Another case of x86-specific shuffle strength reduction:
avoid generating insert*128 instructions with index 0 because
they are slower than their non-lane-changing blend equivalents.
Shuffle lowering already catches most of these cases, but
the zero vector case and some other paths such as in the
modified test in vector-shuffle-256-v32.ll were getting
through.
Differential Revision: http://reviews.llvm.org/D8366
llvm-svn: 232773
There are two main advantages to doing this
* Targets that only need to handle one of the formats specially don't have
to worry about the others. For example, x86 now only registers a
constructor for the COFF streamer.
* Changes to the arguments passed to one format constructor will not impact
the other formats.
llvm-svn: 232699
Currently v2i64 vectors shifts (non-equal shift amounts) are scalarized, costing 4 x extract, 2 x x86-shifts and 2 x insert instructions - and it gets even more awkward on 32-bit targets.
This patch separately shifts the vector by both shift amounts and then shuffles the partial results back together, costing 2 x shuffles and 2 x sse-shifts instructions (+ 2 movs on pre-AVX hardware).
Note - this patch only improves the SHL / LSHR logical shifts as only these are supported in SSE hardware.
Differential Revision: http://reviews.llvm.org/D8416
llvm-svn: 232660
Summary:
This is instead of doing this in target independent code and is the last
non-functional change before targets begin to distinguish between
different memory constraints when selecting code for the ISD::INLINEASM
node.
Next, each target will individually move away from the idea that all
memory constraints behave like 'm'.
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D8173
llvm-svn: 232373
This fixes pr22854.
The core issue on the bug is that there are multiple instructions that
print the same in assembly. In fact, there doesn't seem to be any
syntax for specifying that a constant that fits in 8 bits should use a 32 bit
immediate.
The attached patch changes fast isel to consider i16immSExt8,
i32immSExt8, and i64immSExt8. They were disabled because fastisel didn’t know
to call the predicate back in the day.
llvm-svn: 232223
This patch fixes a bug in the shuffle lowering logic implemented by function
'lowerV2X128VectorShuffle'.
The are few cases where function 'lowerV2X128VectorShuffle' wrongly expands a
shuffle of two v4X64 vectors into a CONCAT_VECTORS of two EXTRACT_SUBVECTOR
nodes. The problematic expansion only occurs when the shuffle mask M has an
'undef' element at position 2, and M is equivalent to mask <0,1,4,5>.
In that case, the algorithm propagates the wrong vector to one of the two
new EXTRACT_SUBVECTOR nodes.
Example:
;;
define <4 x double> @test(<4 x double> %A, <4 x double> %B) {
entry:
%0 = shufflevector <4 x double> %A, <4 x double> %B, <4 x i32><i32 undef, i32 1, i32 undef, i32 5>
ret <4 x double> %0
}
;;
Before this patch, llc (-mattr=+avx) generated:
vinsertf128 $1, %xmm0, %ymm0, %ymm0
With this patch, llc correctly generates:
vinsertf128 $1, %xmm1, %ymm0, %ymm0
Added test lower-vec-shuffle-bug.ll
Differential Revision: http://reviews.llvm.org/D8259
llvm-svn: 232179
The operand flag word for ISD::INLINEASM nodes now contains a 15-bit
memory constraint ID when the operand kind is Kind_Mem. This constraint
ID is a numeric equivalent to the constraint code string and is converted
with a target specific hook in TargetLowering.
This patch maps all memory constraints to InlineAsm::Constraint_m so there
is no functional change at this point. It just proves that using these
previously unused bits in the encoding of the flag word doesn't break
anything.
The next patch will make each target preserve the current mapping of
everything to Constraint_m for itself while changing the target independent
implementation of the hook to return Constraint_Unknown appropriately. Each
target will then be adapted in separate patches to use appropriate
Constraint_* values.
PR22883 was caused the matching operands copying the whole of the operand flags
for the matched operand. This included the constraint id which needed to be
replaced with the operand number. This has been fixed with a conversion
function. Following on from this, matching operands also used the operand
number as the constraint id. This has been fixed by looking up the matched
operand and taking it from there.
llvm-svn: 232165
This should complete the job started in r231794 and continued in r232045:
We want to replace as much custom x86 shuffling via intrinsics
as possible because pushing the code down the generic shuffle
optimization path allows for better codegen and less complexity
in LLVM.
