This unfortunately requires a bunch of bitcasts to be added added to SUBREG_TO_REG, COPY_TO_REGCLASS, and instructions in output patterns. Otherwise tablegen seems to default to picking f128 and then we fail when something tries to get the register class for f128 which isn't always valid.
The test changes are because we were previously mixing fr128 and vr128 due to contrainRegClass finding FR128 first and passes like live range shrinking weren't handling that well.
llvm-svn: 337147
canWidenShuffleElements can do a better job if given a mask with ZeroableElements info. Apparently, ZeroableElements was being only used to identify AllZero candidates, but possibly we could plug it into more shuffle matchers.
Original Patch by Zvi Rackover @zvi
Differential Revision: https://reviews.llvm.org/D42044
llvm-svn: 336903
Noticed while updating D42044, lowerV2X128VectorShuffle can improve the shuffle mask with the zeroable data to create a target shuffle mask to recognise more 'zero upper 128' patterns.
NOTE: lowerV4X128VectorShuffle could benefit as well but the code needs refactoring first to discriminate between SM_SentinelUndef and SM_SentinelZero for negative shuffle indices.
Differential Revision: https://reviews.llvm.org/D49092
llvm-svn: 336900
We now use llvm.fma.f32/f64 or llvm.x86.fmadd.f32/f64 intrinsics that use scalar types rather than vector types. So we don't these special ISD nodes that operate on the lowest element of a vector.
llvm-svn: 336883
This converts them to what clang is now using for codegen. Unfortunately, there seem to be a few kinks to work out still. I'll try to address with follow up patches.
llvm-svn: 336871
Summary:
These changes cover the PR#31399.
Now the ffs(x) function is lowered to (x != 0) ? llvm.cttz(x) + 1 : 0
and it corresponds to the following llvm code:
%cnt = tail call i32 @llvm.cttz.i32(i32 %v, i1 true)
%tobool = icmp eq i32 %v, 0
%.op = add nuw nsw i32 %cnt, 1
%add = select i1 %tobool, i32 0, i32 %.op
and x86 asm code:
bsfl %edi, %ecx
addl $1, %ecx
testl %edi, %edi
movl $0, %eax
cmovnel %ecx, %eax
In this case the 'test' instruction can't be eliminated because
the 'add' instruction modifies the EFLAGS, namely, ZF flag
that is set by the 'bsf' instruction when 'x' is zero.
We now produce the following code:
bsfl %edi, %ecx
movl $-1, %eax
cmovnel %ecx, %eax
addl $1, %eax
Patch by Ivan Kulagin
Reviewers: davide, craig.topper, spatel, RKSimon
Reviewed By: craig.topper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D48765
llvm-svn: 336768
These ISD nodes try to select the MOVLPS and MOVLPD instructions which are special load only instructions. They load data and merge it into the lower 64-bits of an XMM register. They are logically equivalent to our MOVSD node plus a load.
There was only one place in X86ISelLowering that used MOVLPD and no places that selected MOVLPS. The one place that selected MOVLPD had to choose between it and MOVSD based on whether there was a load. But lowering is too early to tell if the load can really be folded. So in isel we have patterns that use MOVSD for MOVLPD if we can't find a load.
We also had patterns that select the MOVLPD instruction for a MOVSD if we can find a load, but didn't choose the MOVLPD ISD opcode for some reason.
So it seems better to just standardize on MOVSD ISD opcode and manage MOVSD vs MOVLPD instruction with isel patterns.
llvm-svn: 336728
Now that rL336250 has landed, we should prefer 2 immediate shifts + a shuffle blend over performing a multiply. Despite the increase in instructions, this is quicker (especially for slow v4i32 multiplies), avoid loads and constant pool usage. It does mean however that we increase register pressure. The code size will go up a little but by less than what we save on the constant pool data.
This patch also adds support for v16i16 to the BLEND(SHIFT(v,c1),SHIFT(v,c2)) combine, and also prevents blending on pre-SSE41 shifts if it would introduce extra blend masks/constant pool usage.
Differential Revision: https://reviews.llvm.org/D48936
llvm-svn: 336642
Summary:
This adds a reverse transform for the instcombine canonicalizations
that were added in D47980, D47981.
As discussed later, that was worse at least for the code size,
and potentially for the performance, too.
https://rise4fun.com/Alive/Zmpl
Reviewers: craig.topper, RKSimon, spatel
Reviewed By: spatel
Subscribers: reames, llvm-commits
Differential Revision: https://reviews.llvm.org/D48768
llvm-svn: 336585
This replaces some asserts in lowerV2F64VectorShuffle with the similar asserts from lowerVIF64VectorShuffle which are more readable. The original asserts mentioned a blend, but there's no guarantee that it is a blend.
Also remove an if that the asserts prove is always true. Mask[0] is always less than 2 and Mask[1] is always at least 2. Therefore (Mask[0] >= 2) + (Mask[1] >= 2) == 1 must wlays be true.
llvm-svn: 336517
Pre-AVX512 (which can perform a quick extend/shift/truncate), extending to 2 v8i16 for the PMULLW and then truncating is more performant than relying on the generic PBLENDVB vXi8 shift path and uses a similar amount of mask constant pool data.
Differential Revision: https://reviews.llvm.org/D48963
llvm-svn: 336513
Splits off isKnownNeverZeroFloat to handle +/- 0 float cases.
This will make it easier to be more aggressive with the integer isKnownNeverZero tests (similar to ValueTracking), use computeKnownBits etc.
Differential Revision: https://reviews.llvm.org/D48969
llvm-svn: 336492
It's a bit neater to write T.isIntOrPtrTy() over `T.isIntegerTy() ||
T.isPointerTy()`.
I used Python's re.sub with this regex to update users:
r'([\w.\->()]+)isIntegerTy\(\)\s*\|\|\s*\1isPointerTy\(\)'
llvm-svn: 336462
The intrinsics can be implemented with a f32/f64 llvm.fma intrinsic and an insert into a zero vector.
There are a couple regressions here due to SelectionDAG not being able to pull an fneg through an extract_vector_elt. I'm not super worried about this though as InstCombine should be able to do it before we get to SelectionDAG.
llvm-svn: 336416
This upgrades all of the intrinsics to use fneg instructions to convert fma into fmsub/fnmsub/fnmadd/fmsubadd. And uses a select instruction for masking.
This matches how clang uses the intrinsics these days.
llvm-svn: 336409
Previously we could only negate the FMADD opcodes. This used to be mostly ok when we lowered FMA intrinsics during lowering. But with the move to llvm.fma from target specific intrinsics, we can combine (fneg (fma)) to (fmsub) earlier. So if we start with (fneg (fma (fneg))) we would get stuck at (fmsub (fneg)).
This patch fixes that so we can also combine things like (fmsub (fneg)).
llvm-svn: 336304
We were only doing this for basic blends, despite shuffle lowering now being good enough to handle more complex blends. This means that the two v8i16 splat shifts are performed in parallel instead of serially as the general shift case.
Reapplied with a fixed (extra null tests) version of rL336113 after reversion in rL336189 - extra test case added at rL336247.
llvm-svn: 336250
We were only doing this for basic blends, despite shuffle lowering now being good enough to handle more complex blends. This means that the two v8i16 splat shifts are performed in parallel instead of serially as the general shift case.
llvm-svn: 336113
I believe all of these are constants so legalizing them should be pretty trivial, but this saves a step.
In one case it looks like we may have been creating a shift amount larger than the shift input itself.
llvm-svn: 336052
This combine runs pretty late and causes us to introduce a shift after the op legalization phase has run. We need to be sure we create the shift with the proper type for the shift amount. If we don't do this, we will still re-legalize the operation properly, but we won't get a chance to fully optimize the truncate that gets inserted.
So this patch adds the necessary truncate when the shift is created. I've also narrowed the subtract that gets created to always be an i32 type. The truncate would have trigered SimplifyDemandedBits to optimize it anyway. But using a more appropriate VT here is free and saves an optimization step.
llvm-svn: 336051
The important part is the creation of the SHLD/SHRD nodes. The compare and the conditional move can use target independent nodes that can be legalized on their own. This gives some opportunities to trigger the optimizations present in the lowering for those things. And its just better to limit the number of places we emit target specific nodes.
The changed test cases still aren't optimal.
Differential Revision: https://reviews.llvm.org/D48619
llvm-svn: 335998
This uses the same technique as for shifts - split the rotation into 4/2/1-bit partial rotations and select those partials based on the amount bit, making use of PBLENDVB if available. This halves the use of PBLENDVB compared to expanding to shifts, which can be a slow op.
Unfortunately I haven't found a decent way to share much of this code with the shift equivalent.
Differential Revision: https://reviews.llvm.org/D48655
llvm-svn: 335957
These opcodes have a fixed type of i8 for their immediate and shouldn't have anything to do with the scalar shift amount used by target independent shift nodes.
llvm-svn: 335578
This recommits r335562 and 335563 as a single commit.
The frontend will surround the intrinsic with the appropriate marshalling to/from a scalar type to match the sigature of the builtin that software expects.
By exposing the vXi1 type directly in the llvm intrinsic we make it available to optimizers much earlier. This can enable the scalar marshalling code to be optimized away.
llvm-svn: 335568
They appear to be untested other than the test case for p37879.ll and I believe we should be using SimplifyDemandedElts here to handle these cases.
llvm-svn: 335436
Changing the logic of scalar mask folding to check for valid input types rather
than against invalid ones, making it more robust and fixing PR37879.
Differential Revision: https://reviews.llvm.org/D48366
llvm-svn: 335323
This should help in lowering the following four intrinsics:
_mm256_cvtepi32_epi8
_mm256_cvtepi64_epi16
_mm256_cvtepi64_epi8
_mm512_cvtepi64_epi8
Differential Revision: https://reviews.llvm.org/D46957
llvm-svn: 335238
I don't believe there is any real reason to have separate X86 specific opcodes for vector compares. Setcc has the same behavior just uses a different encoding for the condition code.
I had to change the CondCodeAction for SETLT and SETLE to prevent some transforms from changing SETGT lowering.
Differential Revision: https://reviews.llvm.org/D43608
llvm-svn: 335173
This patch handles back-end folding of generic patterns created by lowering the
X86 rounding intrinsics to native IR in cases where the instruction isn't a
straightforward packed values rounding operation, but a masked operation or a
scalar operation.
Differential Revision: https://reviews.llvm.org/D45203
llvm-svn: 335037
isVectorClearMaskLegal() is the TLI hook used by the generic
DAGCombiner::XformToShuffleWithZero().
We've grown to accomodate/expect this transform to shuffle
(disabling it more generally results in many regressions).
So I'm narrowly excluding the 256-bit types that clearly
are not worthwhile for AVX1.
I think in most cases we are able to recover by converting
the shuffle back into 'and' ops, but the cases in:
https://bugs.llvm.org/show_bug.cgi?id=37749
...show that there are cracks.
llvm-svn: 334759
This shortcoming was noted in D47330, and the test diffs show we already
had other examples where we failed to fold to a SHRUNKBLEND:
/// Dynamic (non-constant condition) vector blend where only the sign bits
/// of the condition elements are used. This is used to enforce that the
/// condition mask is not valid for generic VSELECT optimizations.
This patch implements an idea from D48043 and would obsolete that patch
because it catches more cases (notable the AVX1 case that was missed there).
All we're doing is allowing the existing transform to fire more often by
removing the post-legalize constraint. All of the relevant feature checks
and other predicates are left as-is.
Differential Revision: https://reviews.llvm.org/D48078
llvm-svn: 334592
Extension to D46954 (PR37426), this patch adds support for v8i16/v16i16 rotations in a similar manner - the conversion of the shift/rotate amount to a multiplication factor and the use of PMULLW to shift left and PMULHUW (ISD::MULHU) to shift the wrapped bits back around to be ORd together.
Differential Revision: https://reviews.llvm.org/D47822
llvm-svn: 334309
Simplify combineVectorTruncationWithPACKUS to mask the upper bits followed by calling truncateVectorWithPACK instead of duplicating with similar code.
This results in the codegen using (V)PACKUSDW on SSE41+ targets for vXi64/vXi32 inputs where before it always used PACKUSWB (along with a lot more bitcasting).
I've raised PR37749 as until we avoid unnecessary concats back to 256-bit for bitwise ops, we can't avoid splitting the input value into 128-bit subvectors for masking.
llvm-svn: 334289
We have some combines/lowerings that attempt to use PACKSS-then-PACKUS and others that use PACKUS-then-PACKSS.
PACKUS is much easier to combine with if we know the upper bits are zero as ComputeKnownBits can easily see through BITCASTs etc. especially now that rL333995 and rL334007 have landed. It also effectively works at byte level which further simplifies shuffle combines.
The only (minor) annoyances are that ComputeKnownBits can sometimes take longer as it doesn't fail as quickly as ComputeNumSignBits (but I'm not seeing any actual regressions in tests) and PACKUSDW only became available after SSE41 so we have more codegen diffs between targets.
llvm-svn: 334276
Simplify combineVectorTruncationWithPACKSS to just a SIGN_EXTEND_INREG followed by using the existing truncateVectorWithPACK instead of duplicating code.
llvm-svn: 334193
Only the bottom 16-bits of BEXTR's control op are required (0:8 INDEX, 15:8 LENGTH).
Differential Revision: https://reviews.llvm.org/D47690
llvm-svn: 334083
Similar to v4i32 SHL, convert v8i16 shift amounts to scale factors instead to improve performance and reduce instruction count. We were already doing this for constant shifts, this adds variable shift support.
Reduces the serial nature of the codegen, which relies on chains of plendvb/pand+pandn+por shifts.
This is a step towards adding support for vXi16 vector rotates.
Differential Revision: https://reviews.llvm.org/D47546
llvm-svn: 334023
Ideally we'd use resolveTargetShuffleInputs to handle faux shuffles as well but:
(a) that code path doesn't handle general/pre-legalized ops/types very well.
(b) I'm concerned about the compute time as they recurse to calls to computeKnownBits/ComputeNumSignBits which would need depth limiting somehow.
llvm-svn: 334007
This is the new version of D46181, allowing setjmp/longjmp
to work correctly with the Intel CET shadow stack by storing
SSP on setjmp and fixing it on longjmp. The patch has been
updated to use the cf-protection-return module flag instead
of HasSHSTK, and the bug that caused D46181 to be reverted
has been fixed with the test expanded to track that fix.
patch by mike.dvoretsky
Differential Revision: https://reviews.llvm.org/D47311
llvm-svn: 333990
Previously we just returned undef, but really we should be returning the pass thru input. We also need to make sure we preserve the chain output that the original intrinsic node had to maintain connectivity in the DAG. So we should just return the incoming chain as the output chain.
llvm-svn: 333804
Noticed while fixing PR37426, for splat rotations (rotation by an uniform value) its better to just expand back to shift ops than performing as a general non-uniform rotation.
llvm-svn: 333661
This improves splat rotations (rotation by an uniform value), to avoid having to use the generic non-uniform shift code (extension to PR37426).
llvm-svn: 333641
Created the IsSplatValue helper from the splat detection code in LowerScalarVariableShift as a first NFC step towards improving support for splat rotations, which is an extension of PR37426.
llvm-svn: 333580
Support for Clang lowering of fused intrinsics. This patch:
1. Removes bindings to clang fma intrinsics.
2. Introduces new LLVM unmasked intrinsics with rounding mode:
int_x86_avx512_vfmadd_pd_512
int_x86_avx512_vfmadd_ps_512
int_x86_avx512_vfmaddsub_pd_512
int_x86_avx512_vfmaddsub_ps_512
supported with a new intrinsic type (INTR_TYPE_3OP_RM).
3. Introduces new x86 fmaddsub/fmsubadd folding.
4. Introduces new tests for code emitted by sequentions introduced in Clang part.
Patch by tkrupa
Reviewers: craig.topper, sroland, spatel, RKSimon
Reviewed By: craig.topper, RKSimon
Differential Revision: https://reviews.llvm.org/D47443
llvm-svn: 333554
There seems to be no real reason to have these separate copies.
The existing implementations just copy each other for x86.
For Mips there is a subtle difference, which is just a bug
since it changes based on the context where which one was called.
Dropping this version, all tests pass. If I try to merge them
to match the removed version, a test fails.
llvm-svn: 333440
1. Introduction of mask scalar TableGen patterns.
2. Introduction of new scalar move TableGen patterns
and refactoring of existing ones.
3. Folding of pattern created by introducing scalar
masking in Clang header files.
Patch by tkrupa
Differential Revision: https://reviews.llvm.org/D47012
llvm-svn: 333419
Summary: We already get this right if the i64 didn't come from a load.
Reviewers: RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D47439
llvm-svn: 333393
We have unmasked intrinsics now and wrap them with a select. This is a net reduction of 36 intrinsics from before the unmasked intrinsics were added.
llvm-svn: 333388
These do the same thing with the first and second sources swapped. They previously came from separate intrinsics that specified different masking behavior. But we can cover that with isel patterns and a single node.
This is a step towards reducing the number of intrinsics needed.
A bunch of tests change because we are now biased to choosing VPERMT over VPERMI when there is nothing to signal that commuting is beneficial.
llvm-svn: 333383
This basically reverts r280696 in favor of using extra patterns as mentioned as an alternative in that commit message. For now I've only added the cases we have test cases for, but it should be easy to add more in the future.
This will help to convert VPERMI2PS/VPERMT2PS intrinsics to use a single ISD node opcode. And hopefully allow some intrinsics to be removed.
llvm-svn: 333365
Summary:
This **appears** to be the last missing piece for the masked merge pattern handling in the backend.
This is [[ https://bugs.llvm.org/show_bug.cgi?id=37104 | PR37104 ]].
[[ https://bugs.llvm.org/show_bug.cgi?id=6773 | PR6773 ]] will introduce an IR canonicalization that is likely bad for the end assembly.
Previously, `andps`+`andnps` / `bsl` would be generated. (see `@out`)
Now, they would no longer be generated (see `@in`), and we need to make sure that they are generated.
Differential Revision: https://reviews.llvm.org/D46528
llvm-svn: 332904
This removes 6 intrinsics since we no longer need separate mask and maskz intrinsics.
Differential Revision: https://reviews.llvm.org/D47124
llvm-svn: 332890
As suggested by Fabian on PR37426, we can use PMULUDQ to perform v4i32 vector rotations as the upper 32bits of the multiply will contain the 'wrapped' bits of the rotation.
v8i16/v16i8 rotations would be straightforward to add to lowerRotate in the future - ideally we'd mostly share code with the vector shifts lowering.
Differential Revision: https://reviews.llvm.org/D46954
llvm-svn: 332832
The intrinsic legalization for masked truncate uses ISD::TRUNCATE which can be constant folded by getNode. This prevents getVectorMaskingNode from seeing the ISD::TRUNCATE special case where it should emit X86ISD::SELECT instead of ISD::VSELECT. This causes a vselect with a v16i1 or v8i1 condition to be emitted during vector legalization. but vector legalization doesn't revisit nodes it creates. DAG combine will then promote this condition to match the result type. Then op legalization will try to legalize it, but the custom lowering hook returned SDValue(). But op legalization doesn't have an Expand for VSELECT because it expects vector legalization to have taken care of it. So the operation sticks around and fails in isel.
This patch adds a custom legalization hook to morph it to a vXi8 vselect instead.
This also simplifies the normal vXi16 vselect handling because vector legalization was normally expanding to AND/ANDN/OR and DAG combine was turning that into VBLENDVB. So we can skip a step by doing it directly.
Fixes PR37499
Differential Revision: https://reviews.llvm.org/D47025
llvm-svn: 332743
As suggested by Fabian on PR37441, use PSHUFLW to extend shift amount types for use with PSRAD/PSRLD to reduce register pressure.
Some of this ideally would be done by combineTargetShuffle but its tricky to do as most of the shuffles are sharing inputs.
Differential Revision: https://reviews.llvm.org/D46959
llvm-svn: 332524
As i64 types are not legal on 32-bit targets, insert these into a suitable zero vector and use the packed vXi64<->FP conversion instructions instead.
Fixes PR3163.
Differential Revision: https://reviews.llvm.org/D43441
llvm-svn: 332498
Summary:
New unsigned saturation downconvert patterns detection was implemented in
X86 Codegen:
(truncate (smin (smax (x, C1), C2)) to dest_type),
where C1 >= 0 and C2 is unsigned max of destination type.
(truncate (smax (smin (x, C2), C1)) to dest_type)
where C1 >= 0, C2 is unsigned max of destination type and C1 <= C2.
These two patterns are equivalent to:
(truncate (umin (smax(x, C1), unsigned_max_of_dest_type)) to dest_type)
Reviewers: RKSimon
Subscribers: llvm-commits, a.elovikov
Differential Revision: https://reviews.llvm.org/D45315
llvm-svn: 332336
With nnan, there's no need for the masked merge / blend
sequence (that probably costs much more than the min/max
instruction).
Somewhere between clang 5.0 and 6.0, we started producing
these intrinsics for fmax()/fmin() in C source instead of
libcalls or fcmp/select. The backend wasn't prepared for
that, so we regressed perf in those cases.
Note: it's possible that other targets have similar problems
as seen here.
Noticed while investigating PR37403 and related bugs:
https://bugs.llvm.org/show_bug.cgi?id=37403
The IR FMF propagation cases still don't work. There's
a proposal that might fix those cases in D46563.
llvm-svn: 331992
Because we create a new kind of debug instruction, DBG_LABEL, we need to
check all passes which use isDebugValue() to check MachineInstr is debug
instruction or not. When expelling debug instructions, we should expel
both DBG_VALUE and DBG_LABEL. So, I create a new function,
isDebugInstr(), in MachineInstr to check whether the MachineInstr is
debug instruction or not.
This patch has no new test case. I have run regression test and there is
no difference in regression test.
Differential Revision: https://reviews.llvm.org/D45342
Patch by Hsiangkai Wang.
llvm-svn: 331844
This is a fix for PR30290: by marking all byval stack slots as being aliased,
the instruction scheduler is more conservative about rescheduling memory
accesses to such stack slots as an LLVM Value* might alias it. This fixes
errors such as in the patched test case, where reads and writes to a data
structure are illegally mixed.
This could be fixed better in the future with better analysis for the
instruction scheduler to know what Values alias what stack slots.
Differential Revision: https://reviews.llvm.org/D45022
llvm-svn: 331749
This patch adds a shadow stack fix when compiling
setjmp/longjmp with the shadow stack enabled. This
allows setjmp/longjmp to work correctly with CET.
Patch by mike.dvoretsky
Differential Revision: https://reviews.llvm.org/D46181
llvm-svn: 331748
Summary:
Split off from D46031.
