Instead of having feedback vector as a subtype of FixedArray with
reserved slots, make it a first-class variable-sized object with a
fixed-size header. This allows us to compress counters to ints in the
header, rather than forcing them to be Smis.
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: Icc5f088ffbc2e2651b845bc71ea42060639e3e48
Reviewed-on: https://chromium-review.googlesource.com/585129
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46935}
Reland of https://chromium-review.googlesource.com/c/544888/.
Instead of counting profiler ticks on the shared function info (which is
shared between native contexts), count them on the feedback vector
(which is not). This allows us to continue pushing optimization
decisions off the SFI, onto the feedback vector.
Note that a side-effect of this is that ICs don't have to walk the stack
to reset profiler ticks, as they can access the feedback vector directly
from their feedback nexus.
Change-Id: I7aa6baed03f726843d1b62629c72b74f05114b48
Reviewed-on: https://chromium-review.googlesource.com/579051
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46868}
This reverts commit a2fcdc7cc8.
Reason for revert: Large regressions in RCS (https://chromeperf.appspot.com/group_report?bug_id=740126)
Original change's description:
> [runtime] Move profiler ticks from SFI to feedback vector
>
> Instead of counting profiler ticks on the shared function info (which is
> shared between native contexts), count them on the feedback vector
> (which is not). This allows us to continue pushing optimization
> decisions off the SFI, onto the feedback vector.
>
> Note that a side-effect of this is that ICs don't have to walk the stack
> to reset profiler ticks, as they can access the feedback vector directly
> from their feedback nexus.
>
> Change-Id: I232ae9e759fca75cd89d393148a4ff42caa2646f
> Reviewed-on: https://chromium-review.googlesource.com/544888
> Reviewed-by: Igor Sheludko <ishell@chromium.org>
> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
> Commit-Queue: Leszek Swirski <leszeks@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#46411}
TBR=rmcilroy@chromium.org,leszeks@chromium.org,ishell@chromium.org
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: Id587e4172e300c420f93c49744a2a0e66696edf8
Reviewed-on: https://chromium-review.googlesource.com/574227
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46702}
Goal of this CL: explicit return from non-async function has position after
return expression as return position (will unblock [1]).
BytecodeArrayBuilder has SetStatementPosition and SetExpressionPosition methods.
If one of these methods is called then next generated bytecode will get passed
position. It's general treatment for most cases.
Unfortunately it doesn't work for Returns:
- debugger requires source positions exactly on kReturn bytecode in stepping
implementation,
- BytecodeGenerator::BuildReturn and BytecodeGenerator::BuildAsyncReturn
generates more then one bytecode and general solution will put return position
on first generated bytecode,
- it's not easy to split BuildReturn function into two parts to allow something
like following in BytecodeGenerator::VisitReturnStatement since generated
bytecodes are actually controlled by execution_control().
..->BuildReturnPrologue();
..->SetReturnPosition(stmt);
..->Return();
In this CL we pass ReturnStatement through ExecutionControl and use it for
position when we emit return bytecode right here.
So this CL only will improve return position for returns inside of non-async
functions, I'll address async functions later.
[1] https://chromium-review.googlesource.com/c/543161/
Change-Id: Iede512c120b00c209990bf50c20e7d23dc0d65db
Reviewed-on: https://chromium-review.googlesource.com/560738
Commit-Queue: Aleksey Kozyatinskiy <kozyatinskiy@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46687}
Instead of counting profiler ticks on the shared function info (which is
shared between native contexts), count them on the feedback vector
(which is not). This allows us to continue pushing optimization
decisions off the SFI, onto the feedback vector.
Note that a side-effect of this is that ICs don't have to walk the stack
to reset profiler ticks, as they can access the feedback vector directly
from their feedback nexus.
Change-Id: I232ae9e759fca75cd89d393148a4ff42caa2646f
Reviewed-on: https://chromium-review.googlesource.com/544888
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46411}
Since the feedback vector is itself a native context structure, why
not store optimized code for a function in there rather than in
a map from native context to code? This allows us to get rid of
the optimized code map in the SharedFunctionInfo, saving a pointer,
and making lookup of any optimized code quicker.