AVX2 introduced proper integer variants of the hacked integer insert/extract
C intrinsics that were created for this same functionality with AVX1.
This should complete the removal of insert/extract128 intrinsics.
The Clang precursor patch for this change was checked in at r232109.
llvm-svn: 232120
This (r232027) has caused PR22883; so it seems those bits might be used by
something else after all. Reverting until we can figure out what else to do.
Original commit message:
The operand flag word for ISD::INLINEASM nodes now contains a 15-bit
memory constraint ID when the operand kind is Kind_Mem. This constraint
ID is a numeric equivalent to the constraint code string and is converted
with a target specific hook in TargetLowering.
This patch maps all memory constraints to InlineAsm::Constraint_m so there
is no functional change at this point. It just proves that using these
previously unused bits in the encoding of the flag word doesn't break anything.
The next patch will make each target preserve the current mapping of
everything to Constraint_m for itself while changing the target independent
implementation of the hook to return Constraint_Unknown appropriately. Each
target will then be adapted in separate patches to use appropriate Constraint_*
values.
llvm-svn: 232093
The permps and permd instructions have their operands swapped compared to the
intrinsic definition. Therefore, they do not fall into the INTR_TYPE_2OP
category.
I did not create a new category for those two, as they are the only one AFAICT
in that case.
<rdar://problem/20108262>
llvm-svn: 232085
Part of the folding logic implemented by function 'PerformISDSETCCCombine'
only worked under the assumption that the condition code in input could have
been either SETNE or SETEQ.
Unfortunately that assumption was incorrect, and in some cases the algorithm
ended up incorrectly folding SETCC nodes.
The incorrect folding only affected SETCC dag nodes where:
- one of the operands was a build_vector of all zeroes;
- the other operand was a SIGN_EXTEND from a vector of MVT:i1 elements;
- the condition code was neither SETNE nor SETEQ.
Example:
(setcc (v4i32 (sign_extend v4i1:%A)), (v4i32 VectorOfAllZeroes), setge)
Before this patch, the entire dag node sequence from the example was
incorrectly folded to node %A.
With this patch, the dag node sequence is folded to a
(xor %A, (v4i1 VectorOfAllOnes)).
Added test setcc-combine.ll.
Thanks to Greg Bedwell for spotting this issue.
llvm-svn: 232046
Summary:
The operand flag word for ISD::INLINEASM nodes now contains a 15-bit
memory constraint ID when the operand kind is Kind_Mem. This constraint
ID is a numeric equivalent to the constraint code string and is converted
with a target specific hook in TargetLowering.
This patch maps all memory constraints to InlineAsm::Constraint_m so there
is no functional change at this point. It just proves that using these
previously unused bits in the encoding of the flag word doesn't break anything.
The next patch will make each target preserve the current mapping of
everything to Constraint_m for itself while changing the target independent
implementation of the hook to return Constraint_Unknown appropriately. Each
target will then be adapted in separate patches to use appropriate Constraint_*
values.
Reviewers: hfinkel
Reviewed By: hfinkel
Subscribers: hfinkel, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D8171
llvm-svn: 232027
Summary:
I don't know why every singled backend had to redeclare its own DataLayout.
There was a virtual getDataLayout() on the common base TargetMachine, the
default implementation returned nullptr. It was not clear from this that
we could assume at call site that a DataLayout will be available with
each Target.
Now getDataLayout() is no longer virtual and return a pointer to the
DataLayout member of the common base TargetMachine. I plan to turn it into
a reference in a future patch.
The only backend that didn't have a DataLayout previsouly was the CPPBackend.
It now initializes the default DataLayout. This commit is NFC for all the
other backends.
Test Plan: clang+llvm ninja check-all
Reviewers: echristo
Subscribers: jfb, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D8243
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 231987
This is a follow-up to r231182. This adds the "vbroadcasti128" instruction
back, but without the intrinsic mapping. Also add a test to check the
instriction encoding.
