CGF/CLGF compares an i64 register with a sign/zero extended loaded i32 value
in memory.
This patch makes such a load considered foldable and so gets a 0 cost.
Review: Ulrich Weigand
https://reviews.llvm.org/D54944
llvm-svn: 347735
AH, SH and MH costs are already covered in the cases where LHS is 32 bits and
RHS is 16 bits of memory sign-extended to i32.
As these instructions are also used when LHS is i16, this patch recognizes
that the loads will get folded then as well.
Review: Ulrich Weigand
https://reviews.llvm.org/D54940
llvm-svn: 347734
Single instructions exist for i8 and i16 comparisons of memory against a
small immediate.
This patch makes sure that if the load in these cases has a single user (the
ICmp), it gets a 0 cost (folded), and also that the ICmp gets a cost of 1.
Review: Ulrich Weigand
https://reviews.llvm.org/D54897
llvm-svn: 347733
Since byte-swapping loads and stores are supported, a 'load -> bswap' or
'bswap -> store' sequence should have the cost of one.
Review: Ulrich Weigand
https://reviews.llvm.org/D54870
llvm-svn: 347732
The check lines marked AVX256 in the zext256/sext256 functions should be closer to the AVX values which would take into account a splitting cost.
llvm-svn: 347722
Our sext/zext cost modeling was somewhat incomplete. And had no coverage for the fact that avx512bw v32i16/v64i8 types return a scalarization cost.
Truncates are a whole different mess because isTruncateFree is returning true for vectors when it shouldn't and that's the fall back for anything not in the tables.
llvm-svn: 347719
This reverts commit r346970.
It was causing PR39774, a crash in slp-vectorizer on a rather simple loop
with just a bunch of 'and's in the body.
llvm-svn: 347541
Implement getIntrinsicInstrCost() and return costs reflecting that bswap can
be done with a vperm per vector register.
Review: Ulrich Weigand
https://reviews.llvm.org/D54789
llvm-svn: 347445
We were adding the entire scalarization extraction cost for reductions, which returns the total cost of extracting every element of a vector type.
For reductions we don't need to do this - we just need to extract the 0'th element after the reduction pattern has completed.
Fixes PR37731
Differential Revision: https://reviews.llvm.org/D54585
llvm-svn: 346970
Add support for the expansion of funnelshift/rotates to getIntrinsicInstrCost.
This also required us to move the X86 fshl/fshr costs to the same place as the rotates to avoid expansion and get correct scalarization vs vectorization costs.
llvm-svn: 346854
When we repeat the 2 shifting operands then this is a bit rotation - annoyingly this has to be done in the other getIntrinsicInstrCost than most intrinsics as we need to check the operands are the same.
llvm-svn: 346688
Improve getCastInstrCost() by respecting the different types of Src and Dst
for vector integer <-> fp conversions.
This means that extracting from integer becomes more expensive (by the
extraction penalty), and the extraction from fp becomes cheaper (no longer
has a false extraction penalty).
Review: Ulrich Weigand
https://reviews.llvm.org/D54423
llvm-svn: 346663
Instead of defaulting to a cost = 1, expand to element extract/insert like we do for other shuffles.
This exposes an issue in LoopVectorize which could call SK_ExtractSubvector with a scalar subvector type.
llvm-svn: 346656
Let i8/i16 uint/sint to fp conversions cost 1 if operand is a load.
Since the load already does the extension, there is no extra cost (previously
returned 2).
Review: Ulrich Weigand
https://reviews.llvm.org/D54028
llvm-svn: 346009
Scalar i1 to fp conversions are done with a branch sequence, so it should
have a higher cost.
Review: Ulrich Weigand
https://reviews.llvm.org/D53924
llvm-svn: 345818
This factors out a new method getBoolVecToIntConversionCost() containing the
code for vector sext/zext of i1, in order to reuse it for i1 to double vector
conversions.
Review: Ulrich Weigand
https://reviews.llvm.org/D53923
llvm-svn: 345817
Correct costings of SK_ExtractSubvector requires the SubTy argument to indicate the type/size of the extracted subvector.
Unlike the rest of the shuffle kinds this means that the main Ty argument represents the source vector type not the destination!
I've done my best to fix a number of vectorizer uses:
SLP - the reduction epilogue costs should be using a SK_PermuteSingleSrc shuffle as these all occur at the hardware vector width - we're not extracting (illegal) subvector types. This is causing the cost model diffs as SK_ExtractSubvector costs are poorly handled and tend to just return 1 at the moment.