In masked merge case, this degrades IPC by decreasing instruction count.
{F6108777}
The next patch should be able to recover and improve this.
This also affects the transform @spatel have added in D27489 / rL289738,
and the test coverage for X86 was missing.
But after i have added it, and looked at the changes in MCA, i'm somewhat confused.
{F6093591} {F6093592} {F6093593}
I'd say this regression is an improvement, since `IPC` increased in that case?
Reviewers: spatel, craig.topper
Reviewed By: spatel
Subscribers: andreadb, llvm-commits, spatel
Differential Revision: https://reviews.llvm.org/D46493
llvm-svn: 331684
Summary:
The legacy VRCPPS/VRSQRTPS instructions aren't available in 512-bit versions. The new increased precision versions are. So we can use those to implement v16f32 reciprocal estimates.
For KNL CPUs we can probably use VRCP28PS/VRSQRT28PS and avoid the NR step altogether, but I leave that for a future patch.
Reviewers: spatel
Reviewed By: spatel
Subscribers: RKSimon, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D46498
llvm-svn: 331606
We've been running doxygen with the autobrief option for a couple of
years now. This makes the \brief markers into our comments
redundant. Since they are a visual distraction and we don't want to
encourage more \brief markers in new code either, this patch removes
them all.
Patch produced by
for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done
Differential Revision: https://reviews.llvm.org/D46290
llvm-svn: 331272
Summary:
Previously the flag intrinsics always used the index instructions even if a mask instruction also exists.
To fix fix this I've created a single ISD node type that returns index, mask, and flags. The SelectionDAG CSE process will merge all flavors of intrinsics with the same inputs to a s ingle node. Then during isel we just have to look at which results are used to know what instruction to generate. If both mask and index are used we'll need to emit two instructions. But for all other cases we can emit a single instruction.
Since I had to do manual isel anyway, I've removed the pseudo instructions and custom inserter code that was working around tablegen limitations with multiple implicit defs.
I've also renamed the recently added sse42.ll test case to sttni.ll since it focuses on that subset of the sse4.2 instructions.
Reviewers: chandlerc, RKSimon, spatel
Reviewed By: chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D46202
llvm-svn: 331091
Previously we only formed MUL_IMM when we split a constant. This blocked load folding on those cases. We should also form MUL_IMM for 3/5/9 to favor LEA over load folding.
Differential Revision: https://reviews.llvm.org/D46040
llvm-svn: 330850
Summary:
If attribute "use-soft-float"="true" is set then X86ISelLowering.cpp sets
'Promote' action for ISD::SINT_TO_FP operation on type i32.
But 'Promote' action is not proper in this case since lib function
__floatsidf is available for casting from signed int to float type.
Thus Expand action is more suitable here.
The Expand action should be set for ISD::UINT_TO_FP for soft float as well.
If function attribute "use-soft-float"="true" is set then infinite looping
can happen in DAG combining, function visitSINT_TO_FP() replaces SINT_TO_FP
node with UINT_TO_FP node and function combineUIntToFP() replace vice versa in cycle.
The fix prevents it.
Patch by vrybalov
Differential Revision: https://reviews.llvm.org/D45572
llvm-svn: 330711
Three new instructions:
umonitor - Sets up a linear address range to be
monitored by hardware and activates the monitor.
The address range should be a writeback memory
caching type.
umwait - A hint that allows the processor to
stop instruction execution and enter an
implementation-dependent optimized state
until occurrence of a class of events.
tpause - Directs the processor to enter an
implementation-dependent optimized state
until the TSC reaches the value in EDX:EAX.
Also modifying the description of the mfence
instruction, as the rep prefix (0xF3) was allowed
before, which would conflict with umonitor during
disassembly.
Before:
$ echo 0xf3,0x0f,0xae,0xf0 | llvm-mc -disassemble
.text
mfence
After:
$ echo 0xf3,0x0f,0xae,0xf0 | llvm-mc -disassemble
.text
umonitor %rax
Reviewers: craig.topper, zvi
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D45253
llvm-svn: 330462
This is the patch that lowers x86 intrinsics to native IR
in order to enable optimizations. The patch also includes folding
of previously missing saturation patterns so that IR emits the same
machine instructions as the intrinsics.
Patch by tkrupa
Differential Revision: https://reviews.llvm.org/D44785
llvm-svn: 330322
Summary:
Add an LLVM intrinsic for type discriminated event logging with XRay.
Similar to the existing intrinsic for custom events, but also accepts
a type tag argument to allow plugins to be aware of different types
and semantically interpret logged events they know about without
choking on those they don't.
Relies on a symbol defined in compiler-rt patch D43668. I may wait
to submit before I can see demo everything working together including
a still to come clang patch.
Reviewers: dberris, pelikan, eizan, rSerge, timshen
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D45633
llvm-svn: 330219
This completes the work started in r329604 and r329605 when we changed clang to no longer use the intrinsics.
We lost some InstCombine SimplifyDemandedBit optimizations through this change as we aren't able to fold 'and', bitcast, shuffle very well.
llvm-svn: 329990
This cleans up a number of operations that only claimed te use EFLAGS
due to using DF. But no instructions which we think of us setting EFLAGS
actually modify DF (other than things like popf) and so this needlessly
creates uses of EFLAGS that aren't really there.
In fact, DF is so restrictive it is pretty easy to model. Only STD, CLD,
and the whole-flags writes (WRFLAGS and POPF) need to model this.
I've also somewhat cleaned up some of the flag management instruction
definitions to be in the correct .td file.
Adding this extra register also uncovered a failure to use the correct
datatype to hold X86 registers, and I've corrected that as necessary
here.
Differential Revision: https://reviews.llvm.org/D45154
llvm-svn: 329673
The key idea is to lower COPY nodes populating EFLAGS by scanning the
uses of EFLAGS and introducing dedicated code to preserve the necessary
state in a GPR. In the vast majority of cases, these uses are cmovCC and
jCC instructions. For such cases, we can very easily save and restore
the necessary information by simply inserting a setCC into a GPR where
the original flags are live, and then testing that GPR directly to feed
the cmov or conditional branch.
However, things are a bit more tricky if arithmetic is using the flags.
This patch handles the vast majority of cases that seem to come up in
practice: adc, adcx, adox, rcl, and rcr; all without taking advantage of
partially preserved EFLAGS as LLVM doesn't currently model that at all.
There are a large number of operations that techinaclly observe EFLAGS
currently but shouldn't in this case -- they typically are using DF.
Currently, they will not be handled by this approach. However, I have
never seen this issue come up in practice. It is already pretty rare to
have these patterns come up in practical code with LLVM. I had to resort
to writing MIR tests to cover most of the logic in this pass already.
I suspect even with its current amount of coverage of arithmetic users
of EFLAGS it will be a significant improvement over the current use of
pushf/popf. It will also produce substantially faster code in most of
the common patterns.
This patch also removes all of the old lowering for EFLAGS copies, and
the hack that forced us to use a frame pointer when EFLAGS copies were
found anywhere in a function so that the dynamic stack adjustment wasn't
a problem. None of this is needed as we now lower all of these copies
directly in MI and without require stack adjustments.
Lots of thanks to Reid who came up with several aspects of this
approach, and Craig who helped me work out a couple of things tripping
me up while working on this.
Differential Revision: https://reviews.llvm.org/D45146
llvm-svn: 329657
LowerIntUnary as its name says has an assert for integer types. But for the bitcast case one side might be an FP type.
Rather than making sure the function really works for fp types and renaming it. Just do really basic splitting directly. The LowerIntUnary has the advantage that it can peek through BUILD_VECTOR because every other call is during Lowering. But these calls are during legalization and will be followed by a DAG combine round.
Revert some change to LowerVectorIntUnary that were originally made just to make these two calls work even in pure integer cases.
This was found purely by compiling the avx512f-builtins.c test from clang so I've copied over the offending function from that.
llvm-svn: 329616
Summary:
r327219 added wrappers to std::sort which randomly shuffle the container before sorting.
This will help in uncovering non-determinism caused due to undefined sorting
order of objects having the same key.
To make use of that infrastructure we need to invoke llvm::sort instead of std::sort.
Note: This patch is one of a series of patches to replace *all* std::sort to llvm::sort.
Refer the comments section in D44363 for a list of all the required patches.
Reviewers: chandlerc, craig.topper, RKSimon
Reviewed By: chandlerc, craig.topper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44874
llvm-svn: 329534
Previously we used a custom lowering for this because of the AVX1 splitting requirement. But we can do the split during DAG combine if we check the types and subtarget
llvm-svn: 329510
This Promote flag was alwasys set to true except in the default case. But in the default case we don't need to set PVT and can just return false.
llvm-svn: 328926
We are re-adding all the bitcasts, constant masks and target shuffles to the work list for no apparent gain.
Found while investigating adding SimplifyDemandedVectorElts to target shuffles.
Differential Revision: https://reviews.llvm.org/D44942
llvm-svn: 328771
Summary:
Re-lands r328386 and r328443, reverting r328482.
Incorporates fixes from @mstorsjo in D44876 (thanks!) so that small
parameters in i8 and i16 do not end up in the SysV register parameters
(EDI, ESI, etc).
I added tests for how we receive small parameters, since that is the
important part. It's always safe to store more bytes than will be read,
but the assumptions you make when loading them are what really matter.
I also tested this by self-hosting clang and it passed tests on win64.
Reviewers: mstorsjo, hans
Subscribers: hiraditya, mstorsjo, llvm-commits
Differential Revision: https://reviews.llvm.org/D44900
llvm-svn: 328570
This broke Chromium (see crbug.com/825748). It looks like mstorsjo's follow-up
patch at D44876 fixes this, but let's revert back to green for now until that's
ready to land.
(Also reverts r328443.)
> Both GCC and MSVC only look at the low byte of a boolean when it is
> passed.
llvm-svn: 328482
These nodes only use the lower 32 bits of their inputs so we can use SimplifyDemandedBits to simplify them.
Differential Revision: https://reviews.llvm.org/D44375
llvm-svn: 328405
Mingw uses the same stack protector functions as GCC provides
on other platforms as well.
Patch by Valentin Churavy!
Differential Revision: https://reviews.llvm.org/D27296
llvm-svn: 328039
We don't need to create an ISD::TRUNCATE node to return, we started with one and can return it. Also remove the call to getExtendInVec, the result is just going to be a getNode of that value passed in.
llvm-svn: 327914
With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that.
llvm-svn: 327869
X86 Supports Indirect Branch Tracking (IBT) as part of Control-Flow Enforcement Technology (CET).
IBT instruments ENDBR instructions used to specify valid targets of indirect call / jmp.
The `nocf_check` attribute has two roles in the context of X86 IBT technology:
1. Appertains to a function - do not add ENDBR instruction at the beginning of the function.
2. Appertains to a function pointer - do not track the target function of this pointer by adding nocf_check prefix to the indirect-call instruction.
This patch implements `nocf_check` context for Indirect Branch Tracking.
It also auto generates `nocf_check` prefixes before indirect branchs to jump tables that are guarded by range checks.
Differential Revision: https://reviews.llvm.org/D41879
llvm-svn: 327767
Previously, we called the same functions twice with a bool flag determining whether we should look for ADDSUB or SUBADD. It would be more efficient to run the code once and detect either pattern with a flag to tell which type it found.
Differential Revision: https://reviews.llvm.org/D44540
llvm-svn: 327730
Previously if getSetccResultType returned an illegal type we just fell back to using the default promoted type. This appears to have been to handle the case where for vectors getSetccResultType returns the input type, but the input type itself isn't legal and will need to be promoted. Without the legality check we would never reach a legal type.
But just picking the promoted type to be the setcc type can create strange setccs where the result type is 128 bits and the operand type is 256 bits. If for example the result type was promoted to v8i16 from v8i1, but the input type was promoted from v8i23 to v8i32. We currently handle this with custom lowering code in X86.
This legality check also caused us reject the getSetccResultType when the input type needed to be widened or split. Even though that result wouldn't have caused legalization to get stuck.
This patch tries to fix this by detecting the getSetccResultType needs to be promoted. If its input type also needs to be promoted we'll try a ask for a new setcc result type based on its eventual promoted value. Otherwise we fall back to default type to promote to.
For any other illegal values we might get back from the initial call to getSetccResultType we just keep and allow it to be re-legalized later via splitting or widening or scalarizing.
llvm-svn: 327683
The FADD part of the addsub/subadd pattern can have its operands commuted, but when checking for fsubadd we were using the fadd as reference and commuting the fsub node.
llvm-svn: 327660
Rather than enumerating all specific types, for the DAG combine we can just use TLI::isTypeLegal and an SSE3 check. For the BUILD_VECTOR version we already know the type is legal so we just need to check SSE3.
llvm-svn: 327649
I had to modify the bswap recognition to allow unshrunk masks to make this work.
Fixes PR36689.
Differential Revision: https://reviews.llvm.org/D44442
llvm-svn: 327530
We now only create recursive concats if we have more than two non-zero values. This keeps our subvector broadcast DAG combine functioning.
llvm-svn: 327457
This better able to detect undef and zeros pieces in the concat. Or cases when only one subvector is non-zero. This allows us to avoid silly things like double inserts into progressively larger undefs.
This still builds 512 bit concats of 128 bits by building up through 256 bits first. But I don't know if that's best.
We probably want to merge this with the vXi1 concat code since they are very similar.
llvm-svn: 327454
Summary: Unless you were intentionally avoiding this syntax? I saw you mentioned makeArrayRef in your commit that added SplitOpsAndApply.
Reviewers: RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44403
llvm-svn: 327418
MVT belongs to the CodeGen layer, but ShuffleDecode is used by the X86 InstPrinter which is part of the MC layer. This only worked because MVT is completely implemented in a header file with no other library dependencies.
Differential Revision: https://reviews.llvm.org/D44353
llvm-svn: 327292
We called MaskedValueIsZero with two different masks, but underneath that calls computeKnownBits before applying the mask. This means we compute the same known bits twice due to the two calls. Instead just call computeKnownBits directly and apply the two masks ourselves.
llvm-svn: 327251
64-bit MMX vector generation usually ends up lowering into SSE instructions before being spilled/reloaded as a MMX type.
This patch creates a MMX vector from MMX source values, taking the lowest element from each source and constructing broadcasts/build_vectors with direct calls to the MMX PUNPCKL/PSHUFW intrinsics.
We're missing a few consecutive load combines that could be handled in a future patch if that would be useful - my main interest here is just avoiding a lot of the MMX/SSE crossover.
Differential Revision: https://reviews.llvm.org/D43618
llvm-svn: 327247
Same as the VPERMILPS/VPERMILPD approach for v8f32/v4f64 cases, rely on PSHUFB using bits[3:0] for indexing - we can ignore the sign bit (zero element) as those index vector values are considered undefined. The select between the lo/hi permute results based on the index size.
llvm-svn: 327242
As VPERMILPS/VPERMILPD only selects elements based on the bits[1:0]/bit[1] then we can permute both the (repeated) lo/hi 128-bit vectors in each case and then select between these results based on whether the index was for for lo/hi.
For v4i64/v4f64 this avoids some rather nasty v4i64 multiples on the AVX2 implementation, which seems to be worse than the extra port5 pressure from the additional shuffles/blends.
llvm-svn: 327239
Helper function to insert a subvector into the bottom elements of a larger zero/undef vector with the same scalar type.
I've converted a couple of INSERT_SUBVECTOR calls to use it, there are plenty more although in some cases I was worried it might make the code more ambiguous.
llvm-svn: 327236
Previously we unpacked the even bytes of each input into the high byte of 16-bit elements then did an v8i16 arithmetic shift right by 8 bits to fill the upper bits of each word with sign bits. Then we did the v8i16 multiply and then masked to zero the upper 8-bits of each result. The similar was done for all the odd bytes. The results are then packed together with packuswb
Since we are masking each multiply result element to 8-bits, and those 8-bits are determined only by the lower 8-bits of each of the inputs, we don't need to fill the upper bits with sign bits. So we can just unpack into the low byte of each element and treat the upper bits as garbage. This is what gcc also does.
Differential Revision: https://reviews.llvm.org/D44267
llvm-svn: 327093
This instruction can be thought of as reading either the even elements of a vXi32 input or the lower half of each element of a vXi64 input. We currently use the vXi32 interpretation, but vXi64 matches better with its broadcast behavior in EVEX.
I'm looking at moving MULDQ/MULUDQ creation to a DAG combine so we can do it when AVX512DQ is enabled without having to go through Custom lowering. But in some of the test cases we failed to use a broadcast load due to the size difference. This should help with that.
I'm also wondering if we can model these instructions in native IR and remove the intrinsics and I think using a vXi64 type will work better with that.
llvm-svn: 326991
The v8i32 conversion on AVX1 targets was only working after LowerMUL splits 256-bit vectors.
While I was there I've also made it so we don't have to check for AVX2 and BWI directly and instead just ask if the type is legal.
Differential Revision: https://reviews.llvm.org/D44190
llvm-svn: 326917
The code checks Level == AfterLegalizeDAG which is the fourth and last of the possible DAG combine stages that we have.
There is a Level called AfterLegalVectorOps, but that's the third DAG combine and it doesn't always run.
A function called isAfterLegalVectorOps should imply it returns true in either of the DAG combines that runs after the legalize vector ops stage, but that's not what this function does.
llvm-svn: 326832
Almost none of these usages were FP specific. And we had no clear guideliness on when to use hasAVX vs hasFP256.
I might also remove hasInt256 too since its an alias for hasAVX2.
llvm-svn: 326682
We were previously doing this with isel patterns. Moving it to op legalization gives us chance to see the required bitcast earlier. And it lets us remove some isel patterns.
llvm-svn: 326669
64-bit MMX constant generation usually ends up lowering into SSE instructions before being spilled/reloaded as a MMX type.
This patch bitcasts the constant to a double value to allow correct loading directly to the MMX register.
I've added MMX constant asm comment support to improve testing, it's better to always print the double values as hex constants as MMX is mainly an integer unit (and even with 3DNow! its just floats).
Differential Revision: https://reviews.llvm.org/D43616
llvm-svn: 326497
Emulated TLS is enabled by llc flag -emulated-tls,
which is passed by clang driver.
When llc is called explicitly or from other drivers like LTO,
missing -emulated-tls flag would generate wrong TLS code for targets
that supports only this mode.
Now use useEmulatedTLS() instead of Options.EmulatedTLS to decide whether
emulated TLS code should be generated.
Unit tests are modified to run with and without the -emulated-tls flag.
Differential Revision: https://reviews.llvm.org/D42999
llvm-svn: 326341
An extract_element where the result type is larger than the scalar element type is semantically an any_extend of from the scalar element type to the result type. If we expect zeroes in the upper bits of the i8/i32 we need to mae sure those zeroes are explicit in the DAG.
For these cases the best way to accomplish this is use an insert_subvector to pad zeroes to the upper bits of the v1i1 first. We extend to either v16i1(for i32) or v8i1(for i8). Then bitcast that to a scalar and finish with a zero_extend up to i32 if necessary. We can't extend past v16i1 because that's the largest mask size on KNL. But isel is smarter enough to know that a zext of a bitcast from v16i1 to i16 can use a KMOVW instruction. The insert_subvectors will be dropped during isel because we can determine that the producing instruction already zeroed the upper bits of the k-register.
llvm-svn: 326308
While the description for the instruction does mention OR, its talking about how the individual classification test results are ORed together.
The incoming mask is used as a zeroing write mask. If the bit is 1 the classification is written to the output. The bit is 0 the output is 0. This equivalent to an AND.
Here is pseudocode from the intrinsics guide
FOR j := 0 to 1
i := j*64
IF k1[j]
k[j] := CheckFPClass_FP64(a[i+63:i], imm8[7:0])
ELSE
k[j] := 0
FI
ENDFOR
k[MAX:2] := 0
llvm-svn: 326306
There's still some shortcoming in our ability to combine binops of constants with different sizes separated by an extend. I'll try to look at that next.
llvm-svn: 326128
Summary:
We have an early DAG combine to turn these patterns into MOVMSK, but that combine doesn't work if the vXi1 type has more elements than the widest legal vXi8 type. Type legalization will eventually split it down to v16i1 or v32i1 and then the bitcast gets legalized to a truncstore and a scalar load. The truncstore will get lowered to a series of extracts and bit math.
This patch adds a custom legalization to use a sign extend and MOVMSK instead. This prevents the eventual scalarization.
Reviewers: spatel, RKSimon, zvi
Reviewed By: RKSimon
Subscribers: mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D43593
llvm-svn: 326119
This code seemed to try to widen to 128, 256, or 512 bit vectors, but we only create X86ISD::AVG with a power of 2 number of elements. This means the only nodes that need to be legalized are less than 128-bits and need to be widened up to 128 bits.
llvm-svn: 326064
Which types are considered 'simple' is a function of the requirements of all targets that LLVM supports. That shouldn't directly affect what types we are able to handle. The remainder of this code checks that the number of elements is a power of 2 and takes care of splitting down to a legal size.
llvm-svn: 326063
Our UMIN/UMAX, vector truncation and shuffle combining is good enough to efficiently handle v8i64 with the number of leading zeros that are necessary for PSUBUS.
llvm-svn: 326034
Now that UMIN etc are Legal/Custom for SSE2+, we can efficiently match SUBUS v8i32 cases from SSSE3 which can perform efficient truncation with PSHUFB.
llvm-svn: 326033
These can be created by type legalization promoting the inputs to select to match scalar boolean contents.
We were trying to pattern match them away during isel, but its better to just remove them from the DAG.
I've cleaned up some patterns to not check for this 'and' anymore. But I suspect this has also opened up opportunities for pattern removal.
llvm-svn: 325949
The test changes you can see are related to the changes in ReplaceNodeResults. Though shuffle-vs-trunc-512.ll does have a test that exercises the code in LowerBITCAST. Looks like the test output didn't change because DAG combining is able to clean up the resulting type legalization. Adding the custom hook just makes type legalization work less hard.
Differential Revision: https://reviews.llvm.org/D43447
llvm-svn: 325933
We won't be able to fold the constant pool load, but its still better than materialing ones and xoring for the invert if we used PCMPEQ.
This will fix another regression from D42948.
llvm-svn: 325845
Previously this code overrode the flags and opcode used by the later code in LowerVSETCC. This makes the code difficult to read and follow.
This patch moves all the SUBUS code into its own function and makes it responsible for creating its own SDNodes on success.
Differential Revision: https://reviews.llvm.org/D43530
llvm-svn: 325827
SimplifyDemandedBits forces the demanded mask to all 1s if the node has multiple uses, unless the AssumeSingleUse flag is set.