This is related to rdar://problem/18742778.
llvm-svn: 231945
Summary:
The generic ELF TargetObjectFile defaults to .ctors, but Linux's
defaults to .init_array by calling InitializeELF with the value of
UseInitArray from TargetMachine. Make NaCl's behavior match.
Reviewers: jvoung
Differential Revision: http://reviews.llvm.org/D8240
llvm-svn: 231934
There were cases where the backend computed a wrong permute mask for a VPERM2X128 node.
Example:
\code
define <8 x float> @foo(<8 x float> %a, <8 x float> %b) {
%shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 undef, i32 undef, i32 6, i32 7, i32 undef, i32 undef, i32 6, i32 7>
ret <8 x float> %shuffle
}
\code end
Before this patch, llc (with -mattr=+avx) emitted the following vperm2f128:
vperm2f128 $0, %ymm0, %ymm0, %ymm0 # ymm0 = ymm0[0,1,0,1]
With this patch, llc emits a vperm2f128 with a correct permute mask:
vperm2f128 $17, %ymm0, %ymm0, %ymm0 # ymm0 = ymm0[2,3,2,3]
Differential Revision: http://reviews.llvm.org/D8119
llvm-svn: 231601
We have an increasing number of cases where we are creating commuted shuffle masks - all implementing nearly the same code.
This patch adds a static helper function - ShuffleVectorSDNode::commuteMask() and replaces a number of cases to use it.
Differential Revision: http://reviews.llvm.org/D8139
llvm-svn: 231581
In theory this allows the compiler to skip materializing the array on
the stack. In practice clang often fails to do that, but that's a
different story. NFC.
llvm-svn: 231571
Add MachO 32-bit (i.e. arm and x86) support for replacing global GOT equivalent
symbol accesses. Unlike 64-bit targets, there's no GOTPCREL relocation, and
access through a non_lazy_symbol_pointers section is used instead.
-- before
_extgotequiv:
.long _extfoo
_delta:
.long _extgotequiv-_delta
-- after
_delta:
.long L_extfoo$non_lazy_ptr-_delta
.section __IMPORT,__pointers,non_lazy_symbol_pointers
L_extfoo$non_lazy_ptr:
.indirect_symbol _extfoo
.long 0
llvm-svn: 231475
Follow up r230264 and add ARM64 support for replacing global GOT
equivalent symbol accesses by references to the GOT entry for the final
symbol instead, example:
-- before
.globl _foo
_foo:
.long 42
.globl _gotequivalent
_gotequivalent:
.quad _foo
.globl _delta
_delta:
.long _gotequivalent-_delta
-- after
.globl _foo
_foo:
.long 42
.globl _delta
Ltmp3:
.long _foo@GOT-Ltmp3
llvm-svn: 231474
We supported forming IMGREL relocations from ConstantExprs involving
__ImageBase if the minuend was a GlobalVariable. Extend this
functionality to all GlobalObjects.
llvm-svn: 231456
This patch reduces code size for all AVX targets and increases speed for some chips.
SSE 4.1 introduced the useless (see code comments) 2-register form of BLENDV and
only in the packed float/double flavors.
AVX subsequently made the instruction useful by adding a 4-register operand form.
So we just need to paper over the lack of scalar forms of this instruction, complicate
the code to choose float or double forms, and use blendv on scalars since all FP is in
xmm registers anyway.
This gives us an approximately 50% speed up for a blendv microbenchmark sequence
on SandyBridge and Haswell:
blendv : 29.73 cycles/iter
logic : 43.15 cycles/iter
No new test cases with this patch because:
1. fast-isel-select-sse.ll tests the positive side for regular X86 lowering and fast-isel
2. sse-minmax.ll and fp-select-cmp-and.ll confirm that we're not firing for scalar selects without AVX
3. fp-select-cmp-and.ll and logical-load-fold.ll confirm that we're not firing for scalar selects with constants.
http://llvm.org/bugs/show_bug.cgi?id=22483
Differential Revision: http://reviews.llvm.org/D8063
llvm-svn: 231408
Added lowering for ISD::CONCAT_VECTORS and ISD::INSERT_SUBVECTOR for i1 vectors,
it is needed to pass all masked_memop.ll tests for SKX.
llvm-svn: 231371
Summary:
In PNaCl, most atomic instructions have their own @llvm.nacl.atomic.* function, each one, with a few exceptions, represents a consistent behaviour across all NaCl-supported targets. Unfortunately, the atomic RMW operations nand, [u]min, and [u]max aren't directly represented by any such @llvm.nacl.atomic.* function. This patch refines shouldExpandAtomicRMWInIR in TargetLowering so that a future `Le32TargetLowering` class can selectively inform the caller how the target desires the atomic RMW instruction to be expanded (ie via load-linked/store-conditional for ARM/AArch64, via cmpxchg for X86/others?, or not at all for Mips) if at all.