LV - I'm not clear on what the SK_ExtractSubvector should represents for recurrences - I've used a <1 x ?> subvector extraction as that seems to match the VF delta.
Differential Revision: https://reviews.llvm.org/D53573
llvm-svn: 345617
Sub, SDiv and UDiv are not commutative, so only the RHS operand can fold a
load. This patch adds a check for this.
Review: Ulrich Weigand
https://reviews.llvm.org/D53791
llvm-svn: 345596
The SystemZ backend can do arithmetic of memory by loading and then extending
one of the operands. Similarly, a load + truncate can be folded into an
operand.
This patch improves the SystemZ TTI cost function to recognize this.
Review: Ulrich Weigand
https://reviews.llvm.org/D52692
llvm-svn: 345327
Enable the DAG optimization that converts vector div/rem with constants into
multiply+shifts sequences by expanding them early. This is needed since
ISD::SMUL_LOHI is 'Custom' lowered on SystemZ, and will therefore not be
available to BuildSDIV after legalization.
Better cost values for these instructions based on how they will be
implemented (a constant divisor is cheaper).
Review: Ulrich Weigand
https://reviews.llvm.org/D53196
llvm-svn: 345321
Non-uniform division/remainder handling was added back at D49248/D50765 - so share the 'mul+sub' costs that already exist for uniform cases.
llvm-svn: 345164
Until mischeduler is clever enough to avoid spilling in a vectorized loop
with many (scalar) DLRs it is better to avoid high vectorization factors (8
and above).
llvm-svn: 344129
A new function getNumVectorRegs() is better to use for the number of needed
vector registers instead of getNumberOfParts(). This is to make sure that the
number of vector registers (and typically operations) required for a vector
type is accurate.
getNumberOfParts() which was previously used works by splitting the vector
type until it is legal gives incorrect results for types with a non
power of two number of elements (rare).
A new static function getScalarSizeInBits() that also checks for a pointer
type and returns 64U for it since otherwise it gets a value of 0). Used in a
few places where Ty may be pointer.
Review: Ulrich Weigand
llvm-svn: 344115
After recent improvements which makes better use of LOC instead of IPM, the
TTI cost functions also needs to be updated to reflect this.
This involves sext, zext and xor of i1.
The tests were updated so that for z13 the new costs are expected, while the
old costs are still checked for on zEC12.
Review: Ulrich Weigand
https://reviews.llvm.org/D51339
llvm-svn: 342207
Summary: This was inheriting the cost from the AVX table, but should be legal under AVX512.
Reviewers: RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D51267
llvm-svn: 340708
We penalize general SDIV/UDIV costs but don't do the same for SREM/UREM.
This patch makes general vector SREM/UREM x20 as costly as scalar, the same approach as we do for SDIV/UDIV. The patch also extends the existing SDIV/UDIV constant costs for SREM/UREM - at the moment this means the additional cost of a MUL+SUB (see D48975).
Differential Revision: https://reviews.llvm.org/D48980
llvm-svn: 336486
This patch adds a custom trunc store lowering for v4i8 vector types.
Since there is not v.4b register, the v4i8 is promoted to v4i16 (v.4h)
and default action for v4i8 is to extract each element and issue 4
byte stores.
A better strategy would be to extended the promoted v4i16 to v8i16
(with undef elements) and extract and store the word lane which
represents the v4i8 subvectores. The construction:
define void @foo(<4 x i16> %x, i8* nocapture %p) {
%0 = trunc <4 x i16> %x to <4 x i8>
%1 = bitcast i8* %p to <4 x i8>*
store <4 x i8> %0, <4 x i8>* %1, align 4, !tbaa !2
ret void
}
Can be optimized from:
umov w8, v0.h[3]
umov w9, v0.h[2]
umov w10, v0.h[1]
umov w11, v0.h[0]
strb w8, [x0, #3]
strb w9, [x0, #2]
strb w10, [x0, #1]
strb w11, [x0]
ret
To:
xtn v0.8b, v0.8h
str s0, [x0]
ret
The patch also adjust the memory cost for autovectorization, so the C
code:
void foo (const int *src, int width, unsigned char *dst)
{
for (int i = 0; i < width; i++)
*dst++ = *src++;
}
can be vectorized to:
.LBB0_4: // %vector.body
// =>This Inner Loop Header: Depth=1
ldr q0, [x0], #16
subs x12, x12, #4 // =4
xtn v0.4h, v0.4s
xtn v0.8b, v0.8h
st1 { v0.s }[0], [x2], #4
b.ne .LBB0_4
Instead of byte operations.
llvm-svn: 335735
AArch64 was only setting costs for SK_Transpose, which meant that many of the simpler shuffles (e.g. SK_Select and SK_PermuteSingleSrc for larger vector elements) was being severely overestimated by the default shuffle expansion.
