This introduces a new truncation bit for truncation of minus-zero to zero.
At the moment it is only used to handle the limit cases of deopt, such as the
one in the Google maps workload (see simplified version below), where the -q
(which is desugared to q * -1.0) currently deoptimizes because the result would
produce minus zero. To handle this situation, we exploit the knowledge that
righthand side of + cannot be -0, so even if lefthand side was -0, the result
would still be 0 (so the + operation cannot distinguish between left hand side
0 and -0).
function f(q) {
q -= 4;
return (-q) + q;
}
f(10);
f(10);
%OptimizeFunctionOnNextCall(f);
f(4);
Review-Url: https://codereview.chromium.org/2734253002
Cr-Commit-Position: refs/heads/master@{#43661}
Don't unconditionally assume that Tagged->Word32 changes are always
truncating independent of the type of the input. Work-around the
remaining issues with for-in by properly renaming the index if it's
not already of appropriate type (happens with generators or OSR).
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2545393002
Cr-Commit-Position: refs/heads/master@{#41484}
This is preparation for using TF to create builtins that handle variable number of
arguments and have to remove these arguments dynamically from the stack upon
return.
The gist of the changes:
- Added a second argument to the Return node which specifies the number of stack
slots to pop upon return in addition to those specified by the Linkage of the
compiled function.
- Removed Tail -> Non-Tail fallback in the instruction selector. Since TF now should
handles all tail-call cases except where the return value type differs, this fallback
was not really useful and in fact caused unexpected behavior with variable
sized argument popping, since it wasn't possible to materialize a Return node
with the right pop count from the TailCall without additional context.
- Modified existing Return generation to pass a constant zero as the additional
pop argument since the variable pop functionality
LOG=N
Review-Url: https://codereview.chromium.org/2446543002
Cr-Commit-Position: refs/heads/master@{#40699}
Reason for revert:
Seems to break arm64 sim debug and blocks roll:
https://build.chromium.org/p/client.v8.ports/builders/V8%20Linux%20-%20arm64%20-%20sim%20-%20debug/builds/3294
Original issue's description:
> [turbofan] Support variable size argument removal in TF-generated functions
>
> This is preparation for using TF to create builtins that handle variable number of
> arguments and have to remove these arguments dynamically from the stack upon
> return.
>
> The gist of the changes:
> - Added a second argument to the Return node which specifies the number of stack
> slots to pop upon return in addition to those specified by the Linkage of the
> compiled function.
> - Removed Tail -> Non-Tail fallback in the instruction selector. Since TF now should
> handles all tail-call cases except where the return value type differs, this fallback
> was not really useful and in fact caused unexpected behavior with variable
> sized argument popping, since it wasn't possible to materialize a Return node
> with the right pop count from the TailCall without additional context.
> - Modified existing Return generation to pass a constant zero as the additional
> pop argument since the variable pop functionality
>
> LOG=N
TBR=bmeurer@chromium.org,mstarzinger@chromium.org,epertoso@chromium.org,danno@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
NOPRESUBMIT=true
Review-Url: https://codereview.chromium.org/2473643002
Cr-Commit-Position: refs/heads/master@{#40691}
This is preparation for using TF to create builtins that handle variable number of
arguments and have to remove these arguments dynamically from the stack upon
return.
The gist of the changes:
- Added a second argument to the Return node which specifies the number of stack
slots to pop upon return in addition to those specified by the Linkage of the
compiled function.
- Removed Tail -> Non-Tail fallback in the instruction selector. Since TF now should
handles all tail-call cases except where the return value type differs, this fallback
was not really useful and in fact caused unexpected behavior with variable
sized argument popping, since it wasn't possible to materialize a Return node
with the right pop count from the TailCall without additional context.