This does not represent a behavioural change and as such no tests were added.
Patch by: Richard Diamond.
Reviewers: jfb
Reviewed By: jfb
Subscribers: jfb, aemerson, t.p.northover, llvm-commits
Differential Revision: http://reviews.llvm.org/D7713
llvm-svn: 231250
The target-independent selection algorithm in FastISel already knows how
to select a SINT_TO_FP if the target is SSE but not AVX.
On targets that have SSE but not AVX, the tablegen'd 'fastEmit' functions
for ISD::SINT_TO_FP know how to select instruction X86::CVTSI2SSrr
(for an i32 to f32 conversion) and X86::CVTSI2SDrr (for an i32 to f64
conversion).
This patch simplifies the logic in method X86SelectSIToFP knowing that
the code would not be reachable if the subtarget doesn't have AVX.
No functional change intended.
llvm-svn: 231243
The intrinsic is no longer generated by the front-end. Remove the intrinsic and
auto-upgrade it to a vector shuffle.
Reviewed by Nadav
This is related to rdar://problem/18742778.
llvm-svn: 231182
Previously we had only Linux using DTPOFF for these; all X86 ELF
targets should. Fixes a side issue mentioned in PR21077.
Differential Revision: http://reviews.llvm.org/D8011
llvm-svn: 231130
This lets us avoid a few copies that are otherwise hard to get rid of.
The way this is done is, the custom-inserter looks at the following
instruction for another CMOV, and replaces both at the same time.
A previous version used a new CMOV2 opcode, but the custom inserter
is expected to be able to return a different basic block anyway, which
means it's OK - though far from ideal - to alter that block's contents.
Explicitly document that, in case it ever makes a difference.
Alternatives welcome!
Follow-up to r231045.
rdar://19767934
Closes http://reviews.llvm.org/D8019
llvm-svn: 231046
Fold and/or of setcc's to double CMOV:
(CMOV F, T, ((cc1 | cc2) != 0)) -> (CMOV (CMOV F, T, cc1), T, cc2)
(CMOV F, T, ((cc1 & cc2) != 0)) -> (CMOV (CMOV T, F, !cc1), F, !cc2)
When we can't use the CMOV instruction, it might increase branch
mispredicts. When we can, or when there is no mispredict, this
improves throughput and reduces register pressure.
These can't be catched by generic combines, because the pattern can
appear when legalizing some instructions (such as fcmp une).
rdar://19767934
http://reviews.llvm.org/D7634
llvm-svn: 231045
With initializer lists there is a really neat idiomatic way to write
this, 'ArrayRef.equals({1, 2, 3, 4, 5})'. Remove the equal method which
always had a hard limit on the number of arguments. I considered
rewriting it with variadic templates but that's not really a good fit
for a function with homogeneous arguments.
'ArrayRef == {1, 2, 3, 4, 5}' would've been even more awesome, but C++11
doesn't allow init lists with binary operators.
llvm-svn: 230907
All of the cases were just appending from random access iterators to a
vector. Using insert/append can grow the vector to the perfect size
directly and moves the growing out of the loop. No intended functionalty
change.
llvm-svn: 230845
Summary:
Until now, we did this (among other things) based on whether or not the
target was Windows. This is clearly wrong, not just for Win64 ABI functions
on non-Windows, but for System V ABI functions on Windows, too. In this
change, we make this decision based on the ABI the calling convention
specifies instead.
Reviewers: rnk
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D7953
llvm-svn: 230793
vectors. This lets us fix the rest of the v16 lowering problems when
pshufb is clearly better.
