When passing the -vector feature to LLVM (or equivalently the
-mno-vx command line argument to clang), the intent is that
generated code must not use any vector features (in particular,
no vector registers must be used).
However, there are some cases where we still could generate
such uses; these are all related to some of the additional
vector features (like +vector-enhancements-1). Since none
of those features are actually usable with -vector, just make
sure we disable them all if -vector is given.
Add support in LegalizerHelper for lowering G_SADDSAT etc. either
using add/subtract-with-overflow or using max/min instructions.
Enable this lowering for AMDGPU so it can be tested. The legalization
rules are still approximate and skips out on using the clamp bit to
treat these as legal, which has never been used before. This also
doesn't yet try to deal with expanding SALU cases.
Summary: We do this already for output operands, but missed it for (non-tied) input operands.
Reviewers: arsenm, Petar.Avramovic
Reviewed By: arsenm
Subscribers: jvesely, wdng, nhaehnle, rovka, hiraditya, llvm-commits, kerbowa
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83763
Currently popFromQueueImpl iterates over all candidates to find the best
one. While the candidate queue is small, this is not a problem. But it
becomes a problem once the queue gets larger. For example, the snippet
below takes 330s to compile with llc -O0, but completes in 3s with this
patch.
define void @test(i4000000* %ptr) {
entry:
store i4000000 0, i4000000* %ptr, align 4
ret void
}
This patch limits the number of candidates to check to 1000. This limit
ensures that it never triggers for test-suite/SPEC2000/SPEC2006 on X86
and AArch64 with -O3, while still drastically limiting the compile-time
in case of very large queues.
It would be even better to use a binary heap to manage to queue
(D83335), but some heuristics change the score of a node in the queue
after another node has been scheduled. I plan to address this for
backends that use the MachineScheduler in the future, but that requires
a more careful evaluation. In the meantime, the limit should help users
impacted by this issue.
The patch includes a slightly smaller version of the motivating example
as test case, to guard against the issue.
Reviewers: efriedma, paquette, niravd
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D84328
SAHF/LAHF instructions are always available in 32-bit mode. Early
64-bit capable CPUs made the undefined opcodes in 64-bit mode. This
was changed on later CPUs.
We have a feature flag to control our usage of these instructions.
This feature flag is hooked up to a clang command line option
-msahf/-mno-sahf specifically to give control of the 64-bit mode
behavior.
In the backend X86Subtarget constructor we were explicitly forcing
+sahf into the feature flag string if we were not compiling for
64-bit mode. This was intended to make the predicates always allow
the instructions outside of 64-bit mode. Unfortunately, the way
it was placed into the string allowed -mno-sahf from clang to disable
SAHF instructions in 32-bit mode. This causes an assertion to fire
if you compile a floating point comparison with something like
"-march=pentium -mno-sahf" as our floating point comparison
handling on CPUs that don't support FCOMI/FUCOMI instructions
requires SAHF.
To fix this, this commit restricts the feature flag to only apply to
64-bit mode by ignoring the flag outside 64-bit mode in
X86Subtarget::hasLAHFSAHF(). This way we don't need to mess with
the feature string at all.
Previously, the vins*vlx instructions were incorrectly defined with i64 as the
second argument. This patches fixes this issue by correcting the second argument
of the vins*vlx instructions/intrinsics to be i32.
Differential Revision: https://reviews.llvm.org/D84277
The implementation of the xvtlsbb builtins/intrinsics were not correct as the
intrinsics previously used i1 as an argument type. This patch changes the i1
argument type used in these intrinsics to be i32 instead, as having the second
as an i1 can lead to issues in the backend.
Differential Revision: https://reviews.llvm.org/D84291
Implementing new functionality tested in this file requires adding new
tests for many IR addressing patterns, which can be a large
maintenance burden. This patch makes adding tests easier by switching
to using autogenerated checks. This patch also removes the testing
mode that has simd128 disabled because it would produce very large
checks and is not particularly interesting.
