v6m cores only have a limited number of registers available. Unrolling
can mean we spend more on stack spills and reloads than we save from the
unrolling. This patch adds an extra heuristic to put a limit on the
unroll count for loops with multiple live out values, as measured from
the LCSSA phi nodes.
Differential Revision: https://reviews.llvm.org/D104659
Added an extra analysis for better choosing of shuffle kind in
getShuffleCost functions for better cost estimation if mask was
provided.
Differential Revision: https://reviews.llvm.org/D100865
Added an extra analysis for better choosing of shuffle kind in
getShuffleCost functions for better cost estimation if mask was
provided.
Differential Revision: https://reviews.llvm.org/D100865
Added cost estimation for switch instruction, updated costs of branches, fixed
phi cost.
Had to increase `-amdgpu-unroll-threshold-if` default value since conditional
branch cost (size) was corrected to higher value.
Test renamed to "control-flow.ll".
Removed redundant code in `X86TTIImpl::getCFInstrCost()` and
`PPCTTIImpl::getCFInstrCost()`.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D96805
This removes the restriction that only Thumb2 targets enable runtime
loop unrolling, allowing it for Thumb1 only cores as well. The existing
T2 heuristics are used (for the time being) to control when and how
unrolling is performed.
Differential Revision: https://reviews.llvm.org/D99588
This UpperBound unrolling was already enabled so long as a series of
conditions in ARMTTIImpl::getUnrollingPreferences pass. This just always
enables it as it can help fully unroll loops that would not otherwise
pass those tests.
Differential Revision: https://reviews.llvm.org/D99174
The scalarization overhead was set deliberately high for MVE, whilst the
codegen was new. It helps protect us against the negative ramifications
of mixing scalar and vector instructions. This decreases that,
especially for floating point where the cost of extracting/inserting
lane elements can be low. For integer the cost is still fairly high due
to the cross-register-bank copy, but is no longer n^2 in the length of
the vector.
In general, this will decrease the cost of scalarizing floats and long
integer vectors. i64 increase in cost, having a high cost before and
after this patch. For floats this allows up to start doing things like
vectorizing fdiv instructions, even if they are scalarized.
Differential Revision: https://reviews.llvm.org/D98245
This uses the shuffle mask cost from D98206 to give a better cost of MVE
VREV instructions. This helps especially in VectorCombine where the cost
of shuffles is used to reorder bitcasts, which this helps keep the phase
ordering test for fp16 reductions producing optimal code. The isVREVMask
has been moved to a header file to allow it to be used across target
transform and isel lowering.
Differential Revision: https://reviews.llvm.org/D98210
This adds an Mask ArrayRef to getShuffleCost, so that if an exact mask
can be provided a more accurate cost can be provided by the backend.
For example VREV costs could be returned by the ARM backend. This should
be an NFC until then, laying the groundwork for that to be added.
Differential Revision: https://reviews.llvm.org/D98206
Recently we improved the lowering of low overhead loops and tail
predicated loops, but concentrated first on the DLS do style loops. This
extends those improvements over to the WLS while loops, improving the
chance of lowering them successfully. To do this the lowering has to
change a little as the instructions are terminators that produce a value
- something that needs to be treated carefully.
Lowering starts at the Hardware Loop pass, inserting a new
llvm.test.start.loop.iterations that produces both an i1 to control the
loop entry and an i32 similar to the llvm.start.loop.iterations
intrinsic added for do loops. This feeds into the loop phi, properly
gluing the values together:
%wls = call { i32, i1 } @llvm.test.start.loop.iterations.i32(i32 %div)
%wls0 = extractvalue { i32, i1 } %wls, 0
%wls1 = extractvalue { i32, i1 } %wls, 1
br i1 %wls1, label %loop.ph, label %loop.exit
...
loop:
%lsr.iv = phi i32 [ %wls0, %loop.ph ], [ %iv.next, %loop ]
..
%iv.next = call i32 @llvm.loop.decrement.reg.i32(i32 %lsr.iv, i32 1)
%cmp = icmp ne i32 %iv.next, 0
br i1 %cmp, label %loop, label %loop.exit
The llvm.test.start.loop.iterations need to be lowered through ISel
lowering as a pair of WLS and WLSSETUP nodes, which each get converted
to t2WhileLoopSetup and t2WhileLoopStart Pseudos. This helps prevent
t2WhileLoopStart from being a terminator that produces a value,
something difficult to control at that stage in the pipeline. Instead
the t2WhileLoopSetup produces the value of LR (essentially acting as a
lr = subs rn, 0), t2WhileLoopStart consumes that lr value (the Bcc).
These are then converted into a single t2WhileLoopStartLR at the same
point as t2DoLoopStartTP and t2LoopEndDec. Otherwise we revert the loop
to prevent them from progressing further in the pipeline. The
t2WhileLoopStartLR is a single instruction that takes a GPR and produces
LR, similar to the WLS instruction.
