The i1 scalar would have been type legalized to i8, but that
doesn't guarantee anything about the upper bits. If we're going
to use it as condition we need to make sure the upper bits are 0.
I've special cased ISD::SETCC conditions since that should
guarantee zero upper bits. We could go further and use
computeKnownBits, but we have no tests that would need that.
Fixes PR43507.
llvm-svn: 373246
ANY_EXTEND of v8i8 is marked Custom on AVX512 for handling extends
from v8i8. But the type legalization infrastructure will call
ReplaceNodeResults for v8i8 results. We should just defer it the
default handling instead of asserting in the default of the switch.
Fixes PR43509.
llvm-svn: 373234
This adds support for lowering variadic musttail calls. To do this, we have
to...
- Detect a musttail call in a variadic function before attempting to lower the
call's formal arguments. This is done in the IRTranslator.
- Compute forwarded registers in `lowerFormalArguments`, and add copies for
those registers.
- Restore the forwarded registers in `lowerTailCall`.
Because there doesn't seem to be any nice way to wrap these up into the outgoing
argument handler, the restore code in `lowerTailCall` is done separately.
Also, irritatingly, you have to make sure that the registers don't overlap with
any passed parameters. Otherwise, the scheduler doesn't know what to do with the
extra copies and asserts.
Add call-translator-variadic-musttail.ll to test this. This is pretty much the
same as the X86 musttail-varargs.ll test. We didn't have as nice of a test to
base this off of, but the idea is the same.
Differential Revision: https://reviews.llvm.org/D68043
llvm-svn: 373226
trigger stack protectors. Fixes PR42238.
Add test coverage for llvm.memset, as proxy for all llvm.mem*
intrinsics. There are two issues here: (1) they could be lowered to a
libc call, which could be intercepted, and do Bad Stuff; (2) with a
non-constant size, they could overwrite the current stack frame.
The test was mostly written by Matt Arsenault in r363169, which was
later reverted; I tweaked what he had and added the llvm.memset part.
Differential Revision: https://reviews.llvm.org/D67845
llvm-svn: 373220
"Captured" and "relevant to Stack Protector" are not the same thing.
This reverts commit f29366b1f5.
aka r363169.
Differential Revision: https://reviews.llvm.org/D67842
llvm-svn: 373216
The VCTP instruction will calculate the predicate masked based upon
the number of elements that need to be processed. I had inserted the
sub before the vctp intrinsic and supplied it as the operand, but
this is incorrect as the phi should directly feed the vctp. The sub
is calculating the value for the next iteration.
Differential Revision: https://reviews.llvm.org/D67921
llvm-svn: 373188
ISD::SADDO uses the suggested sequence described in the section §2.4 of
the RISCV Spec v2.2. ISD::SSUBO uses the dual approach but checking for
(non-zero) positive.
Differential Revision: https://reviews.llvm.org/D47927
llvm-svn: 373187
As we perform a zext on any arguments used in the promoted tree, it
doesn't matter if they're marked as signext. The only permitted
user(s) in the tree which would interpret the sign bits are signed
icmps. For these instructions, their promoted operands are truncated
before the icmp uses them.
Differential Revision: https://reviews.llvm.org/D68019
llvm-svn: 373186
SystemZPostRewrite needs to be run before (it may emit COPYs) the Post-RA
pseudo pass also at -O0, so it should be added in addPostRegAlloc().
Review: Ulrich Weigand
llvm-svn: 373182
There's room from improvement here, but this is a decent
starting point.
There are a few minor regressions in the vector-rotate tests,
where we are now forming a vpternlog from an and before we get
a chance to form it for a bitselect that we were matching
previously. This results in an AND and an ANDN feeding the
vpternlog where previously we just had an AND after the
vpternlog. I think we can probably DAG combine the AND with
the bitselect to get back to similar codegen.
llvm-svn: 373172
This is an attempt to fill in some of the missing instructions from the
Cortex-M4 schedule, and make it easier to do the same for other ARM cpus.
- Some instructions are marked as hasNoSchedulingInfo as they are pseudos or
otherwise do not require scheduling info
- A lot of features have been marked not supported
- Some WriteRes's have been added for cvt instructions.
- Some extra instruction latencies have been added, notably by relaxing the
regex for dsp instruction to catch more cases, and some fp instructions.
This goes a long way to get the CompleteModel working for this CPU. It does not
go far enough as to get all scheduling info for all output operands correct.
Differential Revision: https://reviews.llvm.org/D67957
llvm-svn: 373163
This allows us to reduce the use count on the condition node before
the match. This enables load folding for that operand without
relying on the peephole pass. This will be improved on for
broadcast load folding in a subsequent commit.
This still requires a bunch of isel patterns for vXi16/vXi8 types
though.
llvm-svn: 373156
The fma mutate test will not exercise what it was intended to test
once we simplify those ops immediately, but the test will still
pass with the existing CHECKs, so I'm leaving it in case that
still has minimal value.
llvm-svn: 373149
We need to propagate this information from the IR in order to be able to safely
do tail call optimizations on the intrinsics during legalization. Assuming
it's safe to do tail call opt without checking for the marker isn't safe because
the mem libcall may use allocas from the caller.
This adds an extra immediate operand to the end of the intrinsics and fixes the
legalizer to handle it.
Differential Revision: https://reviews.llvm.org/D68151
llvm-svn: 373140
The majority of the code doesn't run on the X86 nodes today since
its gated by isBeforeLegalizeOps and we don't formm X86 nodes
until after that. Except for a couple special case in type
legalization. But I think we would probably break those if
some of the transforms fire on them.
I want to remove the hardcoded operand numbers and the unusual
use of UpdateNodeOperands. Being able to know which ISD opcodes
are present should help with that.
llvm-svn: 373136
With avx512, the vXi1 type is legal. And we can more easily sign
extend them to vector registers. zext requires a sign extend and
a shift.
If we can easily turn the zext into a sext we should.
llvm-svn: 373131
These two test cases use -march=systemz instead of a triple. In
particular, the used file format is then based on the default host
triple. This leads to different behaviour on different platforms.
The SystemZ implementation uses the integrated assembler for a
long time now. The mature-mc-support test can be fully enabled.
