Rolling v8/buildtools to e84114dbe2b65428951c876349b6a3ff1afbfccd
Rolling v8/tools/clang to 2956eca572ff0e1b181df65f71a045f061a2eb34
TBR=machenbach@chromium.org,vogelheim@chromium.org,hablich@chromium.org
Review URL: https://codereview.chromium.org/1909483004
Cr-Commit-Position: refs/heads/master@{#35690}
The CL #35651 (https://codereview.chromium.org/1858323003) exposed one hiden issue in RunTruncateFloat32ToUint32 test cases and X87 failed at it.
Here is the issue in RunTruncateFloat32ToUint32:
For float input = static_cast<float>(*i), the x87 GCC would optimize the input viariable in float floating register for release build.
The problem is:
SSE float register has single precision rounding semantic While X87 register hasn't when directly use floating register value. It will cause the value of input viariable has
different precision for IA32 and X87 port. So static_cast<uint32_t>(input) will be different for IA32 and X87 port too.
This led to CHECK_EQ(static_cast<uint32_t>(input), m.Call(input)) fail although V8 turbofan JITTed code m.Call(input) has exactly same result in both X87 and IA32 port.
So we add the following sentence to do type cast to keep the single precision for RunTruncateFloat32ToUint32 by forcing the input viariable get value from memory insread of
floating register.
Such as: volatile float input = static_cast<float>(*i).
BUG=
Review URL: https://codereview.chromium.org/1905883002
Cr-Commit-Position: refs/heads/master@{#35689}
This patch provides a new implementation of popcnt and ctz in the case
where the platform does not provide these instructions. Instead of
building a TF graph which implements it we now call a C function.
Additionally I turned on additional tests in test-run-wasm-64.cc
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1857363003
Cr-Commit-Position: refs/heads/master@{#35685}
Port 3518e492c0
Original commit message:
Short external strings do not cache the resource data, and may be used
for compressible strings. The assumptions about their lengths is
invalid and may lead to oob reads.
R=bmeurer@chromium.org
BUG=v8:4923,chromium:604897
LOG=N
Review URL: https://codereview.chromium.org/1902393004
Cr-Commit-Position: refs/heads/master@{#35683}
port 3518e492c0 (r35660)
original commit message:
Short external strings do not cache the resource data, and may be used
for compressible strings. The assumptions about their lengths is
invalid and may lead to oob reads.
BUG=
Review URL: https://codereview.chromium.org/1904003003
Cr-Commit-Position: refs/heads/master@{#35681}
Our previous over-conservative answer caused us to emit hole checks in
full-codegen when eagerly parsing but not when lazily parsing.
With this patch, we use the positions of the BinaryOperations making up
the parameter list (which are the positions of the commas) to determine
the appropriate "end position" for each parameter's initializer. This means
that we get accurate-enough positions for the initializers in the eager
parsing step to get the same answers for hole-check-elimination that we
will later during ParseLazy.
In the included test case, for example:
(function() { ((s = 17, y = s) => s)(); } )();
^2 ^1
The old code would generate a hole check when trying to load
|s| for assignment to |y| (because it treated the closing parentheses
pointed to by "^1" as the "initialization position" of |s|).
The new code uses the comma pointed to by "^2" as the initialization
position of |s|. Since that occurs textually before the load of |s|,
full-codegen knows it can avoid the hole check.
BUG=v8:4908
LOG=n
Review URL: https://codereview.chromium.org/1900343002
Cr-Commit-Position: refs/heads/master@{#35678}
Port 81a1530e6f
Original commit message:
Before frame elision, we finalized the frame shape when assembling the
prologue, which is also when we prepared the frame (saving sp, etc).
The frame finalization only needs to happen once, and happens to be
actually a set of idempotent operations. With frame elision, the logic for
frame finalization was happening every time we constructed the frame.
Albeit idempotent operations, the code would become hard to maintain.
This change separates frame shape finalization from frame
construction. When constructing the CodeGenerator, we finalize the
frame. Subsequent access is to a const Frame*.
Also renamed AssemblePrologue to AssembleConstructFrame, as
suggested in the frame elision CR.
Separating frame setup gave the opportunity to do away with
architecture-independent frame aligning (which is something just arm64
cares about), and also with stack pointer setup (also arm64). Both of
these happen now at frame finalization on arm64.
R=mtrofin@chromium.org, joransiu@ca.ibm.com, bjaideep@ca.ibm.com, michael_dawson@ca.ibm.com, mbrandy@us.ibm.com
BUG=
LOG=N
Review URL: https://codereview.chromium.org/1903403002
Cr-Commit-Position: refs/heads/master@{#35677}
New incoming test262 tests check what happens on detached ("neutered")
ArrayBuffers. This patch makes the test262 infrastructure define
detaching an ArrayBuffer in terms of %ArrayBufferNeuter, passing the
--allow-natives-syntax flag, when it is needed.