We might still be able to improve some of the lowerings by enabling the
other combine-based rewriting to fire for non-128-bit vectors, but this
at least should remove any regressions from using the fancy v16i16
lowering strategy.
llvm-svn: 230753
repeated 128-bit lane shuffles of wider vector types and use it to lower
256-bit v16i16 vector shuffles where applicable.
This should let us perfectly lowering the pattern of pshuflw and pshufhw
even for AVX2 256-bit patterns.
I've not added AVX-512 support, but it should be trivial for someone
working on that to wire up.
Note that currently this generates bad, long shuffle chains because we
don't combine 256-bit target shuffles. The subsequent patches will fix
that.
llvm-svn: 230751
Summary:
This change causes us to actually save non-volatile registers in a Win64
ABI function that calls a System V ABI function, and vice-versa.
Reviewers: rnk
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D7919
llvm-svn: 230714
uses of TM->getSubtargetImpl and propagate to all calls.
This could be a debugging regression in places where we had a
TargetMachine and/or MachineFunction but don't have it as part
of the MachineInstr. Fixing this would require passing a
MachineFunction/Function down through the print operator, but
none of the existing uses in tree seem to do this.
llvm-svn: 230710
a lookup, pass that in rather than use a naked call to getSubtargetImpl.
This involved passing down and around either a TargetMachine or
TargetRegisterInfo. Update all callers/definitions around the targets
and SelectionDAG.
llvm-svn: 230699
blend as legal.
We made the same mistake in two different places. Whenever we are custom
lowering a v32i8 blend we need to check whether we are custom lowering
it only for constant conditions that can be shuffled, or whether we
actually have AVX2 and full dynamic blending support on bytes. Both are
fixed, with comments added to make it clear what is going on and a new
test case.
llvm-svn: 230695
dynamic blends.
This makes it much more clear what is going on. The case we're handling
is that of dynamic conditions, and we're bailing when the nature of the
vector types and subtarget preclude lowering the dynamic condition
vselect as an actual blend.
No functionality changed here, but this will make a subsequent bug-fix
to this code much more clear.
llvm-svn: 230690
The latency for the WriteMULm class was set to 4, which is actually lower than the latency for WriteMULr (5).
A better estimate would be 4 added to WriteMULr, that is, 9.
llvm-svn: 230634
formulaic into the top v8i16 lowering routine.
This makes the generalized lowering a completely general and single path
lowering which will allow generalizing it in turn for multiple 128-bit
lanes.
llvm-svn: 230623
Explanation: This function is in TargetLowering because it uses
RegClassForVT which would need to be moved to TargetRegisterInfo
and would necessitate moving isTypeLegal over as well - a massive
change that would just require TargetLowering having a TargetRegisterInfo
class member that it would use.
llvm-svn: 230585
This required plumbing a TargetRegisterInfo through computeRegisterProperties
and into findRepresentativeClass which uses it for register class
iteration. This required passing a subtarget into a few target specific
initializations of TargetLowering.
llvm-svn: 230583
The Win64 epilogue structure is very restrictive, it permits a very
small number of opcodes and none of them are 'mov'.
This means that given:
mov %rbp, %rsp
pop %rbp
The mov isn't the epilogue, only the pop is. This is problematic unless
a frame pointer is present in which case we are free to do whatever we'd
like in the "body" of the function. If a frame pointer is present,
unwinding will undo the prologue operations in reverse order regardless
of the fact that we are at an instruction which is reseting the stack
pointer.
llvm-svn: 230543
Reapply r230248.
Teach the peephole optimizer to work with MMX instructions by adding
entries into the foldable tables. This covers folding opportunities not
handled during isel.
llvm-svn: 230499
MMX_MOVD64rm zero-extends i32 load results into i64 registers.
The peephole optimizer will try to fold it in other MMX foldable
instructions, the wrong thing to do, since there's no MMX memory
instruction that loads from i32 and does implict zero extension.
Remove 'canFoldAsLoad' from MOVD64rm in order to prevent such folding.
The current MMX tests already test this, but since there are no MMX
instructions in the foldable tables yet, this did not trigger. This
commit prepares the addition of those instructions.
llvm-svn: 230498
Gather and scatter instructions additionally write to one of the source operands - mask register.
In this case Gather has 2 destination values - the loaded value and the mask.