Differential Revision: https://reviews.llvm.org/D84288
Given a vecreduce.add(select(p, x, 0)), we can convert that to a
predicated vaddv, as the else value for the select is the identity
value, a zero. That is what this patch does for the vaddv, vaddva,
vaddlv and vaddlva instructions, copying the existing patterns to also
handle predication through a select.
Differential Revision: https://reviews.llvm.org/D84101
16-bit literals are encoded as 32-bit values. If high 16-bits of the value is 0xFFFF, the decoded instruction cannot be reassembled.
For example, the following code
0xff,0x04,0x04,0x52,0xcd,0xab,0xff,0xff
was decoded as
v_mul_lo_u16_e32 v2, 0xffffabcd, v2
However this literal is actually a 64-bit constant 0x00000000ffffabcd which violates requirements described in the documentation - the truncation is not safe.
This change corrects decoding to make reassembly possible.
Reviewers: arsenm, rampitec
Differential Revision: https://reviews.llvm.org/D84098
A linker optimization is available on PowerPC for GOT indirect PCRelative loads.
The idea is that we can mark a usual GOT indirect load:
pld 3, vec@got@pcrel(0), 1
lwa 3, 4(3)
With a relocation to say that if we don't need to go through the GOT we can let
the linker further optimize this and replace a load with a nop.
pld 3, vec@got@pcrel(0), 1
.Lpcrel1:
.reloc .Lpcrel1-8,R_PPC64_PCREL_OPT,.-(.Lpcrel1-8)
lwa 3, 4(3)
This patch adds the logic that allows the compiler to add the R_PPC64_PCREL_OPT.
Reviewers: nemanjai, lei, hfinkel, sfertile, efriedma, tstellar, grosbach
Reviewed By: nemanjai
Differential Revision: https://reviews.llvm.org/D79864
Summary:
AIX assembly's .set directive is not usable for aliasing purpose.
We need to use extra-label-at-defintion strategy to generate symbol
aliasing on AIX.
Reviewed By: DiggerLin, Xiangling_L
Differential Revision: https://reviews.llvm.org/D83252
For a long time, the InstCombine pass handled target specific
intrinsics. Having target specific code in general passes was noted as
an area for improvement for a long time.
D81728 moves most target specific code out of the InstCombine pass.
Applying the target specific combinations in an extra pass would
probably result in inferior optimizations compared to the current
fixed-point iteration, therefore the InstCombine pass resorts to newly
introduced functions in the TargetTransformInfo when it encounters
unknown intrinsics.
The patch should not have any effect on generated code (under the
assumption that code never uses intrinsics from a foreign target).
This introduces three new functions:
TargetTransformInfo::instCombineIntrinsic
TargetTransformInfo::simplifyDemandedUseBitsIntrinsic
TargetTransformInfo::simplifyDemandedVectorEltsIntrinsic
A few target specific parts are left in the InstCombine folder, where
it makes sense to share code. The largest left-over part in
InstCombineCalls.cpp is the code shared between arm and aarch64.
This allows to move about 3000 lines out from InstCombine to the targets.
Differential Revision: https://reviews.llvm.org/D81728
This is very similar to 243970d03cace2, but handling a slightly
different form of predicated operations. When starting with a pattern of
the form select(p, BinOp(x, y), x), Instcombine will often transform
this to BinOp(x, select(p, y, 0)), where 0 is the identity value of the
binop (0 for adds/subs, 1 for muls, -1 for ands etc). This adds the
patterns that transforms those back into predicated binary operations.
There is also a very minor adjustment to tablegen null_frag in here, to
allow it to also be recognized as a PatLeaf node, so that it can be used
in MVE_TwoOpPattern to easily exclude the cases where we do not need the
alternate transform.
Differential Revision: https://reviews.llvm.org/D84091
Most MVE instructions can be predicated to fold a select into the
instruction, using the predicate and the selects else as a passthough.
This adds tablegen patterns for most two operand instructions using the
newly added TwoOpPattern from 1030e82598da.