So previously we were only really likely to simplify something if the condition had a single use. And on the off chance we did simplify with multiple uses the demanded mask being used was all ones so there was no reason to create a shrunkblend.
This patch now checks that the condition is only used by selects first, and then sets the AssumeSingleUse flag for the simplifcation. Then we convert the selects to shrunkblend, and finally replace condition.
Differential Revision: https://reviews.llvm.org/D43446
llvm-svn: 325604
This allows us to avoid an opsize prefix. And forcing some move immediates to i32 avoids a length changing prefix on those instructions.
This mostly replaces the existing combine we had for zext/sext+cmov of constants. I left in a case for sign extending a 32 bit cmov of constants to 64 bits.
Differential Revision: https://reviews.llvm.org/D43327
llvm-svn: 325601
Previously we used vptestmd, but the scheduling data for SKX says vpmovq2m/vpmovd2m is lower latency. We already used vpmovb2m/vpmovw2m for byte/word truncates. So this is more consistent anyway.
llvm-svn: 325534
We swapped the operands and used setle, but I don't see any reason to do that. I think this is a holdover from SSE where we swap and the invert to use pcmpgt. But with AVX512 we don't want an invert so we won't use pcmpgt. So there's no need to swap.
llvm-svn: 325527
Canonicalize EQ/NE PCMPM to have build vector all zeros on the RHS so we don't have to pattern match it in both locations. This significantly reduces the number of isel patterns needed since we also had to multiply it out with loads being in either operand of the 'and' input node and in the 'and' masking node.
This removes over 24000 bytes from the isel table.
llvm-svn: 325526
We're accidentally checking that the same node is a constant twice instead of checking the other node.
This isn't a functional problem since we didn't do anything below that explicitly requires constants. It just means we may have introduced a sign_extend or zero_extend that won't fold out.
llvm-svn: 325469
Summary:
Currently we convert to shuffles during lowering. This moves it to DAG combine so hopefully we can get it done before type legalization has to extend the condition.
I believe in some cases we're creating SHRUNKBLENDs that end up with constant conditions because we see the extended on the condition and think its a dynamic selelect before DAG combine gets a chance to constant fold the extend. We could add combines to turn SHRUNKBLENDs with constant condition back to vselect. But it seemed like it might be better to just send them to shuffles as early as possible so they never get a chance to become SHRUNKBLENDs. This the reason some tests went from blends controlled by a constant pool load to just move.
Some of the constant pool entries changed because the sign_extend introduced by type legalization turned undef elements in select condition into 0s. While the select->shuffle used -1 in the shuffle mask. So now the shuffle lowering can do what it wants with them.
I'll remove the lowering code as a follow up. We might be able to simplify some of the pre-checks for SHRUNKBLEND as the FIXME there says.
Reviewers: spatel, RKSimon, efriedma, zvi, andreadb
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D43367
llvm-svn: 325417
Undef in select condition means we should pick the element from one side or the other. An undef in a shuffle mask means pick any element from either source or worse.
I suspect by the time we get here most of the undefs in a constant vector have been removed by other things, but doing this for safety.
llvm-svn: 325394
This seems to interfere with a target independent brcond combine that looks for the (srl (and X, C1), C2) pattern to enable TEST instructions. Once we flip, that combine doesn't fire and we end up exposing it to the X86 specific BT combine which causes us to emit a BT instruction. BT has lower throughput than TEST.
We could try to make the brcond combine aware of the alternate pattern, but since the flip was just a code size reduction and not likely to enable other combines, it seemed easier to just delay it until after lowering.
Differential Revision: https://reviews.llvm.org/D43201
llvm-svn: 325371
We already do this for 64-bit when it won't fit into a 64-bit AND/TEST's immediate field. This adds an additional qualifier to do it for any single bit constant larger than 8-bits under optsize
Differential Revision: https://reviews.llvm.org/D43346
llvm-svn: 325290
We can use PACKSS to saturate each stage of the chain: PACKSSDW down to [-32768,32767] and then PACKSSWB to [-128,127].
PACKUS is a little trickier and will be handled in a separate patch.
llvm-svn: 325235
Try to keep PACK*SDW/PACK*SWB as wide as possible, this helps ComputeNumSignBits as it can only peek through bitcasts to wider types, pre-AVX2 codegen was already doing this as it could peek through bitcasts/subvectors more easily than AVX2 could through shuffles.
This shouldn't affect existing results as calls to truncateVectorWithPACK ensure we have enough sign bits to pack to the same value, but it should make it possible to use truncateVectorWithPACK chains to perform saturation in combineTruncateWithSat with a future patch.
llvm-svn: 325149
While the AVX512 VTRUNCS/VTRUNCUS instructions require legal types, truncateVectorWithPACK handles cases with multiples of legal types through splitting/concatenation. So we just need to ensure that the src/dst scalar types are correct and leave truncateVectorWithPACK to handle the rest of it.
llvm-svn: 325127
Summary:
Instead of solving the hard problem of how to pass the callee to the indirect
jump thunk without a register, just use a CSR. At a call boundary, there's
nothing stopping us from using a CSR to hold the callee as long as we save and
restore it in the prologue.
Also, add tests for this mregparm=3 case. I wrote execution tests for
__llvm_retpoline_push, but they never got committed as lit tests, either
because I never rewrote them or because they got lost in merge conflicts.
Reviewers: chandlerc, dwmw2
Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D43214
llvm-svn: 325049
ISD::ADD implies individual vector element addition with no carries between elements. But for a vXi1 type that would be the same as XOR. And we already turn ISD::ADD into ISD::XOR for all vXi1 types during lowering. So the ISD::ADD pattern would never be able to match anyway.
KADD is different, it adds the elements but also propagates a carry between them. This just a way of doing an add in k-register without bitcasting to the scalar domain. There's still no way to match the pattern, but at least its not obviously wrong.
llvm-svn: 324861
Summary:
Currently we only use min/max to help with ule/uge compares because it removes an invert of the result that would otherwise be needed. But we can also use it for ult/ugt compares if it will prevent the need for a sign bit flip needed to use pcmpgt at the cost of requiring an invert after the compare.
I also refactored the code so that the max/min code is self contained and does its own return instead of setting up a flag to manipulate the rest of the function's behavior.
Most of the test cases look ok with this. I did notice that we added instructions when one of the operands being sign flipped is a constant vector that we were able to constant fold the flip into.
I also noticed that sometimes the SSE min/max clobbers a register that is needed after the compare. This resulted in an extra move being inserted before the min/max to preserve the register. We could try to detect this and switch from min to max and change the compare operands to use the operand that gets reused in the compare.
Reviewers: spatel, RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42935
llvm-svn: 324842
This allows us to recognise more saturation patterns and also simplify some MINMAX codegen that was failing to combine CMPGE comparisons to a legal CMPGT.
Differential Revision: https://reviews.llvm.org/D43014
llvm-svn: 324837
This patch adds a new function attribute "required-vector-width" that can be set by the frontend to indicate the maximum vector width present in the original source code. The idea is that this would be set based on ABI requirements, intrinsics or explicit vector types being used, maybe simd pragmas, etc. The backend will then use this information to determine if its save to make 512-bit vectors illegal when the preference is for 256-bit vectors.
For code that has no vectors in it originally and only get vectors through the loop and slp vectorizers this allows us to generate code largely similar to our AVX2 only output while still enabling AVX512 features like mask registers and gather/scatter. The loop vectorizer doesn't always obey TTI and will create oversized vectors with the expectation the backend will legalize it. In order to avoid changing the vectorizer and potentially harm our AVX2 codegen this patch tries to make the legalizer behavior similar.
This is restricted to CPUs that support AVX512F and AVX512VL so that we have good fallback options to use 128 and 256-bit vectors and still get masking.
I've qualified every place I could find in X86ISelLowering.cpp and added tests cases for many of them with 2 different values for the attribute to see the codegen differences.
We still need to do frontend work for the attribute and teach the inliner how to merge it, etc. But this gets the codegen layer ready for it.
Differential Revision: https://reviews.llvm.org/D42724
llvm-svn: 324834
We promote these via a DAG combine now before lowering gets the chance.
Also remove the v2i1 custom handling since it will no longer be triggered.
llvm-svn: 324833
These were added as part of the refactoring for prefer vector width. At the time I thought the hasAVX512 here would be replaced with "allow 512 bit vectors" so that it would read "allow 512 bit vectors OR VLX". But now the plan is to only give the option of disabling 512 bit vectors when VLX is enabled. So we don't need this qualification at all
llvm-svn: 324831
Summary:
This patch changes the signature of the avx512 packed fp compare intrinsics to return a vXi1 vector and no longer take a mask as input. The casts to scalar type will now need to be explicit in the IR. The masking node will now be an explicit and in the IR.
This makes the intrinsic look much more similar to an fcmp instruction that we wish we could use for these but can't. We already use icmp instructions for integer compares.
Previously the lowering step of isel would turn the intrinsic into an X86 specific ISD node and a emit the masking nodes as well as some bitcasts. This means DAG combines can't see the vXi1 type until somewhat late, making it more difficult to combine out gpr<->mask transition sequences. By exposing the vXi1 type explicitly in the IR and initial SelectionDAG we give earlier DAG combines and even InstCombine the chance to see it and optimize it.
This should make any issues with gpr<->mask sequences the same between integer and fp. Meaning we only have to fix them once.
Reviewers: spatel, delena, RKSimon, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D43137
llvm-svn: 324827
Undef VLX, getSetCCResultType returns v2i1/v4i1 for v2f32/v4f32 so default type legalization will end up changing the setcc result type back to vXi1 if it had been extended. The resulting extend gets messed up further by type legalization and is difficult to recombine back to (v4i32 (setcc (v4f32))) after legalization.
I went ahead and enabled this for SSE2 and later since its always the result we want and this helps type legalization get there in less steps.
llvm-svn: 324822
This prevents extends of masks being introduced during lowering where it become difficult to combine them out.
There are a few oddities in here.
We sometimes concatenate two k-registers produced by two compares, sign_extend the combined pair, then extract two halves. This worked better previously because the sign_extend wasn't created until after the fp_to_sint was split which led to a split sign_extend being created.
We probably also need to custom type legalize (v2i32 (sext v2i1)) via widening.
llvm-svn: 324820
This avoids a constant pool load to create 1.
The int->float are showing converts to mask and back. We probably need to widen inputs to sint_to_fp/uint_to_fp before type legalization.
llvm-svn: 324805
Previously we extracted two subvectors and concatenate. But the concatenate will be lowered to two insert subvectors. Then DAG combine will merge once of the inserts and one of the extracts back into the original vector. We might as well just directly use one extract and one insert.
llvm-svn: 324710
This regresses a couple cases in the shuffle combining test. But those cases use intrinsics that InstCombine knows how to turn into a generic shuffle earlier. This should give opportunities to fold this earlier in InstCombine or DAG combine.
llvm-svn: 324709
Most vxi1 constant build vectors have to be implemented in the scalar domain anyway so we'll probably end up with a cast there later. But by then its too late to do the combine to get rid of it.
llvm-svn: 324662
The KTEST instruction sets the C flag if the result of anding both operands together is all 1s. We can use this to lower (icmp eq/ne (bitcast (vXi1 X), -1)
Differential Revision: https://reviews.llvm.org/D42772
llvm-svn: 324577
Summary:
KTEST has weird flag behavior. The Z flag is set for all bits in the AND of the k-registers being 0, and the C flag is set for all bits being 1. All other flags are cleared.
We currently emit this instruction in EmitTEST and don't check the condition code. This can lead to strange things like using the S flag after a KTEST for a signed compare.
The domain reassignment pass can also transform TEST instructions into KTEST and is not protected against the flag usage either. For now I've disabled this part of the domain reassignment pass. I tried to comment out the checks in the mir test so that we could recover them later, but I couldn't figure out how to get that to work.
This patch moves the KTEST handling into LowerSETCC and now creates a ktest+x86setcc. I've chosen this approach because I'd like to add support for the C flag for all ones in a followup patch. To do that requires that I can rewrite the condition code going in the x86setcc to be different than the original SETCC condition code.
This fixes PR36182. I'll file a PR to fix domain reassignment once this goes in. Should this be merged to 6.0?
Reviewers: spatel, guyblank, RKSimon, zvi
Reviewed By: guyblank
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42770
llvm-svn: 324576
We were doing a lot of whitelisting of what we handle in these routines, but setOperationAction constrains what we can get here. So just add some asserts and prune the unreachable paths.
llvm-svn: 324538
that happened to end up in GCC.
This is really unfortunate, as the names don't have much rhyme or reason
to them. Originally in the discussions it seemed fine to rely on aliases
to map different names to whatever external thunk code developers wished
to use but there are practical problems with that in the kernel it turns
out. And since we're discovering this practical problems late and since
GCC has already shipped a release with one set of names, we are forced,
yet again, to blindly match what is there.
Somewhat rushing this patch out for the Linux kernel folks to test and
so we can get it patched into our releases.
Differential Revision: https://reviews.llvm.org/D42998
llvm-svn: 324449
X86 currently has a late DAG combine after cttz/ctlz are turned into BSR+BSF+CMOV to detect this and remove the CMOV. But we should be able to do this much earlier and avoid creating the cmov all together.
For the changed AMDGPU test case it appears that previously the i8 cttz was type legalized to i16 which introduced an OR with 256 in order to limit the result to 8 on the widened type. At this point the result is known to never be zero, but nothing checked that. Then operation legalization is told to promote all i16 cttz to i32. This introduces an extend and a truncate and another OR with 65536 to limit the result to 16. With the DAG combiner change we are able to prevent the creation of the second OR since the opcode will have been changed to cttz_zero_undef after the first OR. I the lack of the OR caused the instruction to change to v_ffbl_b32_sdwa
Differential Revision: https://reviews.llvm.org/D42985
llvm-svn: 324427
Followup to D42544 that matches PACKUSWB cases for non-AVX512, SSE and PACKUSDW cases will have to wait until we can add support for general SMIN/SMAX matching.
llvm-svn: 324347
Followup to D42544 that matches PACKSSWB cases for non-AVX512, SSE and PACKSSDW cases will have to wait until we can add support for general SMIN/SMAX matching.
llvm-svn: 324339
We now allow all signed comparisons and not equal. The complement that needs to be added for this is no worse than the extend. And the vector output forms of pcmpeq/pcmpgt have better latency than the k-register version on SKX.
llvm-svn: 324294
We always created X86ISD::SHUF128 with a 64-bit element type so we can use isel patterns to detect a bitconvert to 32-bit to handle masking.
The test changes are because we also match the bitconvert even if there is no masking. This leads to unnecessary isel pattern, but it requires more multiclass hackery in tablegen to get rid of it.
llvm-svn: 324205
This reduces the number of transitions between k-registers and GPRs, reducing the number of instructions.
There's still some room for improvement to remove more transitions, but this is a good start.
llvm-svn: 324184
Clang already stopped using these a couple months ago.
The test cases aren't great as there is nothing forcing the operations to stay in k-registers so some of them moved back to scalar ops due to the bitcasts being moved around.
llvm-svn: 324177
This is running pre-legalize, we should try to use target independent nodes. This will give the best opportunity for target independent optimizations.
llvm-svn: 324147
Every instruction that has the word TEST in its name seems to have been buried into EmitTest. But that code is largely concerned with trying to reuse the flags from instructions that update flags in a pretty normal way.
PTEST/TESTP/KTEST do not update flags in a normal way. They only update Z and C and the C flag update is non-standard. Rather than try to bend EmitTest's already complex logic to accomodate this, just move the call up to LowerSETCC and replicate the few pre-checks that are needed.
While there add a FIXME for using the C flag for checking for all 1s which we definitely couldn't do from EmitTEST.
llvm-svn: 324029
This allows us to use PSHUFB for v8i16/v4i32 and VPERMD/PERMPS for v4i64/v4f64 variable shuffles.
Differential Revision: https://reviews.llvm.org/D42487
llvm-svn: 323987
Summary: Now that v2i1/v4i1 are legal without VLX. And v32i1 is legalized by splitting rather than widening. And isVectorLoadExtDesirable returns false for vXi1. It appears this handling is dead because the operations simply don't exist.
Reviewers: RKSimon, zvi, guyblank, delena, spatel
Reviewed By: delena
Subscribers: llvm-commits, rengolin
Differential Revision: https://reviews.llvm.org/D42781
llvm-svn: 323983
Summary:
EmitTest sometimes creates X86ISD::AND specifically to hide the AND from DAG combine. But this prevents isel patterns that look for (cmp (and X, Y), 0) from being able to see it. So we end up with an AND and a TEST. The TEST gets removed by compare instruction optimization during the peephole pass.
This patch attempts to fix this by converting X86ISD::AND with no flag users back into ISD::AND during the DAG preprocessing just before isel.
In order to do this correctly I had to make the X86ISD::AND node created by EmitTest in this case really have a flag output. Which arguably it should have had anyway so that the number of operands would be consistent for the opcode in all cases. Then I had to modify the ReplaceAllUsesWith to understand that we might be looking at an instruction with 2 outputs. Though in this case there are no uses to replace since we just created the node, but that's what the code did before so I just made it keep working.
Reviewers: spatel, RKSimon, niravd, deadalnix
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42764
llvm-svn: 323982
Summary:
This change expands the amount of registers stashed by the entry and
`__xray_CustomEvent` trampolines.
We've found that since the `__xray_CustomEvent` trampoline calls can show up in
situations where the scratch registers are being used, and since we don't
typically want to affect the code-gen around the disabled
`__xray_customevent(...)` intrinsic calls, that we need to save and restore the
state of even the scratch registers in the handling of these custom events.
Reviewers: pcc, pelikan, dblaikie, eizan, kpw, echristo, chandlerc
Reviewed By: echristo
Subscribers: chandlerc, echristo, hiraditya, davide, dblaikie, llvm-commits
Differential Revision: https://reviews.llvm.org/D40894
llvm-svn: 323940
This code currently uses isSimple and getSizeInBits in an attempt to prune types. But isSimple will return true for any type that any target supports natively. I don't think that's a good way to prune types. I also don't think the dest element type checks are very robust since we didn't do an isSimple check on the dest type.
This patch adds a check for the input type being legal to the one caller that didn't already check that. Then we explicitly check the element types for the destination are i8, i16, or i32
Differential Revision: https://reviews.llvm.org/D42706
llvm-svn: 323924
Similar to D42437, XOP supports variable shift for v16i8/v8i16/v4i32/v2i64 types.
Differential Revision: https://reviews.llvm.org/D42526
llvm-svn: 323797
We can use the same input for both operands to get a free compare with zero.
We already use this trick in a couple places where we explicitly create PTESTM with the same input twice. This generalizes it.
I'm hoping to remove the ISD opcodes and move this to isel patterns like we do for scalar cmp/test.
llvm-svn: 323605
Legalization is still biased to turn LT compares in to GT by swapping operands to avoid needing extra isel patterns to commute.
I'm hoping to remove TESTM/TESTNM next and this should simplify that by making EQ/NE more similar.
llvm-svn: 323604
If broadcasting from another shuffle, attempt to simplify it.
We can probably generalize this a lot more (embedding in combineX86ShufflesRecursively), but BROADCAST is one of the more troublesome as it accepts inputs of different sizes to the result.
llvm-svn: 323602
X86ISelLowering.cpp:34130:5: error: return type 'llvm::SDValue' must
match previous return type 'const llvm::SDValue' when lambda expression
has unspecified explicit return type
llvm-svn: 323557
We currently coalesce v4i32 extracts from all 4 elements to 2 v2i64 extracts + shifts/sign-extends.
This seems to have been added back in the days when we tended to spill vectors and reload scalars, or ended up with repeated shuffles moving everything down to 0'th index. I don't think either of these are likely these days as we have better EXTRACT_VECTOR_ELT and VECTOR_SHUFFLE handling, and the existing code tends to make it very difficult for various vector and load combines.
Differential Revision: https://reviews.llvm.org/D42308
llvm-svn: 323541
Type legalization would prevent any i64 operands to the build_vector from existing before we get here. The coverage bots show this code as uncovered.
llvm-svn: 323506
The original autoupgrade for kunpck intrinsics used a bitcasted scalar shift, or, and. This combine would turn this into a concat_vectors. Now the kunpck intrinsics are autoupgraded to a vector shuffle that will become a concat_vectors.
llvm-svn: 323504
This listed all legal 128-bit integer types individually, but since we already know we have a legal type and its integer, we can just check is128BitVector.
llvm-svn: 323502
As discussed in D41484, PMADDWD for 'zero extended' vXi32 is nearly always a better option than PMULLD:
On SNB it will result in code that isn't any faster, but not any slower so we may as well keep it.
On KNL it only has half the throughput, so I've disabled it on there - ideally there'd be a better way than this.
Differential Revision: https://reviews.llvm.org/D42258
llvm-svn: 323367
There are a couple tricky things with this patch.
I had to add an override of isVectorLoadExtDesirable to stop DAG combine from combining sign_extend with loads after legalization since we legalize sextload using a load+sign_extend. Overriding this hook actually prevents a lot sextloads from being created in the first place.
I also had to add isel patterns because DAG combine blindly combines sign_extend+truncate to a smaller sign_extend which defeats what legalization was trying to do.
Differential Revision: https://reviews.llvm.org/D42407
llvm-svn: 323301
Minor refactor to make it possible for LowerBUILD_VECTORAsVariablePermute to be used with a wider variety of shuffles op and types.
I'd have liked to add v4i32/v4f32 support as well but we don't see v4i32 index extractions at the moment (which is why I created D42308)
After this I intend to begin adding scaling support for PSHUFB (v8i16, v4i32, v2i64)) and VPERMPS (v4f64, v4i64).
Differential Revision: https://reviews.llvm.org/D42431
llvm-svn: 323260
The existing code was already doing something very similar to subvector insertion so this allows us to remove the nearly duplicate code.
This patch is a little larger than it should be due to differences between the DQI handling between the two today.
llvm-svn: 323212
Summary:
For the most part its better to keep v32i1 as a mask type of a narrower width than trying to promote it to a ymm register.
I had to add some overrides to the methods that get the types for the calling convention so that we still use v32i8 for argument/return purposes.
There are still some regressions in here. I definitely saw some around shuffles. I think we probably should move vXi1 shuffle from lowering to a DAG combine where I think the extend and truncate we have to emit would be better combined.
I think we also need a DAG combine to remove trunc from (extract_vector_elt (trunc))
Overall this removes something like 13000 CHECK lines from lit tests.
Reviewers: zvi, RKSimon, delena, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42031
llvm-svn: 323201
As detailed in rL317463, PSHUFB (like most variable shuffle instructions) uses Op[0] for the source vector and Op[1] for the shuffle index vector, VPERMV works in reverse which is probably where the confusion comes from.
