Tom Stellard [Tue, 20 Jan 2015 19:33:04 +0000 (19:33 +0000)]
R600/SI: Add subtarget feature to enable VGPR spilling for all shader types
This is disabled by default, but can be enabled with the subtarget
feature: 'vgpr-spilling'
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226597
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Tue, 20 Jan 2015 19:33:02 +0000 (19:33 +0000)]
R600/SI: Fix simple-loop.ll test
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226596
91177308-0d34-0410-b5e6-
96231b3b80d8
Jozef Kolek [Tue, 20 Jan 2015 19:29:28 +0000 (19:29 +0000)]
Reverted revision 226577.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226595
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 19:27:58 +0000 (19:27 +0000)]
[PM] Clean up a bunch of the doxygen / API docs on the InstCombiner pass
prior to refactoring it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226594
91177308-0d34-0410-b5e6-
96231b3b80d8
Manman Ren [Tue, 20 Jan 2015 19:24:59 +0000 (19:24 +0000)]
[llvm link] Destroy ConstantArrays in LLVMContext if they are not used.
ConstantArrays constructed during linking can cause quadratic memory
explosion. An example is the ConstantArrays constructed when linking in
GlobalVariables with appending linkage.
Releasing all unused constants can cause a 20% LTO compile-time
slowdown for a large application. So this commit releases unused ConstantArrays
only.
rdar://
19040716. It reduces memory footprint from 20+G to 6+G.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226592
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Tue, 20 Jan 2015 19:24:31 +0000 (19:24 +0000)]
R600/SI: Remove stray debugging code from r226586
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226591
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 18:54:16 +0000 (18:54 +0000)]
[PM] Don't spend time making self moves no-ops. They're allowed to leave
the object in a moved-from state, and its simpler to write the code that
way.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226589
91177308-0d34-0410-b5e6-
96231b3b80d8
Adrian Prantl [Tue, 20 Jan 2015 18:03:37 +0000 (18:03 +0000)]
Add an assertion and prefer a crash over an infinite loop.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226588
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Tue, 20 Jan 2015 17:49:47 +0000 (17:49 +0000)]
R600/SI: Use external symbols for scratch buffer
We were passing the scratch buffer address to the shaders via user sgprs,
but now we use external symbols and have the driver patch the shader
using reloc information.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226586
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Tue, 20 Jan 2015 17:49:45 +0000 (17:49 +0000)]
R600/SI: Add kill flag when copying scratch offset to a register
This allows us to re-use the same register for the scratch offset
when accessing large private arrays.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226585
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Tue, 20 Jan 2015 17:49:43 +0000 (17:49 +0000)]
R600/SI: Don't store scratch buffer frame index in MUBUF offset field
We don't have a good way of legalizing this if the frame index offset
is more than the 12-bits, which is size of MUBUF's offset field, so
now we store the frame index in the vaddr field.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226584
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Tue, 20 Jan 2015 17:49:41 +0000 (17:49 +0000)]
R600/SI: Update SIInstrInfo:verifyInstruction() after r225662
Now that we have our own custom register operand types, we need
to handle them in the verifiier.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226583
91177308-0d34-0410-b5e6-
96231b3b80d8
Aaron Ballman [Tue, 20 Jan 2015 17:10:45 +0000 (17:10 +0000)]
Silencing a -Wunused-variable warning in non-asserts builds; NFC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226581
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 17:04:56 +0000 (17:04 +0000)]
Revert "IR: Specify underlying type instead of r226570, NFC"
This reverts commit r226571. GCC really doesn't like it [1].
[1]: http://bb.pgr.jp/builders/cmake-llvm-x86_64-linux/builds/20260
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226579
91177308-0d34-0410-b5e6-
96231b3b80d8
Jozef Kolek [Tue, 20 Jan 2015 16:45:27 +0000 (16:45 +0000)]
[mips][microMIPS] MicroMIPS 16-bit unconditional branch instruction B
Implement microMIPS 16-bit unconditional branch instruction B.
Implemented 16-bit microMIPS unconditional instruction has real name B16, and
B is an alias which expands to either B16 or BEQ according to the rules:
b 256 --> b16 256 # R_MICROMIPS_PC10_S1
b 12256 --> beq $zero, $zero, 12256 # R_MICROMIPS_PC16_S1
b label --> beq $zero, $zero, label # R_MICROMIPS_PC16_S1
Differential Revision: http://reviews.llvm.org/D3514
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226577
91177308-0d34-0410-b5e6-
96231b3b80d8
Kai Nacke [Tue, 20 Jan 2015 16:14:02 +0000 (16:14 +0000)]
[mips] Add registers and ALL check prefix to octeon test case.
