Loop unrolling did not work properly with floating control. Seeing as
very few spots in the wasm compiler introduced floating control, we
decided to disallow it altogether.
Changes:
- When lowering 64-bit rol/ror/clz/ctz in 32-bit platforms, we use a
diamond operator, which used to introduce floating control. This CL
adds a control edge to these operators so that the diamond can be
chained to that control instead.
- During loop analysis, as an additional safety check, we check that the
explored loop does not have floating control. Exceptionally, floating
control pointing directly do start() is allowed.
- Change wasm-compiler so that generated floating projections point to
start() even after stack check patch-in.
Bug: chromium:1184929, v8:11298
Change-Id: I1ee063f5250037ae6c84d2f16b0bd8fff3923117
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2876851
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#74527}
Add Float32Select and Float64Select as OptionalOperators and insert
these, if supported, when handling a Select expression in the wasm
graph builder. FlagsContinuation have been modified to support the
select operation and code generation support has been added for arm64.
This improves the 'Bullet' physics benchmark by ~2-3%.
Change-Id: I928c3085c9136ad8baeeb34c71c47c1c8338844c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2763871
Commit-Queue: Martyn Capewell <martyn.capewell@arm.com>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#73657}
This is a reland of 4a16305b65
The original CL adjust only one part of the stack check, namely the
comparison of the stack pointer against the stack limit in generated code.
There is a second part: Runtime::kStackGuard repeats this check to
distinguish between a stack overflow and an interrupt request.
This second part in runtime must apply the offset just like in generated
code. It is implemented in this reland by the StackCheckOffset operator
and a new StackGuardWithGap runtime function.
Original change's description:
> [compiler] Optionally apply an offset to stack checks
>
> The motivation behind this change is that the frame size of an optimized
> function and its unoptimized version may differ, and deoptimization
> may thus trigger a stack overflow. The solution implemented in this CL
> is to optionally apply an offset to the stack check s.t. the check
> becomes 'sp - offset > limit'. The offset is applied to stack checks at
> function-entry, and is set to the difference between the optimized and
> unoptimized frame size.
>
> A caveat: OSR may not be fully handled by this fix since we've already
> passed the function-entry stack check. A possible solution would be to
> *not* skip creation of function-entry stack checks for inlinees.
>
> This CL: 1. annotates stack check nodes with the stack check kind, where
> kind is one of {function-entry,iteration-body,unknown}. 2. potentially
> allocates a temporary register to store the result of the 'sp - offset'
> in instruction selection (and switches input registers to 'unique'
> mode). 3. Applies the offset in code generation.
>
> Drive-by: Add src/compiler/globals.h for compiler-specific globals.
>
> Bug: v8:9534,chromium:1000887
> Change-Id: I257191c4a4978ccb60cfa5805ef421f30f0e9826
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1762521
> Commit-Queue: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Georg Neis <neis@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#63701}
Bug: v8:9534, chromium:1000887
Change-Id: I71771c281afd7d57c09aa48ea1b182d01e6dee2a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1822037
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64634}
This reverts commit 4a16305b65.
Reason for revert: Need to revalidate assumptions behind the CHECK.
Original change's description:
> [compiler] Optionally apply an offset to stack checks
>
> The motivation behind this change is that the frame size of an optimized
> function and its unoptimized version may differ, and deoptimization
> may thus trigger a stack overflow. The solution implemented in this CL
> is to optionally apply an offset to the stack check s.t. the check
> becomes 'sp - offset > limit'. The offset is applied to stack checks at
> function-entry, and is set to the difference between the optimized and
> unoptimized frame size.
>
> A caveat: OSR may not be fully handled by this fix since we've already
> passed the function-entry stack check. A possible solution would be to
> *not* skip creation of function-entry stack checks for inlinees.
>
> This CL: 1. annotates stack check nodes with the stack check kind, where
> kind is one of {function-entry,iteration-body,unknown}. 2. potentially
> allocates a temporary register to store the result of the 'sp - offset'
> in instruction selection (and switches input registers to 'unique'
> mode). 3. Applies the offset in code generation.
>
> Drive-by: Add src/compiler/globals.h for compiler-specific globals.