Differential Revision: https://reviews.llvm.org/D42380
llvm-svn: 323190
Summary:
If we can match as a zero extend there's no need to flip the order to get an encoding benefit. As movzx is 3 bytes with independent source/dest registers. The shortest 'and' we could make is also 3 bytes unless we get lucky in the register allocator and its on AL/AX/EAX which have a 2 byte encoding.
This patch was more impressive before r322957 went in. It removed some of the same Ands that got deleted by that patch.
Reviewers: spatel, RKSimon
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42313
llvm-svn: 323175
Add missing patterns for inserting v1i1 into a zero vector. Use insert_subvector to zero upper bits before inserting an element into a vXi1 vector. Replace kshift based isel pattern with insert_subvector based pattern now that code that caused the pattern has been fixed to emit insert_subvector.
llvm-svn: 323173
Summary:
First, we need to explain the core of the vulnerability. Note that this
is a very incomplete description, please see the Project Zero blog post
for details:
https://googleprojectzero.blogspot.com/2018/01/reading-privileged-memory-with-side.html
The basis for branch target injection is to direct speculative execution
of the processor to some "gadget" of executable code by poisoning the
prediction of indirect branches with the address of that gadget. The
gadget in turn contains an operation that provides a side channel for
reading data. Most commonly, this will look like a load of secret data
followed by a branch on the loaded value and then a load of some
predictable cache line. The attacker then uses timing of the processors
cache to determine which direction the branch took *in the speculative
execution*, and in turn what one bit of the loaded value was. Due to the
nature of these timing side channels and the branch predictor on Intel
processors, this allows an attacker to leak data only accessible to
a privileged domain (like the kernel) back into an unprivileged domain.
The goal is simple: avoid generating code which contains an indirect
branch that could have its prediction poisoned by an attacker. In many
cases, the compiler can simply use directed conditional branches and
a small search tree. LLVM already has support for lowering switches in
this way and the first step of this patch is to disable jump-table
lowering of switches and introduce a pass to rewrite explicit indirectbr
sequences into a switch over integers.
However, there is no fully general alternative to indirect calls. We
introduce a new construct we call a "retpoline" to implement indirect
calls in a non-speculatable way. It can be thought of loosely as
a trampoline for indirect calls which uses the RET instruction on x86.
Further, we arrange for a specific call->ret sequence which ensures the
processor predicts the return to go to a controlled, known location. The
retpoline then "smashes" the return address pushed onto the stack by the
call with the desired target of the original indirect call. The result
is a predicted return to the next instruction after a call (which can be
used to trap speculative execution within an infinite loop) and an
actual indirect branch to an arbitrary address.
On 64-bit x86 ABIs, this is especially easily done in the compiler by
using a guaranteed scratch register to pass the target into this device.
For 32-bit ABIs there isn't a guaranteed scratch register and so several
different retpoline variants are introduced to use a scratch register if
one is available in the calling convention and to otherwise use direct
stack push/pop sequences to pass the target address.
This "retpoline" mitigation is fully described in the following blog
post: https://support.google.com/faqs/answer/7625886
We also support a target feature that disables emission of the retpoline
thunk by the compiler to allow for custom thunks if users want them.
These are particularly useful in environments like kernels that
routinely do hot-patching on boot and want to hot-patch their thunk to
different code sequences. They can write this custom thunk and use
`-mretpoline-external-thunk` *in addition* to `-mretpoline`. In this
case, on x86-64 thu thunk names must be:
```
__llvm_external_retpoline_r11
```
or on 32-bit:
```
__llvm_external_retpoline_eax
__llvm_external_retpoline_ecx
__llvm_external_retpoline_edx
__llvm_external_retpoline_push
```
And the target of the retpoline is passed in the named register, or in
the case of the `push` suffix on the top of the stack via a `pushl`
instruction.
There is one other important source of indirect branches in x86 ELF
binaries: the PLT. These patches also include support for LLD to
generate PLT entries that perform a retpoline-style indirection.
The only other indirect branches remaining that we are aware of are from
precompiled runtimes (such as crt0.o and similar). The ones we have
found are not really attackable, and so we have not focused on them
here, but eventually these runtimes should also be replicated for
retpoline-ed configurations for completeness.
For kernels or other freestanding or fully static executables, the
compiler switch `-mretpoline` is sufficient to fully mitigate this
particular attack. For dynamic executables, you must compile *all*
libraries with `-mretpoline` and additionally link the dynamic
executable and all shared libraries with LLD and pass `-z retpolineplt`
(or use similar functionality from some other linker). We strongly
recommend also using `-z now` as non-lazy binding allows the
retpoline-mitigated PLT to be substantially smaller.
When manually apply similar transformations to `-mretpoline` to the
Linux kernel we observed very small performance hits to applications
running typical workloads, and relatively minor hits (approximately 2%)
even for extremely syscall-heavy applications. This is largely due to
the small number of indirect branches that occur in performance
sensitive paths of the kernel.
When using these patches on statically linked applications, especially
C++ applications, you should expect to see a much more dramatic
performance hit. For microbenchmarks that are switch, indirect-, or
virtual-call heavy we have seen overheads ranging from 10% to 50%.
However, real-world workloads exhibit substantially lower performance
impact. Notably, techniques such as PGO and ThinLTO dramatically reduce
the impact of hot indirect calls (by speculatively promoting them to
direct calls) and allow optimized search trees to be used to lower
switches. If you need to deploy these techniques in C++ applications, we
*strongly* recommend that you ensure all hot call targets are statically
linked (avoiding PLT indirection) and use both PGO and ThinLTO. Well
tuned servers using all of these techniques saw 5% - 10% overhead from
the use of retpoline.
We will add detailed documentation covering these components in
subsequent patches, but wanted to make the core functionality available
as soon as possible. Happy for more code review, but we'd really like to
get these patches landed and backported ASAP for obvious reasons. We're
planning to backport this to both 6.0 and 5.0 release streams and get
a 5.0 release with just this cherry picked ASAP for distros and vendors.
This patch is the work of a number of people over the past month: Eric, Reid,
Rui, and myself. I'm mailing it out as a single commit due to the time
sensitive nature of landing this and the need to backport it. Huge thanks to
everyone who helped out here, and everyone at Intel who helped out in
discussions about how to craft this. Also, credit goes to Paul Turner (at
Google, but not an LLVM contributor) for much of the underlying retpoline
design.
Reviewers: echristo, rnk, ruiu, craig.topper, DavidKreitzer
Subscribers: sanjoy, emaste, mcrosier, mgorny, mehdi_amini, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D41723
llvm-svn: 323155
Primarily, this allows us to use the aggressive extraction mechanisms in combineExtractWithShuffle earlier and make use of UNDEF elements that may be lost during lowering.
Reapplied after rL322279 was reverted at rL322335 due to PR35918, underlying issue was fixed at rL322644.
llvm-svn: 323104
1. ReachingDefsAnalysis - Allows to identify for each instruction what is the “closest” reaching def of a certain register. Used by BreakFalseDeps (for clearance calculation) and ExecutionDomainFix (for arbitrating conflicting domains).
2. ExecutionDomainFix - Changes the variant of the instructions in order to minimize domain crossings.
3. BreakFalseDeps - Breaks false dependencies.
4. LoopTraversal - Creatws a traversal order of the basic blocks that is optimal for loops (introduced in revision L293571). Both ExecutionDomainFix and ReachingDefsAnalysis use this to determine the order they will traverse the basic blocks.
This also included the following changes to ExcecutionDepsFix original logic:
1. BreakFalseDeps and ReachingDefsAnalysis logic no longer restricted by a register class.
2. ReachingDefsAnalysis tracks liveness of reg units instead of reg indices into a given reg class.
Additional changes in affected files:
1. X86 and ARM targets now inherit from ExecutionDomainFix instead of ExecutionDepsFix. BreakFalseDeps also was added to the passes they activate.
2. Comments and references to ExecutionDepsFix replaced with ExecutionDomainFix and BreakFalseDeps, as appropriate.
Additional refactoring changes will follow.
This commit is (almost) NFC.
The only functional change is that now BreakFalseDeps will break dependency for all register classes.
Since no additional instructions were added to the list of instructions that have false dependencies, there is no actual change yet.
In a future commit several instructions (and tests) will be added.
This is the first of multiple patches that fix bugzilla https://bugs.llvm.org/show_bug.cgi?id=33869
Most of the patches are intended at refactoring the existent code.
Additional relevant reviews:
https://reviews.llvm.org/D40331https://reviews.llvm.org/D40332https://reviews.llvm.org/D40333https://reviews.llvm.org/D40334
Differential Revision: https://reviews.llvm.org/D40330
Change-Id: Icaeb75e014eff96a8f721377783f9a3e6c679275
llvm-svn: 323087
Summary:
This patch adds an implementation of targetShrinkDemandedConstant that tries to keep shrinkdemandedbits from removing bits that would otherwise have been recognized as a movzx.
We still need a follow patch to stop moving ands across srl if the and could be represented as a movzx before the shift but not after. I think this should help with some of the cases that D42088 ended up removing during isel.
Reviewers: spatel, RKSimon
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42265
llvm-svn: 323048
This change applies to places where we would turn 128/256-bit code into 512-bit in order to get a wider element type through sext/zext. Any 512-bit types that already existed in the IR/DAG will be left that way.
The width preference has no effect on codegen behavior when the target does not have AVX512 enabled. So AVX/AVX2 codegen cannot be limited via this mechanism yet.
If the preference is lower than 256 we may still use a 256 bit type to do the operation. Constraining to 128 bits makes it much more difficult to support some operations. For many of these cases we need to change element width while keeping element count constant which is easiest done by switching between 256 and 128 bit.
The preference is only obeyed when AVX512 and VLX are available. This means the preference is not obeyed for KNL, but is obeyed for SKX, Cannonlake, and Icelake. For KNL, the only way to do masked operation is on 512-bit registers so we would have to completely disable masking to obey the preference. We would also lose support for gather, scatter, ctlz, vXi64 multiplies, etc. This may change in the future, but this simplifies the initial implementation.
Differential Revision: https://reviews.llvm.org/D41895
llvm-svn: 323016
If we are splatting pairs of 32-bit elements, we can use a 64-bit broadcast to get the job done.
We could probably could probably do this with other sizes too, for example four 16-bit elements. Or we could broadcast pairs of 16-bit elements using a 32-bit element broadcast. But I've left that as a future improvement.
I've also restricted this to AVX2 only because we can only broadcast loads under AVX.
Differential Revision: https://reviews.llvm.org/D42086
llvm-svn: 322730
We legalize selects of masks with scalar conditions using a bitcast to an integer type. But if we are in 32-bit mode we can't convert v64i1 to i64. So instead split the v64i1 to v32i1 and concat it back together. Each half will then be legalized by bitcasting to i32 which is fine.
The test case is a little indirect. If we have the v64i1 select in IR it will get legalized by legalize vector ops which has a run of type legalization after it. That type legalization run is able to fix this i64 bitcast. So in order to avoid that we need a build_vector of a splat which legalize vector ops will ignore. Legalize DAG will then turn that into a select via LowerBUILD_VECTORvXi1. And the select will get legalized. In this case there is no type legalizer run to cleanup the bitcast.
This fixes pr35972.
llvm-svn: 322724
The match* functions have the annoying behavior of modifying its inputs.
Save and restore the inputs, just in case the early out for AVX512 is
hit. This is still not great and its only a matter of time this kind of
bug happens again, but I couldn't come up with a better pattern without
rewriting significant chunks of this code. Fixes PR35977.
llvm-svn: 322644
As mentioned on PR35869, (and came up recently on D41517) we don't create a MMX zero register via the PXOR but instead perform a spill to stack from a XMM zero register.
This patch adds support for direct MMX zero vector creation and should make it easier to add better constant vector creation in the future as well.
Differential Revision: https://reviews.llvm.org/D41908
llvm-svn: 322525
We have to take special care to avoid the cases where the result of the truncate would be padded with zero elements.
Ideally we'd just use ISD::TRUNCATE for these cases instead.
llvm-svn: 322454
Extend vXi1 conditions of vXi8/vXi16 selects even before type legalization gets a chance to split wide vectors. Previously we would only extend 128 and 256 bit vectors. But if we start with a 512 bit vector or wider that needs to be split we wouldn't extend until after the split had taken place. By extending early we improve the results of type legalization.
Don't widen condition of 128/256 bit vXi16/vXi8 selects when we have BWI but not VLX. We can still use a mask register by widening the select to 512-bits instead. This is similar to what we do for compares already.
llvm-svn: 322450
In addition to the existing match as part of a loop-reduction, add a
straightforward pattern match for DAG-contained patterns.
Reviewers: RKSimon, craig.topper
Subscribers: llvm-commits
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D41811
llvm-svn: 322446
This avoids having the result type stick around until lowering where we have to extend the setcc and insert a truncate. If we get the types converted early we can do more to optimize it.
llvm-svn: 322432
Summary: This is a preparatory step for D41811: refactoring code for breaking vector operands of binary operation to legal-types.
Reviewers: RKSimon, craig.topper, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41925
llvm-svn: 322296
Primarily, this allows us to use the aggressive extraction mechanisms in combineExtractWithShuffle earlier and make use of UNDEF elements that may be lost during lowering.
llvm-svn: 322279
Summary:
As RKSimon suggested in pr35820, in the case that Src is smaller in
bit-size than Indices, need to widen Src to avoid type mismatch.
Fixes pr35820
Reviewers: RKSimon, craig.topper
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41865
llvm-svn: 322272
If the index is v2i64 we can use the scatter instruction that has v4i32/v4f32 data register, v2i64 index, and v2i1 mask. Similar was already done for gather.
Implement custom widening for v2i32 data to remove the code that reverses type legalization during lowering.
llvm-svn: 322254
Currently we infer the scale at isel time by analyzing whether the base is a constant 0 or not. If it is we assume scale is 1, else we take it from the element size of the pass thru or stored value. This seems a little weird and I think it makes more sense to make it explicit in the DAG rather than doing tricky things in the backend.
Most of this patch is just making sure we copy the scale around everywhere.
Differential Revision: https://reviews.llvm.org/D40055
llvm-svn: 322210
Normally target independent DAG combine would do this combine based on getSetCCResultType, but with VLX getSetCCResultType returns a vXi1 type preventing the DAG combining from kicking in.
But doing this combine can allow us to remove the explicit sign extend that would otherwise be emitted.
This patch adds a target specific DAG combine to combine the sext+setcc when the result type is the same size as the input to the setcc. I've restricted this to FP compares and things that can be represented with PCMPEQ and PCMPGT since we don't have full integer compare support on the older ISAs.
Differential Revision: https://reviews.llvm.org/D41850
llvm-svn: 322101
I had to drop fast-isel-abort from a test because we can't fast isel some of the mask stuff. When we used intrinsics we implicitly fell back to SelectionDAG for the intrinsic call without triggering the abort error. But with native IR that doesn't happen the same way.
llvm-svn: 322050
Summary:
There are few oddities that occur due to v1i1, v8i1, v16i1 being legal without v2i1 and v4i1 being legal when we don't have VLX. Particularly during legalization of v2i32/v4i32/v2i64/v4i64 masked gather/scatter/load/store. We end up promoting the mask argument to these during type legalization and then have to widen the promoted type to v8iX/v16iX and truncate it to get the element size back down to v8i1/v16i1 to use a 512-bit operation. Since need to fill the upper bits of the mask we have to fill with 0s at the promoted type.
It would be better if we could just have the v2i1/v4i1 types as legal so they don't undergo any promotion. Then we can just widen with 0s directly in a k register. There are no real v4i1/v2i1 instructions anyway. Everything is done on a larger register anyway.
This also fixes an issue that we couldn't implement a masked vextractf32x4 from zmm to xmm properly.
We now have to support widening more compares to 512-bit to get a mask result out so new tablegen patterns got added.
I had to hack the legalizer for widening the operand of a setcc a bit so it didn't try create a setcc returning v4i32, extract from it, then try to promote it using a sign extend to v2i1. Now we create the setcc with v4i1 if the original setcc's result type is v2i1. Then extract that and don't sign extend it at all.
There's definitely room for improvement with some follow up patches.
Reviewers: RKSimon, zvi, guyblank
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41560
llvm-svn: 321967
This custom inserter was added in r124272 at which time it added about bunch of Defs for Win64. In r150708, those defs were removed leaving only the "return BB". So I think this means the custom inserter is a NOP these days.
This patch removes the remaining code and stops tagging the instructions for custom insertion
Differential Revision: https://reviews.llvm.org/D41671
llvm-svn: 321747
Currently we use SIGN_EXTEND in lowerMasksToReg as part of calling convention setup, but we don't require a specific value for the upper bits.
This patch changes it to ANY_EXTEND which will be lowered as SIGN_EXTEND if it ends up sticking around.
llvm-svn: 321746
This is an extension of D31156 with the goal that we'll allow memcmp() == 0 expansion
for x86 to use 2 pairs of loads per block.
The memcmp expansion pass (formerly part of CGP) will generate this kind of pattern
with oversized integer compares, so we want to transform these into x86-specific vector
nodes before legalization splits things into scalar chunks.
See PR33325 for more details:
https://bugs.llvm.org/show_bug.cgi?id=33325
Differential Revision: https://reviews.llvm.org/D41618
llvm-svn: 321656
Currently the promotion for these ignores the normal getTypeToPromoteTo and instead just tries to double the element width. This is because the default behavior of getTypeToPromote to just adds 1 to the SimpleVT, which has the affect of increasing the element count while keeping the scalar size the same.
If multiple steps are required to get to a legal operation type, int_to_fp will be promoted multiple times. And fp_to_int will keep trying wider types in a loop until it finds one that works.
getTypeToPromoteTo does have the ability to query a promotion map to get the type and not do the increasing behavior. It seems better to just let the target specify the promotion type in the map explicitly instead of letting the legalizer iterate via widening.
FWIW, it's worth I think for any other vector operations that need to be promoted, we have to specify the type explicitly because the default behavior of getTypeToPromote isn't useful for vectors. The other types of promotion already require either the element count is constant or the total vector width is constant, but neither happens by incrementing the SimpleVT enum.
Differential Revision: https://reviews.llvm.org/D40664
llvm-svn: 321629
The CONCAT_VECTORS will be lowered to INSERT_SUBVECTOR later. In the modified cases this seems to be enough to trick a later DAG combine into running in a different order than allows the ANDs to be removed.
I'll admit this is a bit of a hack that happens to work, but using CONCAT_VECTORS is more consistent with other legalization code anyway.
llvm-svn: 321611
Don't combine buildvector(binop(),binop(),binop(),binop()) -> binop(buildvector(), buildvector()) if its a splat - keep the binop scalar and just splat the result to avoid large vector constants.
llvm-svn: 321607
We end up using an i8 load via an isel pattern from v8i1 anyway. This just makes it more explicit. This seems to improve codgen in some cases and I'd like to kill off some of the load patterns.
llvm-svn: 321598
As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles.
This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles.
By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks.
We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc.
Differential Revision: https://reviews.llvm.org/D38318
llvm-svn: 321553
Previously we used an extend from v8i1 to v8i32/v8i64. Then extracted to the final width. But if we have VLX we should extract first. This way we don't end up with an overly large extend.
This allows us to use vcmpeq to make all ones for the sign extend when DQI isn't available. Otherwise we get a VPTERNLOG.
If we make v2i1/v4i1 legal like proposed in D41560, we could always do this and rely on the lowering of the extend to widen when necessary.
llvm-svn: 321538
-Use MinAlign instead of std::min.
-Use SelectionDAG::getMemBasePlusOffset.
-Apply offset to the pointer info for the second load/store created.
llvm-svn: 321536
If there are 17 or more leading zeros to the v4i32 elements, then we can use PMADD for the integer multiply when PMULLD is unavailable or slow.
The 17 bits need to be zero as the PMADDWD performs a v8i16 signed-mul-extend + pairwise-add - the upper 16 so we're adding a zero pair and the 17th bit so we don't incorrectly sign extend.
Differential Revision: https://reviews.llvm.org/D41484
llvm-svn: 321516
My original implementation ran as a DAG combine post type legalization, but it turns out we don't run that DAG combine step if type legalization didn't change anything. Attempts to make the combine run before type legalization as well hit other issues.
So just do it in LowerMUL where we can catch more cases.
llvm-svn: 321496
Returning SDValue() means nothing changed, SDValue(N,0) means there was a change but the worklist management was taken care of.
I don't know if this has a real effect other than making sure the combine counter in the DAG combiner gets updated, but it is the correct thing to do.
llvm-svn: 321463
Normally we catch this during lowering, but vXi64 mul is considered legal when we have AVX512DQ.
This DAG combine allows us to avoid PMULLQ with AVX512DQ if we can prove its unnecessary. PMULLQ is 3 uops that take 4 cycles each. While pmuldq/pmuludq is only one 4 cycle uop.
llvm-svn: 321437
Previously we extended v2i1 to v2f64 and then tried to use cvtuqq2pd/cvtqq2pd, but that only works with avx512dq. So we ended up scalarizing it. Now we widen to v4i1 first and extend to v4i32.
llvm-svn: 321420
Immediately after it is created we check if its equal to another EVT. Then we inconsistently use one or the other variables in the code below.
Instead do the equality check directly on the getValueType result and remove the variable. Use the origina VT variable throughout the remaining code.
llvm-svn: 321406
getOperand returns an SDValue that contains the node and the result number. There is no guarantee that the result number if 0. By using the -> operator we are calling SDNode::getValueType rather than SDValue::getValueType. This requires supplying a result number and we shouldn't assume it was 0.
I don't have a test case. Just noticed while cleaning up some other code and saw that it occurred in other places.
llvm-svn: 321397
Despite what the comment said there isn't better codegen for 512-bit vectors. The 128/256/512 bit implementation jus stores to memory and loads an element. There's no advantage to doing that with a larger size. In fact in many cases it causes a stack realignment and generates worse code.
llvm-svn: 321369
Previously prefetch was only considered legal if sse was enabled, but it should be supported with 3dnow as well.
The prfchw flag now imply at least some form of prefetch without the write hint is available, either the sse or 3dnow version. This is true even if 3dnow and sse are explicitly disabled.
Similarly prefetchwt1 feature implies availability of prefetchw and the the prefetcht0/1/2/nta instructions. This way we can support _MM_HINT_ET0 using prefetchw and _MM_HINT_ET1 with prefetchwt1. And its assumed that if we have levels for the write hint we would have levels for the non-write hint, thus why we enable the sse prefetch instructions.
