vector types to be legal and a ZERO_EXTEND node is encountered.
When we use widening to legalize vector types, extend nodes are a real
challenge. Either the input or output is likely to be legal, but in many
cases not both. As a consequence, we don't really have any way to
represent this situation and the prior code in the widening legalization
framework would just scalarize the extend operation completely.
This patch introduces a new DAG node to represent doing a zero extend of
a vector "in register". The core of the idea is to allow legal but
different vector types in the input and output. The output vector must
have fewer lanes but wider elements. The operation is defined to zero
extend the low elements of the input to the size of the output elements,
and drop all of the high elements which don't have a corresponding lane
in the output vector.
It also includes generic expansion of this node in terms of blending
a zero vector into the high elements of the vector and bitcasting
across. This in turn yields extremely nice code for x86 SSE2 when we use
the new widening legalization logic in conjunction with the new shuffle
lowering logic.
There is still more to do here. We need to support sign extension, any
extension, and potentially int-to-float conversions. My current plan is
to continue using similar synthetic nodes to model each of these
transitions with generic lowering code for each one.
However, with this patch LLVM already reaches performance parity with
GCC for the core C loops of the x264 code (assuming you disable the
hand-written assembly versions) when compiling for SSE2 and SSE3
architectures and enabling the new widening and lowering logic for
vectors.
Differential Revision: http://reviews.llvm.org/D4405
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212610 91177308-0d34-0410-b5e6-96231b3b80d8
tracks which elements of the build vector are in fact undef.
This should make actually inpsecting them (likely in my next patch)
reasonably pretty. Also makes the output parameter optional as it is
clear now that *most* users are happy with undefs in their splats.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212581 91177308-0d34-0410-b5e6-96231b3b80d8
This patch teaches how to fold a shuffle according to rule:
shuffle (shuffle (x, undef, M0), undef, M1) -> shuffle(x, undef, M2)
We do this only if the resulting mask M2 is legal; this is to avoid introducing
illegal shuffles that are potentially expanded into a sub-optimal sequence
of target specific dag nodes.
This patch has the advantage of being target independent, since it works on ISD
nodes. Therefore, all targets (not only x86) can take advantage of this rule.
The idea behind this patch is that most shuffle pairs can be safely combined
before we run the legalizer on vector operations. This allows us to
combine/simplify dag nodes earlier in the process and not only immediately
before instruction selection stage.
That said. This patch is not meant to replace any existing target specific
combine rules; backends might still introduce new shuffles during legalization
stage. Also, this rule is very simple and avoids to aggressively optimize
shuffles.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212539 91177308-0d34-0410-b5e6-96231b3b80d8
aggressively from the x86 shuffle lowering to the generic SDAG vector
shuffle formation code.
This code already tried to fold away shuffles of splats! It just had
lots of bugs and couldn't handle the case my new x86 shuffle lowering
needed.
First, it failed to correctly compute whether N2 was undef because it
pre-computed this, then did transformations which could *make* N2 undef,
then failed to ever re-consider the precomputed state.
Second, it didn't look through bitcasts at all, even in the safe cases
where they are just element-type bitcasts with no change to the number
of elements.
Third, it didn't handle all-zero bit casts nicely the way my code in the
x86 side of things did, which is essential to getting good zext-shuffle
lowerings.
But all of these are generic. I just ported the code down to this layer
and fixed the surrounding bugs. Tests exercising this in the x86 backend
still pass and some silly code in widen_cast-6.ll gets better. I updated
that test to be a bit more precise but it's still pretty unclear what
the value of the test is in this day and age.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212517 91177308-0d34-0410-b5e6-96231b3b80d8
around the handling of UNDEF lanes in boolean vector content analysis.
The code before my changes here also failed to check for non-constant
splats in a buildvector. I have no idea how to trigger this, I just
spotted by inspection when trying to understand the code. It seems
extremely unlikely to be worth the trouble to teach the only caller of
this code (DAG combining setcc patterns) how to cleverly handle undef
lanes, so I've just commented more thoroughly that we're giving up
there.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212515 91177308-0d34-0410-b5e6-96231b3b80d8
nodes about whether they are splats. This is factored out and improved
from r212324 which got reverted as it was far too aggressive. The new
API should help more conservatively handle buildvectors that are
a mixture of splatted and undef values.