Original patch by Michael Stanton <mvstanton@chromium.org>
BUG=v8:6246,chromium:718891
TBR=yangguo@chromium.org,ulan@chromium.org
Change-Id: I3bb9ec0cfff32e667cca0e1403f964f33a6958a6
Reviewed-on: https://chromium-review.googlesource.com/500134
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45234}
This reverts commit 662aa425ba.
Reason for revert: Crashing on Canary
BUG=chromium:718891
Original change's description:
> Reland: [TypeFeedbackVector] Store optimized code in the vector
>
> Since the feedback vector is itself a native context structure, why
> not store optimized code for a function in there rather than in
> a map from native context to code? This allows us to get rid of
> the optimized code map in the SharedFunctionInfo, saving a pointer,
> and making lookup of any optimized code quicker.
>
> Original patch by Michael Stanton <mvstanton@chromium.org>
>
> BUG=v8:6246
> TBR=yangguo@chromium.org,ulan@chromium.org
>
> Change-Id: Ic83e4011148164ef080c63215a0c77f1dfb7f327
> Reviewed-on: https://chromium-review.googlesource.com/494487
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#45084}
TBR=ulan@chromium.org,rmcilroy@chromium.org,yangguo@chromium.org,jarin@chromium.org
# Not skipping CQ checks because original CL landed > 1 day ago.
BUG=v8:6246
Change-Id: Idab648d6fe260862c2a0e35366df19dcecf13a82
Reviewed-on: https://chromium-review.googlesource.com/498633
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45174}
Since the feedback vector is itself a native context structure, why
not store optimized code for a function in there rather than in
a map from native context to code? This allows us to get rid of
the optimized code map in the SharedFunctionInfo, saving a pointer,
and making lookup of any optimized code quicker.
Original patch by Michael Stanton <mvstanton@chromium.org>
BUG=v8:6246
TBR=yangguo@chromium.org,ulan@chromium.org
Change-Id: Ic83e4011148164ef080c63215a0c77f1dfb7f327
Reviewed-on: https://chromium-review.googlesource.com/494487
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45084}
This reverts commit c5ad9c6d8e.
Reason for revert: Fails on gc stress:
https://build.chromium.org/p/client.v8/builders/V8%20Linux64%20GC%20Stress%20-%20custom%20snapshot/builds/12661
Original change's description:
> [TypeFeedbackVector] Store optimized code in the vector
>
> Since the feedback vector is itself a native context structure, why
> not store optimized code for a function in there rather than in
> a map from native context to code? This allows us to get rid of
> the optimized code map in the SharedFunctionInfo, saving a pointer,
> and making lookup of any optimized code quicker.
>
> Original patch by Michael Stanton <mvstanton@chromium.org>
>
> BUG=v8:6246
>
> Change-Id: I60ff8c408c3001bc272b4b198c9cbaea2872a9e5
> Reviewed-on: https://chromium-review.googlesource.com/476891
> Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#45022}
TBR=ulan@chromium.org,rmcilroy@chromium.org,yangguo@chromium.org,mvstanton@chromium.org,jarin@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:6246
Change-Id: I9cd5735b03898cae6ae7adea0f19d32fceb31619
Reviewed-on: https://chromium-review.googlesource.com/493287
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45027}
Since the feedback vector is itself a native context structure, why
not store optimized code for a function in there rather than in
a map from native context to code? This allows us to get rid of
the optimized code map in the SharedFunctionInfo, saving a pointer,
and making lookup of any optimized code quicker.
Original patch by Michael Stanton <mvstanton@chromium.org>
BUG=v8:6246
Change-Id: I60ff8c408c3001bc272b4b198c9cbaea2872a9e5
Reviewed-on: https://chromium-review.googlesource.com/476891
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45022}
Add a notion of "invocation count" to the baseline compilers, which
increment a special slot in the TypeFeedbackVector for each invocation
of a given function (the optimized code doesn't currently collect this
information).
