ae40dc735d
Remove MipsFPRound. Mips1 is no longer supported.
...
llvm-svn: 140661
2011-09-27 23:55:37 +00:00
30c811246f
Remove X86-dependent stuff from SSEDomainFix.
...
This also enables domain swizzling for AVX code which required a few
trivial test changes.
The pass will be moved to lib/CodeGen shortly.
llvm-svn: 140659
2011-09-27 23:50:46 +00:00
e3e36f80f5
Unbreak CMake build.
...
llvm-svn: 140655
2011-09-27 23:29:59 +00:00
f9b71a2e01
Implement TII::get/setExecutionDomain() for ARM.
...
llvm-svn: 140653
2011-09-27 22:57:21 +00:00
b48c994cc0
Promote the X86 Get/SetSSEDomain functions to TargetInstrInfo.
...
I am going to unify the SSEDomainFix and NEONMoveFix passes into a
single target independent pass. They are essentially doing the same
thing.
llvm-svn: 140652
2011-09-27 22:57:18 +00:00
c63af1b7b6
ARM Thumb2 asm parsing [SU]XT[BH] without rotate but with .w.
...
Add inst alias to handle these assembly forms. Add tests, too.
rdar://10178799
llvm-svn: 140647
2011-09-27 22:18:54 +00:00
354ff9e348
This is the start of the new SjLj EH preparation pass, which will replace the
...
current IR-level pass.
The old SjLj EH pass has some problems, especially with the new EH model. Most
significantly, it violates some of the new restrictions the new model has. For
instance, the 'dispatch' table wants to jump to the landing pad, but we cannot
allow that because only an invoke's unwind edge can jump to a landing pad. This
requires us to mangle the code something awful. In addition, we need to keep the
now dead landingpad instructions around instead of CSE'ing them because the
DWARF emitter uses that information (they are dead because no control flow edge
will execute them - the control flow edge from an invoke's unwind is superceded
by the edge coming from the dispatch).
Basically, this pass belongs not at the IR level where SSA is king, but at the
code-gen level, where we have more flexibility.
llvm-svn: 140646
2011-09-27 22:14:12 +00:00
a5d18f2d7e
Embed patterns in definitions of MFC1 and MTC1 instead of defining them outside
...
of the instruction definitions using Pat<>.
llvm-svn: 140644
2011-09-27 22:01:01 +00:00
7a6e8f2c5d
Remove an invalid assert that is really just asserting when the scheduler emits
...
a suboptimal schedule.
llvm-svn: 140643
2011-09-27 21:59:16 +00:00
af136f71ec
Rename AddSelectionDAGCSEId() to addSelectionDAGCSEId().
...
Naming conventions consistency. No functional change.
llvm-svn: 140636
2011-09-27 20:59:33 +00:00
547b6c5ecd
Stop emitting instructions with the name "tmp" they eat up memory and have to be uniqued, without any benefit.
...
If someone prefers %tmp42 to %42, run instnamer.
llvm-svn: 140634
2011-09-27 20:39:19 +00:00
bf415251df
These symbols appear to be visible by SearchForAddressOfSymbol and no longer
...
require special case handling.
rdar://10117377
llvm-svn: 140629
2011-09-27 20:01:41 +00:00
d3b7b12618
Object: Add archive support.
...
llvm-svn: 140626
2011-09-27 19:36:55 +00:00
68ba81346e
Check that catch clauses have pointer type.
...
llvm-svn: 140625
2011-09-27 19:34:22 +00:00
4f7054e56e
PTX: Fix case where printed alignment could be 0
...
llvm-svn: 140624
2011-09-27 19:25:49 +00:00
e074593498
PTX: Use external symbols to keep track of params and locals. This also fixes
...
a couple of outstanding issues with frame objects occuring as instruction
operands.
llvm-svn: 140616
2011-09-27 18:12:55 +00:00
1c7597693c
Use existing function.
...
llvm-svn: 140615
2011-09-27 17:55:08 +00:00
e41b1d59f0
Fix function MipsRegisterInfo::getRegisterNumbering.
...
Return numbers of 64-bit registers.
llvm-svn: 140609
2011-09-27 17:15:27 +00:00
ff5d0965b0
Do not add the pass that restores $gp if target is Mips64.