Differential Revision: https://reviews.llvm.org/D68129
llvm-svn: 373098
Summary:
An erroneously negated if-statement by an earlier (March 2019) bugfix left phi replacement/simplification under optimizeMemoryInst() in CodeGenPrepare largely inactivated. The error was found when csmith found that the same assert as in the original bug report could still be triggered in a different way. This patch fixes the bugfix. The original bug was:
https://bugs.llvm.org/show_bug.cgi?id=41052
... and the previous fix was D59358.
Reviewers: aprantl, skatkov
Reviewed By: skatkov
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67838
llvm-svn: 373084
This caused severe compile-time regressions, see PR43455.
> Modern processors predict the targets of an indirect branch regardless of
> the size of any jump table used to glean its target address. Moreover,
> branch predictors typically use resources limited by the number of actual
> targets that occur at run time.
>
> This patch changes the semantics of the option `-max-jump-table-size` to limit
> the number of different targets instead of the number of entries in a jump
> table. Thus, it is now renamed to `-max-jump-table-targets`.
>
> Before, when `-max-jump-table-size` was specified, it could happen that
> cluster jump tables could have targets used repeatedly, but each one was
> counted and typically resulted in tables with the same number of entries.
> With this patch, when specifying `-max-jump-table-targets`, tables may have
> different lengths, since the number of unique targets is counted towards the
> limit, but the number of unique targets in tables is the same, but for the
> last one containing the balance of targets.
>
> Differential revision: https://reviews.llvm.org/D60295
llvm-svn: 373060
We can't use short granules with stack instrumentation when targeting older
API levels because the rest of the system won't understand the short granule
tags stored in shadow memory.
Moreover, we need to be able to let old binaries (which won't understand
short granule tags) run on a new system that supports short granule
tags. Such binaries will call the __hwasan_tag_mismatch function when their
outlined checks fail. We can compensate for the binary's lack of support
for short granules by implementing the short granule part of the check in
the __hwasan_tag_mismatch function. Unfortunately we can't do anything about
inline checks, but I don't believe that we can generate these by default on
aarch64, nor did we do so when the ABI was fixed.
A new function, __hwasan_tag_mismatch_v2, is introduced that lets code
targeting the new runtime avoid redoing the short granule check. Because tag
mismatches are rare this isn't important from a performance perspective; the
main benefit is that it introduces a symbol dependency that prevents binaries
targeting the new runtime from running on older (i.e. incompatible) runtimes.
Differential Revision: https://reviews.llvm.org/D68059
llvm-svn: 373035
We have isel patterns that can put an IMPLICIT_DEF on one of
the sources for these instructions. So we should make sure
we break any dependencies there. This should be done by
just using one of the other sources.
llvm-svn: 373025
Similar for f64 and having a non-zero passthru value.
We were previously not trying to fold the load at all. Using
a CodeGenOnly instruction allows us to use FR32X/FR64X as the
register class to avoid a bunch of COPY_TO_REGCLASS.
llvm-svn: 373021
This patch emits the function descriptor csect for functions with definitions
under both 32-bit/64-bit mode on AIX.
Differential Revision: https://reviews.llvm.org/D66724
llvm-svn: 373009
Summary:
This was found during review of https://reviews.llvm.org/D66050.
In the simple test of fdiv, we miss to fold
```
fneg 2, 2
xsmaddasp 3, 2, 0
```
to
```
xsnmsubasp 3, 2, 0
```
We have the patterns for Double Precision and vectors, just missing
Single Precision, the patch add that.
Reviewers: #powerpc, hfinkel, nemanjai, steven.zhang
Reviewed By: #powerpc, steven.zhang
Subscribers: wuzish, hiraditya, kbarton, MaskRay, shchenz, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67595
llvm-svn: 372985
Implement aggregate structure split to simpler types in splitToValueTypes.
splitToValueTypes is used for return values.
According to MipsABIInfo from clang/lib/CodeGen/TargetInfo.cpp,
aggregate structure arguments for O32 always get simplified and thus
will remain unsupported by the MIPS GlobalISel for the time being.
For O32, aggregate structures can be encountered only for complex number
returns e.g. 'complex float' or 'complex double' from <complex.h>.
Differential Revision: https://reviews.llvm.org/D67963
llvm-svn: 372957
With -pg -mfentry -mnop-mcount, a nop is emitted instead of the call to
fentry.
Review: Ulrich Weigand
https://reviews.llvm.org/D67765
llvm-svn: 372950
Summary:
Previously the case
EBB
| \_
| |
| TBB
| /
FBB
was treated as a valid triangle also when TBB and FBB was the same basic
block. This could then lead to an invalid CFG when we removed the edge
from EBB to TBB, since that meant we would also remove the edge from EBB
to FBB.
Since TBB == FBB is quite a degenerated case of a triangle, we now
don't treat it as a valid triangle anymore, and thus we will avoid the
trouble with updating the CFG.
Reviewers: efriedma, dmgreen, kparzysz
Reviewed By: efriedma
Subscribers: bjope, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67832
llvm-svn: 372943
Summary:
The list of indirect labels should ALWAYS have their blockaddresses as
argument operands to the callbr (but not necessarily the other way
around). Add an invariant that checks this.
The verifier catches a bad test case that was added recently in r368478.
I think that was a simple mistake, and the test was made less strict in
regards to the precise addresses (as those weren't specifically the
point of the test).
This invariant will be used to find a reported bug.
Link: https://www.spinics.net/lists/arm-kernel/msg753473.html
Link: https://github.com/ClangBuiltLinux/linux/issues/649
Reviewers: craig.topper, void, chandlerc
Reviewed By: void
Subscribers: ychen, lebedev.ri, javed.absar, kristof.beyls, hiraditya, llvm-commits, srhines
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67196
llvm-svn: 372923
Summary:
Useful in case you want to have control over interrupt vector generation.
For example in Rust language we have an arrangement where all unhandled
ISR vectors gets mapped to a single default handler function. Which is
hard to implement when LLVM tries to generate vectors on its own.
Reviewers: asl, krisb
Subscribers: hiraditya, JDevlieghere, awygle, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67313
llvm-svn: 372910
When checking for tail call eligibility, we should use the correct CCAssignFn
for each argument, rather than just checking if the caller/callee is varargs or
not.
