Backwards search
The sext.w removal pass (before the new patch) checks if the input to sext.w is already in sign-extended form, so it can eliminate it. It does that by checking every definition/source that reaches the sext.w is an instruction that produces a sign-extended value, either by definition (e.g. ADDW), or it propagates sign-extension (e.g. OR) so we check its sources recursively.
Forward search
Sometimes, one of the sources is an instruction that doesn't always produce a sign-extended value, but it has a W-version that does (e.g. ADD / ADDW). If we transform the ADD to ADDW, the sext.w can be removed (assuming other def paths are satisfied), but this transformation is sound only if every use of this ADD/W only reqruires the lower 32-bits either directly (like sll %x, 32) or they propagate dependency (lower word of output only depends on lower word of input) so we check its uses recursively.
When searching backwards, if an instruction that can be replaced with W-variant is encountered, this pass runs the forward search to verify it can be replaced, then adds it to a list of fixable instructions. After verifying all paths, it replaces the instruction and removes the sext.w.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D119928
At some point in instruction selection, A2_tfrsi Constant:i32<...> was
created, where the "Constant" came from SelectAnyInt. Since it wasn't
a TargetConstant, it was selected again, leading to
%vreg = A2_tfrsi ...
... = A2_tfrsi %vreg
which is not a valid code.
This patch adds custom MIR operand comments to VTYPE immediate operands
in VSETVLI instructions and SEW/LMUL operands in vector codegen pseudo
instructions. The result is intended to be more human-readable and
hopefully maintainable when working with MIR, particularly when
writing or reading test cases.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D124187
This is a very specific fold to fix an upstream poor codegen issue.
InstCombine has the much more flexible pushFreezeToPreventPoisonFromPropagating but I don't think we're quite there with DAG/TLI handling for canCreateUndefOrPoison/isGuaranteedNotToBeUndefOrPoison value tracking yet.
Fixes#54911
Differential Revision: https://reviews.llvm.org/D124185
Currently metadata is inserted in a late pass which is lowered
to an AssertZext. The metadata would be more useful if it was
inserted earlier after inlining, but before codegen.
Probably shouldn't change anything now. Just replacing the
late metadata annotation needs more work, since we lose
out on optimizations after these are lowered to CopyFromReg.
Seems to be slightly better than relying on the AssertZext from the
metadata. The test change in cvt_f32_ubyte.ll is a quirk from it using
-start-before=amdgpu-isel instead of running the usual codegen
pipeline.
The most common situation where G_ASSERT_ZEXT appears for AMDGPU is a
copy from a physical register, which happens to use set the actual
register class on the virtual register. After copy coalescing, the
assert's source operand had a vreg with a set class. The verifier was
strictly rejecting cases where the set class/bank weren't an exact
match. Additionally, RegBankSelect was also expecting a register bank
to be set on the register, not a class.
This is much stricter than regular copies so relax this behavior. This
now allows these 2 cases:
1. Source register has either class or bank, and the result does not
2. Source register has a register class, and the result is a register
with a matching bank.
This should avoid needing some kind of special handling to avoid
violating this constraint when folding copies.
This reverts part of https://reviews.llvm.org/D124224 that causes
an assert because the register allocator triggers a pathological
situation where there's no safe way to insert a zeroing MOVPFRX
instruction.
Add DestructiveBinaryComm* patterns for ORR, EOR, AND and BIC.
The above instructions requires that the source and destination registers are
equal, so use movprfx should be beneficial to performance.
note: BIC (i.e. A & ~B) is not a commutative operation.
Reviewed By: paulwalker-arm, david-arm
Differential Revision: https://reviews.llvm.org/D124224
autiasp, autibsp instructions are the counterpart of paciasp/pacibsp instructions
therefore let's emit .cfi_negate_ra_state for these too.
In case of Armv8.3 instruction set the retaa/retbb will do the return and authentication
in one step here we can't emit the . cfi_negate_ra_state because that would be point after
the ret* instruction.