No functionality change at this point. The hope is to slowly
re-introduce the undef-tolerant optimization of splats, but each time
being forced to make a concious decision about how to handle the undefs
in a way that doesn't lead to contradicting assumptions about the
collapsed value.
Hal has pointed out in discussions that this may not end up being the
desired API and instead it may be more convenient to get a mask of the
undef elements or something similar. I'm starting simple and will expand
the API as I adapt actual callers and see exactly what they need.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212514 91177308-0d34-0410-b5e6-96231b3b80d8
lanes in vector splats.
The core problem here is that undef lanes can't *unilaterally* be
considered to contribute to splats. Their handling needs to be more
cautious. There is also a reported failure of the nightly testers
(thanks Tobias!) that may well stem from the same core issue. I'm going
to fix this theoretical issue, factor the APIs a bit better, and then
verify that I don't see anything bad with Tobias's reduction from the
test suite before recommitting.
Original commit message for r212324:
[x86] Generalize BuildVectorSDNode::getConstantSplatValue to work for
any constant, constant FP, or undef splat and to tolerate any undef
lanes in a splat, then replace all uses of isSplatVector in X86's
lowering with it.
This fixes issues where undef lanes in an otherwise splat vector would
prevent the splat logic from firing. It is a touch more awkward to use
this interface, but it is much more accurate. Suggestions for better
interface structuring welcome.
With this fix, the code generated with the widening legalization
strategy for widen_cast-4.ll is *dramatically* improved as the special
lowering strategies for a v16i8 SRA kick in even though the high lanes
are undef.
We also get a slightly different choice for broadcasting an aligned
memory location, and use vpshufd instead of vbroadcastss. This looks
like a minor win for pipelining and domain crossing, but a minor loss
for the number of micro-ops. I suspect its a wash, but folks can
easily tweak the lowering if they want.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212475 91177308-0d34-0410-b5e6-96231b3b80d8
We've been performing the wrong operation on ARM for "atomicrmw nand" for
years, since "a NAND b" is "~(a & b)" rather than ARM's very tempting "a & ~b".
This bled over into the generic expansion pass.
So I assume no-one has ever actually tried to do an atomic nand in the real
world. Oh well.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212443 91177308-0d34-0410-b5e6-96231b3b80d8
any constant, constant FP, or undef splat and to tolerate any undef
lanes in a splat, then replace all uses of isSplatVector in X86's
lowering with it.
This fixes issues where undef lanes in an otherwise splat vector would
prevent the splat logic from firing. It is a touch more awkward to use
this interface, but it is much more accurate. Suggestions for better
interface structuring welcome.
With this fix, the code generated with the widening legalization
strategy for widen_cast-4.ll is *dramatically* improved as the special
lowering strategies for a v16i8 SRA kick in even though the high lanes
are undef.
We also get a slightly different choice for broadcasting an aligned
memory location, and use vpshufd instead of vbroadcastss. This looks
like a minor win for pipelining and domain crossing, but a minor loss
for the number of micro-ops. I suspect its a wash, but folks can easily
tweak the lowering if they want.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212324 91177308-0d34-0410-b5e6-96231b3b80d8
subtarget. This involved having the movt predicate take the current
function - since we care about size in instruction selection for
whether or not to use movw/movt take the function so we can check
the attributes. This required adding the current MachineFunction to
FastISel and propagating through.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212309 91177308-0d34-0410-b5e6-96231b3b80d8
This patch sets the 'KeepReg' bit for any tied and live registers during the PrescanInstruction() phase of the dependency breaking algorithm. It then checks those 'KeepReg' bits during the ScanInstruction() phase to avoid changing any tied registers. For more details, please see comments in:
http://llvm.org/bugs/show_bug.cgi?id=20020
I added two FIXME comments for code that I think can be removed by using register iterators that include self. I don't want to include those code changes with this patch, however, to keep things as small as possible.
The test case is larger than I'd like, but I don't know how to reduce it further and still produce the failing asm.