This is important for tail call lowering with varargs. If we don't check it,
then basically any varargs callee with parameters cannot be tail called on
Darwin, for one thing. If the parameters are all guaranteed to be in registers,
this should be entirely safe.
On top of that, not checking for this could potentially make it so that we have
the wrong stack offsets when checking for tail call eligibility.
Also refactor some of the stuff for CCAssignFnForCall and pull it out into a
helper function.
Update call-translator-tail-call.ll to show that we can now correctly tail call
on Darwin. Also add two extra tail call checks. The first verifies that we still
respect the caller's stack size, and the second verifies that we still don't
tail call when a varargs function has a memory argument.
Differential Revision: https://reviews.llvm.org/D67939
llvm-svn: 372897
Modern processors predict the targets of an indirect branch regardless of
the size of any jump table used to glean its target address. Moreover,
branch predictors typically use resources limited by the number of actual
targets that occur at run time.
This patch changes the semantics of the option `-max-jump-table-size` to limit
the number of different targets instead of the number of entries in a jump
table. Thus, it is now renamed to `-max-jump-table-targets`.
Before, when `-max-jump-table-size` was specified, it could happen that
cluster jump tables could have targets used repeatedly, but each one was
counted and typically resulted in tables with the same number of entries.
With this patch, when specifying `-max-jump-table-targets`, tables may have
different lengths, since the number of unique targets is counted towards the
limit, but the number of unique targets in tables is the same, but for the
last one containing the balance of targets.
Differential revision: https://reviews.llvm.org/D60295
llvm-svn: 372893
We might be able to do better on the example in the test,
but in general, we should not scalarize a splatted vector
binop if there are other uses of the binop. Otherwise, we
can end up with code as we had - a scalar op that is
redundant with a vector op.
llvm-svn: 372886
Merge more Select pseudo instructions in emitSelect() by allowing other
instructions between them as long as they do not clobber CC.
Debug value instructions are now moved down to below the new PHIs instead of
erasing them.
Review: Ulrich Weigand
https://reviews.llvm.org/D67619
llvm-svn: 372873
During legalisation we can end up with some pretty strange nodes, like shifts
of 0. We need to make sure we don't try to make long shifts of these, ending up
with invalid assembly instructions. A long shift with a zero immediate actually
encodes a shift by 32.
Differential Revision: https://reviews.llvm.org/D67664
llvm-svn: 372839
I think we should be able to use shl instead of sshl and ushl for
positive constant shift values, unless I am missing something.
We already have the machinery in place to ensure we only replace
nodes, if the shift value is positive and <= the element width.
This is a generalization of an earlier patch rL372565.
Reviewers: t.p.northover, samparker, dmgreen, anemet
Reviewed By: anemet
Differential Revision: https://reviews.llvm.org/D67955
llvm-svn: 372824
Currently, if an array element type size is 0, the number of
array elements will be set to 0, regardless of what user
specified. This implementation is done in the beginning where
BTF is mostly used to calculate the member offset.
For example,
struct s {};
struct s1 {
int b;
struct s a[2];
};
struct s1 s1;
The BTF will have struct "s1" member "a" with element count 0.
Now BTF types are used for compile-once and run-everywhere
relocations and we need more precise type representation
for type comparison. Andrii reported the issue as there
are differences between original structure and BTF-generated
structure.
This patch made the change to correctly assign "2"
as the number elements of member "a".
Some dead codes related to ElemSize compuation are also removed.
Differential Revision: https://reviews.llvm.org/D67979
llvm-svn: 372785
Similar to rL372717, we can force the splitting of extends of vector loads in
MVE, in order to use the better widening loads as opposed to going through
expensive extends. This adds a combine to early-on detect extends of loads and
split the load in two, from where normal legalisation will kick in and we get a
series of widening loads.
Differential Revision: https://reviews.llvm.org/D67909
llvm-svn: 372721
MVE does not have a simple sign extend instruction that can move elements
across lanes. We currently often end up moving each lane into and out of a GPR,
in order to get elements into the correct places. When we have a store of a
trunc (or a extend of a load), we can instead just split the store/load in two,
using the narrowing/widening load/store instructions from each half of the
vector.
This does that for stores. It happens very early in a store combine, so as to
easily detect the truncates. (It would be possible to do this later, but that
would involve looking through a buildvector of extract elements. Not impossible
but this way seemed simpler).
By enabling store combines we also get a vmovdrr combine for free, helping some
other tests.
Differential Revision: https://reviews.llvm.org/D67828
llvm-svn: 372717
We were miscompiling switch value comparisons with the wrong signedness, which
shows up when we have things like switch case values with i1 types, which end up
being legalized incorrectly.
Fixes PR43383
llvm-svn: 372675
Summary:
Adds the new load_splat instructions as specified at
https://github.com/WebAssembly/simd/blob/master/proposals/simd/SIMD.md#load-and-splat.
DAGISel does not allow matching multiple copies of the same load in a
single pattern, so we use a new node in WebAssemblyISD to wrap loads
that should be splatted.
Depends on D67783.
Reviewers: aheejin
Subscribers: dschuff, sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67784
llvm-svn: 372655
This came up in the x86-specific:
https://bugs.llvm.org/show_bug.cgi?id=43239
...but it is a general problem for the BreakFalseDeps pass.
Dependencies may be broken by adding some other instruction,
so that should be avoided if the overall goal is to minimize size.
Differential Revision: https://reviews.llvm.org/D67363
llvm-svn: 372628
Remove any predicate that we replace with a vctp intrinsic, and try
to remove their operands too. Also look into the exit block to see if
there's any duplicates of the predicates that we've replaced and
clone the vctp to be used there instead.
Differential Revision: https://reviews.llvm.org/D67709
llvm-svn: 372567
Try to generate ushll/sshll for aarch64_neon_ushl/aarch64_neon_sshl,
if their first operand is extended and the second operand is a constant
Also adds a few tests marked with FIXME, where we can further increase
codegen.
Reviewers: t.p.northover, samparker, dmgreen, anemet
Reviewed By: anemet
Differential Revision: https://reviews.llvm.org/D62308
llvm-svn: 372565
Check whether there are any uses or defs between the LoopDec and
LoopEnd. If there's not, then we can use a subs to set the cpsr and
skip generating a cmp.