I believe this behavior is consistent with gcc. I've updated the prefetch.ll to test all of these combinations.
llvm-svn: 321335
This should only affect what we do for v8i16. Previously we went to v8i64, but if we have VLX we only need v8i32. This prevents an unnecessary zmm usage.
llvm-svn: 321303
We should have equally good shuffle options for v8i32 with VLX. This was spotted during my attempts to remove 512-bit vectors from SKX.
We still use 512-bits for v16i1, v32i1, and v64i1. I'm less sure we can handle those well with narrower vectors. i32 and i64 element sizes get the best shuffle support.
llvm-svn: 321291
Gather/scatter can implicitly sign extend from i32->i64 on indices. So if we know the sign bit of the input to a zext is 0 we can use the implicit extension.
llvm-svn: 321209
This patch turns shuffles of fadd/fsub with fmul into fmsubadd.
Patch by Dmitry Venikov
Differential Revision: https://reviews.llvm.org/D40335
llvm-svn: 321200
The gather instruction will implicitly sign extend to the pointer width, we don't need to further extend it. This can prevent unnecessary splitting in some cases.
There's still an issue that lowering on non-VLX can introduce another sign extend that doesn't get combined with shifts from a lowered sign_extend_inreg.
llvm-svn: 321152
Not sure how to test this cause I think the worst that happens is that we don't revisit the node a second time to look for additional combines. We used UpdateNodeOperands so the updating the DAG work was already done.
llvm-svn: 321148
We try to prevent shuffle combining to value types that would stop the folding of masked operations, but by just returning early, we were failing to try different shuffle types.
The TODOs are all still relevant here to improve codegen but we're lacking test examples.
llvm-svn: 321085
As mentioned in D38318 and D40865, modern Intel processors prefer to combine multiple shuffles to a variable shuffle mask (PSHUFB/VPERMPS etc.) instead of having multiple stage 'fixed' shuffles which put more pressure on Port 5 (at the expense of extra shuffle mask loads).
This patch provides a FeatureFastVariableShuffle target flag for Haswell+ CPUs that prefers combining 2 or more fixed shuffles to a single variable shuffle (default is 3 shuffles).
The long term aim is to drive more of this from schedule data (probably via the MC) but we're not close to being ready for that yet.
Differential Revision: https://reviews.llvm.org/D41323
llvm-svn: 321074
Extension to D39729 which performed this for vXi16, with the same bit flipping to handle SMAX/SMIN/UMAX cases, vXi8 UMIN horizontal reductions can be performed.
This makes use of the fact that by performing a pair-wise i8 SHUFFLE/UMIN before PHMINPOSUW, we both get the UMIN of each pair but also zero-extend the upper bits ready for v8i16.
Differential Revision: https://reviews.llvm.org/D41294
llvm-svn: 321070
BWI supports shifting by word amounts. Even if VLX isn't support we can still widen to v32i16 and extract the lower half. For SKX its preferrable to not use 512-bit vector if we can.
llvm-svn: 321059
Previously, we were checking for MVTs with sizes betwen 8 and 64 which only includes i8, i16, i32, and i64 today. But I don't think we should assume that and should list the types that are legal for x86. I also don't think we need i64 since type legalization is guaranteed to split those up.
llvm-svn: 321058
My reading of the SDM says that all bits of the shift amount are used. If the value of the element is larger than the number of bits the result the shift result is zero. So I think we need to zero_extend here to avoid garbage in the upper bits.
In reality we lower any_extend as zero_extend so in most cases it would be hard to hit this.
llvm-svn: 321055
Note:
- X86ISelLowering: setLibcallName(SINCOS) was superfluous as
InitLibcalls() already does it.
- ARMISelLowering: Setting libcallnames for sincos/sincosf seemed
superfluous as in the darwin case it wouldn't be used while for all
other cases InitLibcalls already does it.
llvm-svn: 321036
The block I moved things that need BWI and 512-bit or VLX is incorrectly qualified with just hasBWI || hasVLX. Here I've qualified it with hasBWI && (hasAVX512 || hasVLX) where the hasAVX512 will be replaced with allowing 512-bit vectors in an upcoming patch.
llvm-svn: 320957
Summary:
We had no tests for this and we couldn't do the optimization because of a bad use count check. We need to know how many non-undef pieces of the build vector were filled in and ensure our use count is equal to that. But on the shuffle combine version we need the use count to be 2.
The missing coverage was noticed during the review of D40335.
Reviewers: RKSimon, zvi, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41133
llvm-svn: 320950
Assuming we can safely adjust the broadcast index for the new type to keep it suitably aligned, then peek through BITCASTs when looking for the broadcast source.
Fixes PR32007
llvm-svn: 320933
In those cases, the pass thru operand of the methods isn't used. The calls to the scalar version were passing a MVT::i1 zero, which is an illegal type at the stage this code runs.
llvm-svn: 320928
Previously we promoted to v8i64, but we don't need to go all the way to 512-bits. If we have VLX we can use the 256-bit instruction. And even if we don't have VLX we can widen v8i32 to v16i32 and drop the upper half.
llvm-svn: 320926
The target independent nodes will get legalized to the target specific nodes by their own legalization process. Someday I'd like to stop using a target specific for zero extends and truncates of legal types so the less places we reference the target specific opcode the better.
llvm-svn: 320863
When I wrote it I thought we were missing a potential optimization for KNL. But investigating further shows that for KNL we still do the optimal thing by widening to v4f32 and then using special isel patterns to widen again to zmm a register.
llvm-svn: 320862
Summary:
Currently we don't handle v32i1/v64i1 insert_vector_elt correctly as we fail to look at the number of elements closely and assume it can only be v16i1 or v8i1.
We also can't type legalize v64i1 insert_vector_elt correctly on KNL due to the type not being byte addressable as required by the legalizing through memory accesses path requires.
For the first issue, the patch now tries to pick a 512-bit register with the correct number of elements and promotes to that.
For the second issue, we now extend the vector to a byte addressable type, do the stores to memory, load the two halves, and then truncate the halves back to the original type. Technically since we changed the type, we may not need two loads, but actually checking that is more work and for the v64i1 case we do need them.
Reviewers: RKSimon, delena, spatel, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D40942
llvm-svn: 320849
A couple places didn't use the same SDValue variables to connect everything all the way through.
I don't have a test case for a bug in insert into the lower bits of a non-zero, non-undef vector. Not sure the best way to create that. We don't create the case when lowering concat_vectors which is the main way to get insert_subvectors.
llvm-svn: 320790
We have several instructions that were introduced in AVX512F that are only available in 512-bit form on KNL. We still make use of them for 128/256 by artificially widening and extracting during isel.
This commit separates these operations from the true 512-bit operations. This way we can qualify the normal 512-bit operations with needing 512-bit register support. And these special operations will get qualified with needing 512-bit registers OR VLX.
The 512-bit register qualification will be introduced in a future patch this just gets everything grouped to minimize deltas on that patch.
llvm-svn: 320782
Previously they were sort of interleaved in with XMM/YMM/ZMM action related code.
Trying to separate things so its easier to split 512-bit vectors later.
llvm-svn: 320781
Move it into the separate hasVLX block later in the constructor.
I'm trying to separate 128/256 and 512-bit related code so we can eventually qualify the hasAVX512 block with support for 512-bit vectors required by the prefer-vector-width feature support being talked about in D41096.
llvm-svn: 320779
Most of the -Wsign-compare warnings are due to the fact that
enums are signed by default in the MS ABI, while the
tautological comparison warnings trigger on x86 builds where
sizeof(size_t) is 4 bytes, so N > numeric_limits<unsigned>::max()
is always false.
Differential Revision: https://reviews.llvm.org/D41256
llvm-svn: 320750
Rather than adding more bits to express every
MMO flag you could want, just directly use the
MMO flags. Also fixes using a bunch of bool arguments to
getMemIntrinsicNode.
On AMDGPU, buffer and image intrinsics should always
have MODereferencable set, but currently there is no
way to do that directly during the initial intrinsic
lowering.
llvm-svn: 320746
This doesn't match the semantics of the extract_vector_elt operation. Nothing downstream knows the bits were zeroed so they still get masked or sign extended after the extrat anyway.
llvm-svn: 320723
store operation on a truncated memory (load) of vXi1 is poorly supported by LLVM and most of the time end with an assertion.
This patch fixes this issue.
Differential Revision: https://reviews.llvm.org/D39547
Change-Id: Ida5523dd09c1ad384acc0a27e9e59273d28cbdc9
llvm-svn: 320691
Pass the input vector through SimplifyDemandedBits as we only need the sign bit from each vector element of MOVMSK
We'd probably get more hits if SimplifyDemandedBits was better at handling vectors...
Differential Revision: https://reviews.llvm.org/D41119
llvm-svn: 320570
D40335 was wanting to add FMSUBADD support, but it discovered that there are two pieces of code to make FMADDSUB and only one of those is tested. So I've asked that review to implement the one path until we get tests that test the existing code.
llvm-svn: 320507
Summary:
Simplify and generalize chain handling and search for 64-bit load-store pairs.
Nontemporal test now converts 64-bit integer load-store into f64 which it realizes directly instead of splitting into two i32 pairs.
Reviewers: craig.topper, spatel
Reviewed By: craig.topper
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D40918
llvm-svn: 320505
Recognize constant arrays with the following values:
0x0, 0x1, 0x3, 0x7, 0xF, 0x1F, .... , 2^(size - 1) -1
where //size// is the size of the array.
the result of a load with index //idx// from this array is equivalent to the result of the following:
(0xFFFFFFFF >> (sub 32, idx)) (assuming the array of type 32-bit integer).
And the result of an 'AND' operation on the returned value of such a load and another input, is exactly equivalent to the X86 BZHI instruction behavior.
See test cases in the LIT test for better understanding.
Differential Revision: https://reviews.llvm.org/D34141
llvm-svn: 320481
We may need to widen the vector to make the shifts legal, but if we do that we need to make sure we shift left/right after accounting for the new size. If not we can't guarantee we are shifting in zeros.
The test cases affected actually show cases where we should move the shifts all together, but that's another problem.
llvm-svn: 320248
We were previously using kunpck with zero inputs unnecessarily. And we had cases where we would insert into a zero vector and then insert into larger zero vector incurring two sets of shifts.
llvm-svn: 320244
For narrow sizes we'll widen the zero vector and widen the insert. Then do an extract_subvector to get back down to correct size.
This allows us to remove some patterns from the isel table that had to COPY_TO_REGCLASS to an oversized register, do the shift and then COPY_TO_REGCLASS back to the narrow register. Now this is represented explicitly in the DAG.
This seems to have perturbed the register allocation in one of the tests, but the number of instructions didn't change.
llvm-svn: 320190
These are aliases, but the thing we're checking here is that the target has
vpsllv*, not that the data type is 256-bit. Those instructions exist for
128-bit vectors too...but sadly, not for all element sizes.
llvm-svn: 320170
Previously we only allowed these through if the subvector came from a compare or test instruction which we would again check for during isel.
With this change we only check for the compare and test instructions during isel and have fallback patterns that emit the shifts if needed.
I noticed that in a lot of cases we don't actually see the compare during lowering and rely on an odd legalization of concat_vectors with a zero vector as the second argument. This keeps the concat_vectors around long enough for a later dag combine to expose the compare then we re-legalize the concat_vectors and catch the compare.
llvm-svn: 320134
We previously only supported inserting to the LSB or MSB where it was easy to zero to perform an OR to insert.
This change effectively extracts the old value and the new value, xors them together and then xors that single bit with the correct location in the original vector. This will cancel out the old value in the first xor leaving the new value in the position.
The way I've implemented this uses 3 shifts and two xors and uses an additional register. We can avoid the additional register at the cost of another shift.
llvm-svn: 320120
There's no v2i1 or v4i1 kshift, and v8i1 is only supported with AVXDQ. Isel has fake patterns to extend these types to native shifts, but makes no guarantees about the value of any bits shifted in when shifting right.
This patch promotes the vector to a type that supports a native shift first and only allows inserting into the msb of a native sized shift.
I've constructed this in a way that doesn't do the promotion if we're going to fallback to using a xmm/ymm/zmm shuffle. I think I have a plan to remove the shuffle fall back entirely. In which case we this can be simplified, but I wanted to fix the correctness issue first.
llvm-svn: 320081
Most of the code in these routines is for handling extends from vXi1 types. The 512-bit handling for other extends is very much like the AVX2 code. So make the special routines just do vXi1 types and move the other 512-bit handling to the place that handles AVX2.
llvm-svn: 319878
The patch originally broke Chromium (crbug.com/791714) due to its failing to
specify that the new pseudo instructions clobber EFLAGS. This commit fixes
that.
> Summary: This strengthens the guard and matches MSVC.
>
> Reviewers: hans, etienneb
>
> Subscribers: hiraditya, JDevlieghere, vlad.tsyrklevich, llvm-commits
>
> Differential Revision: https://reviews.llvm.org/D40622
llvm-svn: 319824
This patch, together with a matching clang patch (https://reviews.llvm.org/D39719), implements the lowering of X86 kunpack intrinsics to IR.
Differential Revision: https://reviews.llvm.org/D39720
Change-Id: I4088d9428478f9457f6afddc90bd3d66b3daf0a1
llvm-svn: 319778
Previously we used a wider element type and truncated. But its more efficient to keep the element type and drop unused elements.
If BWI isn't supported and we have a i16 or i8 type, we'll extend it to be i32 and still use a truncate.
llvm-svn: 319740
Previously we used a wider element type and truncated. But its more efficient to keep the element type and drop unused elements.
If BWI isn't supported and we have a i16 or i8 type, we'll extend it to be i32 and still use a truncate.
llvm-svn: 319728
The getConstant function can take care of creating the APInt internally.
getZeroVector will take care of using the correct type for the build vector to avoid re-lowering.
The test change here is because execution domain constraints apparently pass through undef inputs of a zeroing xor. So the different ordering of register allocation here caused the dependency to change.
llvm-svn: 319725
Move the AVX512 code out of LowerAVXExtend. LowerAVXExtend has two callers but one of them pre-checks for AVX-512 so the code is only live from the other caller. So move the AVX-512 checks up to that caller for symmetry.
Move all of the i1 input type code in Lower_AVX512ZeroExend together.
llvm-svn: 319724
These instructions can be used by widening to 512-bits and extracting back to 128/256. We do similar to several other instructions already.
llvm-svn: 319641
We already do this as a DAG combine. The version during lowering can only trigger if known bits changes something that improves known bits analysis. But this means we should be improving known bits analysis to work on the unlowered form instead.
llvm-svn: 319640
The default legalization for v2i32 is promotion to v2i64. This results in a gather that reads 64-bit elements rather than 32. If one of the elements is near a page boundary this can cause an illegal access that can fault.
We also miscalculate the scale for the gather which is an even worse problem, but we probably could have found a separate way to fix that.
llvm-svn: 319521
As part of the unification of the debug format and the MIR format, avoid
printing "vreg" for virtual registers (which is one of the current MIR
possibilities).
Basically:
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E "s/%vreg([0-9]+)/%\1/g"
* grep -nr '%vreg' . and fix if needed
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E "s/ vreg([0-9]+)/ %\1/g"
* grep -nr 'vreg[0-9]\+' . and fix if needed
Differential Revision: https://reviews.llvm.org/D40420
llvm-svn: 319427
Normal type legalization will widen everything. This requires forcing 0s into the mask register. We can instead choose the form that only reads 2 elements without zeroing the mask.
llvm-svn: 319406
Previously we had an isel pattern to add the truncate. Instead use Promote to add the truncate to the DAG before isel.
The Promote legalization code had to be updated to prevent an infinite loop if promotion took multiple steps because it wasn't remembering the previously tried value.
llvm-svn: 319259
These lines all exist identically either under SSE2, AVX2 or AVX512. Given that VLX implies all of those, these aren't providing anything new.
llvm-svn: 319124
Which VTs are considered simple is determined by the superset of the legal types of all targets in LLVM. If we're looking at VTs that are going to be split down to 512-bits we should allow any VT not just simple ones since the simple list changes over time as new targets are added.
llvm-svn: 319110
We don't do this for narrow vectors under AVX or SSE features. We also don't set them to Expand like we do for many vectors op. Nor does TargetLoweringBase.cpp. This leads me to believe these default to Legal.
llvm-svn: 319103
I don't have a good test case for this at the moment. I was playing around with a change in legalizing and triggered this code to produce a PSHUFD with sse1 only.
llvm-svn: 319066
Similar for vXi16/vXi8 with BWI.
Any vector larger than 512 bits will be split to 512 bits during legalization. But without this we will fold sexts with them before that making it difficult to recover leading to scalarization.
llvm-svn: 319059
Summary:
These instructions zero the non-scalar part of the lower 128-bits which makes them different than the FMA3 instructions which pass through the non-scalar part of the lower 128-bits.
I've only added fmadd because we should be able to derive all other variants using operand negation in the intrinsic header like we do for AVX512.
I think there are still some missed negate folding opportunities with the FMA4 instructions in light of this behavior difference that I hadn't noticed before.
I've split the tests so that we can use different intrinsics for scalar testing between the two. I just copied the tests split the RUN lines and changed out the scalar intrinsics.
fma4-fneg-combine.ll is a new test to make sure we negate the fma4 intrinsics correctly though there are a couple TODOs in it.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D39851
llvm-svn: 318984
v4i32 isn't a legal type with sse1 only and would end up getting scalarized otherwise.
This isn't completely ideal as it doesn't handle cases like v8i32 that would get split to v4i32. But it at least helps with code written using the clang intrinsic header.
llvm-svn: 318967
This optimization can occur after type legalization and emit a vselect with v4i32 type. But that type is not legal with sse1. This ultimately gets scalarized by the second type legalization that runs after vector op legalization, but that's really intended to handle the scalar types that might be introduced by legalizing vector ops.
For now just stop this from happening by disabling the optimization with sse1.
llvm-svn: 318965
(V)PHMINPOSUW determines the UMIN element in an v8i16 input, with suitable bit flipping it can also be used for SMAX/SMIN/UMAX cases as well.
This patch matches vXi16 SMAX/SMIN/UMAX/UMIN horizontal reductions and reduces the input down to a v8i16 vector before calling (V)PHMINPOSUW.
A later patch will use this for v16i8 reductions as well (PR32841).
Differential Revision: https://reviews.llvm.org/D39729
llvm-svn: 318917
This makes the fact that X86 needs an explicit mask output not part of the type constraint for the ISD::MSCATTER.
This also gives the X86ISD::MGATHER/MSCATTER nodes a common base class simplifying the address selection code in X86ISelDAGToDAG.cpp
llvm-svn: 318823
Now we consistently represent the mask result without relying on isel ignoring it.
We now have a more general SDNode and type constraints to represent these nodes in isel patterns. This allows us to present both both vXi1 and XMM/YMM mask types with a single set of constraints.
llvm-svn: 318821
All these headers already depend on CodeGen headers so moving them into
CodeGen fixes the layering (since CodeGen depends on Target, not the
other way around).
llvm-svn: 318490
Only do this pre-legalize in case we're using the sign extend to legalize for KNL.
This recovers all of the tests that changed when I stopped SelectionDAGBuilder from deleting sign extends.
There's more work that could be done here particularly to fix the i8->i64 test case that experienced split.
llvm-svn: 318468
The wider element type will normally cause legalize to try to split and scalarize the gather/scatter, but we can't handle that. Instead, truncate the index early so the gather/scatter node is insulated from the legalization.
This really shouldn't happen in practice since InstCombine will normalize index types to the same size as pointers.
llvm-svn: 318452
This allows us to remove extra extend creation during lowering and more accurately reflects the semantics of the instruction.
While there add an extra output VT to X86 masked gather node to better match the isel pattern predicate. Currently we're exploiting the fact that the isel table doesn't count how many output results a node actually has if the result type of any can be inferred from the first result and the type constraints defined in tablegen. I think we might ultimately want to lower all MGATHER/MSCATTER to an X86ISD node with the extra mask result and stop relying on this hole in the isel checking.
llvm-svn: 318278
The VRNDSCALE instructions implement a superset of the (V)ROUND instructions. They are equivalent if the upper 4-bits of the immediate are 0.
This patch lowers the legacy intrinsics to the VRNDSCALE ISD node and masks the upper bits of the immediate to 0. This allows us to take advantage of the larger register encoding space.
We should maybe consider converting VRNDSCALE back to VROUND in the EVEX to VEX pass if the extended registers are not being used.
I notice some load folding opportunities being missed for the VRNDSCALESS/SD instructions that I'll try to fix in future patches.
llvm-svn: 318008
I want to reuse the VRNDSCALE node for the legacy SSE rounding intrinsics so that those intrinsics can use EVEX instructions. All of these nodes share tablegen multiclasses so I split them all so that they all remain similar in their implementations.
llvm-svn: 318007
matchBinOpReduction currently matches against a single opcode, but we already have a case where we repeat calls to try to match against AND/OR and I'll be shortly adding another case for SMAX/SMIN/UMAX/UMIN (D39729).
This NFCI patch alters matchBinOpReduction to try and pattern match against any of the provided list of candidate bin ops at once to save time.
Differential Revision: https://reviews.llvm.org/D39726
llvm-svn: 317985
r317453 added new ISD nodes without rounding modes that were added to an existing if/else chain. But all the previous nodes handled there included a rounding mode. The final code after this if/else chain expected an extra operand that isn't present for the new nodes.
llvm-svn: 317748
The EVEX to VEX pass is already assuming this is true under AVX512VL. We had special patterns to use zmm instructions if VLX and F16C weren't available.
Instead just make AVX512 imply F16C to make the EVEX to VEX behavior explicitly legal and remove the extra patterns.
All known CPUs with AVX512 have F16C so this should safe for now.
llvm-svn: 317521
We still early-out for X86ISD::PEXTRW/X86ISD::PEXTRB so no actual change in behaviour, but it'll make it easier to add support in a future patch.
llvm-svn: 317485
combineExtractWithShuffle can handle more complex shuffles/bitcasts than we can with the equivalent code in XFormVExtractWithShuffleIntoLoad.
Mainly a compile time improvement now (combineExtractWithShuffle combines will have always failed late on inside XFormVExtractWithShuffleIntoLoad), and will let us merge combineExtractVectorElt_SSE in a future commit.
llvm-svn: 317481
Added TESTM and TESTNM to the list of instructions that already zeroing unused upper bits
and does not need the redundant shift left and shift right instructions afterwards.
Added a pattern for TESTM and TESTNM in iselLowering, so now icmp(neq,and(X,Y), 0) goes folds into TESTM
and icmp(eq,and(X,Y), 0) goes folds into TESTNM
This commit is a preparation for lowering the test and testn X86 intrinsics to IR.