%1:gprlr = t2WhileLoopStartLR %0:rgpr, %bb.3
t2B %bb.1
...
bb.2.loop:
%2:gprlr = PHI %1:gprlr, %bb.1, %3:gprlr, %bb.2
...
%3:gprlr = t2LoopEndDec %2:gprlr, %bb.2
t2B %bb.3
The t2WhileLoopStartLR can then be treated similar to the other low
overhead loop pseudos, eventually being lowered to a WLS providing the
branches are within range.
Differential Revision: https://reviews.llvm.org/D97729
As a followup to D95291, getOperandsScalarizationOverhead was still
using a VF as a vector factor if the arguments were scalar, and would
assert on certain matrix intrinsics with differently sized vector
arguments. This patch removes the VF arg, instead passing the Types
through directly. This should allow it to more accurately compute the
cost without having to guess at which operands will be vectorized,
something difficult with more complex intrinsics.
This adjusts one SVE test as it is now calling the wrong intrinsic vs
veccall. Without invalid InstructCosts the cost of the scalarized
intrinsic is too low. This should get fixed when the cost of
scalarization is accounted for with scalable types.
Differential Revision: https://reviews.llvm.org/D96287
getIntrinsicInstrCost takes a IntrinsicCostAttributes holding various
parameters of the intrinsic being costed. It can either be called with a
scalar intrinsic (RetTy==Scalar, VF==1), with a vector instruction
(RetTy==Vector, VF==1) or from the vectorizer with a scalar type and
vector width (RetTy==Scalar, VF>1). A RetTy==Vector, VF>1 is considered
an error. Both of the vector modes are expected to be treated the same,
but because this is confusing many backends end up getting it wrong.
Instead of trying work with those two values separately this removes the
VF parameter, widening the RetTy/ArgTys by VF used called from the
vectorizer. This keeps things simpler, but does require some other
modifications to keep things consistent.
Most backends look like this will be an improvement (or were not using
getIntrinsicInstrCost). AMDGPU needed the most changes to keep the code
from c230965ccf working. ARM removed the fix in
dfac521da1, webassembly happens to get a fixup for an SLP cost
issue and both X86 and AArch64 seem to now be using better costs from
the vectorizer.
Differential Revision: https://reviews.llvm.org/D95291
It appears that pointer types were causing issues for the min/max cost
code in getIntrinsicInstrCost. This makes sure that when matching
icmp/select to a min/max, we only do that for normal int or float types.
A v8i32 compare will produce a v8i1 predicate, but during codegen the
v8i32 will be split into two v4i32, potentially requiring two v4i1
predicates to be merged into a single v8i1. Because this merging of two
v4i1's into a v8i1 is very expensive, we need to make the cost of the
compare equally high.
This patch adds the cost of that to ARMTTIImpl::getCmpSelInstrCost.
Because we don't know whether the user of the predicate can be split,
and the cost model is mostly pre-instruction, we may be pessimistic but
that should only be for larger and legal types. This also adds min/max
detection to the costmodel where it can be detected, to keep those in
line with the cost of simple min/max instructions. Otherwise for the
most part, costs that were already expensive have become more expensive.
Differential Revision: https://reviews.llvm.org/D96692
This adds basic MVE costs for SMIN/SMAX/UMIN/UMAX, as well as MINNUM and
MAXNUM representing fmin and fmax. It tightens up the costs, not using a
ICmp+Select cost.
Differential Revision: https://reviews.llvm.org/D96603
This refactors shouldFavorPostInc() and shouldFavorBackedgeIndex() into
getPreferredAddressingMode() so that we have one interface to steer LSR in
generating the preferred addressing mode.
Differential Revision: https://reviews.llvm.org/D96600
The vector reduction intrinsics started life as experimental ops, so backend support
was lacking. As part of promoting them to 1st-class intrinsics, however, codegen
support was added/improved:
D58015
D90247
So I think it is safe to now remove this complication from IR.
Note that we still have an IR-level codegen expansion pass for these as discussed
in D95690. Removing that is another step in simplifying the logic. Also note that
x86 was already unconditionally forming reductions in IR, so there should be no
difference for x86.
I spot checked a couple of the tests here by running them through opt+llc and did
not see any asm diffs.
If we do find functional differences for other targets, it should be possible
to (at least temporarily) restore the shuffle IR with the ExpandReductions IR
pass.
Differential Revision: https://reviews.llvm.org/D96552
This adds the CostKind to getMVEVectorCostFactor, so that it can
automatically account for CodeSize costs, where it returns a cost of 1
not the MVEFactor used for Throughput/Latency. This helps simplify the
caller code and allows us to get the codesize cost more correct in more
cases.
This changes which of the getScalarizationOverhead overloads is used in
the gather/scatter cost to use the base variant directly, not relying on
the version using heuristics on the number of args with no args
provided. It should still produce the same costs for scalarized
gathers/scatters.
This reverts commit 502a67dd7f.
This expose a failure in test-suite build on PowerPC,
revert to unblock buildbot first,
Dave will re-commit in https://reviews.llvm.org/D96287.