No functional change.
Reviewed by D. Sanders
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226574
91177308-0d34-0410-b5e6-
96231b3b80d8
Kai Nacke [Tue, 20 Jan 2015 16:10:51 +0000 (16:10 +0000)]
[mips] Add octeon branch instructions bbit0/bbit032/bbit1/bbit132
This commits adds the octeon branch instructions bbit0/bbit032/bbit1/bbit132.
It also includes patterns for instruction selection and test cases.
Reviewed by D. Sanders
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226573
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 16:03:09 +0000 (16:03 +0000)]
IR: Specify underlying type instead of r226570, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226571
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 15:51:14 +0000 (15:51 +0000)]
IR: Store StorageType as an unsigned bitfield
Use `unsigned` instead of `StorageType` for the bitfield to prevent MSVC
from treating the top bit of the bitfield as a sign bit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226570
91177308-0d34-0410-b5e6-
96231b3b80d8
Evgeniy Stepanov [Tue, 20 Jan 2015 15:21:35 +0000 (15:21 +0000)]
[msan] Optimize -msan-check-constant-shadow.
The new code does not create new basic blocks in the case when shadow is a
compile-time constant; it generates either an unconditional __msan_warning
call or nothing instead.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226569
91177308-0d34-0410-b5e6-
96231b3b80d8
Mohit K. Bhakkad [Tue, 20 Jan 2015 13:05:42 +0000 (13:05 +0000)]
[MSan][LLVM][MIPS] Shadow and Origin offsets for MIPS
Reviewers: kcc, samsonov, petarj, eugenis
Differential Revision: http://reviews.llvm.org/D6146
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226565
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Tue, 20 Jan 2015 12:15:30 +0000 (12:15 +0000)]
[x86] Add some mayLoad/hasSideEffects flags. Remove one that was already covered by a pattern.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226562
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 10:58:50 +0000 (10:58 +0000)]
[PM] Port LoopInfo to the new pass manager, adding both a LoopAnalysis
pass and a LoopPrinterPass with the expected associated wiring.
I've added a RUN line to the only test case (!!!) we have that actually
prints loops. Everything seems to be working.
This is somewhat exciting as this is the first analysis using another
analysis to go in for the new pass manager. =D I also believe it is the
last analysis necessary for porting instcombine, but of course I may yet
discover more.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226560
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 10:58:38 +0000 (10:58 +0000)]
[PM] Make the LoopInfoBase and LoopInfo objects movable so that they can
be used as results in the new pass manager.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226559
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 10:20:52 +0000 (10:20 +0000)]
[PM] Fix a moderately scary typo in the deleted copy constructor
I noticed when adding move semantics to LoopInfo.
Hopefully not relevant, but still scary. =]
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226556
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 10:02:49 +0000 (10:02 +0000)]
[PM] Use range-based for and auto to clean up some of the LoopInfo code.
No functionality changed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226555
91177308-0d34-0410-b5e6-
96231b3b80d8
Daniel Jasper [Tue, 20 Jan 2015 08:57:44 +0000 (08:57 +0000)]
Factor out a splitSwitchCase() function so that it can be reused.
This is in preparation for a fix to llvm.org/PR22262. One of the ideas
here is to first find a good jump table range first and then split
before and after it. Thereby, we don't need to use the
split-based-on-density heuristic at all, which can make the "binary
tree" deteriorate in various cases.
Also some minor cleanups.
No functional changes.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226551
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 08:35:24 +0000 (08:35 +0000)]
[PM] Move the LoopInfo analysis pointer into the InstCombiner class
along with the other analyses.
The most obvious reason why is because eventually I need to separate out
the pass layer from the rest of the instcombiner. However, it is also
probably a compile time win as every query through the pass manager
layer is pretty slow these days.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226550
91177308-0d34-0410-b5e6-
96231b3b80d8
Karthik Bhat [Tue, 20 Jan 2015 06:11:00 +0000 (06:11 +0000)]
Fix Operandreorder logic in SLPVectorizer to generate longer vectorizable chain.