Differential Revision: https://reviews.llvm.org/D67801
llvm-svn: 372560
CC_Mips doesn't accept vararg functions for O32, so we have to explicitly
use CC_Mips_FixedArg.
For lowerCall we now properly figure out whether callee function is vararg
or not, this has no effect for O32 since we always use CC_Mips_FixedArg.
For lower formal arguments we need to copy arguments in register to stack
and save pointer to start for argument list into MipsMachineFunction
object so that G_VASTART could use it during instruction select.
For vacopy we need to copy content from one vreg to another,
load and store are used for that purpose.
Differential Revision: https://reviews.llvm.org/D67756
llvm-svn: 372555
The attached test case would previous infinite loop after
r365711.
I'm going to move this to X86ISelDAGToDAG.cpp to get the setcc
to match VPTEST in 32-bit mode in a follow up commit.
llvm-svn: 372543
This intrinsics should be shift by immediate, but gcc allows any
i32 scalar and clang needs to match that. So we try to detect the
non-constant case and move the data from an integer register to an
MMX register.
Previously this was done by creating a v2i32 build_vector and
bitcast in SelectionDAGBuilder. This had to be done early since
v2i32 isn't a legal type. The bitcast+build_vector would be DAG
combined to X86ISD::MMX_MOVW2D which isel will turn into a
GPR->MMX MOVD.
This commit just moves the whole thing to lowering and emits
the X86ISD::MMX_MOVW2D directly to avoid the illegal type. The
test changes just seem to be due to nodes being linearized in a
different order.
llvm-svn: 372535
Summary:
PR43381 notes that while we are good at matching `(X >> C1) & C2` as BEXTR/BEXTRI,
we only do that if we either have BEXTRI (TBM),
or if BEXTR is marked as being fast (`-mattr=+fast-bextr`).
In all other cases we don't match.
But that is mainly only true for AMD CPU's.
However, for all the CPU's for which we have sched models,
the BZHI is always fast (or the sched models are all bad.)
So if we decide that it's unprofitable to emit BEXTR/BEXTRI,
we should consider falling-back to BZHI if it is available,
and follow-up with the shift.
While it's really tempting to do something because it's cool
it is wise to first think whether it actually makes sense to do.
We shouldn't just use BZHI because we can, but only it it is beneficial.
In particular, it isn't really worth it if the input is a register,
mask is small, or we can fold a load.
But it is worth it if the mask does not fit into 32-bits.
(careful, i don't know much about intel cpu's, my choice of `-mcpu` may be bad here)
Thus we manage to fold a load:
https://godbolt.org/z/Er0OQz
Or if we'd end up using BZHI anyways because the mask is large:
https://godbolt.org/z/dBJ_5h
But this isn'r actually profitable in general case,
e.g. here we'd increase microop count
(the register renaming is free, mca does not model that there it seems)
https://godbolt.org/z/k6wFoz
Likewise, not worth it if we just get load folding:
https://godbolt.org/z/1M1deGhttps://bugs.llvm.org/show_bug.cgi?id=43381
Reviewers: RKSimon, craig.topper, davezarzycki, spatel
Reviewed By: craig.topper, davezarzycki
Subscribers: andreadb, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67875
llvm-svn: 372532
Recommit: fix asan errors.
The way MachinePipeliner uses these target hooks is stateful - we reduce trip
count by one per call to reduceLoopCount. It's a little overfit for hardware
loops, where we don't have to worry about stitching a loop induction variable
across prologs and epilogs (the induction variable is implicit).
This patch introduces a new API:
/// Analyze loop L, which must be a single-basic-block loop, and if the
/// conditions can be understood enough produce a PipelinerLoopInfo object.
virtual std::unique_ptr<PipelinerLoopInfo>
analyzeLoopForPipelining(MachineBasicBlock *LoopBB) const;
The return value is expected to be an implementation of the abstract class:
/// Object returned by analyzeLoopForPipelining. Allows software pipelining
/// implementations to query attributes of the loop being pipelined.
class PipelinerLoopInfo {
public:
virtual ~PipelinerLoopInfo();
/// Return true if the given instruction should not be pipelined and should
/// be ignored. An example could be a loop comparison, or induction variable
/// update with no users being pipelined.
virtual bool shouldIgnoreForPipelining(const MachineInstr *MI) const = 0;
/// Create a condition to determine if the trip count of the loop is greater
/// than TC.
///
/// If the trip count is statically known to be greater than TC, return
/// true. If the trip count is statically known to be not greater than TC,
/// return false. Otherwise return nullopt and fill out Cond with the test
/// condition.
virtual Optional<bool>
createTripCountGreaterCondition(int TC, MachineBasicBlock &MBB,
SmallVectorImpl<MachineOperand> &Cond) = 0;
/// Modify the loop such that the trip count is
/// OriginalTC + TripCountAdjust.
virtual void adjustTripCount(int TripCountAdjust) = 0;
/// Called when the loop's preheader has been modified to NewPreheader.
virtual void setPreheader(MachineBasicBlock *NewPreheader) = 0;
/// Called when the loop is being removed.
virtual void disposed() = 0;
};
The Pipeliner (ModuloSchedule.cpp) can use this object to modify the loop while
allowing the target to hold its own state across all calls. This API, in
particular the disjunction of creating a trip count check condition and
adjusting the loop, improves the code quality in ModuloSchedule.cpp.
llvm-svn: 372463
Previously we only matched scalar_to_vector and scalar load, but
we should be able to narrow a vector load or match vzload.
Also need to match TargetConstant instead of Constant. The register
patterns were previously updated, but not the memory patterns.
llvm-svn: 372458
The intrinsic has an immarg so its gets created with a TargetConstant
instead of a Constant after r372338. The isel pattern was only
updated for the register form, but not the memory form.
llvm-svn: 372457
I believe all of the uniform/divergent pattern predicates are
redundant and can be removed. The uniformity bit already influences
the register class, and nothhing has broken when I've removed this and
others.
llvm-svn: 372450
We currently always set the HasCalls on MFI during translation and legalization if
we're handling a call or legalizing to a libcall. However, if that call is later
optimized to a tail call then we don't need the flag. The flag being set to true
causes frame lowering to always save and restore FP/LR, which adds unnecessary code.