- Modified existing Return generation to pass a constant zero as the additional
pop argument since the variable pop functionality
LOG=N
Review-Url: https://codereview.chromium.org/2446543002
Cr-Commit-Position: refs/heads/master@{#40678}
The only way to get a minus zero result from subtraction is
(-0) - (+0) = -0, hence checking for minus zero on the RHS is
redundant. This is causing some unnecessary deoptimisations
in Box2D from Octane on 32-bit platforms.
BUG=
Review-Url: https://codereview.chromium.org/2410883003
Cr-Commit-Position: refs/heads/master@{#40207}
Add a dedicated simplified operator to inline the general case for the
ToBoolean conversion. In a follow up CL we will also use the ToBoolean
hints gathered by the baseline compiler.
CQ_INCLUDE_TRYBOTS=master.tryserver.v8:v8_linux_arm64_gc_stress_dbg
R=jarin@chromium.org
BUG=v8:5267
Committed: https://crrev.com/8c50b51ab3d21efcd2f6900d83962159f21e1590
Review-Url: https://codereview.chromium.org/2167593002
Cr-Original-Commit-Position: refs/heads/master@{#37882}
Cr-Commit-Position: refs/heads/master@{#39420}
Reason for revert:
Octane/Mandreel aborts with an exception now:
TypeError: __FUNCTION_TABLE__[(r2 >> 2)] is not a function
Original issue's description:
> [turbofan] Insert dummy values when changing from None type.
>
> Currently we choose the MachineRepresentation::kNone representation for
> values of Type::None, and when converting values from the kNone representation
> we use "impossible" conversions that will crash at runtime. This
> assumes that the impossible conversions should never be hit (the only
> way to produce the impossible values is to perform an always-failing
> runtime check on a value, such as Smi-checking a string). Note that
> this assumes that the runtime check is executed before the impossible
> convesrion.
>
> Introducing BitwiseOr type feedback broke this in two ways:
>
> - we always pick Word32 representation for bitwise-or, so the
> impossible conversion does not trigger (it only triggers with
> None representation), and we could end up with unsupported
> conversions from Word32.
>
> - even if we inserted impossible conversions, they are pure conversions.
> Since untagging, bitwise-or operations are also pure, we could hoist
> all these before the smi check of the inputs and we could hit the
> impossible conversions before we get to the smi check.
>
> This CL addresses this by just providing dummy values for conversions
> from the Type::None type. It also removes the impossible-to-* conversions.
>
> BUG=chromium:638132
>
> Committed: https://crrev.com/c83b21ab755f1420b6da85b3ff43d7e96ead9bbe
> Cr-Commit-Position: refs/heads/master@{#38883}
TBR=mstarzinger@chromium.org,jarin@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:638132
Review-Url: https://codereview.chromium.org/2280613002
Cr-Commit-Position: refs/heads/master@{#38893}
Currently we choose the MachineRepresentation::kNone representation for
values of Type::None, and when converting values from the kNone representation
we use "impossible" conversions that will crash at runtime. This
assumes that the impossible conversions should never be hit (the only
way to produce the impossible values is to perform an always-failing
runtime check on a value, such as Smi-checking a string). Note that
this assumes that the runtime check is executed before the impossible
convesrion.
Introducing BitwiseOr type feedback broke this in two ways:
- we always pick Word32 representation for bitwise-or, so the
impossible conversion does not trigger (it only triggers with
None representation), and we could end up with unsupported
conversions from Word32.
- even if we inserted impossible conversions, they are pure conversions.
Since untagging, bitwise-or operations are also pure, we could hoist
all these before the smi check of the inputs and we could hit the
impossible conversions before we get to the smi check.
This CL addresses this by just providing dummy values for conversions
from the Type::None type. It also removes the impossible-to-* conversions.
BUG=chromium:638132
Review-Url: https://codereview.chromium.org/2266823002
Cr-Commit-Position: refs/heads/master@{#38883}
This leads to a better handling of the Smi case when we introduce a checked truncation from a number or oddbal to a 32 bit word, which we were previously doing by concatenating a Smi to float64 conversion with a float64 to word32 truncation.