Reviewed By: nickdesaulniers, MaskRay
Differential Revision: https://reviews.llvm.org/D111780
We saw a failure caused by unwinding with incomplete CFIs, so we
can't outline CFI instructions when they are needed in EH.
This is a recommit of 0d40688, which was reverted in ce83883 as
related precommit test 360d44e caused some errors.
Reviewed By: luismarques
Differential Revision: https://reviews.llvm.org/D122634
This is a recommit of 360d44e, which was reverted
in b1620d4 because it caused some errors due to no
`nounwind` attrs in `machine-outliner-cfi.mir`.
Reviewed By: luismarques
Differential Revision: https://reviews.llvm.org/D123364
This is to avoid relying on the post-isel hook.
This change also enable the saddr pattern selection for atomic
intrinsics in GlobalISel.
Differential Revision: https://reviews.llvm.org/D123583
This wasn't accounting for the canonicalize of the input, or checking
the output fneg isn't folded as intended. Avoids test failure in
unrelated patch which happens to change register numberings.
Revert "[docs] Fix underline"
Breaks a lot of asan tests in google.
This reverts commit 365c3e85bced1fb56c2d94adc34bff7a94abe4a6.
This reverts commit 78a784bea443cdcecf894155ab37893d7a8e8332.
This emits an `st_size` that represents the actual useable size of an object before the redzone is added.
Reviewed By: vitalybuka, MaskRay, hctim
Differential Revision: https://reviews.llvm.org/D123010
The recently announced IBM z16 processor implements the architecture
already supported as "arch14" in LLVM. This patch adds support for
"z16" as an alternate architecture name for arch14.
If there are fewer than 12 trailing zeros, we'll try to use an ADDI
at the end of the sequence. If we strip trailing zeros and end the
sequence with a SLLI we might find a shorter sequence.
Differential Revision: https://reviews.llvm.org/D124148
Here is the performance data:
```
Using platform: AMD Accelerated Parallel Processing
Using device: gfx900:xnack-
ds_write_b64 aligned by 8: 3.2 sec
ds_write2_b32 aligned by 8: 3.2 sec
ds_write_b16 * 4 aligned by 8: 7.0 sec
ds_write_b8 * 8 aligned by 8: 13.2 sec
ds_write_b64 aligned by 1: 7.3 sec
ds_write2_b32 aligned by 1: 7.5 sec
ds_write_b16 * 4 aligned by 1: 14.0 sec
ds_write_b8 * 8 aligned by 1: 13.2 sec
ds_write_b64 aligned by 2: 7.3 sec
ds_write2_b32 aligned by 2: 7.5 sec
ds_write_b16 * 4 aligned by 2: 7.1 sec
ds_write_b8 * 8 aligned by 2: 13.3 sec
ds_write_b64 aligned by 4: 4.6 sec
ds_write2_b32 aligned by 4: 3.2 sec
ds_write_b16 * 4 aligned by 4: 7.1 sec
ds_write_b8 * 8 aligned by 4: 13.3 sec
ds_read_b64 aligned by 8: 2.3 sec
ds_read2_b32 aligned by 8: 2.2 sec
ds_read_u16 * 4 aligned by 8: 4.8 sec
ds_read_u8 * 8 aligned by 8: 8.6 sec
ds_read_b64 aligned by 1: 4.4 sec
ds_read2_b32 aligned by 1: 7.3 sec
ds_read_u16 * 4 aligned by 1: 14.0 sec
ds_read_u8 * 8 aligned by 1: 8.7 sec
ds_read_b64 aligned by 2: 4.4 sec
ds_read2_b32 aligned by 2: 7.3 sec
ds_read_u16 * 4 aligned by 2: 4.8 sec
ds_read_u8 * 8 aligned by 2: 8.7 sec
ds_read_b64 aligned by 4: 4.4 sec
ds_read2_b32 aligned by 4: 2.3 sec
ds_read_u16 * 4 aligned by 4: 4.8 sec
ds_read_u8 * 8 aligned by 4: 8.7 sec
Using platform: AMD Accelerated Parallel Processing
Using device: gfx1030
ds_write_b64 aligned by 8: 4.4 sec
ds_write2_b32 aligned by 8: 4.3 sec
ds_write_b16 * 4 aligned by 8: 7.9 sec
ds_write_b8 * 8 aligned by 8: 13.0 sec
ds_write_b64 aligned by 1: 23.2 sec
ds_write2_b32 aligned by 1: 23.1 sec
ds_write_b16 * 4 aligned by 1: 44.0 sec
