This tends to increase code size but more importantly it reduces vgpr
usage, and could avoid costly readfirstlanes if the result needs to be
in an sgpr.
Differential Revision: https://reviews.llvm.org/D88245
This reverts commit ca907bfb57.
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
Clustering loads has caching benefits, but as far as I know there is no
advantage to clustering stores on any AMDGPU subtargets.
The disadvantage is that it tends to increase register pressure and
restricts scheduling freedom.
Differential Revision: https://reviews.llvm.org/D85530
Unfortunately this ends up not working as expected on targets with
16-bit operations due to AMDGPUCodeGenPrepare's promotion of uniform
16-bit ops to i32.
The vector case annoyingly requires switching the checked opcode,
since constants for vectors aren't directly handled.
I also need to think more carefully about whether this is valid for i1.
Fix 64-bit copy to SCC by restricting the pattern resulting
in such a copy to subtargets supporting 64-bit scalar compare,
and mapping the copy to S_CMP_LG_U64.
Before introducing the S_CSELECT pattern with explicit SCC
(0045786f14), there was no need
for handling 64-bit copy to SCC ($scc = COPY sreg_64).
The proposed handling to read only the low bits was however
based on a false premise that it is only one bit that matters,
while in fact the copy source might be a vector of booleans and
all bits need to be considered.
The practical problem of mapping the 64-bit copy to SCC is that
the natural instruction to use (S_CMP_LG_U64) is not available
on old hardware. Fix it by restricting the problematic pattern
to subtargets supporting the instruction (hasScalarCompareEq64).
Differential Revision: https://reviews.llvm.org/D85207
Add handling of s_andn2 and mask of 0.
This eliminates redundant instructions from uniform control flow.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D83641
Summary:
Add patterns to select s_cselect in the isel.
Handle more cases of implicit SCC accesses in si-fix-sgpr-copies
to allow new patterns to work.
Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, asbirlea, kerbowa, llvm-commits
Tags: #llvm
Re-commit D81925 with a bugfix D82370.
Differential Revision: https://reviews.llvm.org/D81925
Differential Revision: https://reviews.llvm.org/D82370
Summary:
Make use of both the - (1) clustered bytes and (2) cluster length, to decide on
the max number of mem ops that can be clustered. On an average, when loads
are dword or smaller, consider `5` as max threshold, otherwise `4`. This
heuristic is purely based on different experimentation conducted, and there is
no analytical logic here.
Reviewers: foad, rampitec, arsenm, vpykhtin
Reviewed By: rampitec
Subscribers: llvm-commits, kerbowa, hiraditya, t-tye, Anastasia, tpr, dstuttard, yaxunl, nhaehnle, wdng, jvesely, kzhuravl, thakis
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82393
Summary:
Add patterns to select s_cselect in the isel.
Handle more cases of implicit SCC accesses in si-fix-sgpr-copies
to allow new patterns to work.
Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, asbirlea, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81925
Summary:
Make use of both the - (1) clustered bytes and (2) cluster length, to decide on
the max number of mem ops that can be clustered. On an average, when loads
are dword or smaller, consider `5` as max threshold, otherwise `4`. This heuristic
is purely based on different experimentation conducted, and there is no analytical
logic here.
Reviewers: foad, rampitec, arsenm, vpykhtin
Reviewed By: foad, rampitec
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, Anastasia, t-tye, hiraditya, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D81085
SelectMOVRELOffset prevents peeling of a constant from an index
if final base could be negative. isBaseWithConstantOffset() succeeds
if a value is an "add" or "or" operator. In case of "or" it shall
be an add-like "or" which never changes a sign of the sum given a
non-negative offset. I.e. we can safely allow peeling if operator is
an "or".
Differential Revision: https://reviews.llvm.org/D79898
We can produce such vectors in the Promote Alloca pass,
but we are unable to use movrel to operate it and lower
via scratch. Making it legal makes SI_INDIRECT patterns
work.
There is more work to do in subsequent changes:
1. We initialize m0 twice to access each dword. It shall
be possible to only do it once and increment base register
number instead.
2. We also need v16i64/v16f64 but these first need to be
added to tablegen.
Differential Revision: https://reviews.llvm.org/D79808
Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in
the entry function prologue. This allows us to removes the scratch wave
offset register from the calling convention ABI.