This patch adds costs to help improve SLP performance and avoid a regression in reductions introduced by D48174.
I'm not very knowledgeable about AArch64 shuffle lowering so I've kept the extra costs to a minimum - someone who knows this code can add extra costs which should improve vectorization a lot more.
Differential Revision: https://reviews.llvm.org/D48172
llvm-svn: 335329
These were being over cautious for costs for one/two op general shuffles - VSHUFPD doesn't have to replicate the same shuffle in both lanes like VSHUFPS does.
llvm-svn: 335216
As discussed on PR33744, this patch relaxes ShuffleKind::SK_Alternate which requires shuffle masks to only match an alternating pattern from its 2 sources:
e.g. v4f32: <0,5,2,7> or <4,1,6,3>
This seems far too restrictive as most SIMD hardware which will implement it using a general blend/bit-select instruction, so replaces it with SK_Select, permitting elements from either source as long as they are inline:
e.g. v4f32: <0,5,2,7>, <4,1,6,3>, <0,1,6,7>, <4,1,2,3> etc.
This initial patch just updates the name and cost model shuffle mask analysis, later patch reviews will update SLP to better utilise this - it still limits itself to SK_Alternate style patterns.
Differential Revision: https://reviews.llvm.org/D47985
llvm-svn: 334513
As discussed on D47985, identity shuffle masks should probably be free.
I've limited this to the case where the input and output types all match - but we could probably accept all cases.
Differential Revision: https://reviews.llvm.org/D47986
llvm-svn: 334506
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
This enables us to detect more fast path sdiv cases under cost analysis.
This patch also enables us to handle non-uniform-constant pow2 cases for X86 SDIV costs.
Found while working on D46276
Future patches can then extend the vectorizers to more fully support non-uniform pow2 cases.
Differential Revision: https://reviews.llvm.org/D46637
llvm-svn: 332969
With custom lowering for vector MULLH{S,U}, it is now profitable to
vectorize a divide by constant loop for the custom types (v16i8, v8i16,
and v4i32). The cost if based on TargetLowering::Build{S,U}DIV which
uses a multiply by constant plus adjustment to express a divide by
constant.
Both {u,s}mull{2} are expressed as Instruction::Mul and shifts by
Instruction::AShr.
llvm-svn: 331873
This patch adds a new shuffle kind useful for transposing a 2xn matrix. These
transpose shuffle masks read corresponding even- or odd-numbered vector
elements from two n-dimensional source vectors and write each result into
consecutive elements of an n-dimensional destination vector. The transpose
shuffle kind is meant to model the TRN1 and TRN2 AArch64 instructions. As such,
this patch also considers transpose shuffles in the AArch64 implementation of
getShuffleCost.
Differential Revision: https://reviews.llvm.org/D45982
llvm-svn: 330941
This patch adds a cost model test case for vector shuffles having transpose
masks. The given costs are inaccurate and will be updated in a follow-on patch.
llvm-svn: 330625
The script allows the auto-generation of checks for cost model tests to speed up their creation and help improve coverage, which will help a lot with PR36550.
If the need arises we can add support for other analyze passes as well, but the cost models was the one I needed to get done - at the moment it just warns that any other analysis mode is unsupported.
I've regenerated a couple of x86 test files to show the effect.
Differential Revision: https://reviews.llvm.org/D45272
llvm-svn: 329390
Add fdiv costs for Goldmont using table 16-17 of the Intel Optimization Manual. Also add overrides for FSQRT for Goldmont and Silvermont.
Reviewers: RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44644
llvm-svn: 328451
This patch provides an implementation of getArithmeticReductionCost for
AArch64. We can specialize the cost of add reductions since they are computed
using the 'addv' instruction.
Differential Revision: https://reviews.llvm.org/D44490
llvm-svn: 327702
This patch considers the experimental vector reduce intrinsics in the default
implementation of getIntrinsicInstrCost. The cost of these intrinsics is
computed with getArithmeticReductionCost and getMinMaxReductionCost. This patch
also adds a test case for AArch64 that indicates the costs we currently compute
for vector reduce intrinsics. These costs are inaccurate and will be updated in
a follow-on patch.