...
llvm-svn: 140607
2011-09-27 16:58:43 +00:00
86de1a666d
Have the verifier check that all landingpad operands are constants.
...
llvm-svn: 140606
2011-09-27 16:43:19 +00:00
38b3b83362
Cleanup PromoteIntOp_EXTRACT_VECTOR_ELT and PromoteIntRes_SETCC.
...
Add a new method: getAnyExtOrTrunc and use it to replace the manual check.
llvm-svn: 140603
2011-09-27 11:16:47 +00:00
1b857d2762
Revert r140463; The patch assumes that <4 x i1> is saved to memory as 4 x i8,
...
while the decision is to bit-pack small values.
llvm-svn: 140601
2011-09-27 10:48:29 +00:00
bb050745e7
Mark MipsPseudo isPseudo.
...
llvm-svn: 140598
2011-09-27 04:57:54 +00:00
9f01f89386
PTX: Add support for sitofp in backend
...
llvm-svn: 140593
2011-09-27 01:04:47 +00:00
90f90da156
Split the landing pad basic block with the correct function. Also merge the
...
split landingpad instructions into a PHI node.
PR11016
llvm-svn: 140592
2011-09-27 00:59:31 +00:00
581243919d
Disable LSR retry by default.
...
Disabling aggressive LSR saves compilation time, and with the new
indvars behavior usually improves performance.
llvm-svn: 140590
2011-09-27 00:44:14 +00:00
8868faec63
LSR, one of the new Cost::isLoser() checks did not get merged in the previous checkin.
...
llvm-svn: 140583
2011-09-26 23:35:25 +00:00
b1a9f65487
Remove extraneous commit garbage.
...
llvm-svn: 140581
2011-09-26 23:14:02 +00:00
784729d408
LSR cost metric minor fix and verification.
...
The minor bug heuristic was noticed by inspection. I added the
isLoser/isValid helpers because they will become more
important with subsequent checkins.
llvm-svn: 140580
2011-09-26 23:11:04 +00:00
a6a9c20c23
Set register class of a register according to value of HasMips64.
...
llvm-svn: 140570
2011-09-26 21:55:17 +00:00
7b502920ef
Define variable HasMips64 in MipsTargetLowering.
...
llvm-svn: 140569
2011-09-26 21:47:02 +00:00
e5ce709022
In single float mode, double precision FP arguments are passed in integer
...
registers, so there is no need to check here.
llvm-svn: 140568
2011-09-26 21:37:50 +00:00
f01e2de5e6
ASR #32 is not allowed on Thumb2 USAT and SSAT instructions.
...
llvm-svn: 140560
2011-09-26 21:06:22 +00:00
5c91891cf3
Enhance alias analysis for atomic instructions a bit. Upgrade a couple alias-analysis tests to the new atomic instructions.
...
llvm-svn: 140557
2011-09-26 20:15:28 +00:00
da2919dbd8
PTX: Fix memcpy intrinsic to handle 64-bit pointers
...
llvm-svn: 140556
2011-09-26 19:19:48 +00:00
b40da7f956
PTX: Implement PTXSelectionDAGInfo
...
llvm-svn: 140549
2011-09-26 18:57:27 +00:00
c3edaddfea
PTX: Implement ISD::ANY_EXTEND
...
llvm-svn: 140548
2011-09-26 18:57:24 +00:00
1395cf8423
PTX: Fix detection of stack load/store vs. global load/store, as well as fix the
...
printing of local offsets
llvm-svn: 140547
2011-09-26 18:57:22 +00:00
0ceb8cadd2
Fix emission of debug data for global variables. getContext() on DIGlobalVariables is not valid any more.
...
llvm-svn: 140539
2011-09-26 17:40:42 +00:00
f8dd701bf9
PTX: SM > 2.0 implies +double
...
llvm-svn: 140536
2011-09-26 16:20:36 +00:00
14defde057
PTX: Fix some lingering issues with stack allocation
...
llvm-svn: 140535
2011-09-26 16:20:34 +00:00
37fd87675f
PTX: Split up the TableGen instruction definitions into logical units
...
llvm-svn: 140534
2011-09-26 16:20:31 +00:00
d40f5ababf
PTX: Unify handling of loads/stores
...
llvm-svn: 140533
2011-09-26 16:20:28 +00:00
8c80019352
PTX: Handle FrameIndex nodes
...
llvm-svn: 140532
2011-09-26 16:20:25 +00:00
b1fbf9ff26
PR11004: Inline memcpy to avoid generating nested call sequence. Un-XFAIL 2011-06-09-TailCallByVal and 2010-11-04-BigByval
...
llvm-svn: 140516
2011-09-26 06:13:20 +00:00
45faba98b4
Fix VEX decoding in i386 mode. Fixes PR11008.