As part of this change, allow the use of an inline constant zero for the
SOffset of MUBUF instructions accessing the stack in entry functions
when a frame pointer is not requested/required. Entry functions with
calls still need to set up the calling convention ABI stack pointer
register, and reference it in order to address arguments of called
functions. The ABI stack pointer register remains unswizzled, but is now
wave-relative instead of queue-relative.
Non-entry functions also use an inline constant zero SOffset for
wave-relative scratch access, but continue to use the stack and frame
pointers as before. When the stack or frame pointer is converted to a
swizzled offset it is now scaled directly, as the scratch wave offset no
longer needs to be subtracted first.
Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling
convention.
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D75138
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
Summary:
- `dead-mi-elimination` assumes MIR in the SSA form and cannot be
arranged after phi elimination or DeSSA. It's enhanced to handle the
dead register definition by skipping use check on it. Once a register
def is `dead`, all its uses, if any, should be `undef`.
- Re-arrange the DIE in RA phase for AMDGPU by placing it directly after
`detect-dead-lanes`.
- Many relevant tests are refined due to different register assignment.
Reviewers: rampitec, qcolombet, sunfish
Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72709
Summary:
Mem op clustering adds a weak edge in the DAG between two loads or
stores that should be clustered, but the direction of this edge is
pretty arbitrary (it depends on the sort order of MemOpInfo, which
represents the operands of a load or store). This often means that two
loads or stores will get reordered even if they would naturally have
been scheduled together anyway, which leads to test case churn and goes
against the scheduler's "do no harm" philosophy.
The fix makes sure that the direction of the edge always matches the
original code order of the instructions.
Reviewers: atrick, MatzeB, arsenm, rampitec, t.p.northover
Subscribers: jvesely, wdng, nhaehnle, kristof.beyls, hiraditya, javed.absar, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D72706
"Divergence driven ISel. Assign register class for cross block values
according to the divergence."
that discovered the design flaw leading to several issues that
required to be solved before.
This change reverts AMDGPU specific changes and keeps common part
unaffected.
llvm-svn: 362749
Details: To make instruction selection really divergence driven it is necessary to assign
the correct register classes to the cross block values beforehand. For the divergent targets
same value type requires different register classes dependent on the value divergence.
Reviewers: rampitec, nhaehnle
Differential Revision: https://reviews.llvm.org/D59990
This commit was reverted because of the build failure.
The reason was mlformed patch.
Build failure fixed.
llvm-svn: 361741
Details: To make instruction selection really divergence driven it is necessary to assign
the correct register classes to the cross block values beforehand. For the divergent targets
same value type requires different register classes dependent on the value divergence.
Reviewers: rampitec, nhaehnle
Differential Revision: https://reviews.llvm.org/D59990
llvm-svn: 361644
Summary:
Incorrect code was generated when lowering insertelement operations
for vectors with 8 or 16 bit elements. The value being inserted was
not adjusted for the position of the element within the 32 bit word
and so only the low element within each 32 bit word could receive
the intended value.
Fixed by simply replicating the value to each element of a
congruent vector before the mask and or operation used to
update the intended element.
A number of affected LIT tests have been updated appropriately.
before the mask & or into the intended
Reviewers: arsenm, nhaehnle
Reviewed By: arsenm
Subscribers: llvm-commits, arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D57588
llvm-svn: 352885
I've extended the load/store optimizer to be able to produce dwordx3
loads and stores, This change allows many more load/stores to be combined,
and results in much more optimal code for our hardware.
Differential Revision: https://reviews.llvm.org/D54042
llvm-svn: 348937
This patch adds support for S_ANDN2, S_ORN2 32-bit and 64-bit instructions and adds splits to move them to the vector unit (for which there is no equivalent instruction). It modifies the way that the more complex scalar instructions are lowered to vector instructions by first breaking them down to sequences of simpler scalar instructions which are then lowered through the existing code paths. The pattern for S_XNOR has also been updated to apply inversion to one input rather than the output of the XOR as the result is equivalent and may allow leaving the NOT instruction on the scalar unit.
A new tests for NAND, NOR, ANDN2 and ORN2 have been added, and existing tests now hit the new instructions (and have been modified accordingly).
Differential: https://reviews.llvm.org/D54714
llvm-svn: 347877
This allows to avoid scratch use or indirect VGPR addressing for
small vectors.
Differential Revision: https://reviews.llvm.org/D54606
llvm-svn: 347231
The SchedModel allows the addition of ReadAdvances to express that certain
operands of the instructions are needed at a later point than the others.