ds_write_b8 * 8 aligned by 1: 13.0 sec
ds_write_b64 aligned by 2: 23.2 sec
ds_write2_b32 aligned by 2: 23.1 sec
ds_write_b16 * 4 aligned by 2: 7.9 sec
ds_write_b8 * 8 aligned by 2: 13.1 sec
ds_write_b64 aligned by 4: 13.5 sec
ds_write2_b32 aligned by 4: 4.3 sec
ds_write_b16 * 4 aligned by 4: 7.9 sec
ds_write_b8 * 8 aligned by 4: 13.1 sec
ds_read_b64 aligned by 8: 3.5 sec
ds_read2_b32 aligned by 8: 3.4 sec
ds_read_u16 * 4 aligned by 8: 5.3 sec
ds_read_u8 * 8 aligned by 8: 8.5 sec
ds_read_b64 aligned by 1: 13.1 sec
ds_read2_b32 aligned by 1: 22.7 sec
ds_read_u16 * 4 aligned by 1: 43.9 sec
ds_read_u8 * 8 aligned by 1: 7.9 sec
ds_read_b64 aligned by 2: 13.1 sec
ds_read2_b32 aligned by 2: 22.7 sec
ds_read_u16 * 4 aligned by 2: 5.6 sec
ds_read_u8 * 8 aligned by 2: 7.9 sec
ds_read_b64 aligned by 4: 13.1 sec
ds_read2_b32 aligned by 4: 3.4 sec
ds_read_u16 * 4 aligned by 4: 5.6 sec
ds_read_u8 * 8 aligned by 4: 7.9 sec
```
GFX10 exposes a different pattern for sub-DWORD load/store performance
than GFX9. On GFX9 it is faster to issue a single unaligned load or
store than a fully split b8 access, where on GFX10 even a full split
is better. However, this is a theoretical only gain because splitting
an access to a sub-dword level will require more registers and packing/
unpacking logic, so ignoring this option it is better to use a single
64 bit instruction on a misaligned data with the exception of 4 byte
aligned data where ds_read2_b32/ds_write2_b32 is better.
Differential Revision: https://reviews.llvm.org/D123956
Fix isVCC for register that was assigned register class during
inst-selection. This happens when register has multiple uses.
For wave32, uniform i1 to vcc copy was selected like vcc to vcc
copy when uniform i1 had assigned register class.
Uniform i1 register with assigned register class will have s1 LLT,
be defined using G_TRUNC and class will be SReg_32RegClass.
Vcc i1 register with assigned register class will have s1 LLT,
class will be SReg_32RegClass for wave32 and SReg_64RegClass for
wave64 and register will not be defined by G_TRUNC.
Differential Revision: https://reviews.llvm.org/D124163
This fixes the assertion failure "Loop in the Block Graph!".
SIMachineScheduler groups instructions into blocks (also referred to
as coloring or groups) and then performs a two-level scheduling:
inter-block scheduling, and intra-block scheduling.
This approach requires that the dependency graph on the blocks which
is obtained by contracting the blocks in the original dependency graph
is acyclic. In other words: Whenever A and B end up in the same block,
all vertices on a path from A to B must be in the same block.
When compiling an example consisting of an export followed by
a buffer store, we see a dependency between these two. This dependency
may be false, but that is a different issue.
This dependency was not correctly accounted for by SiMachineScheduler.
A new test case si-scheduler-exports.ll demonstrating this is
also added in this commit.
The problematic part of SiMachineScheduler was a post-optimization of
the block assignment that tried to group all export instructions into
a separate export block for better execution performance. This routine
correctly checked that any paths from exports to exports did not
contain any non-exports, but not vice-versa: In case of an export with
a non-export successor dependency, that single export was moved
to a separate block, which could then be both a successor and a
predecessor block of a non-export block.