...
llvm-svn: 140515
2011-09-26 05:12:43 +00:00
df977fedb6
Add target hook for pseudo instruction expansion.
...
Many targets use pseudo instructions to help register allocation. Like
the COPY instruction, these pseudos can be expanded after register
allocation. The early expansion can make life easier for PEI and the
post-ra scheduler.
This patch adds a hook that is called for all remaining pseudo
instructions from the ExpandPostRAPseudos pass.
llvm-svn: 140472
2011-09-25 19:21:35 +00:00
2279949129
[vector-select] Address one of the issues in pr10902. EXTRACT_VECTOR_ELEMENT
...
SDNodes may return values which are wider than the incoming element types. In
this patch we fix the integer promotion of these nodes.
Fixes spill-q.ll when running -promote-elements.
llvm-svn: 140471
2011-09-25 18:59:42 +00:00
fd719d184e
Clean up code after renaming LowerSubregs -> ExpandPostRAPseudos.
...
No functional change intended.
llvm-svn: 140470
2011-09-25 16:46:08 +00:00
f152df1e6b
Rename LowerSubregs to ExpandPostRAPseudos.
...
I'll fix the file contents in the next commit.
This pass is currently expanding the COPY and SUBREG_TO_REG pseudos. I
am going to add a hook so targets can expand more pseudo-instructions
after register allocation.
Many targets have pseudo-instructions that assist the register
allocator. They can be expanded after register allocation, before PEI
and PostRA scheduling.
llvm-svn: 140469
2011-09-25 16:46:00 +00:00
f550fa9173
Sort CMakeLists.txt.
...
llvm-svn: 140465
2011-09-24 22:06:35 +00:00
c2deabd202
Implement Duncan's suggestion to use the result of getSetCCResultType if it is legal
...
(this is always the case for scalars), otherwise use the promoted result type.
Fix test/CodeGen/X86/vsplit-and.ll when promote-elements is enabled.
llvm-svn: 140464
2011-09-24 19:48:19 +00:00
77426a754b
[Vector-Select] Address one of the problems in 10902.
...
When generating the trunc-store of i1's, we need to use the vector type and not
the scalar type.
This patch fixes the assertion in CodeGen/Generic/bool-vector.ll when
running with -promote-elements.
llvm-svn: 140463
2011-09-24 18:32:19 +00:00
7d7ee0c3ac
Add .td file.
...
llvm-svn: 140446
2011-09-24 01:40:18 +00:00
e96273e75d
Preparation for adding simple Mips64 instructions.
...
llvm-svn: 140443
2011-09-24 01:34:44 +00:00
55cf2ed148
Only run MF.verify() with EXPENSIVE_CHECKS=1.
...
llvm-svn: 140441
2011-09-24 01:11:19 +00:00
9b92e2be30
sys::Process: Add a SetWorkingDirectory method.
...
llvm-svn: 140433
2011-09-23 23:23:36 +00:00
8b2fe2f744
LSR minor bug fix in RateRegister.
...
No test case. Noticed by inspection and I doubt it ever affects the
outcome of the overall heuristic, let alone final codegen.
llvm-svn: 140431
2011-09-23 23:05:19 +00:00
3bb99bc957
Verify that terminators follow non-terminators.
...
This exposes a -segmented-stacks bug.
llvm-svn: 140429
2011-09-23 22:45:39 +00:00
8a15a5aa93
PR10998: It is not legal to sink an instruction past the terminator of a block; make sure we don't do that.
...
llvm-svn: 140428
2011-09-23 22:41:57 +00:00
4916840eb8
Teach the Thumb2 AsmParser to accept pre-indexed loads/stores with an offset of #-0.