Differential Revision: https://reviews.llvm.org/D38732
llvm-svn: 317465
Summary:
Try to lower a BUILD_VECTOR composed of extract-extract chains that can be
reasoned to be a permutation of a vector by indices in a non-constant vector.
We saw this pattern created by ISPC, which resolts to creating it due to the
requirement that shufflevector's mask operand be a *constant* vector.
I didn't check this but we could possibly use this pattern for lowering the X86 permute
C-instrinsics instead of llvm.x86 instrinsics.
This change can be followed by more improvements:
1. Handle vectors with undef elements.
2. Utilize pshufb and zero-mask-blending to support more effiecient
construction of vectors with constant-0 elements.
3. Use smaller-element vectors of same width, and "interpolate" the indices,
when no native operation available.
Reviewers: RKSimon, craig.topper
Reviewed By: RKSimon
Subscribers: chandlerc, DavidKreitzer
Differential Revision: https://reviews.llvm.org/D39126
llvm-svn: 317463
This patch, together with a matching clang patch (https://reviews.llvm.org/D38683), implements the lowering of X86 broadcastm intrinsics to IR.
Differential Revision: https://reviews.llvm.org/D38684
Change-Id: I709ac0b34641095397e994c8ff7e15d1315b3540
llvm-svn: 317458
Next step is to use them for the legacy FMA scalar intrinsics as well. This will enable the legacy intrinsics to use EVEX encoded opcodes and the extended registers.
llvm-svn: 317453
Summary:
AVX512 added RCP14 and RSQRT instructions which improve accuracy over the legacy RCP and RSQRT instruction, but not enough accuracy to remove the need for a Newton Raphson refinement.
Currently we use these new instructions for the legacy packed SSE instrinics, but not the scalar instrinsics. And we use it for fast math optimization of division and reciprocal sqrt.
I think switching the legacy instrinsics maybe surprising to the user since it changes the answer based on which processor you're using regardless of any fastmath settings. It's also weird that we did something different between scalar and packed.
As far at the reciprocal estimation, I think it creates unnecessary deltas in our output behavior (and prevents EVEX->VEX). A little playing around with gcc and icc and godbolt suggest they don't change which instructions they use here.
This patch adds new X86ISD nodes for the RCP14/RSQRT14 and uses those for the new intrinsics. Leaving the old intrinsics to use the old instructions.
Going forward I think our focus should be on
-Supporting 512-bit vectors, which will have to use the RCP14/RSQRT14.
-Using RSQRT28/RCP28 to remove the Newton Raphson step on processors with AVX512ER
-Supporting double precision.
Reviewers: zvi, DavidKreitzer, RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D39583
llvm-svn: 317413
This allows masked operations to be used and allows the register allocator to use YMM16-31 if necessary.
As a follow up I'll look into teaching EVEX->VEX how to turn this back into PERM2X128 if any of the additional features don't work out.
llvm-svn: 317403
Similar to the existing code to lower to PACKSS, we can use PACKUS if the input vector's leading zero bits extend all the way to the packed/truncated value.
We have to account for pre-SSE41 targets not supporting PACKUSDW
llvm-svn: 317315
Similar to the existing code to lower to PACKSS, we can use PACKUS if the input vector's leading zero bits extend all the way to the packed/truncated value.
We have to account for pre-SSE41 targets not supporting PACKUSDW
llvm-svn: 317128
So far we've only been using PACKSS truncations with 'all-bits or zero-bits' patterns (vector comparison results etc.). When really we can safely use it for any case as long as the number of sign bits reach down to the last 16-bits (or 8-bits if we're truncating to bytes).
The next steps after this is add the equivalent support for PACKUS and to support packing to sub-128 bit vectors for truncating stores etc.
Differential Revision: https://reviews.llvm.org/D39476
llvm-svn: 317086
Summary:
INC/DEC don't update the carry flag so we need to make sure we don't try to use it.
This patch introduces new X86ISD opcodes for locked INC/DEC. Teaches lowerAtomicArithWithLOCK to emit these nodes if INC/DEC is not slow or the function is being optimized for size. An additional flag is added that allows the INC/DEC to be disabled if the caller determines that the carry flag is being requested.
The test_sub_1_cmp_1_setcc_ugt test is currently showing this bug. The other test case changes are recovering cases that were regressed in r316860.
This should fully fix PR35068 finishing the fix started in r316860.
Reviewers: RKSimon, zvi, spatel
Reviewed By: zvi
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D39411
llvm-svn: 316913
If the carry flag is being used, this transformation isn't safe.
This does prevent some test cases from using DEC now, but I'll try to look into that separately.
Fixes PR35068.
llvm-svn: 316860
This code attempted to say that v8i16/v16i16 VSELECT is legal if BWI and VLX are enabled, but the only way we could reach this point is if the condition was not a vXi1 type. Which means it really wasn't legal.
We don't have any tests that exercise this code. So I'm hoping it wasn't really reachable.
llvm-svn: 316851
If the extend type is 64-bits, emit a 32-bit -> 64-bit extend after the UDIVREM8_ZEXT_HREG/UDIVREM8_SEXT_HREG operation.
This gives a shorter encoding for the second extend in the sext case, and allows us to completely remove the second extend in the zext case.
This also adds known bit and num sign bits support for UDIVREM8_ZEXT_HREG/SDIVREM8_SEXT_HREG.
Differential Revision: https://reviews.llvm.org/D38275
llvm-svn: 316702
Instead of loading (a potential ton of) scalar constants, load those as a vector and then insert into it.
Differential Revision: https://reviews.llvm.org/D38756
llvm-svn: 316685
By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits.
llvm-svn: 316448
Remove AssertZext and instead add PEXTRW/PEXTRB support to computeKnownBitsForTargetNode to simplify instruction selection.
Differential Revision: https://reviews.llvm.org/D39169
llvm-svn: 316336
Summary: __multi3 is not available on x86 (32-bit). Setting lib call name for MULI_128 to nullptr forces DAGTypeLegalizer::ExpandIntRes_MUL to generate instructions for 128-bit multiply instead of a call to an undefined function. This fixes PR20871 though it may be worth looking at why licm and indvars combine to generate 65-bit multiplies in that test.
Patch by Riyaz V Puthiyapurayil
Reviewers: craig.topper, schweitz
Reviewed By: craig.topper, schweitz
Subscribers: RKSimon, llvm-commits
Differential Revision: https://reviews.llvm.org/D38668
llvm-svn: 316254
x86 has its own copy of integer absolute pattern matching to combine directly to a SUB+CMOV.
This patch removes the x86 combine and adds custom lowering support for ISD::ABS instead, allowing us to use the DAGCombiner version.
Additional test cases are already covered by iabs.ll (rL315706 and rL315711).
Differential Revision: https://reviews.llvm.org/D38895
llvm-svn: 316162
Summary:
This was impeding our ability to combine the extending shuffles with other shuffles as you can see from the test changes.
There's one special case that needed to be added to use VZEXT directly for v8i8->v8i64 since the custom lowering requires v64i8.
Reviewers: RKSimon, zvi, delena
Reviewed By: delena
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38714
llvm-svn: 315860
Summary:
It's better to use our shuffle lowering code to handle these than loading an immediate into a k-register.
It really feels like this should be a DAG combine optimization rather than a lowering operation, but that's a problem for another day.
Reviewers: RKSimon, delena, zvi
Reviewed By: delena
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38932
llvm-svn: 315849
If we are applying a byte mask to a value extracted from a shuffle, see if we can combine the mask into shuffle.
Fixes the last issue with PR22415
llvm-svn: 315807
This is particularly important for AVX512VL where we are better able to recognize the VBROADCAST loads to fold with other operations.
For AVX512VL we now use X86ISD::VBROADCAST for all of the patterns and remove the 128-bit X86ISD::VMOVDDUP.
We may be able to use this for AVX1 as well which would allow us to remove more isel patterns.
I also had to add X86ISD::VBROADCAST as a node to call combineShuffle for so that we treat it similar to X86ISD::MOVDDUP.
Differential Revision: https://reviews.llvm.org/D38836
llvm-svn: 315768
Summary: We seem to inconsistently create CMOV nodes some with a Glue result and some without. But I can't find any cases that use the Glue result. So I've tried to remove all the place that did this.
Reviewers: RKSimon, spatel, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38664
llvm-svn: 315686
Legalization of fp128 assumes things that we should have asserts for,
so that's another potential improvement.
Differential Revision: https://reviews.llvm.org/D38771
llvm-svn: 315485
Summary:
On behalf of julia.koval@intel.com
The patch transforms canonical version of unsigned saturation, which is sub(max(a,b),a) or sub(a,min(a,b)) to special psubus insturuction on targets, which support it(8bit and 16bit uints).
umax(a,b) - b -> subus(a,b)
a - umin(a,b) -> subus(a,b)
There is also extra case handled, when right part of sub is 32 bit and can be truncated, using UMIN(this transformation was discussed in https://reviews.llvm.org/D25987).
The example of special case code:
```
void foo(unsigned short *p, int max, int n) {
int i;
unsigned m;
for (i = 0; i < n; i++) {
m = *--p;
*p = (unsigned short)(m >= max ? m-max : 0);
}
}
```
Max in this example is truncated to max_short value, if it is greater than m, or just truncated to 16 bit, if it is not. It is vaid transformation, because if max > max_short, result of the expression will be zero.
Here is the table of types, I try to support, special case items are bold:
| Size | 128 | 256 | 512
| ----- | ----- | ----- | -----
| i8 | v16i8 | v32i8 | v64i8
| i16 | v8i16 | v16i16 | v32i16
| i32 | | **v8i32** | **v16i32**
| i64 | | | **v8i64**
Reviewers: zvi, spatel, DavidKreitzer, RKSimon
Reviewed By: zvi
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37534
llvm-svn: 315237
We believe that despite AMD's documentation, that they really do support all 32 comparision predicates under AVX.
Differential Revision: https://reviews.llvm.org/D38609
llvm-svn: 315201
Return the combined shuffle from combineX86ShufflesRecursively and perform the combineTo in the caller.
Makes it easier for future patches to use this in functions that aren't actually shuffles themselves.
llvm-svn: 315195
Summary:
We currently disable some converting of shuffles to MOVSS/MOVSD during legalization if SSE41 is enabled. But later during shuffle combining we go back to prefering MOVSS/MOVSD.
Additionally we have patterns that look for BLENDIs to detect scalar arithmetic operations. I believe due to the combining using MOVSS/MOVSD these are unnecessary.
Interestingly, we still codegen blend instructions even though lowering/isel emit movss/movsd instructions. Turns out machine CSE commutes them to blend, and then commuting those blends back into blends that are equivalent to the original movss/movsd.
This patch fixes the inconsistency in legalization to prefer MOVSS/MOVSD. The one test change was caused by this change. The problem is that we have integer types and are mostly selecting integer instructions except for the shufps. This shufps forced the execution domain, but the vpblendw couldn't have its domain changed with a naive instruction swap. We could fix this by special casing VPBLENDW based on the immediate to widen the element type.
The rest of the patch is removing all the excess scalar patterns.
Long term we should probably add isel patterns to make MOVSS/MOVSD emit blends directly instead of relying on the double commute. We may also want to consider emitting movss/movsd for optsize. I also wonder if we should still use the VEX encoded blendi instructions even with AVX512. Blends have better throughput, and that may outweigh the register constraint.
Reviewers: RKSimon, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38023
llvm-svn: 315181
Recognise cases when we can merge the shuffles with their horizontal (HADD/HSUB/PACK) instruction inputs.
Replaces an older implementation which performed some of this during lowering, expanding an existing target shuffle combine stage instead.
Differential Revision: https://reviews.llvm.org/D38506
llvm-svn: 315150
The SjLj intrinsics in the X86 backend are intended for use with
SjLj exception handling as well, since SVN r271244.
Differential Revision: https://reviews.llvm.org/D38532
llvm-svn: 315146
The code which lowers BUILD_VECTOR of consecutive loads into a single vector
load doesn't update chains properly. As a result the vector load can be
reordered with the store to the same location.
The current code in EltsFromConsecutiveLoads only updates the chain following
the first load. The fix is to update the chains following all the loads
comprising the vector.
This is a fix for PR10114.
Reviewed By: niravd
Differential Revision: https://reviews.llvm.org/D38547
llvm-svn: 314988
Early out from vector shift by immediates that will exceed eltsize - don't bother making an unnecessary ComputeNumSignBits recursive call.
llvm-svn: 314903
The previous version didn't work if the jump table base address didn't
fit in 32 bit, since it was encoded as an immediate offset. And in case
the jump table is encoded as 32 bit label differences, we need to
load and add them to the table base first.
This solves the first half of the issues mentioned in PR34720.
Also fix some of the errors pointed out by -verify-machineinstrs, by
using GR32_NOSPRegClass.
Differential Revision: https://reviews.llvm.org/D38333
llvm-svn: 314876
If the upper bits of a truncation shuffle patterns have at least the minimum number of sign/zero bits on their inputs then we can safely use PACKSS/PACKUS as shuffles.
Partial fix for https://bugs.llvm.org/show_bug.cgi?id=34773
Differential Revision: https://reviews.llvm.org/D38472
llvm-svn: 314788
This makes sure the LSDA pointer isn't truncated to 32 bit.
Make LowerINTRINSIC_WO_CHAIN a member function instead of a static
function, so that it can use the getGlobalWrapperKind method.
This solves the second half of the issues mentioned in PR34720.
Differential Revision: https://reviews.llvm.org/D38343
llvm-svn: 314767
The refactoring in
"[X86][SSE] Add createPackShuffleMask helper function. NFCI."
resulted in warning when compiling the code (seen in build bots).
This patch restores some types from int to unsigned to avoid
those warnings.
llvm-svn: 314667
Implemented by splitting into two v32i8 mulhu/mulhs and concatenating the results.
Differential Revision: https://reviews.llvm.org/D38307
llvm-svn: 314584
If we have BWI, we can truncate in a much simpler way by using vpmovwb. This even works without VLX by using the wider zmm->ymm truncate with a subvector extract.
Differential Revision: https://reviews.llvm.org/D38375
llvm-svn: 314457
We aren't do any in register extends here so we should be able to just the target independent nodes directly and allow them to be lowered as necessary.
llvm-svn: 314447
Summary: If we have BWI instructions we can widen to v32i16 to do the multiply instead of splitting.
Reviewers: RKSimon, spatel, zvi
Reviewed By: zvi
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38305
llvm-svn: 314432
Previously we were using one of the subvector indices twice. The included test case causes an assert without this change.
Thanks to Simon Pilgrim for catching this.
llvm-svn: 314429
We already have zeroable bits in an APInt. We might as well use that instead of checking for an all zero BUILD_VECTOR.
Differential Revision: https://reviews.llvm.org/D37950
llvm-svn: 314332
In some cases the result psadbw is smaller than the type of the add that started the match. Currently in these cases we are using a smaller add and inserting the result.
If we instead combine the psadbw with zeros and use the full size add we can take advantage of implicit zeroing we get if we emit a narrower move before the add.
In a future patch, I want to make isel aware that the psadbw itself already zeroed the upper bits and remove the move entirely.
Differential Revision: https://reviews.llvm.org/D37453
llvm-svn: 314331
As commented on D37849 and rL313547, AVX1 targets were missing a chance to use vmovmskpd for v4f64/v4i64 results for bool vector bitcasts
llvm-svn: 314293
This is necessary, but not sufficient, for having working SJLJ exception
handling on x86_64.
Differential Revision: https://reviews.llvm.org/D38254
llvm-svn: 314277
The callsite value is already stored indexed from 0 in
the _Unwind_Context struct. When accessed via the functions
_Unwind_GetIP and _Unwind_SetIP, the value is indexed from 1,
but those functions handle the offseting. When reading directly
from the struct here, we shouldn't subtract 1.
This matches the code generated by the ARM target, where SJLJ
exception handling is used by default on iOS.
This makes clang-built object files for 32 bit x86 mingw work when
linked with libgcc/libstdc++.
Differential Revision: https://reviews.llvm.org/D38251
llvm-svn: 314276
The XOP rotations act as ROTL with +ve values and ROTR with -ve values, which means that we can treat them all as ROTL with unsigned modulo. We already check that we're only trying to lower as ROTL for XOP rotations.
Differential Revision: https://reviews.llvm.org/D37949
llvm-svn: 314207
This required changing the ISD opcode for these instructions to have the commutable operands first and the addend last. This way tablegen can autogenerate the additional patterns for us.
llvm-svn: 314083
This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type.
Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases.
Differential Revision: https://reviews.llvm.org/D35320
llvm-svn: 314076
This is a follow-up from D38181 (r314023). We have to put 64-bit
constants into a register using a separate instruction, so we
should try harder to avoid that.
From what I see, we're not likely to encounter this pattern in the
DAG because the upstream setcc combines from this don't (usually?)
produce this pattern. If we fix that, then this will become more
relevant. Since the cost of handling this case is just loosening
the predicate of the existing fold, we might as well do it now.
llvm-svn: 314064
The (non-)obvious win comes from saving 3 bytes by using the 0x83 'and' opcode variant instead of 0x81.
There are also better improvements based on known-bits that allow us to eliminate the mask entirely.
As noted, this could be extended. There are potentially other wins from always shifting first, but doing
that reveals a tangle of problems in other pattern matching. We do this transform generically in
instcombine, but we often have icmp IR that doesn't match that pattern, so we must account for this
in the backend.
Differential Revision: https://reviews.llvm.org/D38181
llvm-svn: 314023
Combine CMOV[i16]<-[SIGN,ZERO,ANY]_EXTEND to [i32,i64] into CMOV[i32,i64].
One example of where it is useful is:
before (20 bytes)
<foo>:
test $0x1,%dil
mov $0x307e,%ax
mov $0xffff,%cx
cmovne %ax,%cx
movzwl %cx,%eax
retq
after (18 bytes)
<foo>:
test $0x1,%dil
mov $0x307e,%ecx
mov $0xffff,%eax
cmovne %ecx,%eax
retq
Reviewers: craig.topper, aaboud, spatel, RKSimon, zvi
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36711
llvm-svn: 313982
The shuffle combining and lowerVectorShuffleAsLanePermuteAndBlend were both still trying to use VPERM2XF128 for unary shuffles when AVX2 is enabled. VPERM2X128 takes two inputs meaning when we use it for a unary shuffle one of those inputs is left undefined creating a false dependency on whatever register gets allocated there.
If we have VPERMQ/PD we should prefer those since they only have a single input.
Differential Revision: https://reviews.llvm.org/D37947
llvm-svn: 313542
The same code appears earlier in the function. This represents an earlier version of what became r313373 that I still had sitting in my local repo.
llvm-svn: 313465
I've moved the test cases from the InstCombine optimizations to the backend to keep the coverage we had there. It covered every possible immediate so I've preserved the resulting shuffle mask for each of those immediates.
llvm-svn: 313450
The early out for AVX2 in lowerV2X128VectorShuffle is positioned in a weird spot below some shuffle mask equivalency checks.
But I think we want to allow VPERMQ for any unary shuffle.
Differential Revision: https://reviews.llvm.org/D37893
llvm-svn: 313373
When handling a v64i1 build vector of constants on 32-bit targets we were creating an illegal i64 constant that we then bitcasted back to v64i1. We need to instead create two 32-bit constants, bitcast them to v32i1 and concat the result. We should also take care to handle the halves being all zeros/ones after the split.
This patch splits the build vector and then recursively lowers the two pieces. This allows us to handle the all ones and all zeros cases with minimal effort. Ideally we'd just do the split and concat, and let lowering get called again on the new nodes, but getNode has special handling for CONCAT_VECTORS that reassembles the pieces back into a single BUILD_VECTOR. Hopefully the two temporary BUILD_VECTORS we had to create to do this that don't get returned don't cause any issues.
Fixes PR34605.
Differential Revision: https://reviews.llvm.org/D37858
llvm-svn: 313366
Load with zero-extend and sign-extend from v2i8 to v2i32 is "Legal" since SSE4.1 and may be performed using PMOVZXBD , PMOVSXBD instructions.
llvm-svn: 313121
Fuchsia's lowest API layer has been renamed from Magenta to Zircon.
In LLVM proper, this is only mentioned in comments.
Patch by Roland McGrath
Differential Revision: https://reviews.llvm.org/D37763
llvm-svn: 313105
The masked store instruction only cares about the sign-bit of each mask element,
so the compare s<0 isn't needed.
As noted in PR11210:
https://bugs.llvm.org/show_bug.cgi?id=11210
...fixing this should allow us to eliminate x86-specific masked store intrinsics in IR.
(Although more testing will be needed to confirm that.)
I filed a bug to track improvements for AVX512:
https://bugs.llvm.org/show_bug.cgi?id=34584
Differential Revision: https://reviews.llvm.org/D37446
llvm-svn: 313089
Recognizing this pattern during DAG combine hides information about the 'and' and the shift from other combines. I think it should be recognized at isel so its as late as possible. But it can't be done with table based isel because you need to be able to look at both immediates. This patch moves it to custom isel in X86ISelDAGToDAG.cpp.
This does break a couple tests in tbm_patterns because we are now emitting an and_flag node or (cmp and, 0) that we dont' recognize yet. We already had this problem for several other TBM patterns so I think this fine and we can address of them together.
I've also fixed a bug where the combine to BEXTR was preventing us from using a trick of zero extending AH to handle extracts of bits 15:8. We might still want to use BEXTR if it enables load folding. But honestly I hope we narrowed the load instead before got to isel.
I think we should probably also support matching BEXTR from (srl/srl (and mask << C), C). But that should be a different patch.
Differential Revision: https://reviews.llvm.org/D37592
llvm-svn: 313054
Summary:
r275950 added support for turning (trunc (X >> N) to i1) into BT(X, N). But that's no longer necessary now that i1 isn't legal.
This patch removes the support for that, but preserves some of the refactorings done in that commit.
Reviewers: guyblank, RKSimon, spatel, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37673
llvm-svn: 312925
Helps improve combineLogicBlendIntoPBLENDV support by allowing us to peek into through PACKSS truncations of vector comparison results.
Differential Revision: https://reviews.llvm.org/D37680
llvm-svn: 312916
First step towards making it possible to use the shuffle combines for cases where we don't want to call DCI.CombineTo() with the result.
llvm-svn: 312884
Summary:
Just because INC/DEC is a little slow on some processors doesn't mean we shouldn't prefer it when optimizing for size.
This appears to match gcc behavior.