This change does the same thing as SelectionDAG and ports over some code that scans
instructions after selection, using TargetInstrInfo to determine if target opcodes
are known calls.
Code size geomean improvements on CTMark:
-O0 : 0.1%
-Os : 0.3%
Differential Revision: https://reviews.llvm.org/D67868
llvm-svn: 372443
The recently announced IBM z15 processor implements the architecture
already supported as "arch13" in LLVM. This patch adds support for
"z15" as an alternate architecture name for arch13.
The patch also uses z15 in a number of places where we used arch13
as long as the official name was not yet announced.
llvm-svn: 372435
This patch adds the patterns to select the dot product instructions.
Tested on aarch64-linux with make check-all.
Differential Revision: https://reviews.llvm.org/D67645
llvm-svn: 372408
According to the documentation method returns predecessor
if the given loop's header has exactly one unique predecessor
outside the loop. Otherwise return null.
In reality it asserts if there is no predecessor outside of
the loop.
The testcase has the loop where predecessors outside of the
loop were not identified as analyzeBranch() was unable to
process the mask branch and returned true. That is also not
correct to assert for the truly dead loops.
Differential Revision: https://reviews.llvm.org/D67634
llvm-svn: 372405
The insertion of an unconditional branch during FastISel can differ depending on
building with or without debug information. This happens because FastISel::fastEmitBranch
emits an unconditional branch depending on the size of the current basic block
without distinguishing between debug and non-debug instructions.
This patch fixes this issue by ignoring debug instructions when getting the size
of the basic block.
Reviewers: aprantl
Reviewed By: aprantl
Subscribers: ormris, aprantl, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67703
llvm-svn: 372389
This reverts commit 52621307bc.
Tests have been failing all night with
[0/2] ACTION //llvm/test:check-llvm(//llvm/utils/gn/build/toolchain:unix)
-- Testing: 33647 tests, 64 threads --
Testing: 0 .. 10..
UNRESOLVED: LLVM :: CodeGen/AMDGPU/GlobalISel/isel-blendi-gettargetconstant.ll (6943 of 33647)
******************** TEST 'LLVM :: CodeGen/AMDGPU/GlobalISel/isel-blendi-gettargetconstant.ll' FAILED ********************
Test has no run line!
********************
Since there were other concerns on https://reviews.llvm.org/D67785,
I'm just reverting for now.
llvm-svn: 372383
The way MachinePipeliner uses these target hooks is stateful - we reduce trip
count by one per call to reduceLoopCount. It's a little overfit for hardware
loops, where we don't have to worry about stitching a loop induction variable
across prologs and epilogs (the induction variable is implicit).
This patch introduces a new API:
/// Analyze loop L, which must be a single-basic-block loop, and if the
/// conditions can be understood enough produce a PipelinerLoopInfo object.
virtual std::unique_ptr<PipelinerLoopInfo>
analyzeLoopForPipelining(MachineBasicBlock *LoopBB) const;
The return value is expected to be an implementation of the abstract class:
/// Object returned by analyzeLoopForPipelining. Allows software pipelining
/// implementations to query attributes of the loop being pipelined.
class PipelinerLoopInfo {
public:
virtual ~PipelinerLoopInfo();
/// Return true if the given instruction should not be pipelined and should
/// be ignored. An example could be a loop comparison, or induction variable
/// update with no users being pipelined.
virtual bool shouldIgnoreForPipelining(const MachineInstr *MI) const = 0;
/// Create a condition to determine if the trip count of the loop is greater
/// than TC.
///
/// If the trip count is statically known to be greater than TC, return
/// true. If the trip count is statically known to be not greater than TC,
/// return false. Otherwise return nullopt and fill out Cond with the test
/// condition.
virtual Optional<bool>
createTripCountGreaterCondition(int TC, MachineBasicBlock &MBB,
SmallVectorImpl<MachineOperand> &Cond) = 0;
/// Modify the loop such that the trip count is
/// OriginalTC + TripCountAdjust.
virtual void adjustTripCount(int TripCountAdjust) = 0;
/// Called when the loop's preheader has been modified to NewPreheader.
virtual void setPreheader(MachineBasicBlock *NewPreheader) = 0;
/// Called when the loop is being removed.
virtual void disposed() = 0;
};
The Pipeliner (ModuloSchedule.cpp) can use this object to modify the loop while
allowing the target to hold its own state across all calls. This API, in
particular the disjunction of creating a trip count check condition and
adjusting the loop, improves the code quality in ModuloSchedule.cpp.
llvm-svn: 372376
Summary: This fixes a crasher introduced by r372338.
Reviewers: echristo, arsenm
Subscribers: jvesely, wdng, nhaehnle, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67785
Tighten up the test case.
llvm-svn: 372366
If an instruction had multiple subregister defs, and one of them was
undef, this would improperly conclude all other lanes are
killed. There could still be other defs of those read-undef lanes in
other operands. This would improperly remove register uses from
CurrentVRegUses, so the visitation of later operands would not find
the necessary register dependency. This would also mean this would
fail or not depending on how different subregister def operands were
ordered.
On an undef subregister def, scan the instruction for other
subregister defs and avoid killing those.
This possibly should be deferring removing anything from
CurrentVRegUses until the entire instruction has been processed
instead.
llvm-svn: 372362
This reverts r372314, reapplying r372285 and the commits which depend
on it (r372286-r372293, and r372296-r372297)
This was missing one switch to getTargetConstant in an untested case.
llvm-svn: 372338
This patch converts the DAGCombine isNegatibleForFree/GetNegatedExpression into overridable TLI hooks and includes a demonstration X86 implementation.
The intention is to let us extend existing FNEG combines to work more generally with negatible float ops, allowing it work with target specific combines and opcodes (e.g. X86's FMA variants).
Unlike the SimplifyDemandedBits, we can't just handle target nodes through a Target callback, we need to do this as an override to allow targets to handle generic opcodes as well. This does mean that the target implementations has to duplicate some checks (recursion depth etc.).