BUG=
Review-Url: https://codereview.chromium.org/2191503002
Cr-Commit-Position: refs/heads/master@{#38091}
Add a dedicated simplified operator to inline the general case for the
ToBoolean conversion. In a follow up CL we will also use the ToBoolean
hints gathered by the baseline compiler.
Review-Url: https://codereview.chromium.org/2167593002
Cr-Commit-Position: refs/heads/master@{#37882}
This introduces optimized number operations based on type feedback.
Summary of changes:
1. Typed lowering produces SpeculativeNumberAdd/Subtract for JSAdd/Subtract if
there is suitable feedback. The speculative nodes are connected to both the
effect chain and the control chain and they retain the eager frame state.
2. Simplified lowering now executes in three phases:
a. Propagation phase computes truncations by traversing the graph from uses to
definitions until checkpoint is reached. It also records type-check decisions
for later typing phase, and computes representation.
b. The typing phase computes more precise types base on the speculative types (and recomputes
representation for affected nodes).
c. The lowering phase performs lowering and inserts representation changes and/or checks.
3. Effect-control linearization lowers the checks to machine graphs.
Notes:
- SimplifiedLowering will be refactored to have handling of each operation one place and
with clearer input/output protocol for each sub-phase. I would prefer to do this once
we have more operations implemented, and the pattern is clearer.
- The check operations (Checked<A>To<B>) should have some flags that would affect
the kind of truncations that they can handle. E.g., if we know that a node produces
a number, we can omit the oddball check in the CheckedTaggedToFloat64 lowering.
- In future, we want the typer to reuse the logic from OperationTyper.
BUG=v8:4583
LOG=n
Review-Url: https://codereview.chromium.org/1921563002
Cr-Commit-Position: refs/heads/master@{#36674}
This prevents the compiler from optimizing
f64-to-tagged(tagged-to-f64(x)) ==> x
for non-number x (such as undefined).
Review-Url: https://codereview.chromium.org/2027593002
Cr-Commit-Position: refs/heads/master@{#36613}
Now ChangeLowering is only concerned with lowering memory access and
allocation operations, and all changes are consistently lowered during
the effect/control linearization pass. The next step is to move the
left over lowerings to a pass dedicated to eliminate redundant loads and
stores, eliminate write barriers, fold and inline allocations.
Drive-by-fix: Rename ChangeBitToBool to ChangeBitToTagged,
ChangeBoolToBit to ChangeTaggedToBit, and ChangeInt31ToTagged to
ChangeInt31ToTaggedSigned for consistency.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux64_tsan_rel
Committed: https://crrev.com/ceca5ae308bddda166651c654f96d71d74f617d0
Cr-Commit-Position: refs/heads/master@{#35924}
Review-Url: https://codereview.chromium.org/1941673002
Cr-Commit-Position: refs/heads/master@{#35929}
Reason for revert:
[Sheriff] Breaks mac gc stress:
https://build.chromium.org/p/client.v8/builders/V8%20Mac%20GC%20Stress/builds/5821
Original issue's description:
> [turbofan] Remove left-over change bits from ChangeLowering.
>
> Now ChangeLowering is only concerned with lowering memory access and
> allocation operations, and all changes are consistently lowered during
> the effect/control linearization pass. The next step is to move the
> left over lowerings to a pass dedicated to eliminate redundant loads and
> stores, eliminate write barriers, fold and inline allocations.
>
> Also remove the atomic regions now that we wire everything into the
> effect chain properly. This is an important step towards allocation
> inlining.
>
> Drive-by-fix: Rename ChangeBitToBool to ChangeBitToTagged,
> ChangeBoolToBit to ChangeTaggedToBit, and ChangeInt31ToTagged to
> ChangeInt31ToTaggedSigned for consistency.