This patch fixes 2 issues in reorderInputsAccordingToOpcode
1) AllSameOpcodeLeft and AllSameOpcodeRight was being calculated incorrectly resulting in code not being vectorized in few cases.
2) Adds logic to reorder operands if we get longer chain of consecutive loads enabling vectorization. Handled the same for cases were we have AltOpcode.
Thanks Michael for inputs and review.
Review: http://reviews.llvm.org/D6677
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226547
91177308-0d34-0410-b5e6-
96231b3b80d8
David Majnemer [Tue, 20 Jan 2015 05:58:07 +0000 (05:58 +0000)]
Bitcode: Don't create comdats when autoupgrading macho bitcode
Don't infer COMDAT groups from older bitcode if the target is macho,
it doesn't have COMDATs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226546
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 05:02:42 +0000 (05:02 +0000)]
Reapply "IR: Simplify DIBuilder's HeaderBuilder API, NFC"
This reverts commit r226542, effectively reapplying r226540. This time,
initialize `IsEmpty` in the copy and move constructors as well.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226545
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 03:01:27 +0000 (03:01 +0000)]
Revert "IR: Simplify DIBuilder's HeaderBuilder API, NFC"
This reverts commit r226540, since I hit an unexpected bot failure [1].
I'll investigate.
[1]: http://bb.pgr.jp/builders/cmake-llvm-x86_64-linux/builds/20244
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226542
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 02:56:57 +0000 (02:56 +0000)]
IR: Move MDNode clone() methods from ValueMapper to MDNode, NFC
Now that the clone methods used by `MapMetadata()` don't do any
remapping (and return a temporary), they make more sense as member
functions on `MDNode` (and subclasses).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226541
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 02:54:07 +0000 (02:54 +0000)]
IR: Simplify DIBuilder's HeaderBuilder API, NFC
Change `HeaderBuilder` API to work well even when it's not starting with
a tag. There's already one case like this, and the tag is moving
elsewhere as part of PR22235.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226540
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 02:42:29 +0000 (02:42 +0000)]
AsmParser: PARSE_MD_FIELD() => ParseMDField(), NFC
Extract most of `PARSE_MD_FIELD()` into a function.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226539
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 02:39:21 +0000 (02:39 +0000)]
AsmParser: Refactor duplicate code, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226538
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Tue, 20 Jan 2015 01:37:09 +0000 (01:37 +0000)]
[PM] Replace the Pass argument in MergeBasicBlockIntoOnlyPred with
a DominatorTree argument as that is the analysis that it wants to
update.
This removes the last non-loop utility function in Utils/ which accepts
a raw Pass argument.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226537
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 01:18:32 +0000 (01:18 +0000)]
IR: Delete GenericDwarfNode during teardown
Fix a leak in `LLVMContextImpl` teardown that the leak sanitizer tracked
down [1]. I've just switched to automatic dispatch here (since I'll
inevitably forget again with the next class).
[1]: http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fast/builds/811/steps/check-llvm%20asan/logs/stdio
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226536
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 01:03:09 +0000 (01:03 +0000)]
Bitcode: Simplify MDNode subclass dispatch, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226535
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 01:01:53 +0000 (01:01 +0000)]
Bitcode: WriteMDNode() => WriteMDTuple(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226534
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 01:00:23 +0000 (01:00 +0000)]
Bitcode: Add ValueEnumerator::getMetadataOrNullID(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226533
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 00:58:46 +0000 (00:58 +0000)]
IR: Canonicalize GenericDwarfNode empty headers to null
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226532
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 00:57:33 +0000 (00:57 +0000)]
IR: Detect whether to call recalculateHash() via SFINAE, NFC
Rather than relying on updating switch statements correctly, detect
whether `setHash()` exists in the subclass. If so, call
`recalculateHash()` and `setHash(0)` appropriately.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226531
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Tue, 20 Jan 2015 00:01:43 +0000 (00:01 +0000)]
IR: Introduce GenericDwarfNode
As part of PR22235, introduce `DwarfNode` and `GenericDwarfNode`. The
former is a metadata node with a DWARF tag. The latter matches our
current (generic) schema of a header with string (and stringified
integer) data and an arbitrary number of operands.