>
> Bug: v8:9534,chromium:1000887
> Change-Id: I257191c4a4978ccb60cfa5805ef421f30f0e9826
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1762521
> Commit-Queue: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Georg Neis <neis@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#63701}
TBR=neis@chromium.org,sigurds@chromium.org,jgruber@chromium.org
Change-Id: Iebf46d5256b6dee13451741781ef85a5fe9b1628
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:9534, chromium:1000887
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1800565
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#63706}
The motivation behind this change is that the frame size of an optimized
function and its unoptimized version may differ, and deoptimization
may thus trigger a stack overflow. The solution implemented in this CL
is to optionally apply an offset to the stack check s.t. the check
becomes 'sp - offset > limit'. The offset is applied to stack checks at
function-entry, and is set to the difference between the optimized and
unoptimized frame size.
A caveat: OSR may not be fully handled by this fix since we've already
passed the function-entry stack check. A possible solution would be to
*not* skip creation of function-entry stack checks for inlinees.
This CL: 1. annotates stack check nodes with the stack check kind, where
kind is one of {function-entry,iteration-body,unknown}. 2. potentially
allocates a temporary register to store the result of the 'sp - offset'
in instruction selection (and switches input registers to 'unique'
mode). 3. Applies the offset in code generation.
Drive-by: Add src/compiler/globals.h for compiler-specific globals.
Bug: v8:9534,chromium:1000887
Change-Id: I257191c4a4978ccb60cfa5805ef421f30f0e9826
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1762521
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#63701}
Now that all uses of LoadStackPointer have been removed, this CL cleans
up related code:
- Removed LoadStackPointer.
- Removed ArchStackPointer.
- Removed IA32StackCheck.
- Removed X64StackCheck.
- Removed StackCheckMatcher.
All stack checks now follow a simple path without matchers or special
register constraints: they load the limit and pass it to
StackPointerGreaterThan, which is finally handled by code generation.
Bug: v8:9534
Change-Id: Ib1d7be1502a471541d6441f3261aac0c949525fb
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1748737
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#63166}
This replaces all typedefs that define types and not functions by the
equivalent "using" declaration.
This was done mostly automatically using this command:
ag -l '\btypedef\b' src test | xargs -L1 \
perl -i -p0e 's/typedef ([^*;{}]+) (\w+);/using \2 = \1;/sg'
Patchset 2 then adds some manual changes for typedefs for pointer types,
where the regular expression did not match.
R=mstarzinger@chromium.orgTBR=yangguo@chromium.org, jarin@chromium.org
Bug: v8:9183
Change-Id: I6f6ee28d1793b7ac34a58f980b94babc21874b78
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1631409
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#61849}
Googletest is (at last) converging with industry-standard terminology
[1]. We previously called test suites "test cases", which was rather
confusing for folks coming from any other testing framework.
Chrome now has a googletest version that supports _TEST_SUITE_ macros
instead of _TEST_CASE_, so this CL cleans up some of the outdated usage.
[1] https://github.com/google/googletest/blob/master/googletest/docs/primer.md#beware-of-the-nomenclature
Bug: chromium:925652
Change-Id: I3cd02b9fa6dbece1594bbfd50a21ad7503c2aab9
Reviewed-on: https://chromium-review.googlesource.com/c/1475654
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Victor Costan <pwnall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#59666}
This was removed from gtest and is necessary to roll gtest in
Chromium.
TBR=adamk@chromium.org
Bug: chromium:893369
Change-Id: I21762aa65ab2fc3f52731e7e812f0bf155f285e6
Reviewed-on: https://chromium-review.googlesource.com/c/1310598
Commit-Queue: Fabrice de Gans-Riberi <fdegans@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57193}
So far we don't have a useful way to inline Math.max or Math.min in
TurboFan optimized code. This adds new operators NumberMax and NumberMin
and changes the Float64Max/Float64Min operators to have JavaScript
semantics instead of the C++ semantics that it had previously.
This also removes support for recognizing the tenary case in the
CommonOperatorReducer, since that doesn't seem to have any positive
impact (and actually doesn't show up in regular JavaScript, where
people use Math.max/Math.min instead).