Use this invocation count to relativize the call counts on the call
sites within the function, so that the inlining heuristic has a view
of relative importance of a call site rather than some absolute numbers
with unclear meaning for the current function. Also apply the call site
frequency as a factor to all frequencies in the inlinee by passing this
to the graph builders so that the importance of a call site in an
inlinee is relative to the topmost optimized function.
Note that all functions that neither have literals nor need type
feedback slots will share a single invocation count cell in the
canonical empty type feedback vector, so their invocation count is
meaningless, but that doesn't matter since we only use the invocation
count to relativize call counts within the function, which we only have
if we have at least one type feedback vector (the CallIC slot).
See the design document for additional details on this change:
https://docs.google.com/document/d/1VoYBhpDhJC4VlqMXCKvae-8IGuheBGxy32EOgC2LnT8
BUG=v8:5267,v8:5372
R=mvstanton@chromium.org,rmcilroy@chromium.org,mstarzinger@chromium.org
Review-Url: https://codereview.chromium.org/2337123003
Cr-Commit-Position: refs/heads/master@{#39410}
For historical reasons, the interpreter's bytecode expectations tests
required a type for the constant pool. This had two disadvantages:
1. Strings and numbers were not visible in mixed pools, and
2. Mismatches of pool types (e.g. when rebaselining) would cause parser
errors
This removes the pool types, making everything 'mixed', but appending
the values to string and number valued constants. Specifying a pool type
in the *.golden header now prints a warning (for backwards compatibility).
BUG=v8:5350
Review-Url: https://codereview.chromium.org/2310103002
Cr-Commit-Position: refs/heads/master@{#39216}
Adds TestResultScope and uses it to directly jump/fall through to the
correct branch in expressions used as branch conditions.
Should enable nicer TurboFan-graphs for easier control-flow
transformations in the future.
BUG=v8:4280
LOG=n
Review-Url: https://codereview.chromium.org/2242463002
Cr-Commit-Position: refs/heads/master@{#38634}
Prints source position information alongside bytecode.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1963663002
Cr-Commit-Position: refs/heads/master@{#36171}
This change introduces wide prefix bytecodes to support wide (16-bit)
and extra-wide (32-bit) operands. It retires the previous
wide-bytecodes and reduces the number of operand types.
Operands are now either scalable or fixed size. Scalable operands
increase in width when a bytecode is prefixed with wide or extra-wide.
The bytecode handler table is extended to 256*3 entries. The
first 256 entries are used for bytecodes with 8-bit operands,
the second 256 entries are used for bytecodes with operands that
scale to 16-bits, and the third group of 256 entries are used for
bytecodes with operands that scale to 32-bits.
LOG=N
BUG=v8:4747,v8:4280
Review URL: https://codereview.chromium.org/1783483002
Cr-Commit-Position: refs/heads/master@{#34955}
Bytecode expectations have been moved to external (.golden) files,
one per test. Each test in the suite builds a representation of the
the compiled bytecode using BytecodeExpectationsPrinter. The output is
then compared to the golden file. If the comparision fails, a textual
diff can be used to identify the discrepancies.
Only the test snippets are left in the cc file, which also allows to
make it more compact and meaningful. Leaving the snippets in the cc
file was a deliberate choice to allow keeping the "truth" about the
tests in the cc file, which will rarely change, as opposed to golden
files.
Golden files can be generated and kept up to date using
generate-bytecode-expectations, which also means that the test suite
can be batch updated whenever the bytecode or golden format changes.
The golden format has been slightly amended (no more comments about
`void*`, add size of the bytecode array) following the consideration
made while converting the tests.
There is also a fix: BytecodeExpectationsPrinter::top_level_ was left
uninitialized, leading to undefined behaviour.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1717293002
Cr-Commit-Position: refs/heads/master@{#34285}