Till now we did not support code gen pattern for gather - the instruction was generated from
intrinsic only and machine node was hardcoded.
When we introduce the masked_gather node, we need to select instruction automatically,
in the standard way.
I added a flag "hasTwoExplicitDefs" that allows to handle 2 destination operands.
(Some code in the X86InstrFragmentsSIMD.td is commented out, just to split one big
patch in many small patches)
llvm-svn: 230471
Prologue emission, in some cases, requires calls to a stack probe helper
function. The amount of stack to probe is passed as a register
argument in the Win64 ABI but the instruction sequence used is
pessimistic: it assumes that the number of bytes to probe is greater
than 4 GB.
Instead, select a more appropriate opcode depending on the number of
bytes we are going to probe.
llvm-svn: 230270
Front-ends could use global unnamed_addr to hold pointers to other
symbols, like @gotequivalent below:
@foo = global i32 42
@gotequivalent = private unnamed_addr constant i32* @foo
@delta = global i32 trunc (i64 sub (i64 ptrtoint (i32** @gotequivalent to i64),
i64 ptrtoint (i32* @delta to i64))
to i32)
The global @delta holds a data "PC"-relative offset to @gotequivalent,
an unnamed pointer to @foo. The darwin/x86-64 assembly output for this follows:
.globl _foo
_foo:
.long 42
.globl _gotequivalent
_gotequivalent:
.quad _foo
.globl _delta
_delta:
.long _gotequivalent-_delta
Since unnamed_addr indicates that the address is not significant, only
the content, we can optimize the case above by replacing pc-relative
accesses to "GOT equivalent" globals, by a PC relative access to the GOT
entry of the final symbol instead. Therefore, "delta" can contain a pc
relative relocation to foo's GOT entry and we avoid the emission of
"gotequivalent", yielding the assembly code below:
.globl _foo
_foo:
.long 42
.globl _delta
_delta:
.long _foo@GOTPCREL+4
There are a couple of advantages of doing this: (1) Front-ends that need
to emit a great deal of data to store pointers to external symbols could
save space by not emitting such "got equivalent" globals and (2) IR
constructs combined with this opt opens a way to represent GOT pcrel
relocations by using the LLVM IR, which is something we previously had
no way to express.
Differential Revision: http://reviews.llvm.org/D6922
rdar://problem/18534217
llvm-svn: 230264
Teach the peephole optimizer to work with MMX instructions by adding
entries into the foldable tables. This covers folding opportunities not
handled during isel.
llvm-svn: 230226
I made the templates general, no need to define pattern separately for each instruction/intrinsic.
Now only need to add r_Int pattern for AVX.
llvm-svn: 230221
Synthesizing a call directly using the MI layer would confuse the frame
lowering code. This is problematic as frame lowering is highly
sensitive the particularities of calls, etc.
llvm-svn: 230129
Everyone except R600 was manually passing the length of a static array
at each callsite, calculated in a variety of interesting ways. Far
easier to let ArrayRef handle that.
There should be no functional change, but out of tree targets may have
to tweak their calls as with these examples.
llvm-svn: 230118
Stack realignment occurs after the prolog, not during, for Win64.
Because of this, don't factor in the maximum stack alignment when
establishing a frame pointer.
This fixes PR22572.
llvm-svn: 230113
This patch teaches X86FastISel how to select intrinsic 'convert_from_fp16' and
intrinsic 'convert_to_fp16'.
If the target has F16C, we can select VCVTPS2PHrr for a float-half conversion,
and VCVTPH2PSrr for a half-float conversion.
Differential Revision: http://reviews.llvm.org/D7673
llvm-svn: 230043
This canonicalization step saves us 3 pattern matching possibilities * 4 math ops
for scalar FP math that uses xmm regs. The backend can re-commute the operands
post-instruction-selection if that makes register allocation better.
The tests in llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll cover this scenario already,
so there are no new tests with this patch.
Differential Revision: http://reviews.llvm.org/D7777
llvm-svn: 230024
the wrong answer. We also got initializer lists which are *way* cleaner
for this kind of thing. Let's use those and make this a normal, boring
functionn accepting ArrayRef.
llvm-svn: 230004
The new shuffle lowering has been the default for some time. I've
enabled the new legality testing by default with no really blocking
regressions. I've fuzz tested this very heavily (many millions of fuzz
test cases have passed at this point). And this cleans up a ton of code.
