If we have a variable shift amount and the demanded mask has leading
zeros, we can propagate those leading zeros to not demand those bits
from operand 0. This can allow zero_extend/sign_extend to become
any_extend. This pattern can occur due to C integer promotion rules.
This transform is already done by InstCombineSimplifyDemanded.cpp where
sign_extend can be turned into zero_extend for example.
Reviewed By: spatel, foad
Differential Revision: https://reviews.llvm.org/D121833
Since the divergence-driven instruction selection has been enabled for AMDGPU,
all the uniform instructions are expected to be selected to SALU form, except those not having one.
VGPR to SGPR copies appear in MIR to connect values producers and consumers. This change implements an algorithm
that evolves a reasonable tradeoff between the profit achieved from keeping the uniform instructions in SALU form
and overhead introduced by the data transfer between the VGPRs and SGPRs.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D128252
Fold immediates regardless of how many uses they have. This is expected
to increase overall code size, but decrease register usage.
Differential Revision: https://reviews.llvm.org/D114644
Use shufflevector to do the subvector extracts. This allows a lot more
load merging on AMDGPU and also on NVPTX when <2 x half> is involved.
Differential Revision: https://reviews.llvm.org/D117219
Using a BufferSize of one for memory ProcResources will result in better
ILP since it more accurately models the dependencies between memory ops
and their consumers on an in-order processor. After this change, the
scheduler will treat the data edges from loads as blocking so that
stalls are guaranteed when waiting for data to be retreaved from memory.
Since we don't actually track waitcnt here, this should do a better job
at modeling their behavior.
Practically, this means that the scheduler will trigger the 'STALL'
heuristic more often.
This type of change needs to be evaluated experimentally. Preliminary
results are positive.
Fixes: SWDEV-282962
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D114777
Change VOP_PAT_GEN to default to not generating an instruction selection
pattern for the VOP2 (e32) form of an instruction, only for the VOP3
(e64) form. This allows SIFoldOperands maximum freedom to fold copies
into the operands of an instruction, before SIShrinkInstructions tries
to shrink it back to the smaller encoding.
This affects the following VOP2 instructions:
v_min_i32
v_max_i32
v_min_u32
v_max_u32
v_and_b32
v_or_b32
v_xor_b32
v_lshr_b32
v_ashr_i32
v_lshl_b32
A further cleanup could simplify or remove VOP_PAT_GEN, since its
optional second argument is never used.
Differential Revision: https://reviews.llvm.org/D114252
The existing constrained shift PatFrags only dealt with masked shift
from OpenCL front-ends. This change copies the
X86DAGToDAGISel::isUnneededShiftMask() function to AMDGPU and uses it in
the shift PatFrag predicates.
Differential Revision: https://reviews.llvm.org/D113448
This simple heuristic uses the estimated live range length combined
with the number of registers in the class to switch which heuristic to
use. This was taking the raw number of registers in the class, even
though not all of them may be available. AMDGPU heavily relies on
dynamically reserved numbers of registers based on user attributes to
satisfy occupancy constraints, so the raw number is highly misleading.
There are still a few problems here. In the original testcase that
made me notice this, the live range size is incorrect after the
scheduler rearranges instructions, since the instructions don't have
the original InstrDist offsets. Additionally, I think it would be more
appropriate to use the number of disjointly allocatable registers in
the class. For the AMDGPU register tuples, there are a large number of
registers in each tuple class, but only a small fraction can actually
be allocated at the same time since they all overlap with each
other. It seems we do not have a query that corresponds to the number
of independently allocatable registers. Relatedly, I'm still debugging
some allocation failures where overlapping tuples seem to not be
handled correctly.
The test changes are mostly noise. There are a handful of x86 tests
that look like regressions with an additional spill, and a handful
that now avoid a spill. The worst looking regression is likely
test/Thumb2/mve-vld4.ll which introduces a few additional
spills. test/CodeGen/AMDGPU/soft-clause-exceeds-register-budget.ll
shows a massive improvement by completely eliminating a large number
of spills inside a loop.
Use GCNHazardRecognizer in postra sched.
Updated tests for the new schedules.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D109536
Change-Id: Ia86ba2ae168f12fb34b4d8efdab491f84d936cde
This is to allow 64 bit constant rematerialization. If a constant
is split into two separate moves initializing sub0 and sub1 like
now RA cannot rematerizalize a 64 bit register.
This gives 10-20% uplift in a set of huge apps heavily using double
precession math.
Fixes: SWDEV-292645
Differential Revision: https://reviews.llvm.org/D104874
Based off a discussion on D89281 - where the AARCH64 implementations were being replaced to use funnel shifts.
Any target that has efficient funnel shift lowering can handle the shift parts expansion using the same expansion, avoiding a lot of duplication.
I've generalized the X86 implementation and moved it to TargetLowering - so far I've found that AARCH64 and AMDGPU benefit, but many other targets (ARM, PowerPC + RISCV in particular) could easily use this with a few minor improvements to their funnel shift lowering (or the folding of their target ops that funnel shifts lower to).