...
llvm-svn: 140426
2011-09-23 22:25:02 +00:00
2056d15bd9
Also match negative offsets for addrmode3 and addrmode5.
...
Math is hard, and isScaledConstantInRange() always returned false for
negative constants. It was doing unsigned division of negative numbers
before casting back to signed.
llvm-svn: 140425
2011-09-23 22:10:33 +00:00
b0b865d658
Add more fixed bits to USAT16 encoding to filter out incorrect decodings.
...
llvm-svn: 140422
2011-09-23 21:57:50 +00:00
737beaf86d
Post-index loads/stores in still need to print the post-indexed immediate, even if it's zero, to distinguish them from non-post-indexed instructions.
...
llvm-svn: 140420
2011-09-23 21:26:40 +00:00
987a878946
Reapply r140412 (Thumb2 reg-reg loads cannot target SP or PC), with invalid testcases updated.
...
llvm-svn: 140415
2011-09-23 21:07:25 +00:00
ffa8428acf
Revert r140412. This affects more instructions than intended.
...
llvm-svn: 140413
2011-09-23 21:02:01 +00:00
7591d0c363
Thumb2 register-shifted-register loads cannot target the PC or the SP.
...
llvm-svn: 140412
2011-09-23 21:00:32 +00:00
d6af2c62b4
Implement N32/64 calling convention. Patch by Liu.
...
llvm-svn: 140401
2011-09-23 19:08:15 +00:00
ceb55e72de
Make FGR64RegisterClass available if target is Mips64.
...
llvm-svn: 140397
2011-09-23 18:28:39 +00:00
77709a6793
Add definitions of 64-bit register files. Add code for returning Mips64's sets of
...
callee-saved registers and reserved registers.
llvm-svn: 140395
2011-09-23 18:11:56 +00:00
71d32c980d
PTX: Fix parameter order bug
...
llvm-svn: 140394
2011-09-23 17:59:11 +00:00
24e45cabbc
Fix a couple of 80 column violations.
...
patch contributed by Jia Liu!
llvm-svn: 140391
2011-09-23 17:24:41 +00:00
6e84a68023
PTX: Cleanup unused code in PTXMachineFunctionInfo
...
llvm-svn: 140390
2011-09-23 17:15:53 +00:00
0f1af22183
PTX: Fix another 80-column violation
...
llvm-svn: 140387
2011-09-23 16:50:35 +00:00
37f35f0083
PTX: Handle function call return values
...
llvm-svn: 140386
2011-09-23 16:48:41 +00:00
ae191ef63b
Fix 80 column violations.
...
Original patch by Liu.
llvm-svn: 140385
2011-09-23 16:28:10 +00:00
a54fd541c2
Implement Chris's suggestion of legalizing the various SSE and AVX
...
hadd/hsub intrinsics into the new fhadd/fhsub X86 node.
llvm-svn: 140383
2011-09-23 16:10:22 +00:00
6c23d2ee55
PTX: Start fixing function calls
...
llvm-svn: 140378
2011-09-23 14:31:12 +00:00
edc6bf474d
PTX: Remove PTX calling convention files
...
llvm-svn: 140377
2011-09-23 14:18:27 +00:00
f2b540e815
[PATCH 2/2] PTXInstrInfo.td PTXIntrinsicInstrInfo.td 80 columns
...
From 5936c03172e251f12a0332d1033de5718e6e2091 Mon Sep 17 00:00:00 2001
---
lib/Target/PTX/PTXInstrInfo.td | 165 ++++++++++++++++++++----------
lib/Target/PTX/PTXIntrinsicInstrInfo.td | 88 +++++++++++------
2 files changed, 167 insertions(+), 86 deletions(-)
llvm-svn: 140376
2011-09-23 14:18:24 +00:00
b823e41bf4
PTX: Generalize handling of .param types
...
llvm-svn: 140375
2011-09-23 14:18:22 +00:00
2f82cc61af
PTX: Cleanup unused code in the PTXMFInfoExtract pass
...
llvm-svn: 140374
2011-09-23 14:18:19 +00:00
b461176cfb
Tweak the handling of MERGE_VALUES nodes: remove the need for
...