Thanks Dave.
getIntrinsicInstrCost takes a IntrinsicCostAttributes holding various
parameters of the intrinsic being costed. It can either be called with a
scalar intrinsic (RetTy==Scalar, VF==1), with a vector instruction
(RetTy==Vector, VF==1) or from the vectorizer with a scalar type and
vector width (RetTy==Scalar, VF>1). A RetTy==Vector, VF>1 is considered
an error. Both of the vector modes are expected to be treated the same,
but because this is confusing many backends end up getting it wrong.
Instead of trying work with those two values separately this removes the
VF parameter, widening the RetTy/ArgTys by VF used called from the
vectorizer. This keeps things simpler, but does require some other
modifications to keep things consistent.
Most backends look like this will be an improvement (or were not using
getIntrinsicInstrCost). AMDGPU needed the most changes to keep the code
from c230965ccf working. ARM removed the fix in
dfac521da1, webassembly happens to get a fixup for an SLP cost
issue and both X86 and AArch64 seem to now be using better costs from
the vectorizer.
Differential Revision: https://reviews.llvm.org/D95291
The MVE VLD2/4 and VST2/4 instructions require the pointer to be aligned
to at least the size of the element type. This adds a check for that
into the ARM lowerInterleavedStore and lowerInterleavedLoad functions,
not creating the intrinsics if they are invalid for the alignment of
the load/store.
Unfortunately this is one of those bug fixes that does effect some
useful codegen, as we were able to sometimes do some nice lowering of
q15 types. But they can cause problem with low aligned pointers.
Differential Revision: https://reviews.llvm.org/D95319
This adds cost modelling for the inloop vectorization added in
745bf6cf44. Up until now they have been modelled as the original
underlying instruction, usually an add. This happens to works OK for MVE
with instructions that are reducing into the same type as they are
working on. But MVE's instructions can perform the equivalent of an
extended MLA as a single instruction:
%sa = sext <16 x i8> A to <16 x i32>
%sb = sext <16 x i8> B to <16 x i32>
%m = mul <16 x i32> %sa, %sb
%r = vecreduce.add(%m)
->
R = VMLADAV A, B
There are other instructions for performing add reductions of
v4i32/v8i16/v16i8 into i32 (VADDV), for doing the same with v4i32->i64
(VADDLV) and for performing a v4i32/v8i16 MLA into an i64 (VMLALDAV).
The i64 are particularly interesting as there are no native i64 add/mul
instructions, leading to the i64 add and mul naturally getting very
high costs.
Also worth mentioning, under NEON there is the concept of a sdot/udot
instruction which performs a partial reduction from a v16i8 to a v4i32.
They extend and mul/sum the first four elements from the inputs into the
first element of the output, repeating for each of the four output
lanes. They could possibly be represented in the same way as above in
llvm, so long as a vecreduce.add could perform a partial reduction. The
vectorizer would then produce a combination of in and outer loop
reductions to efficiently use the sdot and udot instructions. Although
this patch does not do that yet, it does suggest that separating the
input reduction type from the produced result type is a useful concept
to model. It also shows that a MLA reduction as a single instruction is
fairly common.
This patch attempt to improve the costmodelling of in-loop reductions
by:
- Adding some pattern matching in the loop vectorizer cost model to
match extended reduction patterns that are optionally extended and/or
MLA patterns. This marks the cost of the reduction instruction correctly
and the sext/zext/mul leading up to it as free, which is otherwise
difficult to tell and may get a very high cost. (In the long run this
can hopefully be replaced by vplan producing a single node and costing
it correctly, but that is not yet something that vplan can do).
- getExtendedAddReductionCost is added to query the cost of these
extended reduction patterns.
- Expanded the ARM costs to account for these expanded sizes, which is a
fairly simple change in itself.
- Some minor alterations to allow inloop reduction larger than the highest
vector width and i64 MVE reductions.
- An extra InLoopReductionImmediateChains map was added to the vectorizer
for it to efficiently detect which instructions are reductions in the
cost model.
- The tests have some updates to show what I believe is optimal
vectorization and where we are now.
Put together this can greatly improve performance for reduction loop
under MVE.
Differential Revision: https://reviews.llvm.org/D93476
It turns out the vectorizer calls the getIntrinsicInstrCost functions
with a scalar return type and vector VF. This updates the costmodel to
handle that, still producing the correct vector costs.
A vectorizer test is added to show it vectorizing at the correct factor
again.
This adds some basic MVE sadd_sat/ssub_sat/uadd_sat/usub_sat costs,
based on when the instruction is legal. With smaller than legal types
that are promoted we generate shr(qadd(shl, shl)), so the cost is 4
appropriately.
Differential Revision: https://reviews.llvm.org/D94958
We did not have specific costs for larger than legal truncates that were
not otherwise cheap (where they were next to stores, for example). As
MVE does not have a dedicated instruction for them (and we do not use
loads/stores yet), they should be expensive as they get expanded to a
series of lane moves.
Differential Revision: https://reviews.llvm.org/D94260
David Green