I've only begun to replace X86's FNEG handling here, handling FMADDSUB/FMSUBADD negation and some low impact codegen changes (some FMA negatation propagation). We can build on this in future patches.
Differential Revision: https://reviews.llvm.org/D67557
llvm-svn: 372333
This broke the Chromium build, causing it to fail with e.g.
fatal error: error in backend: Cannot select: t362: v4i32 = X86ISD::VSHLI t392, Constant:i8<15>
See llvm-commits thread of r372285 for details.
This also reverts r372286, r372287, r372288, r372289, r372290, r372291,
r372292, r372293, r372296, and r372297, which seemed to depend on the
main commit.
> Encode them directly as an imm argument to G_INTRINSIC*.
>
> Since now intrinsics can now define what parameters are required to be
> immediates, avoid using registers for them. Intrinsics could
> potentially want a constant that isn't a legal register type. Also,
> since G_CONSTANT is subject to CSE and legalization, transforms could
> potentially obscure the value (and create extra work for the
> selector). The register bank of a G_CONSTANT is also meaningful, so
> this could throw off future folding and legalization logic for AMDGPU.
>
> This will be much more convenient to work with than needing to call
> getConstantVRegVal and checking if it may have failed for every
> constant intrinsic parameter. AMDGPU has quite a lot of intrinsics wth
> immarg operands, many of which need inspection during lowering. Having
> to find the value in a register is going to add a lot of boilerplate
> and waste compile time.
>
> SelectionDAG has always provided TargetConstant for constants which
> should not be legalized or materialized in a register. The distinction
> between Constant and TargetConstant was somewhat fuzzy, and there was
> no automatic way to force usage of TargetConstant for certain
> intrinsic parameters. They were both ultimately ConstantSDNode, and it
> was inconsistently used. It was quite easy to mis-select an
> instruction requiring an immediate. For SelectionDAG, start emitting
> TargetConstant for these arguments, and using timm to match them.
>
> Most of the work here is to cleanup target handling of constants. Some
> targets process intrinsics through intermediate custom nodes, which
> need to preserve TargetConstant usage to match the intrinsic
> expectation. Pattern inputs now need to distinguish whether a constant
> is merely compatible with an operand or whether it is mandatory.
>
> The GlobalISelEmitter needs to treat timm as a special case of a leaf
> node, simlar to MachineBasicBlock operands. This should also enable
> handling of patterns for some G_* instructions with immediates, like
> G_FENCE or G_EXTRACT.
>
> This does include a workaround for a crash in GlobalISelEmitter when
> ARM tries to uses "imm" in an output with a "timm" pattern source.
llvm-svn: 372314
We needn't BFI each lane individually into a predicate register when each lane
in the same. A simple sign extend and a vmsr will do.
Differential Revision: https://reviews.llvm.org/D67653
llvm-svn: 372313
This needs special handling due to some subtargets that have a
nonstandard register layout for f16 vectors
Also reject some illegal types on other targets.
llvm-svn: 372293
Encode them directly as an imm argument to G_INTRINSIC*.
Since now intrinsics can now define what parameters are required to be
immediates, avoid using registers for them. Intrinsics could
potentially want a constant that isn't a legal register type. Also,
since G_CONSTANT is subject to CSE and legalization, transforms could
potentially obscure the value (and create extra work for the
selector). The register bank of a G_CONSTANT is also meaningful, so
this could throw off future folding and legalization logic for AMDGPU.
This will be much more convenient to work with than needing to call
getConstantVRegVal and checking if it may have failed for every
constant intrinsic parameter. AMDGPU has quite a lot of intrinsics wth
immarg operands, many of which need inspection during lowering. Having
to find the value in a register is going to add a lot of boilerplate
and waste compile time.
SelectionDAG has always provided TargetConstant for constants which
should not be legalized or materialized in a register. The distinction
between Constant and TargetConstant was somewhat fuzzy, and there was
no automatic way to force usage of TargetConstant for certain
intrinsic parameters. They were both ultimately ConstantSDNode, and it
was inconsistently used. It was quite easy to mis-select an
instruction requiring an immediate. For SelectionDAG, start emitting
TargetConstant for these arguments, and using timm to match them.
Most of the work here is to cleanup target handling of constants. Some
targets process intrinsics through intermediate custom nodes, which
need to preserve TargetConstant usage to match the intrinsic
expectation. Pattern inputs now need to distinguish whether a constant
is merely compatible with an operand or whether it is mandatory.
The GlobalISelEmitter needs to treat timm as a special case of a leaf
node, simlar to MachineBasicBlock operands. This should also enable
handling of patterns for some G_* instructions with immediates, like
G_FENCE or G_EXTRACT.
This does include a workaround for a crash in GlobalISelEmitter when
ARM tries to uses "imm" in an output with a "timm" pattern source.
llvm-svn: 372285
Summary:
Large slowdowns were observed in Rust due to many small, constant
sized copies in conjunction with poorly-optimized memory.copy
implementations. Since memory.copy cannot be expected to be inlined
efficiently by engines at this time, stop using it for the smallest
copies. We continue to lower all memcpy intrinsics to memory.copy,
though.
Reviewers: aheejin, alexcrichton
Subscribers: dschuff, sbc100, jgravelle-google, hiraditya, JDevlieghere, sunfish, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67639
llvm-svn: 372275
Since we now lower most tail calls, it makes sense to support musttail.
Instead of always falling back to SelectionDAG, only fall back when a musttail
call was not able to be emitted as a tail call. Once we can handle most
incoming and outgoing arguments, we can change this to a `report_fatal_error`
like in ISelLowering.
Remove the assert that we don't have varargs and a musttail, and replace it
with a return false. Implementing this requires that we implement
`saveVarArgRegisters` from AArch64ISelLowering, which is an entirely different
patch.
Add GlobalISel lines to vararg-tallcall.ll to make sure that we produce correct
code. Right now we only fall back, but eventually this will be relevant.
Differential Revision: https://reviews.llvm.org/D67681
llvm-svn: 372273
DIFlagBlockByRefStruct is an unused DIFlag that originally was used by
clang to express (Objective-)C block captures in debug info. For the
last year Clang has been emitting complex DIExpressions to describe
block captures instead, which makes all the code supporting this flag
redundant.