Differential Revision: https://reviews.llvm.org/D83222
In fixupIsDeadOrKill, we assume StartMI and EndMI not exist in same
basic block, so we add an assertion in that function. This is wrong
before RA, as before RA the true definition may exist in another
block through copy like instructions.
Reviewed By: nemanjai
Differential Revision: https://reviews.llvm.org/D83365
This patch addresses two issues:
* Forces the availability of the base-pointer (x19) when the frame has
both scalable vectors and variable-length arrays. Otherwise it will
be expensive to access non-SVE locals.
* In presence of SVE stack objects, it will allocate the emergency
scavenging slot close to the SP, so that they can be accessed from
the SP or BP if available. If accessed from the frame-pointer, it will
otherwise need an extra register to access the scavenging slot because
of mixed scalable/non-scalable addressing modes.
Reviewers: efriedma, ostannard, cameron.mcinally, rengolin, david-arm
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D70174
Summary:
[Thumb] set code alignment for 16-bit load from constant pool
LLVM miscompiles this code when compiling for a target with v8.2-A FP16 and the Thumb ISA at -O0:
extern void bar(__fp16 P5);
int main() {
__fp16 P5 = 1.96875;
bar(P5);
}
The code section containing main has 2 byte alignment.
It needs to have 4 byte alignment,
because the load literal instruction has an offset from the
load address with the low 2 bits zeroed.
I do not include a test case in this check-in.
llc and llvm-mc do not exhibit this bug. They do not set code section alignment
in the same manner as clang.
Reviewers: dnsampaio
Reviewed By: dnsampaio
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D84169
Range that f16 can represent fits into i32.
Lower as f16->i32->i64 instead of f16->f32->i64
since f32->i64 has long expansion.
Differential Revision: https://reviews.llvm.org/D84166
Current powerpc backend generates wrong code sequence if stack pointer
has to realign if `-fstack-clash-protection` enabled. When probing
dynamic stack allocation, current `PREPARE_PROBED_ALLOCA` takes
`NegSizeReg` as input and returns
`FinalStackPtr`. `FinalStackPtr=StackPtr+ActualNegSize` is calculated
correctly, however code following `PREPARE_PROBED_ALLOCA` still uses
value of `NegSizeReg`, which does not contain `ActualNegSize` if
`MaxAlign > TargetAlign`, to calculate loop trip count and residual
number of bytes.
This patch is part of fix of
https://bugs.llvm.org/show_bug.cgi?id=46759.
Differential Revision: https://reviews.llvm.org/D84152
Current powerpc backend generates wrong code sequence if stack pointer
has to realign if -fstack-clash-protection enabled. When probing in
prologue, backend should generate a subtraction instruction rather
than a `stux` instruction to realign the stack pointer.
This patch is part of fix of
https://bugs.llvm.org/show_bug.cgi?id=46759.
Differential Revision: https://reviews.llvm.org/D84218
Summary:
In the function `PPCInstrInfo::PredicateInstruction()`, we will replace
non-Predicate Instructions to Predicate Instruction. But we forget add
the new implicit operands the new Predicate Instruction needed. This
patch is to fix this.
Reviewed By: jsji, efriedma
Differential Revision: https://reviews.llvm.org/D82390
This was happening because the BLR didn't have a use of the X0 arg register,
which would end up being re-used in high reg pressure situations.
The change also avoids hard coding the use of X0 for the sequence except to
copy the value for the call. ld64 should still be able to optimize it.
rdar://65438258
There are a few questionable things about this intrinsic and existing
DAG implementation. For some reason the intrinsic hardcodes the second
operand to be scalar-only i32, and SelectionDAG builder makes a
legalization decision based on whether the operand is constant.
These are treated identically to value aggregates placed in the kernel
argument list. A %struct.foo or %struct.foo addrspace(4)*
byref(sizeof(%struct.foo)) align(alignof(%struct.foo)) argument should
produce the same offsets and argument metadata.
This handles all 3 kernel ABI implementations, and the two HSA
metadata emission paths.
getTargetShuffleMask is used by the various "SimplifyDemanded" folds so we can't assume that the bypassed extract_subvector can be safely simplified - getFauxShuffleMask performs a more general decode that allows us to more safely catch many of these cases so the impact is minimal.