NOTE: I'm trying to avoid adding full SHIFT_PARTS legalizer handling as I think it might actually be possible to remove these opcodes in the medium-term and use funnel shift / libcall expansion directly.
Differential Revision: https://reviews.llvm.org/D101987
Treat a non-atomic volatile load and store as a relaxed atomic at
system scope for the address spaces accessed. This will ensure all
relevant caches will be bypassed.
A volatile atomic is not changed and still only bypasses caches upto
the level specified by the SyncScope operand.
Differential Revision: https://reviews.llvm.org/D94214
This reverts commit ca907bfb57d8ad3ec3bcc2cff2abab7b1b933af6.
According to michel.daenzer,
> This completely broke the Mesa radeonsi driver on Navi 14. Xorg +
> xterm come up with major corruption & psychedelic colours.
When memory operations are outstanding on function calls, either the
caller or the callee can insert a waitcnt to ensure that all reads are
finished.
Calls need some time to be executed, so if the callee inserts the
waitcnt, filling the instruction buffer and waiting for memory will be
interleaved, hiding some latency. This comes at the cost of having a
waitcnt inside functions that may not be needed as no memory operations
are outstanding.
For function calls, this is already implemented. The same principal
applies to returns: If the caller inserts a waitcnt after the call, the
callee does not have to wait and the return and memory operation can be
run in parallel.
This commit implements waiting in the caller after returning from a
function call.
Differential Revision: https://reviews.llvm.org/D87674
tryLatency compares two sched candidates. For the top zone it prefers
the one with lesser depth, but only if that depth is greater than the
total latency of the instructions we've already scheduled -- otherwise
its latency would be hidden and there would be no stall.
Unfortunately it only tests the depth of one of the candidates. This can
lead to situations where the TopDepthReduce heuristic does not kick in,
but a lower priority heuristic chooses the other candidate, whose depth
*is* greater than the already scheduled latency, which causes a stall.
The fix is to apply the heuristic if the depth of *either* candidate is
greater than the already scheduled latency.
All this also applies to the BotHeightReduce heuristic in the bottom
zone.
Differential Revision: https://reviews.llvm.org/D72392
VMEM loads of the same type (sampler vs no sampler) are guaranteed to
write their result registers in order, so there is no need for an
s_waitcnt even if they write to overlapping vgprs.
Differential Revision: https://reviews.llvm.org/D79176
Summary:
pickNodeBidirectional tried to compare the best top candidate and the
best bottom candidate by examining TopCand.Reason and BotCand.Reason.
This is unsound because, after calling pickNodeFromQueue, Cand.Reason
does not reflect the most important reason why Cand was chosen. Rather
it reflects the most recent reason why it beat some other potential
candidate, which could have been for some low priority tie breaker
reason.
I have seen this cause problems where TopCand is a good candidate, but
because TopCand.Reason is ORDER (which is very low priority) it is
repeatedly ignored in favour of a mediocre BotCand. This is not how
bidirectional scheduling is supposed to work.
To fix this I changed the code to always compare TopCand and BotCand
directly, like the generic implementation of pickNodeBidirectional does.
This removes some uncommented AMDGPU-specific logic; if this logic turns
out to be important then perhaps it could be moved into an override of
tryCandidate instead.
Graphics shader benchmarking on gfx10 shows a lot more positive than
negative effects from this change.
Reviewers: arsenm, tstellar, rampitec, kzhuravl, vpykhtin, dstuttard, tpr, atrick, MatzeB
Subscribers: jvesely, wdng, nhaehnle, yaxunl, t-tye, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D68338
We are relying on atrificial DAG edges inserted by the
MemOpClusterMutation to keep loads and stores together in the
post-RA scheduler. This does not work all the time since it
allows to schedule a completely independent instruction in the
middle of the cluster.
Removed the DAG mutation and added pass to bundle already
clustered instructions. These bundles are unpacked before the
memory legalizer because it does not work with bundles but also
because it allows to insert waitcounts in the middle of a store
cluster.
Removing artificial edges also allows a more relaxed scheduling.
Differential Revision: https://reviews.llvm.org/D72737
As detailed in https://blog.regehr.org/archives/1709 we don't make use of the known leading/trailing zeros for shifted values in cases where we don't know the shift amount value.
This patch adds support to SelectionDAG::ComputeKnownBits to use KnownBits::countMinTrailingZeros and countMinLeadingZeros to set the minimum guaranteed leading/trailing known zero bits.
Differential Revision: https://reviews.llvm.org/D72573
See https://reviews.llvm.org/D47106 for details.
Reviewed By: probinson
Differential Revision: https://reviews.llvm.org/D47171
This commit drops that patch's changes to:
llvm/test/CodeGen/NVPTX/f16x2-instructions.ll
llvm/test/CodeGen/NVPTX/param-load-store.ll
For some reason, the dos line endings there prevent me from commiting
via the monorepo. A follow-up commit (not via the monorepo) will
finish the patch.
llvm-svn: 336843
This replaces most argument uses with loads, but for
now not all.