RegAlloc may add pseudo operands that are not part of the instruction
descriptor, and therefore cannot have any read advance entries. This meant
that in some cases the desired read advance was nullified by such a pseudo
operand, which still had the original latency.
This patch fixes this by making sure that such pseudo operands get a zero
latency during DAG construction.
Review: Matthias Braun, Ulrich Weigand.
https://reviews.llvm.org/D49671
llvm-svn: 345606
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
Summary:
Also explicitly port over some tests in llvm.amdgcn.image.* that were
missing. Some tests are removed because they no longer apply (i.e.
explicitly testing building an address vector via insertelement).
This is in preparation for the eventual removal of the old-style
intrinsics.
Some additional notes:
- constant-address-space-32bit.ll: change some GCN-NEXT to GCN because
the instruction schedule was subtly altered
- insert_vector_elt.ll: the old test didn't actually test anything,
because %tmp1 was not used; remove the load, because it doesn't work
(Because of the amdgpu_ps calling convention? In any case, it's
orthogonal to what the test claims to be testing.)
Change-Id: Idfa99b6512ad139e755e82b8b89548ab08f0afcf
Reviewers: arsenm, rampitec
Subscribers: MatzeB, qcolombet, kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D48018
llvm-svn: 335229
This has two main components. First, widen
widen short constant loads in DAG when they have
the correct alignment. This is already done a bit in
AMDGPUCodeGenPrepare, since that has access to
DivergenceAnalysis. This can't help kernarg loads
created in the DAG. Start to use DAG divergence analysis
to help this case.
The second part is to avoid kernel argument lowering
breaking the alignment of short vector elements because
calling convention lowering wants to split everything
into legal register types.
When loading a split type, load the nearest 4-byte aligned
segment and shift to get the desired bits. This extra
load of the earlier argument piece ends up merging,
and the bit extract hopefully folds out.
There are a number of improvements and regressions with
this, but I think as-is this is a better compromise between
several of the worst parts of SelectionDAG.
Particularly when i16 is legal, this produces worse code
for i8 and i16 element vector kernel arguments. This is
partially due to the very weak load merging the DAG does.
It only looks for fairly specific combines between pairs
of loads which no longer appear. In particular this
causes v4i16 loads to be split into 2 components when
previously the two halves were merged.
Worse, because of the newly introduced shifts, there
is a lot more unnecessary vector packing and unpacking code
emitted. At least some of this is due to reporting
false for isTypeDesirableForOp for i16 as a workaround for
the lack of divergence information in the DAG. The cases
where this happens it doesn't actually matter, but the
relevant code in SimplifyDemandedBits doens't have the context
to know to ignore this.
The use of the scalar cache is probably more important
than the mess of mostly scalar instructions doing this packing
and unpacking. Future work can fix this, possibly by making better
use of the new DAG divergence information for controlling promotion
decisions, or adding another version of shift + trunc + shift
combines that doesn't only know about the used types.
llvm-svn: 334180
This usually results in better code. Fixes using
inline asm with short2, and also fixes having a different
ABI for function parameters between VI and gfx9.
Partially cleans up the mess used for lowering of the d16
operations. Making v4f16 legal will help clean this up more,
but this requires additional work.
llvm-svn: 332953
Two issues found when doing codegen for splitting vector with non-zero alloca addr space:
DAGTypeLegalizer::SplitVecRes_INSERT_VECTOR_ELT/SplitVecOp_EXTRACT_VECTOR_ELT uses dummy pointer info for creating
SDStore. Since one pointer operand contains multiply and add, InferPointerInfo is unable to
infer the correct pointer info, which ends up with a dummy pointer info for the target to lower
store and results in isel failure. The fix is to introduce MachinePointerInfo::getUnknownStack to
represent MachinePointerInfo which is known in alloca address space but without other information.
TargetLowering::getVectorElementPointer uses value type of pointer in addr space 0 for
multiplication of index and then add it to the pointer. However the pointer may be in an addr
space which has different size than addr space 0. The fix is to use the pointer value type for
index multiplication.
Differential Revision: https://reviews.llvm.org/D39758
llvm-svn: 319622
Try to avoid mutually exclusive features. Don't use
a real default GPU, and use a fake "generic". The goal
is to make it easier to see which set of features are
incompatible between feature strings.
Most of the test changes are due to random scheduling changes
from not having a default fullspeed model.
llvm-svn: 310258
Mircea Trofin