This doesn't move it into place yet; that change will require a large
number of testcase updates.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226529
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:44:41 +0000 (23:44 +0000)]
AsmParser: Abstract more of MDLocation parser, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226527
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:39:32 +0000 (23:39 +0000)]
AsmParser: Split up ParseMDFieldsImpl(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226526
91177308-0d34-0410-b5e6-
96231b3b80d8
Frederic Riss [Mon, 19 Jan 2015 23:33:14 +0000 (23:33 +0000)]
[dsymutil] Add the detected target triple to the debug map.
It will be needed to instantiate the Target object that we will
use to create all the MC objects for the dwarf emission.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226525
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:32:36 +0000 (23:32 +0000)]
AsmParser: Fix error location for missing fields
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226524
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:18:34 +0000 (23:18 +0000)]
IR: Cleanup MDNode field use, NFC
Swap usage of `SubclassData32` and `MDNodeSubclassData`, and rename
`MDNodeSubclassData` to `NumUnresolved`. Small drive-by cleanup to
`countUnresolvedOperands()` since otherwise the name clash with local
vars named `NumUnresolved` would be confusing.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226523
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:17:09 +0000 (23:17 +0000)]
IR: Move replaceWithUniqued(), etc., to source file, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226522
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:15:21 +0000 (23:15 +0000)]
IR: Cleanup MDNode::MDNode(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226521
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 23:13:14 +0000 (23:13 +0000)]
IR: Merge UniquableMDNode back into MDNode, NFC
As pointed out in r226501, the distinction between `MDNode` and
`UniquableMDNode` is confusing. When we need subclasses of `MDNode`
that don't use all its functionality it might make sense to break it
apart again, but until then this makes the code clearer.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226520
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:53:18 +0000 (22:53 +0000)]
IR: Extract MDNodeOpsKey, NFC
Make the MDTuple operand hashing logic reusable.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226519
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:52:07 +0000 (22:52 +0000)]
IR: Simplify uniquifyImpl(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226518
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:47:08 +0000 (22:47 +0000)]
IR: Simplify erasing from uniquing store, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226517
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:45:41 +0000 (22:45 +0000)]
Remove dead code, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226516
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:44:32 +0000 (22:44 +0000)]
Utils: Simplify MapMetadata(), NFC
Extract out the operand remapping loops, which are now very similar.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226515
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:41:14 +0000 (22:41 +0000)]
Skip upcast, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226514
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Mon, 19 Jan 2015 22:40:45 +0000 (22:40 +0000)]
[X86][AVX] Missing AVX1 memory folding float instructions
Now that we can create much more exhaustive X86 memory folding tests, this patch adds the missing AVX1/F16C floating point instruction stack foldings we can easily test for including the scalar intrinsics (add, div, max, min, mul, sub), conversions float/int to double, half precision conversions, rounding, dot product and bit test. The patch also adds a couple of obviously missing SSE instructions (more to follow once we have full SSE testing).
Now that scalar folding is working it broke a very old test (2006-10-07-ScalarSSEMiscompile.ll) - this test appears to make no sense as its trying to ensure that a scalar subtraction isn't folded as it 'would zero the top elts of the loaded vector' - this test just appears to be wrong to me.
Differential Revision: http://reviews.llvm.org/D7055
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226513
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:40:25 +0000 (22:40 +0000)]
Fix whitespace, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226512
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:39:07 +0000 (22:39 +0000)]
Utils: Simplify MapMetadata(), NFC
Take advantage of the new ability of temporary nodes to mutate to
distinct and uniqued nodes to greatly simplify the `MapMetadata()`
helper functions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226511
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:24:52 +0000 (22:24 +0000)]
IR: Allow temporary nodes to become uniqued or distinct
Add `MDNode::replaceWithUniqued()` and `MDNode::replaceWithDistinct()`,
which mutate temporary nodes to become uniqued or distinct. On uniquing
collisions, the unique version is returned and the node is deleted.
This takes advantage of temporary nodes being folded back in, and should
let me clean up some awkward logic in `MapMetadata()`.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226510
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:21:15 +0000 (22:21 +0000)]
IR: Remove templates from TempMDNodeDeleter, NFC
r226504 added `TempMDNodeDeleter` to help with `std::unique_ptr<>`-izing
the `MDNode::getTemporary()` interface. It doesn't need to be
templated, though.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226509
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:18:29 +0000 (22:18 +0000)]
IR: Split out countUnresolvedOperands(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226508
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 22:16:01 +0000 (22:16 +0000)]
Cleanup whitespace, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226507
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 21:30:18 +0000 (21:30 +0000)]
IR: Return unique_ptr from MDNode::getTemporary()
Change `MDTuple::getTemporary()` and `MDLocation::getTemporary()` to
return (effectively) `std::unique_ptr<T, MDNode::deleteTemporary>`, and
clean up call sites. (For now, `DIBuilder` call sites just call
`release()` immediately.)