Differential Revision: http://reviews.llvm.org/D4351
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212275 91177308-0d34-0410-b5e6-96231b3b80d8
The PowerPC 128-bit long double data type (ppcf128 in LLVM) is in fact a
pair of two doubles, where one is considered the "high" or
more-significant part, and the other is considered the "low" or
less-significant part. When a ppcf128 value is stored in memory or a
register pair, the high part always comes first, i.e. at the lower
memory address or in the lower-numbered register, and the low part
always comes second. This is true both on big-endian and little-endian
PowerPC systems. (Similar to how with a complex number, the real part
always comes first and the imaginary part second, no matter the byte
order of the system.)
This was implemented incorrectly for little-endian systems in LLVM.
This commit fixes three related issues:
- When printing an immediate ppcf128 constant to assembler output
in emitGlobalConstantFP, emit the high part first on both big-
and little-endian systems.
- When lowering a ppcf128 type to a pair of f64 types in SelectionDAG
(which is used e.g. when generating code to load an argument into a
register pair), use correct low/high part ordering on little-endian
systems.
- In a related issue, because lowering ppcf128 into a pair of f64 must
operate differently from lowering an int128 into a pair of i64,
bitcasts between ppcf128 and int128 must not be optimized away by the
DAG combiner on little-endian systems, but must effect a word-swap.
Reviewed by Hal Finkel.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212274 91177308-0d34-0410-b5e6-96231b3b80d8
This operation was classified as a binary operation in the widening
logic for some reason (clearly, untested). It is in fact a unary
operation. Add a RUN line to a test to exercise this for x86.
Note that again the vector widening strategy doesn't regress anything
and in one case removes a totally unecessary instruction that we
couldn't avoid when promoting the element type.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212257 91177308-0d34-0410-b5e6-96231b3b80d8
vector type legalization strategies in a more fine grained manner, and
change the legalization of several v1iN types and v1f32 to be widening
rather than scalarization on AArch64.
This fixes an assertion failure caused by scalarizing nodes like "v1i32
trunc v1i64". As v1i64 is legal it will fail to scalarize v1i32.
This also provides a foundation for other targets to have more granular
control over how vector types are legalized.
Patch by Hao Liu, reviewed by Tim Northover. I'm committing it to allow
some work to start taking place on top of this patch as it adds some
really important hooks to the backend that I'd like to immediately start
using. =]
http://reviews.llvm.org/D4322
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212242 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r212205.
Reverting this again, still seeing crashes when building compiler-rt...
Sorry for the continued noise, not sure why I'm failing to reproduce
this locally.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212226 91177308-0d34-0410-b5e6-96231b3b80d8
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
heuristic.
By default, no functionality change.
This is a follow-up of r212099.
This hook provides a finer grain to control the optimization.
<rdar://problem/17444599>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212204 91177308-0d34-0410-b5e6-96231b3b80d8
If a function isn't actually in a CU's subprogram list in the debug info
metadata, ignore all the DebugLocs and don't try to build scopes, track
variables, etc.
While this is possibly a minor optimization, it's also a correctness fix
for an incoming patch that will add assertions to LexicalScopes and the
debug info verifier to ensure that all scope chains lead to debug info
for the current function.
Fix up a few test cases that had broken/incomplete debug info that could
violate this constraint.
Add a test case where this occurs by design (inlining a
debug-info-having function in an attribute nodebug function - we want
this to work because /if/ the nodebug function is then inlined into a
debug-info-having function, it should be fine (and will work fine - we
just stitch the scopes up as usual), but should the inlining not happen
we need to not assert fail either).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212203 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r212109, which reverted r212088.
However, disable the assert as it's not necessary for correctness. There are
several corner cases that the assert needed to handle better for in-order
scheduling, but none of them are incorrect scheduler behavior. The assert is
mainly there to collect good unit tests like this and ensure that the
target-independent scheduler is working as expected with the various machine
models.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212187 91177308-0d34-0410-b5e6-96231b3b80d8
switches) into a switch, and sink them into a dispatch function that can
return the result rather than awkward variable setting with breaks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212166 91177308-0d34-0410-b5e6-96231b3b80d8
This doesn't add any information for methods in the VectorLegalizer
class that clearly take SDAG operations to legalize.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212157 91177308-0d34-0410-b5e6-96231b3b80d8
The argument list vector is never used after it has been passed to the
CallLoweringInfo and moving it to the CallLoweringInfo is cleaner and
pretty much as cheap as keeping a pointer to it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212135 91177308-0d34-0410-b5e6-96231b3b80d8
This macro is sometimes defined manually but isn't (and doesn't need to be) in
llvm-config.h so shouldn't appear in the headers, likewise NDEBUG.