DecomposeMERGE_VALUES to "know" that results are legalized in
a particular order, by passing it the number of the result
being legalized (the type legalization core provides this, it
just needs to be passed on).
llvm-svn: 140373
2011-09-23 13:59:22 +00:00
57e30726ad
Vector-Select: Address one of the problems in pr10902. Add handling for the
...
integer-promotion of CONCAT_VECTORS.
Test: test/CodeGen/X86/widen_shuffle-1.ll
This patch fixes the above tests (when running in with -promote-elements).
llvm-svn: 140372
2011-09-23 09:33:24 +00:00
42fe6bd5f2
Add definitions of 64-bit int registers.
...
llvm-svn: 140366
2011-09-23 02:33:15 +00:00
61bbcce84a
Do not rely on the enum values of argument registers A0-A3 being consecutive.
...
Define function getNextIntArgReg, which takes a register as a parameter and
returns the next O32 argument integer register. Use this function when double
precision floating point arguments are passed in two integer registers.
llvm-svn: 140363
2011-09-23 00:58:33 +00:00
64a4bf1788
PR10989: Don't print .hidden on Windows.
...
llvm-svn: 140356
2011-09-23 00:13:02 +00:00
87c844cdf8
PR10991: make fast-isel correctly check whether accessing a global through an alias involves thread-local storage. (I'm not entirely sure how this is supposed to work, but this patch makes fast-isel consistent with the normal isel path.)
...
llvm-svn: 140355
2011-09-22 23:41:28 +00:00
f25c37e384
Make changes in instruction and pattern definitions so that tablegen does not
...
complain it cannot infer types in patterns. Fix a mistake in definition of
SDT_MipsExtractElementF64.
llvm-svn: 140354
2011-09-22 23:31:54 +00:00
adea3f0c01
Add new files to CMake.
...
llvm-svn: 140352
2011-09-22 23:20:48 +00:00
e83e1b2d2c
Fix SimplifySelectCC to add newly created nodes to the DAGCombiner
...
worklist, as it may be possible to perform further optimization on them.
llvm-svn: 140349
2011-09-22 23:01:29 +00:00
f05864ad7d
Add support for GR32 <-> FR32 cross class copies.
...
We already support GR64 <-> VR128 copies. All of these copies break
partial register dependencies by zeroing the high part of the target
register.
llvm-svn: 140348
2011-09-22 22:45:24 +00:00
fbf0fba9f4
Update CMake build.
...
llvm-svn: 140347
2011-09-22 22:38:34 +00:00
6cca67fc4a
Start stubbing out MCModule and MCAtom, which provide an API for accessing the rich disassembly of a complete object or executable.
...
These are very much a work in progress, and not really useful yet.
llvm-svn: 140345
2011-09-22 22:32:22 +00:00
e92e5ee81f
Constrain register classes instead of emitting copies.
...
Sometimes register class constraints are trivial, like GR32->GR32_NOSP,
or GPR->rGPR. Teach InstrEmitter to simply constrain the virtual
register instead of emitting a copy in these cases.
Normally, these copies are handled by the coalescer. This saves some
coalescer work.
llvm-svn: 140340
2011-09-22 21:39:34 +00:00
0f36544c08
Add a MinNumRegs argument to MRI::constrainRegClass().
...
The function will refuse to use a register class with fewer registers
than MinNumRegs. This can be used by clients to avoid accidentally
increase register pressure too much.
The default value of MinNumRegs=0 doesn't affect how constrainRegClass()
works.
llvm-svn: 140339
2011-09-22 21:39:31 +00:00
0e4fcb8e3b
Synthesize SSE3/AVX 128 bit horizontal add/sub instructions from
...
floating point add/sub of appropriate shuffle vectors. Does not
synthesize the 256 bit AVX versions because they work differently.
llvm-svn: 140332
2011-09-22 20:15:48 +00:00
f9b785f185
PR10987: add a missed safety check to isSafePHIToSpeculate in scalarrepl.
...
llvm-svn: 140327
2011-09-22 18:56:30 +00:00
56acf840f1
Print parentheses in next line.
...
llvm-svn: 140325
2011-09-22 18:29:29 +00:00
c021a4b8b4
Change subreg index of AFPR64 from sub_fpeven to sub_32 per Jakob's comment.
...
llvm-svn: 140324
2011-09-22 18:24:21 +00:00
Akira Hatanaka