There's an accompanying change in each of clang and polly to use the new
API.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226504
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Mon, 19 Jan 2015 21:11:14 +0000 (21:11 +0000)]
Add r224985 back with fixes.
The fixes are to note that AArch64 has additional restrictions on when local
relocations can be used. In particular, ld64 requires that relocations to
cstring/cfstrings use linker visible symbols.
Original message:
In an assembly expression like
bar:
.long L0 + 1
the intended semantics is that bar will contain a pointer one byte past L0.
In sections that are merged by content (strings, 4 byte constants, etc), a
single position in the section doesn't give the linker enough information.
For example, it would not be able to tell a relocation must point to the
end of a string, since that would look just like the start of the next.
The solution used in ELF to use relocation with symbols if there is a non-zero
addend.
In MachO before this patch we would just keep all symbols in some sections.
This would miss some cases (only cstrings on x86_64 were implemented) and was
inefficient since most relocations have an addend of 0 and can be represented
without the symbol.
This patch implements the non-zero addend logic for MachO too.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226503
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 20:37:44 +0000 (20:37 +0000)]
IR: Add MDLocation::getTemporary()
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226502
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 20:36:39 +0000 (20:36 +0000)]
IR: Remove MDNodeFwdDecl
Remove `MDNodeFwdDecl` (as promised in r226481). Aside from API
changes, there's no real functionality change here.
`MDNode::getTemporary()` now forwards to `MDTuple::getTemporary()`,
which returns a tuple with `isTemporary()` equal to true.
The main point is that we can now add temporaries of other `MDNode`
subclasses, needed for PR22235 (I introduced `MDNodeFwdDecl` in the
first place because I didn't recognize this need, and thought they were
only needed to handle forward references).
A few things left out of (or highlighted by) this commit:
- I've had to remove the (few) uses of `std::unique_ptr<>` to deal
with temporaries, since the destructor is no longer public.
`getTemporary()` should probably return the equivalent of
`std::unique_ptr<T, MDNode::deleteTemporary>`.
- `MDLocation::getTemporary()` doesn't exist yet (worse, it actually
does exist, but does the wrong thing: `MDNode::getTemporary()` is
inherited and returns an `MDTuple`).
- `MDNode` now only has one subclass, `UniquableMDNode`, and the
distinction between them is actually somewhat confusing.
I'll fix those up next.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226501
91177308-0d34-0410-b5e6-
96231b3b80d8
Colin LeMahieu [Mon, 19 Jan 2015 20:31:18 +0000 (20:31 +0000)]
[Hexagon] Updating muxir/ri/ii intrinsics. Setting predicate registers as compatible with i32 rather than doing custom type conversion.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226500
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 20:18:13 +0000 (20:18 +0000)]
IR: Extract out and reuse `storeImpl()`, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226499
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 20:16:50 +0000 (20:16 +0000)]
IR: Extract out getUniqued(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226498
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 20:14:15 +0000 (20:14 +0000)]
IR: Reuse `getImpl()` for `getDistinct()`, NFC
Merge `getDistinct()`'s implementation with those of `get()` and
`getIfExists()` for both `MDTuple` and `MDLocation`. This will make it
easier to scale to supporting temporaries.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226497
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:35:21 +0000 (19:35 +0000)]
IR: Move replaceAllUsesWith() back up to MDNode
Move `MDNodeFwdDecl::replaceAllUsesWith()` back up to `MDNode`, but
assert `isTemporary()`.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226495
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:32:27 +0000 (19:32 +0000)]
IR: Be specific about what MDNode::get() returns, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226494
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:31:09 +0000 (19:31 +0000)]
Remove dead code, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226493
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:29:25 +0000 (19:29 +0000)]
IR: Simplify MDNode::setOperand(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226492
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:28:28 +0000 (19:28 +0000)]
IR: Simplify handleChangedOperand() fast path, NFC
Use `isUniqued()` instead of `isStoredDistinctInContext()`, and remove
an assertion that won't be valid once temporaries are merged back in.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226491