This patch removes the flag and all supporting "dead" code, so we can
reuse the bit for something else in the future.
Since this only affects debug info generated by Clang with the block
extension this mostly affects Apple platforms and I don't have any
bitcode compatibility concerns for removing this. The Verifier will
reject debug info that uses the bit and thus degrade gracefully when
LTO'ing older bitcode with a newer compiler.
rdar://problem/44304813
Differential Revision: https://reviews.llvm.org/D67453
llvm-svn: 372272
Summary:
`DAGCombiner::visitADDLikeCommutative()` already has a sibling fold:
`(add X, Carry) -> (addcarry X, 0, Carry)`
This fold, as suggested by @efriedma, helps recover from //some//
of the regressions of D62266
Reviewers: efriedma, deadalnix
Subscribers: javed.absar, kristof.beyls, llvm-commits, efriedma
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62392
llvm-svn: 372259
Summary:
AArch64 GlobalISel doesn't support MachO's large code model, so this patch
adds a check for that combination before implicitly enabling it.
Reviewers: paquette
Subscribers: kristof.beyls, ributzka, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D67724
llvm-svn: 372256
Summary:
As discussed in https://reviews.llvm.org/D62341#1515637,
for MIPS `add %x, -1` isn't optimal. Unlike X86 there
are no fastpaths to matearialize such `-1`/`1` vector constants,
and `sub %x, 1` results in better codegen,
so undo canonicalization
Reviewers: atanasyan, Petar.Avramovic, RKSimon
Reviewed By: atanasyan
Subscribers: sdardis, arichardson, hiraditya, jrtc27, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66805
llvm-svn: 372254
We need "xgot" flag in the MipsAsmParser to implement correct expansion
of some pseudo instructions in case of using 32-bit GOT (XGOT).
MipsAsmParser does not have reference to MipsSubtarget but has a
reference to "feature bit set".
llvm-svn: 372220
This patch fixes a bug exposed by D65653 where a subsequent invocation
of `determineCalleeSaves` ends up with a different size for the callee
save area, leading to different frame-offsets in debug information.
In the invocation by PEI, `determineCalleeSaves` tries to determine
whether it needs to spill an extra callee-saved register to get an
emergency spill slot. To do this, it calls 'estimateStackSize' and
manually adds the size of the callee-saves to this. PEI then allocates
the spill objects for the callee saves and the remaining frame layout
is calculated accordingly.
A second invocation in LiveDebugValues causes estimateStackSize to return
the size of the stack frame including the callee-saves. Given that the
size of the callee-saves is added to this, these callee-saves are counted
twice, which leads `determineCalleeSaves` to believe the stack has
become big enough to require spilling an extra callee-save as emergency
spillslot. It then updates CalleeSavedStackSize with a larger value.
Since CalleeSavedStackSize is used in the calculation of the frame
offset in getFrameIndexReference, this leads to incorrect offsets for
variables/locals when this information is recalculated after PEI.
Reviewers: omjavaid, eli.friedman, thegameg, efriedma
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D66935
llvm-svn: 372204
This generates worse code, but matches what is done for avx2 and
prevents crashes when more arguments are passed than we have
registers for.
llvm-svn: 372200
Currently, not all user specified relocations
(with clang intrinsic __builtin_preserve_access_index())
will turn into relocations.
In the current implementation, a __builtin_preserve_access_index()
chain is turned into relocation only if the result of the clang
intrinsic is used in a function call or a nonzero offset computation
of getelementptr. For all other cases, the relocatiion request
is ignored and the __builtin_preserve_access_index() is turned
into regular getelementptr instructions.
The main reason is to mimic bpf_probe_read() requirement.
But there are other use cases where relocatable offset is
generated but not used for bpf_probe_read(). This patch
relaxed previous constraints when to generate relocations.
Now, all user __builtin_preserve_access_index() will have
relocations generated.
Differential Revision: https://reviews.llvm.org/D67688
llvm-svn: 372198
r361845 changed the way we handle "D16" vs. "D32" targets; there used to
be a negative "d16" which removed instructions from the instruction set,
and now there's a "d32" feature which adds instructions to the
instruction set. This is good, but there was an oversight in the
implementation: the behavior of VFPv2 was changed. In particular, the
"vfp2" feature was changed to imply "d32". This is wrong: VFPv2 only
supports 16 D registers.
In practice, this means if you specify -mfpu=vfpv2, the compiler will
generate illegal instructions.
This patch gets rid of "vfp2d16" and "vfp2d16sp", and fixes "vfp2" and
"vfp2sp" so they don't imply "d32".
Differential Revision: https://reviews.llvm.org/D67375
llvm-svn: 372186
This adds support for `-tailcallopt` tail calls to CallLowering. This
piggy-backs off the changes from D67577, since doing it without a bit of
refactoring gets extremely ugly.
Support is basically ported from AArch64ISelLowering. The main difference here
is that tail calls in `-tailcallopt` change the ABI, so there's some extra
bookkeeping for the stack.
Show that we are correctly lowering these by updating tail-call.ll.
Also show that we don't do anything strange in general by updating
fastcc-reserved.ll, which passes `-tailcallopt`, but doesn't emit any tail
calls.
Differential Revision: https://reviews.llvm.org/D67580
llvm-svn: 372177
We currently produce a load, followed by (possibly a move for integers and) a
splat as separate instructions. VSX has always had a splatting load for
doublewords, but as of Power9, we have it for words as well. This patch just
exploits these instructions.
Differential revision: https://reviews.llvm.org/D63624
llvm-svn: 372139
We were previously using the SelectT2AddrModeImm7 for both normal and narrowing
MVE loads/stores. As the narrowing instructions do not accept sp as a register,
it makes little sense to optimise a FrameIndex into the load, only to have to
recover that later on. This adds a SelectTAddrModeImm7 which does not do that
folding, and uses it for narrowing load/store patterns.
Differential Revision: https://reviews.llvm.org/D67489
llvm-svn: 372134
Similar to D67327, but this time for the FP16 VLDR and VSTR instructions that
use the AddrMode5FP16 addressing mode. We need to reserve an emergency spill
slot for instructions that will be out of range to use sp directly.