BUG=v8:4193
LOG=N
R=adamk,machenbach
Review URL: https://codereview.chromium.org/1897203003
Cr-Commit-Position: refs/heads/master@{#35676}
Port 81a1530e6f
Original commit message:
Before frame elision, we finalized the frame shape when assembling the
prologue, which is also when we prepared the frame (saving sp, etc).
The frame finalization only needs to happen once, and happens to be
actually a set of idempotent operations. With frame elision, the logic for
frame finalization was happening every time we constructed the frame.
Albeit idempotent operations, the code would become hard to maintain.
This change separates frame shape finalization from frame
construction. When constructing the CodeGenerator, we finalize the
frame. Subsequent access is to a const Frame*.
Also renamed AssemblePrologue to AssembleConstructFrame, as
suggested in the frame elision CR.
Separating frame setup gave the opportunity to do away with
architecture-independent frame aligning (which is something just arm64
cares about), and also with stack pointer setup (also arm64). Both of
these happen now at frame finalization on arm64.
R=mtrofin@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, michael_dawson@ca.ibm.com, mbrandy@us.ibm.com
BUG=
LOG=N
Review URL: https://codereview.chromium.org/1903343002
Cr-Commit-Position: refs/heads/master@{#35674}
Port 59546149c6
Original commit message:
Now that all 'const' declarations are of the ES2015 variety, the only
use of CONST_LEGACY is for function name bindings in sloppy mode
named function expressions.
This patch aims to delete all code meant to handle other cases, which
mostly had to do with hole initialization/hole checks. Since function
name bindings are initialized at entry to a function, it's impossible
to ever observe one in an uninitialized state.
To simplify the patch further, it removes the `IMPORT` VariableMode,
as it's not likely to be needed (IMPORT is identical to CONST for
the purpose of VariableMode).
R=adamk@chromium.org, joransiu@ca.ibm.com, bjaideep@ca.ibm.com, michael_dawson@ca.ibm.com, mbrandy@us.ibm.com
BUG=
LOG=N
Review URL: https://codereview.chromium.org/1901423004
Cr-Commit-Position: refs/heads/master@{#35673}
This causes an incoming test262 test to pass, as part of the next
test262 roll.
R=adamk,machenbach
BUG=v8:1569
LOG=N
Review URL: https://codereview.chromium.org/1896293003
Cr-Commit-Position: refs/heads/master@{#35667}
Before, just a string was thrown, so no stack trace was attached there.
Generated code from wasm does not grow by this change, we just pass a
message id to the respective (new) runtime function.
R=mstarzinger@chromium.org, titzer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1874383002
Cr-Commit-Position: refs/heads/master@{#35664}
clang assumes 16-byte stack alignment, but incoming stack alignment isn't
always guaranteed to be that way. It looks like v8 was lucky to not hit
this so far.
See https://crbug.com/418554 -- this makes v8's standalone config match
Chromium. See also https://llvm.org/bugs/show_bug.cgi?id=21414
Maybe it's possible to change the caller of OnEntryHook() to guarantee
the right alignment, but matching Chromium's build flags here seems like
a good idea in general.
BUG=v8:4928
LOG=n
Committed: https://crrev.com/3afb3324941625559635380ef98a2ee73e370a0a
Cr-Commit-Position: refs/heads/master@{#35597}
Review URL: https://codereview.chromium.org/1899783002
Cr-Commit-Position: refs/heads/master@{#35662}
This is just a pure renaming because "baseline" will be the code name
for our upcoming middle tier within the compilation pipeline. It makes
sure the name "baseline" remains unused.
R=rmcilroy@chromium.org
Review URL: https://codereview.chromium.org/1904463002
Cr-Commit-Position: refs/heads/master@{#35661}
Short external strings do not cache the resource data, and may be used
for compressible strings. The assumptions about their lengths is
invalid and may lead to oob reads.
R=jkummerow@chromium.org
BUG=v8:4923,chromium:604897
LOG=N
Review URL: https://codereview.chromium.org/1901573003
Cr-Commit-Position: refs/heads/master@{#35660}
Port 978ad03b92
Original commit message:
Fix and re-enable the flexible representation for Math.floor (which is used to
implement Math.ceil) and Math.round, which allows Math.floor and Math.round to
return double results instead of int32, and therefore allows values outside
the int32 range, especially -0 is now a valid result, which doesn't deopt.
Also port this feature to x64 and ia32 when the CPU supports the SSE4.1
extension.
This addresses all the known deoptimization loops related to Math.round
in the Kraken benchmark suite, and seems to also address most of the
deoptimization loops related to Math.floor in the Oort Online benchmark.
Drive-by-fix: Import the regression tests for the broken HMathFloorOfDiv
optimization that caused the initial revert of the feature (for arm64 only
back then).