>
> CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux64_tsan_rel
>
> Committed: https://crrev.com/ceca5ae308bddda166651c654f96d71d74f617d0
> Cr-Commit-Position: refs/heads/master@{#35924}
TBR=ishell@chromium.org,bmeurer@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review-Url: https://codereview.chromium.org/1942733002
Cr-Commit-Position: refs/heads/master@{#35927}
Now ChangeLowering is only concerned with lowering memory access and
allocation operations, and all changes are consistently lowered during
the effect/control linearization pass. The next step is to move the
left over lowerings to a pass dedicated to eliminate redundant loads and
stores, eliminate write barriers, fold and inline allocations.
Also remove the atomic regions now that we wire everything into the
effect chain properly. This is an important step towards allocation
inlining.
Drive-by-fix: Rename ChangeBitToBool to ChangeBitToTagged,
ChangeBoolToBit to ChangeTaggedToBit, and ChangeInt31ToTagged to
ChangeInt31ToTaggedSigned for consistency.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux64_tsan_rel
Review-Url: https://codereview.chromium.org/1941673002
Cr-Commit-Position: refs/heads/master@{#35924}
This allows us to get rid of the "push TruncateFloat64ToInt32 into Phi"
trick that was used in the MachineOperatorReducer to combine the
ChangeTaggedToFloat64 and TruncateFloat64ToInt32 operations. Instead of
doing that later, we can just introduce the proper operator during the
representation selection directly.
Also separate the TruncateFloat64ToInt32 machine operator, which had two
different meanings depending on a flag (either JavaScript truncation or
C++ style round to zero). Now there's a TruncateFloat64ToWord32 which
represents the JavaScript truncation (implemented via TruncateDoubleToI
macro + code stub) and the RoundFloat64ToInt32, which implements the C++
round towards zero operation (in the same style as the other WebAssembly
driven Round* machine operators).
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1919513002
Cr-Commit-Position: refs/heads/master@{#35743}
Get rid of further typing checks from ChangeLowering and put them into
the representation selection pass instead (encoding the information in
the operator instead).
Drive-by-change: Rename ChangeSmiToInt32 to ChangeTaggedSignedToInt32
for consistency about naming Tagged, TaggedSigned and TaggedPointer.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1909343002
Cr-Commit-Position: refs/heads/master@{#35723}
If we have to convert a float64 value to tagged representation and we
already know that the value is either in Signed31/Signed32 or
Unsigned32 range, then we can just convert the float64 to word32 and
use the fast word32 to tagged conversion. Doing this in
ChangeLowering (or the effect linearization pass) would be unsound, as
the types on the nodes are no longer usable.
This removes all Type uses from effect linearization. There's still some
work to be done for ChangeLowering tho.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1908093002
Cr-Commit-Position: refs/heads/master@{#35713}
Instead of using CheckFloatEq and CheckDoubleEq directly, I introduced
a macro which first stores the expected result in a volatile variable.
Here are some comments of previous CLs:
The reason is same as the CL #31808 (issue 1430943002, X87: Change the test case for X87 float operations), please refer: https://codereview.chromium.org/1430943002/.
Here is the key comments from CL #31808
Some new test cases use CheckFloatEq(...) and CheckDoubleEq(...) function for result check. When GCC compiling the CheckFloatEq() and CheckDoubleEq() function,
those inlined functions has different behavior comparing with GCC ia32 build and x87 build.
The major difference is sse float register still has single precision rounding semantic. While X87 register has no such rounding precsion semantic when directly use register value.
The V8 turbofan JITTed has exactly same result in both X87 and IA32 port.
So we add the following sentence to do type cast to keep the same precision for RunCallInt64ToFloat32/RunCallInt64ToFloat64. Such as: volatile double expect = static_cast<float>(*i).
R=titzer@chromium.org, weiliang.lin@intel.com
Review URL: https://codereview.chromium.org/1773513002
Cr-Commit-Position: refs/heads/master@{#34534}
TryTruncateFloat32ToUint64 converts a float32 to a uint64. Additionally it
provides an optional second return value which indicates whether the conversion
succeeded (i.e. float32 value was within uint64 range) or not.