The code in SelectionDAG for calling convention lowering
is actively harmful for amdgpu_kernel. It attempts to
split the argument types into register legal types, which
results in low quality code for arbitary types. Since
all kernel arguments are passed in memory, we just want the
raw types.
I've tried a couple of methods of mitigating this in SelectionDAG,
but it's easier to just bypass this problem alltogether. It's
possible to hack around the problem in the initial lowering,
but the real problem is the DAG then expects to be able to use
CopyToReg/CopyFromReg for uses of the arguments outside the block.
Exposing the argument loads in the IR also has the advantage
that the LoadStoreVectorizer can merge them.
I'm not sure the best approach to dealing with the IR
argument list is. The patch as-is just leaves the IR arguments
in place, so all the existing code will still compute the same
kernarg size and pointlessly lowers the arguments.
Arguably the frontend should emit kernels with an empty argument
list in the first place. Alternatively a dummy array could be
inserted as a single argument just to reserve space.
This does have some disadvantages. Local pointer kernel arguments can
no longer have AssertZext placed on them as the equivalent !range
metadata is not valid on pointer typed loads. This is mostly bad
for SI which needs to know about the known bits in order to use the
DS instruction offset, so in this case this is not done.
More importantly, this skips noalias arguments since this pass
does not yet convert this to the equivalent !alias.scope and !noalias
metadata. Producing this metadata correctly seems to be tricky,
although this logically is the same as inlining into a function which
doesn't exist. Additionally, exposing these loads to the vectorizer
may result in degraded aliasing information if a pointer load is
merged with another argument load.
I'm also not entirely sure this is preserving the current clover
ABI, although I would greatly prefer if it would stop widening
arguments and match the HSA ABI. As-is I think it is extending
< 4-byte arguments to 4-bytes but doesn't align them to 4-bytes.
llvm-svn: 335650
r600 uses dummy pointer info for lowering load/store. Since dummy pointer info
assumes address space 0, this causes isel failure when temporary load/store SDNodes
are generated for amdgiz environment.
Since the offest is not constant, FixedStack pseudo source value cannot be used
to create the pointer info. This patch creates pointer info using llvm undef value.
At least this provides correct address space so that isel can be done correctly.
Differential Revision: https://reviews.llvm.org/D39698
llvm-svn: 317862
We already have a combine for this pattern when the input to shl is add, so we just need to enable the transformation when the input is or.
Original patch by @tstellar
Differential Revision: https://reviews.llvm.org/D19325
llvm-svn: 313251
This fixes 17 of the 41 -verify-machineinstrs test failures identified in PR33045
Differential Revision: https://reviews.llvm.org/D33451
llvm-svn: 303691
If workgroup size is known inform llvm about range returned by local
id and local size queries.
Differential Revision: https://reviews.llvm.org/D31804
llvm-svn: 300102
Currently the default C calling convention functions are treated
the same as compute kernels. Make this explicit so the default
calling convention can be changed to a non-kernel.
Converted with perl -pi -e 's/define void/define amdgpu_kernel void/'
on the relevant test directories (and undoing in one place that actually
wanted a non-kernel).
llvm-svn: 298444
This switches to the workaround that HSA defaults to
for the mesa path.
This should be applied to the 4.0 branch.
Patch by Vedran Miletić <vedran@miletic.net>
llvm-svn: 292982
This breaks the tests that were meant for testing
64-bit inline immediates, so move those to shl where
they won't be broken up.
This should be repeated for the other related bit ops.
llvm-svn: 258095
Re-comitting with a change that avoids undefined uses getting put into
the VRegUses list.
The new algorithm remembers the uses encountered while walking backwards
until a matching def is found. Contrary to the previous version this:
- Works without LiveIntervals being available
- Allows to increase the precision to subregisters/lanemasks
(not used for now)
The changes in the AMDGPU tests are necessary because the R600 scheduler
is not stable with respect to the order of nodes in the ready queues.
Differential Revision: http://reviews.llvm.org/D9068
llvm-svn: 254683
This works mostly fine but breaks some stage 1 builders when compiling
compiler-rt on i386. Revert for further investigation as I can't see an
obvious cause/fix.
This reverts commit r254577.
llvm-svn: 254586
The new algorithm remembers the uses encountered while walking backwards
until a matching def is found. Contrary to the previous version this:
- Works without LiveIntervals being available
- Allows to increase the precision to subregisters/lanemasks
(not used for now)
The changes in the AMDGPU tests are necessary because the R600 scheduler
is not stable with respect to the order of nodes in the ready queues.
Differential Revision: http://reviews.llvm.org/D9068
llvm-svn: 254577
This can be done only with moves which theoretically
will optimize better later.
Although this transform increases the instruction count,
it should be code size / cycle count neutral in the worst
VALU case. It also seems to slightly improve a couple
of testcases due to other DAG combines this exposes.
This is probably slightly worse for the SALU case, so
it might be better to handle this during moveToVALU,
although then you lose some simplifications like
the load width reducing in the simple testcase.
llvm-svn: 242177