Differential Revision: https://reviews.llvm.org/D44489
llvm-svn: 327698
Since there is no instruction for integer vector division, factor in the
cost of singling out each element to be used with the scalar division
instruction.
Differential revision: https://reviews.llvm.org/D43974
llvm-svn: 326955
Agner's tables indicate that for SSE42+ targets (Core2 and later) we can reduce the FADD/FSUB/FMUL costs down to 1, which should fix the Himeno benchmark.
Note: the AVX512 FDIV costs look rather dodgy, but this isn't part of this patch.
Differential Revision: https://reviews.llvm.org/D43733
llvm-svn: 326133
There are too many perf regressions resulting from this, so we need to
investigate (and add tests for) targets like ARM and AArch64 before
trying to reinstate.
llvm-svn: 325658
This change was mentioned at least as far back as:
https://bugs.llvm.org/show_bug.cgi?id=26837#c26
...and I found a real program that is harmed by this:
Himeno running on AMD Jaguar gets 6% slower with SLP vectorization:
https://bugs.llvm.org/show_bug.cgi?id=36280
...but the change here appears to solve that bug only accidentally.
The div/rem costs for x86 look very wrong in some cases, but that's already true,
so we can fix those in follow-up patches. There's also evidence that more cost model
changes are needed to solve SLP problems as shown in D42981, but that's an independent
problem (though the solution may be adjusted after this change is made).
Differential Revision: https://reviews.llvm.org/D43079
llvm-svn: 325515
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
Based on the names of the check lines, features seems more appropriate that cpu.
Spotted while prototyping my patch to make 512-bit vectors illegal on SKX sometimes.
llvm-svn: 320959
This patch contains more accurate cost of interelaved load\store of stride 2 for the types int64\double on AVX2.
Reviewers: delena, RKSimon, craig.topper, dorit
Reviewed By: dorit
Differential Revision: https://reviews.llvm.org/D40008
llvm-svn: 318385
Recommit:
This patch contains update of the costs of interleaved loads of v8f32 of stride 3 and 8.
fixed the location of the lit test it works with make check-all.
Differential Revision: https://reviews.llvm.org/D39403
llvm-svn: 317471
reverted my changes will be committed later after fixing the failure
This patch contains update of the costs of interleaved loads of v8f32 of stride 3 and 8.
Differential Revision: https://reviews.llvm.org/D39403
llvm-svn: 317433
This patch contains update of the costs of interleaved loads of v8f32 of stride 3 and 8.
Differential Revision: https://reviews.llvm.org/D39403
llvm-svn: 317432
This patch adds accurate instructions cost.
The formula presents two cases(stride 3 and stride 4) and calculates the cost according to the VF and stride.
Reviewers:
1. delena
2. Farhana
3. zvi
4. dorit
5. Ayal
Differential Revision: https://reviews.llvm.org/D38762
Change-Id: If4cfbd4ac0e63694e8144cb78c7fa34850647ff7
llvm-svn: 316072
Significantly reduces performancei (~30%) of gipfeli
(https://github.com/google/gipfeli)
I have not yet managed to reproduce this regression with the open-source
version of the benchmark on github, but will work with others to get a
reproducer to you later today.
llvm-svn: 315680
Summary:
If the pointer width is 32 bits and the calculated GEP offset is
negative, we call APInt::getLimitedValue(), which does a
*zero*-extension of the offset. That's wrong -- we should do an sext.
Fixes a bug introduced in rL314362 and found by Evgeny Astigeevich.
Reviewers: efriedma
Subscribers: sanjoy, javed.absar, llvm-commits, eastig
Differential Revision: https://reviews.llvm.org/D38557
llvm-svn: 314935
Function isLoweredToCall can only accept non-null function pointer, but a function pointer can be null for indirect function call. So check it before calling isLoweredToCall from getInstructionLatency.
Differential Revision: https://reviews.llvm.org/D38204
llvm-svn: 314927
Recommitting r314517 with the fix for handling ConstantExpr.
