
anything to the hazard detection capability (and so modeling them just wastes
compile time).

No functionality change intended.

llvm-svn: 195946
2013-11-29 05:58:38 +00:00
Lang Hames 39609996d9 Refactor a lot of patchpoint/stackmap related code to simplify and make it
target independent.

Most of the x86 specific stackmap/patchpoint handling was necessitated by the
use of the native address-mode format for frame index operands. PEI has now
been modified to treat stackmap/patchpoint similarly to DEBUG_INFO, allowing
us to use a simple, platform independent register/offset pair for frame
indexes on stackmap/patchpoints.

Notes:
  - Folding is now platform independent and automatically supported.
  - Emiting patchpoints with direct memory references now just involves calling
    the TargetLoweringBase::emitPatchPoint utility method from the target's
    XXXTargetLowering::EmitInstrWithCustomInserter method. (See
    X86TargetLowering for an example).
  - No more ugly platform-specific operand parsers.

This patch shouldn't change the generated output for X86. 

llvm-svn: 195944
2013-11-29 03:07:54 +00:00
Hao Liu ba38eee8ac AArch64: The pattern match should check the range of the immediate value.
Or we can generate some illegal instructions.
E.g. shrn2 v0.4s, v1.2d, #35. The legal range should be in [1, 16].

llvm-svn: 195941
2013-11-29 02:11:22 +00:00
Jiangning Liu c429c00f3b Add missing pattern for supporting intrinsic function vbsl_f64 with
argument double floating point.

llvm-svn: 195938
2013-11-29 01:37:15 +00:00
Kevin Qin 337cfcc83c [AArch64 NEON]Fix a assertion failure when disassemble SHLL instruction.
llvm-svn: 195936
2013-11-29 01:29:16 +00:00
Rafael Espindola d5bd5a4716 Refactor to remove a bit of duplication. No functionality change.
llvm-svn: 195933
2013-11-28 20:12:44 +00:00
Benjamin Kramer ea1982aff9 Silence sign-compare warning and reduce nesting.
No functionality change.

llvm-svn: 195932
2013-11-28 19:58:56 +00:00
NAKAMURA Takumi 226e10edff [CMake] Let add_public_tablegen_target() provide intrinsics_gen, too.
I think, in principle, intrinsics_gen may be added explicitly.
That said, it can be added incidentally, since each target already has dependencies to llvm-tblgen.
Almost all source files depend on both CommonTaleGen and intrinsics_gen.

Explicit add_dependencies() have been pruned under lib/Target.

llvm-svn: 195929
2013-11-28 17:04:31 +00:00
NAKAMURA Takumi ce746c6c49 [CMake] Let add_public_tablegen_target responsible to provide dependency to CommonTableGen.
add_public_tablegen_target adds *CommonTableGen to LLVM_COMMON_DEPENDS.
LLVM_COMMON_DEPENDS affects add_llvm_library (and other add_target stuff) within its scope.

llvm-svn: 195927
2013-11-28 17:04:04 +00:00
Rafael Espindola 848493d886 The global prefix is always one char. Don't use a string for it.
llvm-svn: 195926
2013-11-28 17:00:49 +00:00
NAKAMURA Takumi b2abd160b3 [CMake] Prune include_directories() in llvm/lib/Target, take #2.
I forgot to commit them. They were staging in my local repo.

llvm-svn: 195924
2013-11-28 15:30:37 +00:00
Daniel Sanders 063b74ad4e [mips] Revert test commit r195922.
llvm-svn: 195923
2013-11-28 15:26:33 +00:00
Daniel Sanders eb16443fca [mips] A test commit to test my Herald and Audit workflow
Will be reverted in the next commit

llvm-svn: 195922
2013-11-28 15:25:43 +00:00
NAKAMURA Takumi 413518f1f8 [CMake] Prune include_directories() in llvm/lib/Target. add_llvm_target() sets them.
llvm-svn: 195921
2013-11-28 14:53:30 +00:00
NAKAMURA Takumi 979e604d8c Add newline at eof.
llvm-svn: 195920
2013-11-28 14:52:52 +00:00
Rafael Espindola 3e3a3f1f85 Use the mangler consistently instead of using getGlobalPrefix directly.
llvm-svn: 195911
2013-11-28 08:59:52 +00:00
Hal Finkel 92720ab1b2 Don't share functional units among the PPC itineraries
Instead of sharing functional unit names between the various PPC itineraries,
give each core its own unit names prefixed with the core name.  This follows
the convention used by other backends (such as ARM), and removes a non-obvious
ordering dependency between the various PPCSchedule*.td files.