AddrMode5FP16 is 8 bits with a scale of 2.
Differential Revision: https://reviews.llvm.org/D67483
llvm-svn: 372132
Remove setPreservesCFG from ARMConstantIslandPass and add a couple
of -verify-machine-dom-info instances into the existing codegen
tests.
llvm-svn: 372126
MVE loads and stores have a 7 bit immediate range, scaled by the length of the type. This needs to be taught to the stack estimation code to ensure that an emergency spill slot is reserved in case we run out of registers when materialising stack indices.
Also the narrowing loads/stores can be created with frame indices even though they do not accept SP as a register. We need in those cases to make sure we have an emergency register to use as the frame base, as SP can never be used.
Differential Revision: https://reviews.llvm.org/D67327
llvm-svn: 372114
Converting the *LoopStart pseudo instructions into DLS/WLS results in
LR being defined. These instructions were inserted on the assumption
that LR would already contain the loop counter because a mov is
introduced during ISel as the the consumers in the loop can only use
LR. That assumption proved wrong!
So perform a safety check, finding an appropriate place to insert the
DLS/WLS instructions or revert if this isn't possible.
Differential Revision: https://reviews.llvm.org/D67539
llvm-svn: 372111
Most of the test changes are trivial instruction reorderings and differing
register allocations, without any obvious performance impact.
Differential Revision: https://reviews.llvm.org/D66973
llvm-svn: 372106
The low-overhead branch extension provides a loop-end 'LE' instruction
that performs no decrement nor compare, it just jumps backwards. This
patch modifies the constant islands pass to try to insert LE
instructions in place of a Thumb2 conditional branch, instead of
shrinking it. This only happens if a cmp can be converted to a cbn/z
and used to exit the loop.
Differential Revision: https://reviews.llvm.org/D67404
llvm-svn: 372085
Previously we tried to split them into narrower v64i1 or v16i1
pieces that each got promoted to vXi8 and then passed in a zmm
or xmm register. But this crashes when you need to pass more
pieces than available registers reserved for argument passing.
The scalarizing done here generates much longer and slower code,
but is consistent with the behavior of avx2 and earlier targets
for these types.
Fixes PR43323.
llvm-svn: 372069
The BLENDM instructions allow an 2 sources and an independent
destination while masked VBROADCAST has the destination tied
to the source.
llvm-svn: 372068
r371901 was overeager and widenScalarDst() and the like in the legalizer
attempt to increment the insert point given in order to add new instructions
after the currently legalizing inst. In cases where the insertion point is not
exactly the current instruction, then callers need to de-compensate for the
behaviour by decrementing the insertion iterator before calling them. It's not
a nice state of affairs, for now just undo the problematic parts of the change.
llvm-svn: 372050
This is a follow up patch from https://reviews.llvm.org/D57857 to handle
extract_subvector v4f32. For cases where we fpext of v2f32 to v2f64 from
extract_subvector we currently generate on P9 the following:
lxv 0, 0(3)
xxsldwi 1, 0, 0, 1
xscvspdpn 2, 0
xxsldwi 3, 0, 0, 3
xxswapd 0, 0
xscvspdpn 1, 1
xscvspdpn 3, 3
xscvspdpn 0, 0
xxmrghd 0, 0, 3
xxmrghd 1, 2, 1
stxv 0, 0(4)
stxv 1, 0(5)
This patch custom lower it to the following sequence:
lxv 0, 0(3) # load the v4f32 <w0, w1, w2, w3>
xxmrghw 2, 0, 0 # Produce the following vector <w0, w0, w1, w1>
xxmrglw 3, 0, 0 # Produce the following vector <w2, w2, w3, w3>
xvcvspdp 2, 2 # FP-extend to <d0, d1>
xvcvspdp 3, 3 # FP-extend to <d2, d3>
stxv 2, 0(5) # Store <d0, d1> (%vecinit11)
stxv 3, 0(4) # Store <d2, d3> (%vecinit4)
Differential Revision: https://reviews.llvm.org/D61961
llvm-svn: 372029
The adds some very basic folding of PREDICATE_CASTS, removing cases when they
are chained together. These would already be removed eventually, as these are
lowered to copies. This just allows it to happen earlier, which can help other
simplifications.
Differential Revision: https://reviews.llvm.org/D67591
llvm-svn: 372012
Lower CTTZ on MVE using VBRSR and VCLS which will reverse the bits and
count the leading zeros, equivalent to a count trailing zeros (CTTZ).
llvm-svn: 372000
MVE has VPT instructions, which perform the duties of both a VCMP and a VPST in
a single instruction, performing the compare and starting the VPT block in one.
This teaches the MVEVPTBlockPass to fold them, searching back through the
basicblock for a valid VCMP and creating the VPT from its operands.
There are some changes to the VPT instructions to accommodate this, altering
the order of the operands to match the VCMP better, and changing P0 register
defs to be VPR defs, as is used in other places.
Differential Revision: https://reviews.llvm.org/D66577
llvm-svn: 371982
Summary:
Adds the following inline asm constraints for SVE:
- Upl: One of the low eight SVE predicate registers, P0 to P7 inclusive
- Upa: SVE predicate register with full range, P0 to P15
Reviewers: t.p.northover, sdesmalen, rovka, momchil.velikov, cameron.mcinally, greened, rengolin
Reviewed By: rovka
Subscribers: javed.absar, tschuett, rkruppe, psnobl, cfe-commits, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D66524
llvm-svn: 371967
After our previous machinecombiner exercises (rL371321, rL371818, rL371833), we
were still missing a few FP16 FMA patterns.
Differential Revision: https://reviews.llvm.org/D67576
llvm-svn: 371960
Masked loads and store fit naturally with MVE, the instructions being easily
predicated. This adds lowering for the simple cases of masked loads and stores.
It does not yet deal with widening/narrowing or pre/post inc, and so is
currently behind an option.
The llvm masked load intrinsic will accept a "passthru" value, dictating the
values used for the zero masked lanes. In MVE the instructions write 0 to the
zero predicated lanes, so we need to match a passthru that isn't 0 (or undef)
with a select instruction to pull in the correct data after the load.
Differential Revision: https://reviews.llvm.org/D67186
llvm-svn: 371932
Craig Topper