Reviewers: chandlerc, zvi, RKSimon, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37177
llvm-svn: 312866
Summary:
Add patterns for
fptoui <16 x float> to <16 x i8>
fptoui <16 x float> to <16 x i16>
Reviewers: igorb, delena, craig.topper
Reviewed By: craig.topper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37505
llvm-svn: 312704
Intrinsic handling is still creating these nodes with 32-bit elements as well. But at least this gets rid of 8 and 16.
Ideally, someday we'll convert the intrinsics to generic vector shuffles and remove the intrinsics.
llvm-svn: 312702
Summary:
Most instructions in AVX work “in-lane”, that is, each source element is applied only to other
elements of the same lane, thus a cross lane permutation is costly and needs more than one instrution.
AVX2 includes instructions to perform any-to-any permutation of words over a 256-bit register
and vectorized table lookup.
This should also Fix PR34369
Differential Revision: https://reviews.llvm.org/D37388
llvm-svn: 312608
In a future patch, I plan to teach isel to use a small vector move with implicit zeroing of the upper elements when it sees the (insert_subvector zero, X, 0) pattern.
llvm-svn: 312448
This is limited to a set of patterns based on the example in PR34111:
https://bugs.llvm.org/show_bug.cgi?id=34111
...but as I was investigating this, I see that horizontal patterns can go wrong in many,
many other ways that would not be handled by this patch. Each data type may even go
different in the DAG after starting with the same basic IR pattern, so even proper IR
canonicalization won't fix it all.
Differential Revision: https://reviews.llvm.org/D37357
llvm-svn: 312379
Prior to this patch we had a DAG combine that tried to bypass an X86ISD::ADD with -1 being added to the carry flag of some previous operation. We would then pass the carry flag directly to user.
But this is only safe if the user is looking for the carry flag and not the zero flag.
So we need to only do this combine in a context where we know what flag the consumer is using.
Fixes PR34381.
Differential Revision: https://reviews.llvm.org/D37317
llvm-svn: 312285
Summary:
This patch adjusts the patterns to make the result type of the broadcast node vXf64/vXi64. Then adds a bitcast to vXi32 after that. Intrinsic lowering was also adjusted to generate this new pattern.
Fixes PR34357
We should probably just drop the intrinsic entirely and use native IR, but I'll leave that for a future patch.
Any idea what instruction we should be lowering the floating point 128-bit result version of this pattern to? There's a 128-bit v2i32 integer broadcast but not an fp one.
Reviewers: aymanmus, zvi, igorb
Reviewed By: aymanmus
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37286
llvm-svn: 312101
EXTRACT_SUBVECTOR was marked Custom solely so we could combine it with BUILD_VECTOR operations to create smaller BUILD_VECTORS during Legalization. But that sort of combining should really be done by the DAG combiner.
This patch adds the last piece of needed supported DAG combine to handle this. Once that's done we can make the EXTRACT_SUBVECTOR operations Legal.
Differential Revision: https://reviews.llvm.org/D37197
llvm-svn: 311893
We used to do a late DAG combine to move the bitcasts out of the way, but I'm starting to think that it's better to canonicalize extract_subvector's type to match the type of its input. I've seen some cases where we've formed two different extract_subvector from the same node where one had a bitcast and the other didn't.
Add some more test cases to ensure we've also got most of the zero masking covered too.
llvm-svn: 311837
The comment for this code indicated that it should work similar to our
handling of add lowering above: if we see uses of an instruction other
than flag usage and store usage, it tries to avoid the specialized
X86ISD::* nodes that are designed for flag+op modeling and emits an
explicit test.
Problem is, only the add case actually did this. In all the other cases,
the logic was incomplete and inverted. Any time the value was used by
a store, we bailed on the specialized X86ISD node. All of this appears
to have been historical where we had different logic here. =/
Turns out, we have quite a few patterns designed around these nodes. We
should actually form them. I fixed the code to match what we do for add,
and it has quite a positive effect just within some of our test cases.
The only thing close to a regression I see is using:
notl %r
testl %r, %r
instead of:
xorl -1, %r
But we can add a pattern or something to fold that back out. The
improvements seem more than worth this.
I've also worked with Craig to update the comments to no longer be
actively contradicted by the code. =[ Some of this still remains
a mystery to both Craig and myself, but this seems like a large step in
the direction of consistency and slightly more accurate comments.
Many thanks to Craig for help figuring out this nasty stuff.
Differential Revision: https://reviews.llvm.org/D37096
llvm-svn: 311737
This goes back to a discussion about IR canonicalization. We'd like to preserve and convert
more IR to 'select' than we currently do because that's likely the best choice in IR:
http://lists.llvm.org/pipermail/llvm-dev/2016-September/105335.html
...but that's often not true for codegen, so we need to account for this pattern coming in
to the backend and transform it to better DAG ops.
Steps in this patch:
1. Add an EVT param to the existing convertSelectOfConstantsToMath() TLI hook to more finely
enable this transform. Other targets will probably want that anyway to distinguish scalars
from vectors. We're using that here to exclude AVX512 targets, but it may not be necessary.
2. Convert a vselect to ext+add. This eliminates a constant load/materialization, and the
vector ext is often free.
Implementing a more general fold using xor+and can be a follow-up for targets that don't have
a legal vselect. It's also possible that we can remove the TLI hook for the special case fold
implemented here because we're eliminating a constant, but it needs to be tested on other
targets.
Differential Revision: https://reviews.llvm.org/D36840
llvm-svn: 311731
This patch is intended to enable the use of basic double letter constraints used in GCC extended inline asm {Yi Y2 Yz Y0 Ym Yt}.
Supersedes D35204
Clang counterpart: D36371
Differential Revision: https://reviews.llvm.org/D36369
llvm-svn: 311644
There are no 512-bit blend instructions so we shouldn't create SHRUNKBLEND for them.
On a side note, it looks like there may be a missed opportunity for constant folding TESTM when LHS and RHS are equal.
This fixes PR34139.
Differential Revision: https://reviews.llvm.org/D36992
llvm-svn: 311572
There's no reason to have a target specific node with the same semantics as a target independent opcode.
This should simplify D36335 so that it doesn't need to touch X86ISelDAGToDAG.cpp
Differential Revision: https://reviews.llvm.org/D36983
llvm-svn: 311568
This partially reverts r311429 in favor of making ISD::isConstantSplatVector do something not confusing. Turns out the only other user of it was also having to deal with the weird property of it returning a smaller size.
So rather than continue to deal with this quirk everywhere, just make the interface do something sane.
Differential Revision: https://reviews.llvm.org/D37039
llvm-svn: 311510
ISD::isConstantSplatVector can shrink to the smallest splat width. But we don't check the size of the resulting APInt at all. This can cause us to misinterpret the results.
This patch just adds a flag to prevent the APInt from changing width.
Fixes PR34271.
Differential Revision: https://reviews.llvm.org/D36996
llvm-svn: 311429
rather than doing a separate comparison.
This both saves an explicit comparision and avoids the use of `xadd`
which introduces register constraints and other challenges to the
generated code.
The motivating case is from atomic reference counts where `1` is the
sentinel rather than `0` for whatever reason. This can and should be
lowered efficiently on x86 by just using a different flag, however the
x86 code only handled the `0` case.
There remains some further opportunities here that are currently hidden
due to canonicalization. I've included test cases that show these and
FIXMEs. However, I don't at the moment have any production use cases and
they seem substantially harder to address.
Differential Revision: https://reviews.llvm.org/D36945
llvm-svn: 311317
According to the X86ISelLowering.h, UMUL results are low, high, and flags. But this place was treating result 1 or 2 as flags.
Differential Revision: https://reviews.llvm.org/D36654
llvm-svn: 310846
Summary:
The flag result is an i32 type. But its only really used for connectivity. I don't think anything even assumes a particular format. We don't ever do any real operations on it. So known bits don't help us optimize anything.
My main motivation is that the UMUL behavior is actually wrong. I was going to fix this in D36654, but then realized there was just no reason for it to be here.
Reviewers: RKSimon, zvi, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36657
llvm-svn: 310845
Previously it would not return true for extracting either of the upper quarters of a 512-bit registers.
For mask registers we support extracting anything from index 0. And otherwise we only support extracting the upper half of a register.
Differential Revision: https://reviews.llvm.org/D36638
llvm-svn: 310794
Summary:
Without the SrcVT its hard to know what is really being asked for. For example if your target has 128, 256, and 512 bit vectors. Maybe extracting 128 from 256 is cheap, but maybe extracting 128 from 512 is not.
For x86 we do support extracting a quarter of a 512-bit register. But for i1 vectors we don't have isel patterns for extracting arbitrary pieces. So we need this to have a correct implementation of isExtractSubvectorCheap for mask vectors.
Reviewers: RKSimon, zvi, efriedma
Reviewed By: RKSimon
Subscribers: aemerson, javed.absar, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D36649
llvm-svn: 310793
Add an X86 combine for TESTM when one of the operands is a BUILD_VECTOR(0,0,...).
TESTM op0, BUILD_VECTOR(0,0,...) -> BUILD_VECTOR(0,0,...)
TESTM BUILD_VECTOR(0,0,...), op1 -> BUILD_VECTOR(0,0,...)
Differential Revision:
https://reviews.llvm.org/D36536
llvm-svn: 310787
Summary:
Previously we were creating the flag result with MVT::Other which is interpretted as a Chain node. If we used a memory form of the instruction we would end up with a copyToReg that consumed the chain result of the adcx instruction instead of the flag result.
Pretty sure we should be using MVT::i32 here, that's what we do other places we create these node types.
We should probably consider this for 5.0 as well.
Reviewers: RKSimon, zvi, spatel
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36645
llvm-svn: 310784
Summary:
Previously we would use these instructions if sse was disabled and fastmath was enabled.
As mentioned in D28335, this is a bad idea.
Reviewers: efriedma, scanon, DavidKreitzer
Reviewed By: DavidKreitzer
Subscribers: zvi, llvm-commits
Differential Revision: https://reviews.llvm.org/D36344
llvm-svn: 310762
Summary:
This autoupgrades most of the broadcast intrinsics. They've been unused in clang for some time.
This leaves the 32x2 intrinsics because they are still used in clang.
Reviewers: RKSimon, zvi, igorb
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36606
llvm-svn: 310725
Summary:
This teaches 512-bit shuffles to detect unused halfs in order to reduce shuffle size.
We may need to refine the 512-bit exit point. I couldn't remember if we had good cross lane shuffles for 8/16 bit with AVX-512 or not.
I believe this is step towards being able to handle D36454 without a special case.
From here we need to improve our ability to combine extract_subvector with insert_subvector and other extract_subvectors. And we need to support narrowing binary operations where we don't demand all elements. This may be improvements to DAGCombiner::narrowExtractedVectorBinOp(by recognizing an insert_subvector in addition to concat) or we may need a target specific combiner.
Reviewers: RKSimon, zvi, delena, jbhateja
Reviewed By: RKSimon, jbhateja
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36601
llvm-svn: 310724
The previous rev (r310208) failed to account for overflow when subtracting the
constants to see if they're suitable for shift/lea. This version add a check
for that and more test were added in r310490.
We can convert any select-of-constants to math ops:
http://rise4fun.com/Alive/d7d
For this patch, I'm enhancing an existing x86 transform that uses fake multiplies
(they always become shl/lea) to avoid cmov or branching. The current code misses
cases where we have a negative constant and a positive constant, so this is just
trying to plug that hole.
The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start
with a select in IR, create a select DAG node, convert it into a sext, convert it
back into a select, and then lower it to sext machine code.
Some notes about the test diffs:
1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR.
2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. We could avoid the
push/pop in some cases if we used 'movzbl %al' instead of an xor on a different reg? That's a
post-DAG problem though.
3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if
that's a regression, but those would always be nearly equivalent.
4. pr22338.ll and sext-i1.ll - These tests have undef operands, so we don't actually care about these diffs.
5. sbb.ll - This shows a win for what is likely a common case: choose -1 or 0.
6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again.
7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops.
Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass.
Differential Revision: https://reviews.llvm.org/D35340
llvm-svn: 310717
Move store merge to happen after intrinsic lowering to allow lowered
stores to be merged.
Some regressions due in MergeConsecutiveStores to missing
insert_subvector that are addressed in follow up patch.
Reviewers: craig.topper, efriedma, RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D34559
llvm-svn: 310710
Summary:
Preserve chain dependecies between old and new loads constructed to
prevent loads from reordering below later stores.
Fixes PR34088.
Reviewers: craig.topper, spatel, RKSimon, efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36528
llvm-svn: 310604
a legal cond operand.
When scalarizing the result of a vselect, the legalizer currently expects
to already have scalarized the operands. While this is true for the true/false
operands (which have the same type as the result), it is not case for the
condition operand. On X86 AVX512, v1i1 is legal - this leads to operations such
as '< N x type> vselect < N x i1> < N x type> < N x type>' where < N x type > is
illegal to hit an assertion during the scalarization.
The handling is similar to r205625.
This also exposes the fact that (v1i1 extract_subvector) should be legal
and selectable on AVX512 - We do this by custom lowering to vector_extract_elt.
This still leaves us in some cases with redundant dag nodes which will be
combined in a separate soon to come patch.
This fixes pr33349.
Differential revision: https://reviews.llvm.org/D36511
llvm-svn: 310552
We can convert any select-of-constants to math ops:
http://rise4fun.com/Alive/d7d
For this patch, I'm enhancing an existing x86 transform that uses fake multiplies
(they always become shl/lea) to avoid cmov or branching. The current code misses
cases where we have a negative constant and a positive constant, so this is just
trying to plug that hole.
The DAGCombiner diff prevents us from hitting a terrible inefficiency: we can start
with a select in IR, create a select DAG node, convert it into a sext, convert it
back into a select, and then lower it to sext machine code.
Some notes about the test diffs:
1. 2010-08-04-MaskedSignedCompare.ll - We were creating control flow that didn't exist in the IR.
2. memcmp.ll - Choose -1 or 1 is the case that got me looking at this again. I
think we could avoid the push/pop in some cases if we used 'movzbl %al' instead of an xor on
a different reg? That's a post-DAG problem though.
3. mul-constant-result.ll - The trade-off between sbb+not vs. setne+neg could be addressed if
that's a regression, but I think those would always be nearly equivalent.
4. pr22338.ll and sext-i1.ll - These tests have undef operands, so I don't think we actually care about these diffs.
5. sbb.ll - This shows a win for what I think is a common case: choose -1 or 0.
6. select.ll - There's another borderline case here: cmp+sbb+or vs. test+set+lea? Also, sbb+not vs. setae+neg shows up again.
7. select_const.ll - These are motivating cases for the enhancement; replace cmov with cheaper ops.
Assembly differences between movzbl and xor to avoid a partial reg stall are caused later by the X86 Fixup SetCC pass.
Differential Revision: https://reviews.llvm.org/D35340
llvm-svn: 310208
This patch is in 2 parts:
1 - replace combineBT's use of SimplifyDemandedBits (hasOneUse only) with SelectionDAG::GetDemandedBits to more aggressively determine the lower bits used by BT.
2 - update SelectionDAG::GetDemandedBits to support ANY_EXTEND - if the demanded bits are only in the non-extended portion, then peek through and demand from the source value and then ANY_EXTEND that if we found a match.
Differential Revision: https://reviews.llvm.org/D35896
llvm-svn: 309486
The X86 tail call eligibility logic was correct when it was written, but
the addition of inalloca and argument copy elision broke its
assumptions. It was assuming that fixed stack objects were immutable.
Currently, we aim to emit a tail call if no arguments have to be
re-arranged in memory. This code would trace the outgoing argument
values back to check if they are loads from an incoming stack object.
If the stack argument is immutable, then we won't need to store it back
to the stack when we tail call.
Fortunately, stack objects track their mutability, so we can just make
the obvious check to fix the bug.
This was http://crbug.com/749826
llvm-svn: 309343
Like r309323, X86 had a typo where it passed the wrong flags to TLO.
Found by inspection; I haven't been able to tickle this into having
observable behavior. I don't think it does, given that X86 doesn't have
custom demanded bits logic, and the generic logic doesn't have a lot of
exposure to illegal constructs.
llvm-svn: 309325
Assign all concat elements to zero and then just replace the first element, instead of setting them all to null and copying everything in.
llvm-svn: 309261
Changing mask argument type from const SmallVectorImpl<int>& to
ArrayRef<int>.
This came up in D35700 where a mask is received as an ArrayRef<int> and
we want to pass it to TargetLowering::isShuffleMaskLegal().
Also saves a few lines of code.
llvm-svn: 309085
splitting patch D34601 into two part. This part changes the location of two functions.
The second part will be based on that patch. This was requested by @RKSimon.
Reviewers:
1. dorit
2. Farhana
3. RKSimon
4. guyblank
5. DavidKreitzer
llvm-svn: 309084
D35067/rL308322 attempted to support up to 4 load pairs for memcmp inlining which resulted in regressions for some optimized libc memcmp implementations (PR33914).
Until we can match these more optimal cases, this patch reduces the memcmp expansion to a maximum of 2 load pairs (which matches what we do for -Os).
This patch should be considered for the 5.0.0 release branch as well
Differential Revision: https://reviews.llvm.org/D35830
llvm-svn: 308986
This patch makes LSR generate better code for SystemZ in the cases of memory
intrinsics, Load->Store pairs or comparison of immediate with memory.
In order to achieve this, the following common code changes were made:
* New TTI hook: LSRWithInstrQueries(), which defaults to false. Controls if
LSR should do instruction-based addressing evaluations by calling
isLegalAddressingMode() with the Instruction pointers.
* In LoopStrengthReduce: handle address operands of memset, memmove and memcpy
as address uses, and call isFoldableMemAccessOffset() for any LSRUse::Address,
not just loads or stores.
SystemZ changes:
* isLSRCostLess() implemented with Insns first, and without ImmCost.
* New function supportedAddressingMode() that is a helper for TTI methods
looking at Instructions passed via pointers.
Review: Ulrich Weigand, Quentin Colombet
https://reviews.llvm.org/D35262https://reviews.llvm.org/D35049
llvm-svn: 308729
Currently we only support (i32 bitcast(v32i1)) using the AVX2 VPMOVMSKB ymm instruction.
This patch adds support for splitting pre-AVX2 targets into 2 x (V)PMOVMSKB xmm instructions and merging the integer results.
In future we could probably generalize this to handle more cases.
Differential Revision: https://reviews.llvm.org/D35303
llvm-svn: 308723
It should be a win to avoid going out to the system lib for all small memcmp() calls using scalar ops. For x86 32-bit, this means most everything up to 16 bytes. For 64-bit, that doubles because we can do 8-byte loads.
Notes:
Reduced from 4 to 2 loads for -Os behavior, which might not be optimal in all cases. It's effectively a question of how much do we trust the system implementation. Linux and macOS (and Windows I assume, but did not test) have optimized memcmp() code for x86, so it's probably not bad either way? PPC is using 8/4 for defaults on these. We do not expand at all for -Oz.
There are still potential improvements to make for the CGP expansion IR and/or lowering such as avoiding select-of-constants (D34904) and not doing zexts to the max load type before doing a compare.
We have special-case SSE/AVX codegen for (memcmp(x, y, 16/32) == 0) that will no longer be produced after this patch. I've shown the experimental justification for that change in PR33329:
https://bugs.llvm.org/show_bug.cgi?id=33329#c12
TLDR: While the vector code is a likely winner, we can't guarantee that it's a winner in all cases on all CPUs, so I'm willing to sacrifice it for the greater good of expanding all small memcmp(). If we want to resurrect that codegen, it can be done by adjusting the CGP params or poking a hole to let those fall-through the CGP expansion.
Committed on behalf of Sanjay Patel
Differential Revision: https://reviews.llvm.org/D35067
llvm-svn: 308322
This isn't legal code, but we shouldn't crash on it. Now we just don't convert the gather intrinsic if the scale isn't constant and let it go through to isel where we'll report an isel failure.
Fixes PR33772.
llvm-svn: 308267
Rename the enum value from X86_64_Win64 to plain Win64.
The symbol exposed in the textual IR is changed from 'x86_64_win64cc'
to 'win64cc', but the numeric value is kept, keeping support for
old bitcode.
Differential Revision: https://reviews.llvm.org/D34474
llvm-svn: 308208
OpenCL 2.0 introduces the notion of memory scopes in atomic operations to
global and local memory. These scopes restrict how synchronization is
achieved, which can result in improved performance.
This change extends existing notion of synchronization scopes in LLVM to
support arbitrary scopes expressed as target-specific strings, in addition to
the already defined scopes (single thread, system).
The LLVM IR and MIR syntax for expressing synchronization scopes has changed
to use *syncscope("<scope>")*, where <scope> can be "singlethread" (this
replaces *singlethread* keyword), or a target-specific name. As before, if
the scope is not specified, it defaults to CrossThread/System scope.
Implementation details:
- Mapping from synchronization scope name/string to synchronization scope id
is stored in LLVM context;
- CrossThread/System and SingleThread scopes are pre-defined to efficiently
check for known scopes without comparing strings;
- Synchronization scope names are stored in SYNC_SCOPE_NAMES_BLOCK in
the bitcode.
Differential Revision: https://reviews.llvm.org/D21723
llvm-svn: 307722
x86 scalar select-of-constants (Cond ? C1 : C2) combining/lowering is a mess
with missing optimizations. We handle some patterns, but miss logical variants.
To clean that up, we should convert all select-of-constants to logic/math and
enhance the combining for the expected patterns from that. Selecting 0 or -1
needs extra attention to produce the optimal code as shown here.
Attempt to verify that all of these IR forms are logically equivalent:
http://rise4fun.com/Alive/plxs
Earlier steps in this series:
rL306040
rL306072
rL307404 (D34652)
As acknowledged in the earlier review, there's a possibility that some Intel
uarch would prefer to produce an xor to clear the fake register operand with
sbb %eax, %eax. This will likely need to be addressed in a separate pass.
llvm-svn: 307471
x86 scalar select-of-constants (Cond ? C1 : C2) combining/lowering is a mess
with missing optimizations. We handle some patterns, but miss logical variants.
To clean that up, we should convert all select-of-constants to logic/math and
enhance the combining for the expected patterns from that. DAGCombiner already
has the foundation to allow the transforms, so we just need to fill in the holes
for x86 math op lowering. Selecting 0 or -1 needs extra attention to produce the
optimal code as shown here.
Attempt to verify that all of these IR forms are logically equivalent:
http://rise4fun.com/Alive/plxs
Earlier steps in this series:
rL306040
rL306072
Differential Revision: https://reviews.llvm.org/D34652
llvm-svn: 307404
We are combining shuffles to bit shifts before unary permutes, which means we can't fold loads plus the destination register is destructive
llvm-svn: 306978
We are combining shuffles to bit shifts before unary permutes, which means we can't fold loads plus the destination register is destructive
The 32-bit shuffles are a bit tricky and will be dealt with in a later patch
llvm-svn: 306977
[X86][AVX512] Improve lowering of AVX512 compare intrinsics (remove redundant shift left+right instructions).