No functionality change intended.

llvm-svn: 195908
2013-11-28 06:05:59 +00:00
Jiangning Liu 4bc9dbd846 Remove the variable only used by assert to avoid the build failure
caused by build options [-Werror,-Wunused-variable].

llvm-svn: 195905
2013-11-28 01:34:55 +00:00
Hao Liu f9f468abee AArch64: Fix a bug about disassembling post-index load single element to 4 vectors
llvm-svn: 195903
2013-11-28 01:07:45 +00:00
Reed Kotler 0d409e2dfe Check in conditional branches for constant islands. Still need to finish
conditional branches for very large targets. That will be the next small
patch. Everything now should in principle work as good (functionality
wise) as without constant islands so we decided at Mips/Imagination to
make constant islands the default for Mips16 now so that it will get
excercised a lot and this port is still experimentatl though hopefully soon
we will change the status. Some more cleanup and code review is in order
but things are converging fast.

llvm-svn: 195902
2013-11-28 00:56:37 +00:00
Akira Hatanaka f6109e4ad7 [mips] Redefine TAILCALL as a pseudo instruction.
No functionality change.

llvm-svn: 195896
2013-11-27 23:58:32 +00:00
Akira Hatanaka f9a0ec4fc4 Add MipsOptimizePICCall.cpp to CMakeLists.txt.
llvm-svn: 195894
2013-11-27 23:47:25 +00:00
Akira Hatanaka 168d4e5b20 [mips] Implement the following optimizations using dominance information to
make PIC calls a little more efficient:

1. Remove instructions setting up $gp if it is known that a function has been
   called at least once.
2. Save the address of a called function in a register instead of loading
   it from the GOT at every call site.

llvm-svn: 195892
2013-11-27 23:38:42 +00:00
Hal Finkel 3e5a360ba3 Add IIC_ prefix to PPC instruction-class names
This adds the IIC_ prefix to the instruction itinerary class names, giving the
PPC backend a naming convention for itinerary classes that is more consistent
with that used by the X86 and ARM backends.

Instruction scheduling in the PPC backend needs a bunch of cleanup and
improvement (especially for the ooo cores). This is just a preliminary step.

No functionality change intended.

llvm-svn: 195890
2013-11-27 23:26:09 +00:00
Rafael Espindola c90584b6f6 Don't set GlobalPrefix to the default value.
llvm-svn: 195884
2013-11-27 21:57:54 +00:00
Rafael Espindola 429e3fb068 The R600 has its own asm printer which doesn't use GlobalPrefix. Drop it.
llvm-svn: 195883
2013-11-27 21:52:37 +00:00
Tom Stellard 175e7a8c97 R600: Expand vector FABS
NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195881
2013-11-27 21:23:39 +00:00
Tom Stellard c149dc02d3 R600/SI: Implement spilling of SGPRs v5
SGPRs are spilled into VGPRs using the {READ,WRITE}LANE_B32 instructions.

v2:
  - Fix encoding of Lane Mask
  - Use correct register flags, so we don't overwrite the low dword
    when restoring multi-dword registers.

v3:
  - Register spilling seems to hang the GPU, so replace all shaders
    that need spilling with a dummy shader.

v4:
  - Fix *LANE definitions
  - Change destination reg class for 32-bit SMRD instructions

v5:
  - Remove small optimization that was crashing Serious Sam 3.

https://bugs.freedesktop.org/show_bug.cgi?id=68224
https://bugs.freedesktop.org/show_bug.cgi?id=71285

NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195880
2013-11-27 21:23:35 +00:00
Tom Stellard 859199dad8 R600/SI: Use SGPR_32 register class for 32-bit SMRD outputs
Writing to the M0 register from an SMRD instruction hangs the GPU, so
we need to use the SGPR_32 register class, which does not include M0.

NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195879
2013-11-27 21:23:29 +00:00
Tom Stellard 4d566b2edf R600: Add support for ISD::FROUND
NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195878
2013-11-27 21:23:20 +00:00
Rafael Espindola 3dc549dbe3 Remove dead code.
MO_ExternalSymbol and MO_JumpTableIndex don't show up in inline asm.

llvm-svn: 195861
2013-11-27 18:38:14 +00:00
Rafael Espindola 52434f9673 Convert two if sequences to switches.
llvm-svn: 195859
2013-11-27 18:26:51 +00:00
Rafael Espindola ed20f478bc Use a switch.
llvm-svn: 195857
2013-11-27 18:18:24 +00:00
Rafael Espindola c5c7bb6b20 Remove more dead code now that this is only used for inline asm.
MO_ConstantPoolIndex is handled in printLeaMemReference.
MO_JumpTableIndex and MO_ExternalSymbol don't show up in inline asm.

llvm-svn: 195847
2013-11-27 15:13:06 +00:00
Jiangning Liu 97aa8cf8b7 Fix the AArch64 NEON bug exposed by checking constant integer argument range of ACLE intrinsics.
llvm-svn: 195843
2013-11-27 14:02:25 +00:00
Rafael Espindola e370147b8c Convert more methods in static helpers.
llvm-svn: 195826
2013-11-27 07:34:09 +00:00
Rafael Espindola 7caa135677 Convert these methods into static functions.
llvm-svn: 195825
2013-11-27 07:14:26 +00:00