=]
Thanks again to the many folks that helped with this transition. There
was a lot of work by others that went into the new shuffle lowering to
make it really excellent.
In case you aren't using a diff algorithm that can handle this:
X86ISelLowering.cpp: 22 insertions(+), 2940 deletions(-)
llvm-svn: 229964
is going well, remove the flag and the code for the old legality tests.
This is the first step toward removing the entire old vector shuffle
lowering. *Much* more code to delete coming up next.
llvm-svn: 229963
reflects the fact that the x86 backend can in fact lower any shuffle you
want it to with reasonably high code quality.
My recent work on the new vector shuffle has made this regress *very*
little. The diff in the test cases makes me very, very happy.
llvm-svn: 229958
systematic lowering of v8i16.
This required a slight strategy shift to prefer unpack lowerings in more
places. While this isn't a cut-and-dry win in every case, it is in the
overwhelming majority. There are only a few places where the old
lowering would probably be a touch faster, and then only by a small
margin.
In some cases, this is yet another significant improvement.
llvm-svn: 229859
addition to lowering to trees rooted in an unpack.
This saves shuffles and or registers in many various ways, lets us
handle another class of v4i32 shuffles pre SSE4.1 without domain
crosses, etc.
llvm-svn: 229856
terribly complex partial blend logic.
This code path was one of the more complex and bug prone when it first
went in and it hasn't faired much better. Ultimately, with the simpler
basis for unpack lowering and support bit-math blending, this is
completely obsolete. In the worst case without this we generate
different but equivalent instructions. However, in many cases we
generate much better code. This is especially true when blends or pshufb
is available.
This does expose one (minor) weakness of the unpack lowering that I'll
try to address.
In case you were wondering, this is actually a big part of what I've
been trying to pull off in the recent string of commits.
llvm-svn: 229853
needed, and significantly improve the SSSE3 path.
This makes the new strategy much more clear. If we can blend, we just go
with that. If we can't blend, we try to permute into an unpack so
that we handle cases where the unpack doing the blend also simplifies
the shuffle. If that fails and we've got SSSE3, we now call into
factored-out pshufb lowering code so that we leverage the fact that
pshufb can set up a blend for us while shuffling. This generates great
code, especially because we *know* we don't have a fast blend at this
point. Finally, we fall back on decomposing into permutes and blends
because we do at least have a bit-math-based blend if we need to use
that.
This pretty significantly improves some of the v8i16 code paths. We
never need to form pshufb for the single-input shuffles because we have
effective target-specific combines to form it there, but we were missing
its effectiveness in the blends.
llvm-svn: 229851
them into permutes and a blend with the generic decomposition logic.
This works really well in almost every case and lets the code only
manage the expansion of a single input into two v8i16 vectors to perform
the actual shuffle. The blend-based merging is often much nicer than the
pack based merging that this replaces. The only place where it isn't we
end up blending between two packs when we could do a single pack. To
handle that case, just teach the v2i64 lowering to handle these blends
by digging out the operands.
With this we're down to only really random permutations that cause an
explosion of instructions.
llvm-svn: 229849
v16i8 shuffles, and replace it with new facilities.
This uses precise patterns to match exact unpacks, and the new
generalized unpack lowering only when we detect a case where we will
have to shuffle both inputs anyways and they terminate in exactly
a blend.
This fixes all of the blend horrors that I uncovered by always lowering
blends through the vector shuffle lowering. It also removes *sooooo*
much of the crazy instruction sequences required for v16i8 lowering
previously. Much cleaner now.
The only "meh" aspect is that we sometimes use pshufb+pshufb+unpck when
it would be marginally nicer to use pshufb+pshufb+por. However, the
difference there is *tiny*. In many cases its a win because we re-use
the pshufb mask. In others, we get to avoid the pshufb entirely. I've
left a FIXME, but I'm dubious we can really do better than this. I'm
actually pretty happy with this lowering now.