Instead switch them over to LLVM_DUMP_METHOD on the definitions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212130 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r212088, which is causing a number of spec
failures. Will provide reduced test cases shortly.
PR20057
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212109 91177308-0d34-0410-b5e6-96231b3b80d8
copies.
This patch extends the peephole optimization introduced in r190713 to produce
register-coalescer friendly copies when possible.
This extension taught the existing cross-bank copy optimization how to deal
with the instructions that generate cross-bank copies, i.e., insert_subreg,
extract_subreg, reg_sequence, and subreg_to_reg.
E.g.
b = insert_subreg e, A, sub0 <-- cross-bank copy
...
C = copy b.sub0 <-- cross-bank copy
Would produce the following code:
b = insert_subreg e, A, sub0 <-- cross-bank copy
...
C = copy A <-- same-bank copy
This patch also introduces a new helper class for that: ValueTracker.
This class implements the logic to look through the copy related instructions
and get the related source.
For now, the advanced rewriting is disabled by default as we are lacking the
semantic on target specific instructions to catch the motivating examples.
Related to <rdar://problem/12702965>.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212100 91177308-0d34-0410-b5e6-96231b3b80d8
By default, no functionality change.
Before evicting a local variable, this heuristic tries to find another (set of)
local(s) that can be reassigned to a free color.
In some extreme cases (large basic blocks with tons of local variables), the
compilation time is dominated by the local interference checks that this
heuristic must perform, with no code gen gain.
E.g., the motivating example takes 4 minutes to compile with this heuristic, 12
seconds without.
Improving the situation will likely require to make drastic changes to the
register allocator and/or the interference check framework.
For now, provide this flag to better understand the impact of that heuristic.
<rdar://problem/17444599>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212099 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r212085.
This breaks the sanitizer bot... & I thought I'd tried pretty hard not
to do that. Guess I need to try harder.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212089 91177308-0d34-0410-b5e6-96231b3b80d8
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
and I now believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212085 91177308-0d34-0410-b5e6-96231b3b80d8
This exception format is not specific to Windows x64. A similar approach is
taken on nearly all architectures. Generalise the name to reflect reality.
This will eventually be used for Windows on ARM data emission as well.
Switch the enum and namespace into an enum class.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212000 91177308-0d34-0410-b5e6-96231b3b80d8
Rename the routines to reflect the reality that they are more related to call
frame information than to Win64 EH. Although EH is implemented in an intertwined
manner by augmenting with an exception handler and an associated parameter, the
majority of these routines emit information required to unwind the frames. This
also helps identify that these routines are generic for most windows platforms
(they apply equally to nearly all architectures except x86) although the
encoding of the information is architecture dependent.
Unwinding data is emitted via EmitWinCFI* and exception handling information via
EmitWinEH*.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211994 91177308-0d34-0410-b5e6-96231b3b80d8
This new IR facility allows us to represent the object-file semantic of
a COMDAT group.
COMDATs allow us to tie together sections and make the inclusion of one
dependent on another. This is required to implement features like MS
ABI VFTables and optimizing away certain kinds of initialization in C++.
This functionality is only representable in COFF and ELF, Mach-O has no
similar mechanism.
Differential Revision: http://reviews.llvm.org/D4178
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211920 91177308-0d34-0410-b5e6-96231b3b80d8
Reverting this again, didn't mean to commit it - while r211872 fixes one
of the issues here, there are still others to figure out and address.
This reverts commit r211871.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211873 91177308-0d34-0410-b5e6-96231b3b80d8
Fixe for Bug 20057 - Assertion failied in llvm::SUnit* llvm::SchedBoundary::pickOnlyChoice(): Assertion `i <= (HazardRec->getMaxLookAhead() + MaxObservedStall) && "permanent hazard"'
Thanks to Chad for the test case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211865 91177308-0d34-0410-b5e6-96231b3b80d8
There is no need to calculate the liveness information for stackmaps. The
liveness information is still available for the patchpoint intrinsic and
that is also the intended usage model.
Related to <rdar://problem/17473725>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211816 91177308-0d34-0410-b5e6-96231b3b80d8