AVX512 compare instructions return v*i1 types.
In cases where the number of elements in the returned value are less than 8, clang adds zeroes to get a mask of v8i1 type.
Later on it's replaced with CONCAT_VECTORS, which then is lowered to many DAG nodes including insert/extract element and shift right/left nodes.
The fact that AVX512 compare instructions put the result in a k register and zeroes all its upper bits allows us to remove the extra nodes simply by copying the result to the required register class.
When lowering, identify these cases and transform them into an INSERT_SUBVECTOR node (marked legal), then catch this pattern in instructions selection phase and transform it into one avx512 cmp instruction.
Differential Revision: https://reviews.llvm.org/D33188
llvm-svn: 306402
Convert vector increment or decrement to sub/add with an all-ones constant:
add X, <1, 1...> --> sub X, <-1, -1...>
sub X, <1, 1...> --> add X, <-1, -1...>
The all-ones vector constant can be materialized using a pcmpeq instruction that is
commonly recognized as an idiom (has no register dependency), so that's better than
loading a splat 1 constant.
AVX512 uses 'vpternlogd' for 512-bit vectors because there is apparently no better
way to produce 512 one-bits.
The general advantages of this lowering are:
1. pcmpeq has lower latency than a memop on every uarch I looked at in Agner's tables,
so in theory, this could be better for perf, but...
2. That seems unlikely to affect any OOO implementation, and I can't measure any real
perf difference from this transform on Haswell or Jaguar, but...
3. It doesn't look like it from the diffs, but this is an overall size win because we
eliminate 16 - 64 constant bytes in the case of a vector load. If we're broadcasting
a scalar load (which might itself be a bug), then we're replacing a scalar constant
load + broadcast with a single cheap op, so that should always be smaller/better too.
4. This makes the DAG/isel output more consistent - we use pcmpeq already for padd x, -1
and psub x, -1, so we should use that form for +1 too because we can. If there's some
reason to favor a constant load on some CPU, let's make the reverse transform for all
of these cases (either here in the DAG or in a later machine pass).
This should fix:
https://bugs.llvm.org/show_bug.cgi?id=33483
Differential Revision: https://reviews.llvm.org/D34336
llvm-svn: 306289
This is very similar to the transform in:
https://reviews.llvm.org/rL306040
...but in this case, we use cmp X, 1 to set the carry bit as needed.
Again, we can show that all of these are logically equivalent (although
InstCombine currently canonicalizes to a form not seen here), and if
we believe IACA, then this is the smallest/fastest code. Eg, with SNB:
| Num Of | Ports pressure in cycles | |
| Uops | 0 - DV | 1 | 2 - D | 3 - D | 4 | 5 | |
---------------------------------------------------------------------
| 1 | 1.0 | | | | | | | cmp edi, 0x1
| 2 | | 1.0 | | | | 1.0 | CP | sbb eax, eax
The larger motivation is to clean up all select-of-constants combining/lowering
because we're missing some common cases.
llvm-svn: 306072
Our handling of select-of-constants is lumpy in IR (https://reviews.llvm.org/D24480),
lumpy in DAGCombiner, and lumpy in X86ISelLowering. That's why we only had the 'sbb'
codegen in 1 out of the 4 tests. This is a step towards smoothing that out.
First, show that all of these IR forms are equivalent:
http://rise4fun.com/Alive/mx
Second, show that the 'sbb' version is faster/smaller. IACA output for SandyBridge
(later Intel and AMD chips are similar based on Agner's tables):
This is the "obvious" x86 codegen (what gcc appears to produce currently):
| Num Of | Ports pressure in cycles | |
| Uops | 0 - DV | 1 | 2 - D | 3 - D | 4 | 5 | |
---------------------------------------------------------------------
| 1* | | | | | | | | xor eax, eax
| 1 | 1.0 | | | | | | CP | test edi, edi
| 1 | | | | | | 1.0 | CP | setnz al
| 1 | | 1.0 | | | | | CP | neg eax
This is the adc version:
| 1* | | | | | | | | xor eax, eax
| 1 | 1.0 | | | | | | CP | cmp edi, 0x1
| 2 | | 1.0 | | | | 1.0 | CP | adc eax, 0xffffffff
And this is sbb:
| 1 | 1.0 | | | | | | | neg edi
| 2 | | 1.0 | | | | 1.0 | CP | sbb eax, eax
If IACA is trustworthy, then sbb became a single uop in Broadwell, so this will be
clearly better than the alternatives going forward.
llvm-svn: 306040
Masked gather for vector length 2 is lowered incorrectly for element type i32.
The type <2 x i32> was automatically extended to <2 x i64> and we generated VPGATHERQQ instead of VPGATHERQD.
The type <2 x float> is extended to <4 x float>, so there is no bug for this type, but the sequence may be more optimal.
In this patch I'm fixing <2 x i32>bug and optimizing <2 x float> sequence for GATHERs only. The same fix should be done for Scatters as well.
Differential revision: https://reviews.llvm.org/D34343
llvm-svn: 305987
There are a couple of potential improvements as seen in the IR and asm:
1. We're unnecessarily extending to a larger type to compare values.
2. The codegen for (select cond, 1, -1) could avoid a cmov.
(or we could change the order of the compares, so we have a select with 0 operand)
llvm-svn: 305802
Target shuffle combining now supports the matching of INSERT_VECTOR_ELT/PINSRW/PINSRB for merging multiple insertions into shuffles/bitmasks.
llvm-svn: 305788
AVX512 compare instructions return v*i1 types.
In cases where the number of elements in the returned value are less than 8, clang adds zeroes to get a mask of v8i1 type.
Later on it's replaced with CONCAT_VECTORS, which then is lowered to many DAG nodes including insert/extract element and shift right/left nodes.
The fact that AVX512 compare instructions put the result in a k register and zeroes all its upper bits allows us to remove the extra nodes simply by copying the result to the required register class.
When lowering, identify these cases and transform them into an INSERT_SUBVECTOR node (marked legal), then catch this pattern in instructions selection phase and transform it into one avx512 cmp instruction.
Differential Revision: https://reviews.llvm.org/D33188
llvm-svn: 305465
We know that shuffle masks are power-of-2 sizes, but there's no way (?) for LLVM to know that,
so hack combineX86ShufflesRecursively() to be much faster by replacing div/rem with shift/mask.
This makes the motivating compile-time bug in PR32037 ( https://bugs.llvm.org/show_bug.cgi?id=32037 )
about 9% faster overall.
Differential Revision: https://reviews.llvm.org/D34174
llvm-svn: 305398
Much of PR32037's compile time regression is due to getTargetConstantBitsFromNode always creating large (>64bit) APInts during the bitcasting from the source data to the destination bitwidth.
This commit avoids this bitcast stage if the data is already the correct bitwidth.
llvm-svn: 305284
This step is just intended to reduce code duplication rather than change any functionality.
A follow-up would be to replace PPCTargetLowering::spliceIntoChain() usage with this new helper.
Differential Revision: https://reviews.llvm.org/D33649
llvm-svn: 305192
I was looking closer at the x86 test diffs in D33866, and the first change seems like it
shouldn't happen in the first place. So this patch will resolve that.
Using Agner's tables and AMD docs, vperm2f128 and vinsertf128 have identical timing for
any given CPU model, so we should be able to interchange those without affecting perf.
But as we can see in some of the diffs here, using vperm2f128 allows load folding, so
we should take that opportunity to reduce code size and register pressure.
A secondary advantage is making AVX1 and AVX2 codegen more similar. Given that vperm2f128
was introduced with AVX1, we should be selecting it in all of the same situations that we
would with AVX2. If there's some reason that an AVX1 CPU would not want to use this
instruction, that should be fixed up in a later pass.
Differential Revision: https://reviews.llvm.org/D33938
llvm-svn: 305171
If we know that both operands of an unsigned integer vector comparison are non-negative,
then it's safe to directly use a signed-compare-greater-than instruction (the only non-equality
integer vector compare predicate provided by SSE/AVX).
We're intentionally not changing the condition code to signed in order to preserve the
existing transforms that use min/max/psubus below here.
This should solve PR33276:
https://bugs.llvm.org/show_bug.cgi?id=33276
Differential Revision: https://reviews.llvm.org/D33862
llvm-svn: 304909
We currently generate BUILD_VECTOR as a tree of UNPCKL shuffles of the same type:
e.g. for v4f32:
Step 1: unpcklps 0, 2 ==> X: <?, ?, 2, 0>
: unpcklps 1, 3 ==> Y: <?, ?, 3, 1>
Step 2: unpcklps X, Y ==> <3, 2, 1, 0>
The issue is because we are not placing sequential vector elements together early enough, we fail to recognise many combinable patterns - consecutive scalar loads, extractions etc.
Instead, this patch unpacks progressively larger sequential vector elements together:
e.g. for v4f32:
Step 1: unpcklps 0, 2 ==> X: <?, ?, 1, 0>
: unpcklps 1, 3 ==> Y: <?, ?, 3, 2>
Step 2: unpcklpd X, Y ==> <3, 2, 1, 0>
This does mean that we are creating UNPCKL shuffle of different value types, but the relevant combines that benefit from this are quite capable of handling the additional BITCASTs that are now included in the shuffle tree.
Differential Revision: https://reviews.llvm.org/D33864
llvm-svn: 304688
Since r288804, we try to lower build_vectors on AVX using broadcasts of
float/double. However, when we broadcast integer values that happen to
have a NaN float bitpattern, we lose the NaN payload, thereby changing
the integer value being broadcast.
This is caused by ConstantFP::get, to which we pass the splat i32 as
a float (by bitcasting it using bitsToFloat). ConstantFP::get takes
a double parameter, so we end up lossily converting a single-precision
NaN to double-precision.
Instead, avoid any kinds of conversions by directly building an APFloat
from the splatted APInt.
Note that this also fixes another piece of code (broadcast of
subvectors), that currently isn't susceptible to the same problem.
Also note that we could really just use APInt and ConstantInt
throughout: the constant pool type doesn't matter much. Still, for
consistency, use the appropriate type.
llvm-svn: 304590
This might give a few better opportunities to optimize these to memcpy
rather than loops - also a few minor cleanups (StringRef-izing,
templating (to avoid std::function indirection), etc).
The SmallVector::assign(iter, iter) could be improved with the use of
SFINAE, but the (iter, iter) ctor and append(iter, iter) need it to and
don't have it - so, workaround it for now rather than bothering with the
added complexity.
(also, as noted in the added FIXME, these assign ops could potentially
be optimized better at least for non-trivially-copyable types)
llvm-svn: 304566
Summary:
Add an early combine to match patterns such as:
(i16 bitcast (v16i1 x))
->
(i16 movmsk (v16i8 sext (v16i1 x)))
This combine needs to happen early enough before
type-legalization scalarizes the result of the setcc.
Reviewers: igorb, craig.topper, RKSimon
Subscribers: delena, llvm-commits
Differential Revision: https://reviews.llvm.org/D33311
llvm-svn: 304406
Summary:
This is a continuation of the work started in D29872 . Passing the carry down as a value rather than as a glue allows for further optimizations. Introducing setcccarry makes the use of addc/subc unecessary and we can start the removal process.
This patch only introduce the optimization strictly required to get the same level of optimization as was available before nothing more.
Reviewers: jyknight, nemanjai, mkuper, spatel, RKSimon, zvi, bkramer
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D33374
llvm-svn: 304404
Summary:
Currently FPOWI defaults to Legal and LegalizeDAG.cpp turns Legal into Expand for this opcode because Legal is a "lie".
This patch changes the default for this opcode to Expand and removes the hack from LegalizeDAG.cpp. It also removes all the code in the targets that set this opcode to Expand themselves since they can just rely on the default.
Reviewers: spatel, RKSimon, efriedma
Reviewed By: RKSimon
Subscribers: jfb, dschuff, sbc100, jgravelle-google, nemanjai, javed.absar, andrew.w.kaylor, llvm-commits
Differential Revision: https://reviews.llvm.org/D33530
llvm-svn: 304215
AVX512_VPOPCNTDQ is a new feature set that was published by Intel.
The patch represents the LLVM side of the addition of two new intrinsic based instructions (vpopcntd and vpopcntq).
Differential Revision: https://reviews.llvm.org/D33169
llvm-svn: 303858
This patch defines the i1 type as illegal in the X86 backend for AVX512.
For DAG operations on <N x i1> types (build vector, extract vector element, ...) i8 is used, and should be truncated/extended.
This should produce better scalar code for i1 types since GPRs will be used instead of mask registers.
Differential Revision: https://reviews.llvm.org/D32273
llvm-svn: 303421
- '-verify-mahcineinstrs' starts to complain allocatable live-in physical
registers on non-entry or non-landing-pad basic blocks.
- Refactor the XBEGIN translation to define EAX on a dedicated fallback code
path due to XABORT. Add a pseudo instruction to define EAX explicitly to
avoid add physical register live-in.
Differential Revision: https://reviews.llvm.org/D33168
llvm-svn: 303306
This function gives the wrong answer on some non-ELF platforms in some
cases. The function that does the right thing lives in Mangler.h. To try to
discourage people from using this function, give it a different name.
Differential Revision: https://reviews.llvm.org/D33162
llvm-svn: 303134
Replace SelectionDAG::getNode(ISD::SELECT, ...)
and SelectionDAG::getNode(ISD::VSELECT, ...)
with SelectionDAG::getSelect(...)
Saves a few lines of code and in some cases saves the need to explicitly
check the type of the desired node.
llvm-svn: 303024
This patch adds min/max population count, leading/trailing zero/one bit counting methods.
The min methods return answers based on bits that are known without considering unknown bits. The max methods give answers taking into account the largest count that unknown bits could give.
Differential Revision: https://reviews.llvm.org/D32931
llvm-svn: 302925
Avoid using report_fatal_error, because it will ask the user to file a
bug. If the user attempts to disable SSE on x86_64 and them use floating
point, that's a bug in their code, not a bug in the compiler.
This is just a start. There are other ways to crash the backend in this
configuration, but they should be updated to follow this pattern.
Differential Revision: https://reviews.llvm.org/D27522
llvm-svn: 302835
manages to form a VSELECT with a non-i1 element type condition. Those
are technically allowed in SDAG (at least, the generic type legalization
logic will form them and I wouldn't want to try to audit everything te
preclude forming them) so we need to be able to lower them.
This isn't too hard to implement. We mark VSELECT as custom so we get
a chance in C++, add a fast path for i1 conditions to get directly
handled by the patterns, and a fallback when we need to manually force
the condition to be an i1 that uses the vptestm instruction to turn
a non-mask into a mask.
This, unsurprisingly, generates awful code. But it at least doesn't
crash. This was actually impacting open source packages built with LLVM
for AVX-512 in the wild, so quickly landing a patch that at least stops
the immediate bleeding.
I think I've found where to fix the codegen quality issue, but less
confident of that change so separating it out from the thing that
doesn't change the result of any existing test case but causes mine to
not crash.
llvm-svn: 302785
Use variadic templates instead of relying on <cstdarg> + sentinel.
This enforces better type checking and makes code more readable.
Differential Revision: https://reviews.llvm.org/D32541
llvm-svn: 302571
for scalar masked instructions only the lower bit of the mask is relevant. so for constant masks we should either do an unmasked operation or no operation, depending on the value of the lower bit.
This patch handles cases where the lower bit is '1'.
Differential Revision: https://reviews.llvm.org/D32805
llvm-svn: 302546
Using arguments with attribute inalloca creates problems for verification
of machine representation. This attribute instructs the backend that the
argument is prepared in stack prior to CALLSEQ_START..CALLSEQ_END
sequence (see http://llvm.org/docs/InAlloca.htm for details). Frame size
stored in CALLSEQ_START in this case does not count the size of this
argument. However CALLSEQ_END still keeps total frame size, as caller can
be responsible for cleanup of entire frame. So CALLSEQ_START and
CALLSEQ_END keep different frame size and the difference is treated by
MachineVerifier as stack error. Currently there is no way to distinguish
this case from actual errors.
This patch adds additional argument to CALLSEQ_START and its
target-specific counterparts to keep size of stack that is set up prior to
the call frame sequence. This argument allows MachineVerifier to calculate
actual frame size associated with frame setup instruction and correctly
process the case of inalloca arguments.
The changes made by the patch are:
- Frame setup instructions get the second mandatory argument. It
affects all targets that use frame pseudo instructions and touched many
files although the changes are uniform.
- Access to frame properties are implemented using special instructions
rather than calls getOperand(N).getImm(). For X86 and ARM such
replacement was made previously.
- Changes that reflect appearance of additional argument of frame setup
instruction. These involve proper instruction initialization and
methods that access instruction arguments.
- MachineVerifier retrieves frame size using method, which reports sum of
frame parts initialized inside frame instruction pair and outside it.
The patch implements approach proposed by Quentin Colombet in
https://bugs.llvm.org/show_bug.cgi?id=27481#c1.
It fixes 9 tests failed with machine verifier enabled and listed
in PR27481.
Differential Revision: https://reviews.llvm.org/D32394
llvm-svn: 302527
Similar to what we do for vXi8 ASHR(X, 7), use SSE42's PCMPGTQ to splat the sign instead of using the PSRAD+PSHUFD.
Avoiding bitcasts this improves combines that utilize computeNumSignBits, permits memory folding and reduces pipe pressure. Although it does require a second register, given that this is a (cheap) zero register the impact is minimal.
Differential Revision: https://reviews.llvm.org/D32973
llvm-svn: 302525
Currently combineLogicBlendIntoPBLENDV can only match ASHR to detect sign splatting of a bit mask, this patch generalises this to use computeNumSignBits instead.
This is a first step in several things we can do to improve PBLENDV support:
* Better matching of X86ISD::ANDNP patterns.
* Handle floating point cases.
* Better vector and bitcast support in computeNumSignBits.
* Recognise that PBLENDV only uses the sign bit of the mask, we should be able strip away sign splats (ASHR, PCMPGT isNeg tests etc.).
Differential Revision: https://reviews.llvm.org/D32953
llvm-svn: 302424
This patch introduces an LLVM intrinsic and a target opcode for custom event
logging in XRay. Initially, its use case will be to allow users of XRay to log
some type of string ("poor man's printf"). The target opcode compiles to a noop
sled large enough to enable calling through to a runtime-determined relative
function call. At runtime, when X-Ray is enabled, the sled is replaced by
compiler-rt with a trampoline to the logic for creating the custom log entries.
Future patches will implement the compiler-rt parts and clang-side support for
emitting the IR corresponding to this intrinsic.
Reviewers: timshen, dberris
Subscribers: igorb, pelikan, rSerge, timshen, echristo, dberris, llvm-commits
Differential Revision: https://reviews.llvm.org/D27503
llvm-svn: 302405
rL294581 broke unnecessary register dependencies on partial v16i8/v8i16 BUILD_VECTORs, but on SSE41 we (currently) use insertion for full BUILD_VECTORs as well. By allowing full insertion to occur on SSE41 targets we can break register dependencies here as well.
llvm-svn: 302355
This adds routines for reseting KnownBits to unknown, making the value all zeros or all ones. It also adds methods for querying if the value is zero, all ones or unknown.
Differential Revision: https://reviews.llvm.org/D32637
llvm-svn: 302262
This patch adds zext, sext, and trunc methods to KnownBits and uses them where possible.
Differential Revision: https://reviews.llvm.org/D32784
llvm-svn: 302088
This patch adds support for the the LightWeight Profiling (LWP) instructions which are available on all AMD Bulldozer class CPUs (bdver1 to bdver4).
Reapplied - this time without changing line endings of existing files.
Differential Revision: https://reviews.llvm.org/D32769
llvm-svn: 302041
This patch adds support for the the LightWeight Profiling (LWP) instructions which are available on all AMD Bulldozer class CPUs (bdver1 to bdver4).
Differential Revision: https://reviews.llvm.org/D32769
llvm-svn: 302028
Summary: As per discution on how to get better codegen an large int legalization, it became clear that using a glue for the carry was preventing several desirable optimizations. Passing the carry down as a value allow for more flexibility.
Reviewers: jyknight, nemanjai, mkuper, spatel, RKSimon, zvi, bkramer
Subscribers: igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D29872
llvm-svn: 301775
Adds a new method finalizeLowering to TargetLoweringBase. This is in
preparation for an upcoming commit.
This function is meant for target specific adjustments to
MachineFrameInfo or register reservations.
Move the freezeRegisters() and the hasCopyImplyingStackAdjustment()
handling into the new function to prove the concept. As an added bonus
GlobalISel no longer missed the hasCopyImplyingStackAdjustment()
handling with this.
Differential Revision: https://reviews.llvm.org/D32621
llvm-svn: 301679
This patch replaces the separate APInts for KnownZero/KnownOne with a single KnownBits struct. This is similar to what was done to ValueTracking's version recently.
This is largely a mechanical transformation from KnownZero to Known.Zero.
Differential Revision: https://reviews.llvm.org/D32569
llvm-svn: 301620
This patch uses various APInt methods to reduce the number of temporary APInts. These were all found while working through converting SelectionDAG's computeKnownBits to also use the KnownBits struct recently added to the ValueTracking version.
llvm-svn: 301618
This reverts commit r301105, 4, 3 and 1, as a follow up of the previous
revert, which broke even more bots.
For reference:
Revert "[APInt] Use operator<<= where possible. NFC"
Revert "[APInt] Use operator<<= instead of shl where possible. NFC"
Revert "[APInt] Use ashInPlace where possible."
PR32754.
llvm-svn: 301111
immediate operands.
This commit adds an AArch64 dag-combine that optimizes code generation
for logical instructions taking immediate operands. The optimization
uses demanded bits to change a logical instruction's immediate operand
so that the immediate can be folded into the immediate field of the
instruction.
This recommits r300932 and r300930, which was causing dag-combine to
loop forever. The problem was that optimizeLogicalImm was returning
true even when there was no change to the immediate node (which happened
when the immediate was all zeros or ones), which caused dag-combine to
push and pop the same node to the work list over and over again without
making any progress.
This commit fixes the bug by returning false early in optimizeLogicalImm
if the immediate is all zeros or ones. Also, it changes the code to
compare the immediate with 0 or Mask rather than calling
countPopulation.
rdar://problem/18231627
Differential Revision: https://reviews.llvm.org/D5591
llvm-svn: 301019
Craig Topper