The AMDGPU handling of f16 vectors is terrible still since it gets
scalarized even when the vector operation is legal.
The code is is essentially duplicated between the non-strict and
strict case. Apparently no other expansions are currently trying to do
this. This is mostly because I found the behavior of
getStrictFPOperationAction to be confusing. In the ARM case, it would
expand strict_fsub even though it shouldn't due to the later check. At
that point, the logic required to check for legality was more complex
than just duplicating the 2 instruction expansion.
SUMMARY:
when we call memset, memcopy,memmove etc(this are llvm intrinsic function) in the c source code. the llvm will generate IR
like call call void @llvm.memset.p0i8.i32(i8* align 4 bitcast (%struct.S* @s to i8*), i8 %1, i32 %2, i1 false)
for c source code
bash> cat test_memset.call
struct S{
int a;
int b;
};
extern struct S s;
void bar() {
memset(&s, s.b, s.b);
}
like
%struct.S = type { i32, i32 }
@s = external global %struct.S, align 4
; Function Attrs: noinline nounwind optnone
define void @bar() #0 {
entry:
%0 = load i32, i32* getelementptr inbounds (%struct.S, %struct.S* @s, i32 0, i32 1), align 4
%1 = trunc i32 %0 to i8
%2 = load i32, i32* getelementptr inbounds (%struct.S, %struct.S* @s, i32 0, i32 1), align 4
call void @llvm.memset.p0i8.i32(i8* align 4 bitcast (%struct.S* @s to i8*), i8 %1, i32 %2, i1 false)
ret void
}
declare void @llvm.memset.p0i8.i32(i8* nocapture writeonly, i8, i32, i1 immarg) #1
If we want to let the aix as assembly compile pass without -u
it need to has following assembly code.
.extern .memset
(we do not output extern linkage for llvm instrinsic function.
even if we output the extern linkage for llvm intrinsic function, we should not out .extern llvm.memset.p0i8.i32,
instead of we should emit .extern memset)
for other llvm buildin function floatdidf . even if we do not call these function floatdidf in the c source code(the generated IR also do not the call __floatdidf . the function call
was generated in the LLVM optimized.
the function is not in the functions list of Module, but we still need to emit extern .__floatdidf
The solution for it as :
We record all the lllvm intrinsic extern symbol when transformCallee(), and emit all these symbol in the AsmPrinter::doFinalization(Module &M)
Reviewers: jasonliu, Sean Fertile, hubert.reinterpretcast,
Differential Revision: https://reviews.llvm.org/D78929
Current tail duplication in machine block placement pass uses block frequency
information in cost model. But frequency number has only relative meaning
compared to other basic blocks in the same function. A large frequency number
doesn't mean it is hot and a small frequency number doesn't mean it is cold.
To overcome this problem, this patch uses profile count in cost model if it's
available. So we can tail duplicate real hot basic blocks.
Differential Revision: https://reviews.llvm.org/D83265
The default calling convention needs to save/restore the SVE callee
saves according to the SVE PCS when the function takes or returns
scalable types, even when the `aarch64_sve_vector_pcs` CC is not
specified for the function.
Reviewers: efriedma, paulwalker-arm, david-arm, rengolin
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D84041
Summary:
Teach LLVM to recognize the above pattern, where the operands are
either signed or unsigned types.
Subscribers: kristof.beyls, hiraditya, danielkiss, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D83777
These extra vcvt instructions were missed from 74ca67c109 because they
live in a different Domain, but should be treated in the same way.
Differential Revision: https://reviews.llvm.org/D83204
This isn't a natively supported operation, so convert it to a
mask+compare.
In addition to the operation itself, fix up some surrounding stuff to
make the testcase work: we need concat_vectors on i1 vectors, we need
legalization of i1 vector truncates, and we need to fix up all the
relevant uses of getVectorNumElements().