I implemented the new operator on x64, arm64, and mips64. @v8-ppc-ports, can you
please take care of the ppc64 implementation of the second output?
Additionally I fixed a bug on x64 and mips64 in the implementation of
TryTruncateFloat64ToUint64. Cases where the input value was between -1 and 0
were handled incorrectly.
R=titzer@chromium.org, v8-arm-ports@googlegroups.com, v8-mips-ports@googlegroups.com
Review URL: https://codereview.chromium.org/1512023002
Cr-Commit-Position: refs/heads/master@{#32796}
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}
This change replaces the bitwise masking of uses by storing the most general truncation for all uses.
Review URL: https://codereview.chromium.org/1464763003
Cr-Commit-Position: refs/heads/master@{#32248}
If the input type does not help us, we are conservative and truncate (rather than guessing signed).
Review URL: https://codereview.chromium.org/1455103002
Cr-Commit-Position: refs/heads/master@{#32075}
This moves all cctest files for the compiler to live in the same
namespace as the components they are testing. Hence we can avoid the
forbidden using directives pulling in entire namespaces.
From the Google C++ style guide: "You may not use a using-directive to
make all names from a namespace available". This would be covered by
presubmit linter checks if build/namespaces were not blacklisted.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1424943004
Cr-Commit-Position: refs/heads/master@{#31671}
This fixes the lifetime of nodes created by JSGlobalSpecialization that
contain a simplified operator. In the case where this reducer runs as
part of the inliner, the SimplifiedOperatorBuilder was instantiated with
the wrong zone. This led to use-after-free of simplified operators.
To avoid such situations in the future, we decided to move this operator
builder into the JSGraph and make the situation uniform with all other
operator builders.
R=bmeurer@chromium.org
BUG=chromium:543528
LOG=n
Review URL: https://codereview.chromium.org/1409993002
Cr-Commit-Position: refs/heads/master@{#31334}
The usage of Unique<T> throughout the TurboFan IR does not have any
advantage. There is no single point in time when they are initialized
and most use-sites looked through to the underlying Handle<T> anyways.
Also there already was a mixture of Handle<T> versus Unique<T> in the
graph and this unifies the situation to use Handle<T> everywhere.
R=bmeurer@chromium.org,titzer@chromium.org
Review URL: https://codereview.chromium.org/1314473007
Cr-Commit-Position: refs/heads/master@{#30458}
When compiling on a laptop I like to concatenate the small test files.
This makes a big difference to compile times. These changes make that
easier.
R=ulan@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1163803002
Cr-Commit-Position: refs/heads/master@{#28742}
Along the way:
- Thread isolate parameter explicitly through code that used to
rely on getting it from the zone.
- Canonicalize the parameter position of isolate and zone for
affected code
- Change Hydrogen New<> instruction templates to automatically
pass isolate
R=mstarzinger@chromium.org
LOG=N
Review URL: https://codereview.chromium.org/868883002
Cr-Commit-Position: refs/heads/master@{#26252}
- Extend verifier to check types of JS and Simplified nodes.
- Untyped nodes now contain NULL as types, enforcing hard failure.
- Typer immediately installs itself as a decorator; remove explicit decorator installation.
- Decorator eagerly types all nodes that have typed inputs
(subsumes typing of constant cache, removing its typing
side-channel and various spurious dependencies on the typer).
- Cut down typer interface to prevent inconsistently typed graphs.
- Remove verification from start, since it caused too much trouble
with semi-wellformed nodes.
- Fix a couple of bugs on the way that got uncovered.
To do: verifying machine operators. Also, various conditions in the
verifier are currently commented out, because they don't yet hold.
BUG=
R=jarin@chromium.org,titzer@chromium.org
Review URL: https://codereview.chromium.org/658543002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24626 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