Drive-by-fix: Also nuke the unused Float32Max/Float32Min operators.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2170343002
Cr-Commit-Position: refs/heads/master@{#37971}
Add control dependencies to Projection and Int32Add/SubWithOverflow
operators, to prevent the scheduler from moving the Projection nodes
into the wrong place. This way the instruction selection can combine
the Int32Add/SubWithOverflow operations with the DeoptimizeIf and/or
DeoptimizeUnless nodes. This needs new operators CheckedInt32Add and
CheckedInt32Sub so that we can delay the actual lowering until the
effect/control linearizer.
This also makes CheckIf operator obsolete, so we can drop it.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2082993002
Cr-Commit-Position: refs/heads/master@{#37148}
Adding optional operators for FNeg for WebAssembly, as the current implementation was significantly suboptimal for ARM.
Review-Url: https://codereview.chromium.org/2011303002
Cr-Commit-Position: refs/heads/master@{#36544}
MachineType is now a class with two enum fields:
- MachineRepresentation
- MachineSemantic
Both enums are usable on their own, and this change switches some places from using MachineType to use just MachineRepresentation. Most notably:
- register allocator now uses just the representation.
- Phi and Select nodes only refer to representations.
Review URL: https://codereview.chromium.org/1513543003
Cr-Commit-Position: refs/heads/master@{#32738}
We actually need round to zero truncation to implement the counterpart
of LDoubleToI in TurboFan, which tries to convert a double to an integer
as required for keyed load/store optimizations.
Drive-by-cleanup: Reduce some code duplication in the InstructionSelector
implementations.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1225993002
Cr-Commit-Position: refs/heads/master@{#29527}
This makes usage of the MachineOperatorBuilder more robust, as it will be
an error to request an unsupported operator.
Along the way, I noticed that all 7 platforms support Float32Abs and
Float64Abs. Should make them non-optional in another CL?
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1128133003
Cr-Commit-Position: refs/heads/master@{#29223}
These operators compute the absolute floating point value of some
arbitrary input, and are implemented without any branches (i.e. using
vabs on arm, and andps/andpd on x86).
R=svenpanne@chromium.org
Review URL: https://codereview.chromium.org/1066393002
Cr-Commit-Position: refs/heads/master@{#27662}
This adds the basics necessary to support float32 operations in TurboFan.
The actual functionality required to detect safe float32 operations will
be added based on this later. Therefore this does not affect production
code except for some cleanup/refactoring.
In detail, this patchset contains the following features:
- Add support for float32 operations to arm, arm64, ia32 and x64
backends.
- Add float32 machine operators.
- Add support for float32 constants to simplified lowering.
- Handle float32 representation for phis in simplified lowering.
In addition, contains the following (related) cleanups:
- Fix/unify naming of backend instructions.
- Use AVX comparisons when available.
- Extend ArchOpcodeField to 9 bits (required for arm64).
- Refactor some code duplication in instruction selectors.
BUG=v8:3589
LOG=n
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/1044793002
Cr-Commit-Position: refs/heads/master@{#27509}
Basically recognize certain x < y ? x : y constructs and turn that into
Float64Min/Float64Max operations, if the target machine supports that.
On x86 we lower to (v)minsd/(v)maxsd.
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/998283002
Cr-Commit-Position: refs/heads/master@{#27160}
Provide an intrinsic %MathFloor / %_MathFloor that is used to optimize
both Math.ceil and Math.floor, and use the JS inlining mechanism to
inline Math.ceil into TurboFan code. Although we need to touch code
outside of TurboFan to make this work, this does not affect the way we
handle Math.ceil and/or Math.floor in CrankShaft, because for CrankShaft
the old-style builtin function id based inlining still kicks in first.
Once this solution is stabilized, we can use it for Math.floor as well.
And once that is settled, we can establish it as the unified way to
inline builtins, and get rid of the specialized builtin function id
based inlining at some point.
Note that "builtin" applies to basically every piece of internal
JavaScript/intrinsics based code, so this also applies to the yet to be
defined JavaScript based code stubs and handlers.
BUG=v8:3953
LOG=n
R=yangguo@chromium.org,svenpanne@chromium.org
Review URL: https://codereview.chromium.org/990963003
Cr-Commit-Position: refs/heads/master@{#27086}
This adds support for the double bits intrinsics to TurboFan, and is
a first step towards fast Math functions inlined into TurboFan code
or even compiled by themselves with TurboFan.
Review URL: https://codereview.chromium.org/974313002
Cr-Commit-Position: refs/heads/master@{#27006}