Differential Revision: https://reviews.llvm.org/D83811
Its effect could be achieved by
`-stop-after`,`-print-after`,`-print-after-all`. But a few tests need to
print MIR after ISel which could not be done with
`-print-after`/`-stop-after` since isel pass does not have commandline name.
That's the reason `--print-machineinstrs` is downgraded to
`--print-after-isel` in this patch. `--print-after-isel` could be
removed after we switch to new pass manager since isel pass would have a
commandline text name to use `print-after` or equivalent switches.
The motivation of this patch is to reduce tests dependency on
would-be-deprecated feature.
Reviewed By: arsenm, dsanders
Differential Revision: https://reviews.llvm.org/D83275
This was failing to add the size of LDS globals that weren't directly
used by an instruction. They could be used by constant expressions
which are transitively used by the function. This requires a better
search, but just abort on this for now for correctness.
This allows tracking the in-memory type of a pointer argument to a
function for ABI purposes. This is essentially a stripped down version
of byval to remove some of the stack-copy implications in its
definition.
This includes the base IR changes, and some tests for places where it
should be treated similarly to byval. Codegen support will be in a
future patch.
My original attempt at solving some of these problems was to repurpose
byval with a different address space from the stack. However, it is
technically permitted for the callee to introduce a write to the
argument, although nothing does this in reality. There is also talk of
removing and replacing the byval attribute, so a new attribute would
need to take its place anyway.
This is intended avoid some optimization issues with the current
handling of aggregate arguments, as well as fixes inflexibilty in how
frontends can specify the kernel ABI. The most honest representation
of the amdgpu_kernel convention is to expose all kernel arguments as
loads from constant memory. Today, these are raw, SSA Argument values
and codegen is responsible for turning these into loads.
Background:
There currently isn't a satisfactory way to represent how arguments
for the amdgpu_kernel calling convention are passed. In reality,
arguments are passed in a single, flat, constant memory buffer
implicitly passed to the function. It is also illegal to call this
function in the IR, and this is only ever invoked by a driver of some
kind.
It does not make sense to have a stack passed parameter in this
context as is implied by byval. It is never valid to write to the
kernel arguments, as this would corrupt the inputs seen by other
dispatches of the kernel. These argumets are also not in the same
address space as the stack, so a copy is needed to an alloca. From a
source C-like language, the kernel parameters are invisible.
Semantically, a copy is always required from the constant argument
memory to a mutable variable.
The current clang calling convention lowering emits raw values,
including aggregates into the function argument list, since using
byval would not make sense. This has some unfortunate consequences for
the optimizer. In the aggregate case, we end up with an aggregate
store to alloca, which both SROA and instcombine turn into a store of
each aggregate field. The optimizer never pieces this back together to
see that this is really just a copy from constant memory, so we end up
stuck with expensive stack usage.
This also means the backend dictates the alignment of arguments, and
arbitrarily picks the LLVM IR ABI type alignment. By allowing an
explicit alignment, frontends can make better decisions. For example,
there's real no advantage to an aligment higher than 4, so a frontend
could choose to compact the argument layout. Similarly, there is a
high penalty to using an alignment lower than 4, so a frontend could
opt into more padding for small arguments.
Another design consideration is when it is appropriate to expose the
fact that these arguments are all really passed in adjacent
memory. Currently we have a late IR optimization pass in codegen to
rewrite the kernel argument values into explicit loads to enable
vectorization. In most programs, unrelated argument loads can be
merged together. However, exposing this property directly from the
frontend has some disadvantages. We still need a way to track the
original argument sizes and alignments to report to the driver. I find
using some side-channel, metadata mechanism to track this
unappealing. If the kernel arguments were exposed as a single buffer
to begin with, alias analysis would be unaware that the padding bits
betewen arguments are meaningless. Another family of problems is there
are still some gaps in replacing all of the available parameter
attributes with metadata equivalents once lowered to loads.
The immediate plan is to start using this new attribute to handle all
aggregate argumets for kernels. Long term, it makes sense to migrate
all kernel arguments, including scalars, to be passed indirectly in
the same manner.
Additional context is in D79744.