R=bmeurer@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, michael_dawson@ca.ibm.com
BUG=chromium:476477,v8:2890,v8:4059
LOG=n
Review URL: https://codereview.chromium.org/1839643007
Cr-Commit-Position: refs/heads/master@{#35659}
In the long run we do not want to rely on compilation results being
available on the CompilationInfo. This removes the accessor for the
abstract code, which is very inviting to be used outside of compilation
pipeline.
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1903723002
Cr-Commit-Position: refs/heads/master@{#35653}
Due to cross-compilation from simulator-builds into non-simulator with snapshot it's complicated to conditionally include an extra field. Given that the memory overhead should be minimal, just always include a separate field instead.
BUG=
Review URL: https://codereview.chromium.org/1901423002
Cr-Commit-Position: refs/heads/master@{#35652}
Also factor out test cases from test-run-machops.cc into test-run-load-store.cc
BUG=chromium:599717
LOG=Y
Review URL: https://codereview.chromium.org/1858323003
Cr-Commit-Position: refs/heads/master@{#35651}
In a full mark-compact GC, instead of copying memory to old space for
pages that have more than X% live bytes, we just move the whole page over to old
space.
X=70 (default value)
BUG=chromium:581412
LOG=N
TBR=hpayer@chromium.org
This reverts commit 8b3337278f.
Review URL: https://codereview.chromium.org/1901093002
Cr-Commit-Position: refs/heads/master@{#35649}
port 81a1530e6f (r35642)
original commit message:
Before frame elision, we finalized the frame shape when assembling the
prologue, which is also when we prepared the frame (saving sp, etc).
The frame finalization only needs to happen once, and happens to be
actually a set of idempotent operations. With frame elision, the logic for
frame finalization was happening every time we constructed the frame.
Albeit idempotent operations, the code would become hard to maintain.
This change separates frame shape finalization from frame
construction. When constructing the CodeGenerator, we finalize the
frame. Subsequent access is to a const Frame*.
Also renamed AssemblePrologue to AssembleConstructFrame, as
suggested in the frame elision CR.
Separating frame setup gave the opportunity to do away with
architecture-independent frame aligning (which is something just arm64
cares about), and also with stack pointer setup (also arm64). Both of
these happen now at frame finalization on arm64.
additional message:
This CL also removed the temporary workaround for CL #35139 (53d51c52f3)
BUG=
Review URL: https://codereview.chromium.org/1900213004
Cr-Commit-Position: refs/heads/master@{#35648}
This moves the installation of the scope info object on the shared
function info into a separate helper to share common code. This is
preparatory work in order to reuse existing scope info objects.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1894023004
Cr-Commit-Position: refs/heads/master@{#35647}
This removes some obsolete conditions checking whether we are performing
concurrent OSR compilation. This feature has been removed some time ago.
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1901163003
Cr-Commit-Position: refs/heads/master@{#35646}
Previously we had to run SimplifiedLowering, ChangeLowering and
JSGenericLowering independently and exactly in this order to
achieve great performance for the common case of JSToNumber (i.e.
input is already a Smi or a HeapNumber). This phase ordering
problem already causes trouble with not being able to run the
generic lowering phase earlier, but also blocks proper plain
primitive ToNumber optimizations. So this properly integrates
JSToNumber into the truncation analysis and optimizes according
to the truncation (either Word32 or Float64).
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1901803002
Cr-Commit-Position: refs/heads/master@{#35643}
Before frame elision, we finalized the frame shape when assembling the
prologue, which is also when we prepared the frame (saving sp, etc).
The frame finalization only needs to happen once, and happens to be
actually a set of idempotent operations. With frame elision, the logic for
frame finalization was happening every time we constructed the frame.
Albeit idempotent operations, the code would become hard to maintain.
This change separates frame shape finalization from frame
construction. When constructing the CodeGenerator, we finalize the
frame. Subsequent access is to a const Frame*.
Also renamed AssemblePrologue to AssembleConstructFrame, as
suggested in the frame elision CR.
Separating frame setup gave the opportunity to do away with
architecture-independent frame aligning (which is something just arm64
cares about), and also with stack pointer setup (also arm64). Both of
these happen now at frame finalization on arm64.
BUG=
Review URL: https://codereview.chromium.org/1843143002
Cr-Commit-Position: refs/heads/master@{#35642}
port 59546149c6 (r35632)
original commit message:
Now that all 'const' declarations are of the ES2015 variety, the only
use of CONST_LEGACY is for function name bindings in sloppy mode
named function expressions.
This patch aims to delete all code meant to handle other cases, which
mostly had to do with hole initialization/hole checks. Since function
name bindings are initialized at entry to a function, it's impossible
to ever observe one in an uninitialized state.
To simplify the patch further, it removes the `IMPORT` VariableMode,
as it's not likely to be needed (IMPORT is identical to CONST for
the purpose of VariableMode).
BUG=
Review URL: https://codereview.chromium.org/1902943003
Cr-Commit-Position: refs/heads/master@{#35641}