As fix, we now skip export grouping if there are exports with direct
non-export successor dependencies. This fixes the issue at hand,
but is slightly pessimistic:
We *could* group all exports into a separate block that have neither
direct nor indirect export successor dependencies.
We will review the potential performance impact and potentially
revisit with a more sophisticated implementation.
Note that just grouping all exports without direct non-export successor
dependencies could still lead to illegal blocks, since non-export A
could depend on export B that depends on export C. In that case,
export C has no non-export successor, but still may not be grouped
into an export block.
The related instructions are:
VPERMD/Q/PS/PD
VRANGEPD/PS/SD/SS
VGETMANTSS/SD/SH
VGETMANDPS/PD - mem version only
VPMULLQ
VFMULCSH/PH
VFCMULCSH/PH
Differential Revision: https://reviews.llvm.org/D116072
We saw a failure caused by unwinding with incomplete CFIs, so we
can't outline CFI instructions when they are needed in EH.
Reviewed By: luismarques
Differential Revision: https://reviews.llvm.org/D122634
We can't shift-right negative numbers to divide them, so avoid emitting
such sequences. Use negative numerators as a proxy for this situation, since
the indices are always non-negative.
An alternative strategy could be to add a compiler flag to emit division
instructions, which would at least allow us to test the VID sequence
matching itself.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D123796
This test shows a (contrived) BUILD_VECTOR which is correctly identified
as a sequence of ((vid * -3) / 8) + 5. However, the issue is that using
shift-right for the divide is invalid as the step values are negative.
This patch just adds the test: the fix is added in D123796.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D123989
Based on available register budget, reserve highest available VGPR for
AGPR copy before RA. After RA, shift it to lowest unused VGPR if the one
exist.
Fixes SWDEV-330006.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D123525
[1]. Move the test which reject the usage of agpr before gfx908 into a separate file - reject-agpr-usage-before-gfx908.ll.
[2]. Move those tests which are applicable to both gfx900 and gfx908 into a separate file - spill-vgpr.ll.
[3]. Keep those tests which are specific to only gfx908 in the file spill-vgpr-to-agpr.ll.
Above split is required to properly update the tests in D123525.
Reviewed By: rampitec
Differential Revision: https://reviews.llvm.org/D123973
Current stack size diagnostics ignore the size of the unsafe stack.
This patch attaches the size of the static portion of the unsafe stack
to the function as metadata, which can be used by the backend to emit
diagnostics regarding stack usage.
Reviewed By: phosek, mcgrathr
Differential Revision: https://reviews.llvm.org/D119996
We can process the long shuffles (working across several actual
vector registers) in the best way if we take the actual register
represantion into account. We can build more correct representation of
register shuffles, improve number of recognised buildvector sequences.
Also, same function can be used to improve the cost model for the
shuffles. in future patches.
Part of D100486
Differential Revision: https://reviews.llvm.org/D115653
This is x86 specific, and adds statefulness to
MachineModuleInfo. Instead of explicitly tracking this, infer if we
need to declare the symbol based on the reference previously inserted.
This produces a small change in the output due to the move from
AsmPrinter::doFinalization to X86's emitEndOfAsmFile. This will now be
moved relative to other end of file fields, which I'm assuming doesn't
matter (e.g. the __morestack_addr declaration is now after the
.note.GNU-split-stack part)
This also produces another small change in code if the module happened
to define/declare __morestack_addr, but I assume that's invalid and
doesn't really matter.
When folding a COPY of exec into another COPY, the call to
TII->isOperandLegal would crash because COPYs don't have defined
register classes for their operands.
Differential Revision: https://reviews.llvm.org/D122737
We can process the long shuffles (working across several actual
vector registers) in the best way if we take the actual register
represantion into account. We can build more correct representation of
register shuffles, improve number of recognised buildvector sequences.
Also, same function can be used to improve the cost model for the
shuffles. in future patches.
Part of D100486
Differential Revision: https://reviews.llvm.org/D115653