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:26:24 +0000 (19:26 +0000)]
IR: Remove direct comparisons against Metadata::Storage, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226490
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:25:33 +0000 (19:25 +0000)]
IR: Assert that resolve() is only called on uniqued nodes, NFC
Add an assertion in `UniquableMDNode::resolve()` to prevent temporaries
from being resolved (once they're merged back in). Needed to shuffle
order of `resolve()` and `storeDistinctInContext()` to prevent it from
firing.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226489
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:10:14 +0000 (19:10 +0000)]
IR: Remove isa<UniquableMDNode>, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226488
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:09:14 +0000 (19:09 +0000)]
IR: Simplify DIBuilder::trackIfUnresolved(), NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226487
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:06:41 +0000 (19:06 +0000)]
IR: Remove isa<MDNodeFwdDecl>, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226486
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:03:18 +0000 (19:03 +0000)]
IR: Unify code for MDNode::isResolved(), NFC
Unify the definitions of `MDNode::isResolved()` and
`UniquableMDNode::isResolved()`. Previously, `UniquableMDNode` could
answer this question more efficiently, but now that RAUW support has
been unified with `MDNodeFwdDecl`, `MDNode` doesn't need any casts to
figure out the answer.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226485
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 19:02:06 +0000 (19:02 +0000)]
IR: Store RAUW support and Context in the same pointer, NFC
Add an `LLVMContext &` to `ReplaceableMetadataImpl`, create a class that
either holds a reference to an `LLVMContext` or owns a
`ReplaceableMetadataImpl`, and use the new class in `MDNode`.
- This saves a pointer in `UniquableMDNode` at the cost of a pointer
in `ValueAsMetadata` (which didn't used to store the `LLVMContext`).
There are far more of the former.
- Unifies RAUW support between `MDNodeFwdDecl` (which is going away,
see r226481) and `UniquableMDNode`.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226484
91177308-0d34-0410-b5e6-
96231b3b80d8
Colin LeMahieu [Mon, 19 Jan 2015 18:56:19 +0000 (18:56 +0000)]
[Hexagon] Converting intrinsics combine imm/imm, simple shifts and extends.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226483
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 18:45:35 +0000 (18:45 +0000)]
IR: Add isUniqued() and isTemporary()
Change `MDNode::isDistinct()` to only apply to 'distinct' nodes (not
temporaries), and introduce `MDNode::isUniqued()` and
`MDNode::isTemporary()` for the other two possibilities.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226482
91177308-0d34-0410-b5e6-
96231b3b80d8
Duncan P. N. Exon Smith [Mon, 19 Jan 2015 18:36:18 +0000 (18:36 +0000)]
IR: Use an enum to describe Metadata storage, NFC
More clearly describe the type of storage used for `Metadata`.
- `Uniqued`: uniqued, stored in the context.
- `Distinct`: distinct, stored in the context.
- `Temporary`: not owned by anyone.
This is the first in a series of commits to fix a design problem with
`MDNodeFwdDecl` that I need to solve for PR22235. While `MDNodeFwdDecl`
works well as a forward declaration, we use `MDNode::getTemporary()` for
more than forward declarations -- we also need to create early versions
of nodes (with fields not filled in) that we'll fill out later (see
`DIBuilder::finalize()` and `CGDebugInfo::finalize()` for examples).
This was a blind spot I had when I introduced `MDNodeFwdDecl` (which
David Blaikie (indirectly) highlighted in an unrelated review [1]).
[1]: http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-
20150112/252381.html
In general, we need `MDTuple::getTemporary()` to give a temporary tuple
(like `MDNodeFwdDecl`), `MDLocation::getTemporary()` to give a temporary
location, and (the problem at hand) `GenericDebugMDNode::getTemporary()`
to give a temporary generic debug node.
So I need to fold the idea of "temporary" nodes back into
`UniquableMDNode`. (More commits to follow as I refactor.)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226481
91177308-0d34-0410-b5e6-
96231b3b80d8
Colin LeMahieu [Mon, 19 Jan 2015 18:33:58 +0000 (18:33 +0000)]
[Hexagon] Converting remaining ALU32/ALU intrinsics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226480
91177308-0d34-0410-b5e6-
96231b3b80d8
Colin LeMahieu [Mon, 19 Jan 2015 18:22:19 +0000 (18:22 +0000)]
[Hexagon] Converting ALU32/ALU intrinsics to new patterns.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226478
91177308-0d34-0410-b5e6-
96231b3b80d8
Adrian Prantl [Mon, 19 Jan 2015 17:57:29 +0000 (17:57 +0000)]
Remove support for DIVariable's FlagIndirectVariable and expect
frontends to use a DIExpression with a DW_OP_deref instead.