Original commit message:
Currently, getGEPCost() returns TCC_FREE whenever a GEP is a legal addressing
mode in the target. However, since it doesn't check its actual users, it will
return FREE even in cases where the GEP cannot be folded away as a part of
actual addressing mode. For example, if an user of the GEP is a call
instruction taking the GEP as a parameter, then the GEP may not be folded in
isel.
llvm-svn: 314923
Summary:
Currently, getGEPCost() returns TCC_FREE whenever a GEP is a legal addressing mode in the target.
However, since it doesn't check its actual users, it will return FREE even in cases
where the GEP cannot be folded away as a part of actual addressing mode.
For example, if an user of the GEP is a call instruction taking the GEP as a parameter,
then the GEP may not be folded in isel.
Reviewers: hfinkel, efriedma, mcrosier, jingyue, haicheng
Reviewed By: hfinkel
Subscribers: javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D38085
llvm-svn: 314517
Summary:
This avoids C++ UB if the GEP is weird and the calculation overflows
int64_t, and it's also observable in the cost model's results.
Such GEPs are almost surely not valid pointers, but LLVM nonetheless
generates them sometimes.
Reviewers: sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38337
llvm-svn: 314362
Usually an intrinsic is a simple target instruction, it should have a small latency. A real function call has much larger latency. So handle the intrinsic call in function getInstructionLatency().
Differential Revision: https://reviews.llvm.org/D38104
llvm-svn: 314003
Static alloca usually doesn't generate any machine instructions, so it has 0 cost.
Differential Revision: https://reviews.llvm.org/D37879
llvm-svn: 313410
For instructions that unlikely generate machine instructions, they should also have 0 latency.
Differential Revision: https://reviews.llvm.org/D37833
llvm-svn: 313288
Current TargetTransformInfo can support throughput cost model and code size model, but sometimes we also need instruction latency cost model in different optimizations. Hal suggested we need a single public interface to query the different cost of an instruction. So I proposed following interface:
enum TargetCostKind {
TCK_RecipThroughput, ///< Reciprocal throughput.
TCK_Latency, ///< The latency of instruction.
TCK_CodeSize ///< Instruction code size.
};
int getInstructionCost(const Instruction *I, enum TargetCostKind kind) const;
All clients should mainly use this function to query the cost of an instruction, parameter <kind> specifies the desired cost model.
This patch also provides a simple default implementation of getInstructionLatency.
The default getInstructionLatency provides latency numbers for only small number of instruction classes, those latency numbers are only reasonable for modern OOO processors. It can be extended in following ways:
Add more detail into this function.
Add getXXXLatency function and call it from here.
Implement target specific getInstructionLatency function.
Differential Revision: https://reviews.llvm.org/D37170
llvm-svn: 312832
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
Add missing SK_PermuteSingleSrc costs for AVX2 targets and earlier, also added some of the simpler SK_PermuteTwoSrc costs to support splitting of SK_PermuteSingleSrc shuffles
llvm-svn: 310632
This adds support for the new 32-bit vector float instructions of z14.
This includes:
- Enabling the instructions for the assembler/disassembler.
- CodeGen for the instructions, including new LLVM intrinsics.
- Scheduler description support for the instructions.
- Update to the vector cost function calculations.
In general, CodeGen support for the new v4f32 instructions closely
matches support for the existing v2f64 instructions.
llvm-svn: 308195
this patch updates the cost of addq\subq (add\subtract of vectors of 64bits)
based on the performance numbers of SLM arch.
Differential Revision: https://reviews.llvm.org/D33983
llvm-svn: 306974
The cost of an interleaved access was only implemented for AVX512. For other
X86 targets an overly conservative Base cost was returned, resulting in
avoiding vectorization where it is actually profitable to vectorize.
This patch starts to add costs for AVX2 for most prominent cases of
interleaved accesses (stride 3,4 chars, for now).
Note1: Improvements of up to ~4x were observed in some of EEMBC's rgb
workloads; There is also a known issue of 15-30% degradations on some of these
workloads, associated with an interleaved access followed by type
promotion/widening; the resulting shuffle sequence is currently inefficient and
will be improved by a series of patches that extend the X86InterleavedAccess pass
(such as D34601 and more to follow).
Note 2: The costs in this patch do not reflect port pressure penalties which can
be very dominant in the case of interleaved accesses since most of the shuffle
operations are restricted to a single port. Further tuning, that may incorporate
these considerations, will be done on top of the upcoming improved shuffle
sequences (that is, along with the abovementioned work to extend
X86InterleavedAccess pass).
Differential Revision: https://reviews.llvm.org/D34023
llvm-svn: 306238
Jonas Paulsson