For SSE2 this exposes some horrors that were really already there. Those
will have to fixed by changing a different path through the v16i8
lowering.
llvm-svn: 229846
on things not being marked as either custom or legal, but we now do
custom lowering of more VSELECT nodes. To cope with this, manually
replicate the legality tests here. These have to stay in sync with the
set of tests used in the custom lowering of VSELECT.
Ideally, we wouldn't do any of this combine-based-legalization when we
have an actual custom legalization step for VSELECT, but I'm not going
to be able to rewrite all of that today.
I don't have a test case for this currently, but it was found when
compiling a number of the test-suite benchmarks. I'll try to reduce
a test case and add it.
This should at least fix the test-suite fallout on build bots.
llvm-svn: 229844
lowering paths. I'm going to be leveraging this to simplify a lot of the
overly complex lowering of v8 and v16 shuffles in pre-SSSE3 modes.
Sadly, this isn't profitable on v4i32 and v2i64. There, the float and
double blending instructions for pre-SSE4.1 are actually pretty good,
and we can't beat them with bit math. And once SSE4.1 comes around we
have direct blending support and this ceases to be relevant.
Also, some of the test cases look odd because the domain fixer
canonicalizes these to floating point domain. That's OK, it'll use the
integer domain when it matters and some day I may be able to update
enough of LLVM to canonicalize the other way.
This restores almost all of the regressions from teaching x86's vselect
lowering to always use vector shuffle lowering for blends. The remaining
problems are because the v16 lowering path is still doing crazy things.
I'll be re-arranging that strategy in more detail in subsequent commits
to finish recovering the performance here.
llvm-svn: 229836
First, don't combine bit masking into vector shuffles (even ones the
target can handle) once operation legalization has taken place. Custom
legalization of vector shuffles may exist for these patterns (making the
predicate return true) but that custom legalization may in some cases
produce the exact bit math this matches. We only really want to handle
this prior to operation legalization.
However, the x86 backend, in a fit of awesome, relied on this. What it
would do is mark VSELECTs as expand, which would turn them into
arithmetic, which this would then match back into vector shuffles, which
we would then lower properly. Amazing.
Instead, the second change is to teach the x86 backend to directly form
vector shuffles from VSELECT nodes with constant conditions, and to mark
all of the vector types we support lowering blends as shuffles as custom
VSELECT lowering. We still mark the forms which actually support
variable blends as *legal* so that the custom lowering is bypassed, and
the legal lowering can even be used by the vector shuffle legalization
(yes, i know, this is confusing. but that's how the patterns are
written).
This makes the VSELECT lowering much more sensible, and in fact should
fix a bunch of bugs with it. However, as you'll see in the test cases,
right now what it does is point out the *hilarious* deficiency of the
new vector shuffle lowering when it comes to blends. Fortunately, my
very next patch fixes that. I can't submit it yet, because that patch,
somewhat obviously, forms the exact and/or pattern that the DAG combine
is matching here! Without this patch, teaching the vector shuffle
lowering to produce the right code infloops in the DAG combiner. With
this patch alone, we produce terrible code but at least lower through
the right paths. With both patches, all the regressions here should be
fixed, and a bunch of the improvements (like using 2 shufps with no
memory loads instead of 2 andps with memory loads and an orps) will
stay. Win!
There is one other change worth noting here. We had hilariously wrong
vectorization cost estimates for vselect because we fell through to the
code path that assumed all "expand" vector operations are scalarized.
However, the "expand" lowering of VSELECT is vector bit math, most
definitely not scalarized. So now we go back to the correct if horribly
naive cost of "1" for "not scalarized". If anyone wants to add actual
modeling of shuffle costs, that would be cool, but this seems an
improvement on its own. Note the removal of 16 and 32 "costs" for doing
a blend. Even in SSE2 we can blend in fewer than 16 instructions. ;] Of
course, we don't right now because of OMG bad code, but I'm going to fix
that. Next patch. I promise.
llvm-svn: 229835
Previously, subtarget features were a bitfield with the underlying type being uint64_t.
Since several targets (X86 and ARM, in particular) have hit or were very close to hitting this bound, switching the features to use a bitset.
No functional change.
Differential Revision: http://reviews.llvm.org/D7065
llvm-svn: 229831
Sanjay Patel