This is not only a much more natural place for this informationl; there
is also a technical reason: The FlagIndirectVariable is used to mark a
variable that is turned into a reference by virtue of the calling
convention; this happens for example to aggregate return values.
The inliner, for example, may actually need to undo this indirection to
correctly represent the value in its new context. This is impossible to
implement because the DIVariable can't be safely modified. We can however
safely construct a new DIExpression on the fly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226476
91177308-0d34-0410-b5e6-
96231b3b80d8
Greg Fitzgerald [Mon, 19 Jan 2015 17:40:05 +0000 (17:40 +0000)]
[AArch64] Implement GHC calling convention
Original patch by Luke Iannini. Minor improvements and test added by
Erik de Castro Lopo.
Differential Revision: http://reviews.llvm.org/D6877
From: Erik de Castro Lopo <erikd@mega-nerd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226473
91177308-0d34-0410-b5e6-
96231b3b80d8
Colin LeMahieu [Mon, 19 Jan 2015 17:36:32 +0000 (17:36 +0000)]
[Hexagon] Converting halfword to double accumulating multiply intrinsics.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226472
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Mon, 19 Jan 2015 17:30:24 +0000 (17:30 +0000)]
Produce errors when an assignment expression would use a common symbol.
An assignment will produce a symbol with a given section and offset. There is
no way to represent something like "1 byte after a common symbol".
This matches the behavior of GNU as.
Part of PR22217.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226470
91177308-0d34-0410-b5e6-
96231b3b80d8
Bradley Smith [Mon, 19 Jan 2015 16:37:17 +0000 (16:37 +0000)]
[ARM] SSAT/USAT with an 'asr #32' shift should result in an undefined encoding rather than unpredictable
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226469
91177308-0d34-0410-b5e6-
96231b3b80d8
Bradley Smith [Mon, 19 Jan 2015 16:36:02 +0000 (16:36 +0000)]
[ARM] Fixup sign extend instruction availability w.r.t. DSP extension
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226468
91177308-0d34-0410-b5e6-
96231b3b80d8
Rafael Espindola [Mon, 19 Jan 2015 15:16:06 +0000 (15:16 +0000)]
Bring r226038 back.
No change in this commit, but clang was changed to also produce trivial comdats when
needed.
Original message:
Don't create new comdats in CodeGen.
This patch stops the implicit creation of comdats during codegen.
Clang now sets the comdat explicitly when it is required. With this patch clang and gcc
now produce the same result in pr19848.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226467
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 19 Jan 2015 12:36:53 +0000 (12:36 +0000)]
[PM] Replace the Pass argument to SplitEdge with specific analyses used
and updated.
This may appear to remove handling for things like alias analysis when
splitting critical edges here, but in fact no callers of SplitEdge
relied on this. Similarly, all of them wanted to preserve LCSSA if there
was any update of the loop info. That makes the interface much simpler.
With this, all of BasicBlockUtils.h is free of Pass arguments and
prepared for the new pass manager. This is tho majority of utilities
that relied on pass arguments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226459
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 19 Jan 2015 12:12:00 +0000 (12:12 +0000)]
[PM] Cleanup a dead option to critical edge splitting that I noticed
while refactoring this API for the new pass manager.
No functionality changed here, the code didn't actually support this
option.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226457
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Mon, 19 Jan 2015 12:09:11 +0000 (12:09 +0000)]
[PM] Remove the Pass argument from all of the critical edge splitting
APIs and replace it and numerous booleans with an option struct.
The critical edge splitting API has a really large surface of flags and
so it seems worth burning a small option struct / builder. This struct
can be constructed with the various preserved analyses and then flags
can be flipped in a builder style.
The various users are now responsible for directly passing along their
analysis information. This should be enough for the critical edge
splitting to work cleanly with the new pass manager as well.
This API is still pretty crufty and could be cleaned up a lot, but I've
focused on this change just threading an option struct rather than
a pass through the API.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226456
91177308-0d34-0410-b5e6-
96231b3b80d8