To enable executing code in a context of a particular time or date (e.g. when
codepath depends on whether it's say evening or New Year) there is a need for
a way to provide it bypassing actual system time.
Bug: chromium:751993
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: Iee35d97b74345f63fff814a65a6f134d7c970341
Reviewed-on: https://chromium-review.googlesource.com/598666
Commit-Queue: Sergei Datsenko <dats@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47700}
The predicate in question used to report true on both, Crankshaft and
TurboFan code. It has hence become obsolete and can be replaced by the
existing {Code::is_turbofanned} predicate. This also frees up a bit in
the second kind-specific bit field.
R=jarin@chromium.org
BUG=v8:6408
Change-Id: I204d7dd78a639c752c9749fd305c7006c6b6aca3
Reviewed-on: https://chromium-review.googlesource.com/599868
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47125}
The unwinding information we emit wrongly encodes code locations as relative
offsets. If we look at the .eh_frame section of shared object generated by "perf
inject" using "objdump -g":
~~~
00000000 0000000000000018 00000000 CIE
(snip)
0000001c 0000000000000028 00000020 FDE cie=00000000 pc=fffffffffffffee8..00000000000017f8
(snip)
00000048 ZERO terminator
~~~
We can see the range that the FDE entry covers is incorrect, it should point to
where the .text section is, at address 0x40 on a 64-bit architecture.
The reason for this was that the PerfJitLogger logs a code size that is
different from the one we've used when encoding the unwinding information. The
logger will ignore the safepoint table while the unwinding info assumes it is
part of the code.
BUG=
Review-Url: https://codereview.chromium.org/2790403002
Cr-Commit-Position: refs/heads/master@{#44378}
The "perf inject" command will place the generated function into a .text
section, placed directly after the ELF header. As a result, source position
addresses need to be adjusted according to the size of the ELF header, which is
0x40 for 64 bit architectures and 0x34 on 32 bit architectures.
We would previously adjust the addresses with 0x40 regardless of the
architecture.
BUG=
Review-Url: https://codereview.chromium.org/2783203005
Cr-Commit-Position: refs/heads/master@{#44325}
Logging for --perf-prof is not GC safe. Now, we are going to
emit source position info for optimized code when we are
profiling, logging, or debugging, and under the same condition,
pre-compute the line ends array for line number computation.
R=tebbi@chromium.org
BUG=v8:5730
Review-Url: https://codereview.chromium.org/2562973002
Cr-Commit-Position: refs/heads/master@{#41619}
SourcePosition::InliningId() refers to a the new table DeoptimizationInputData::InliningPositions(), which provides the following data for every inlining id:
- The inlined SharedFunctionInfo as an offset into DeoptimizationInfo::LiteralArray
- The SourcePosition of the inlining. Recursively, this yields the full inlining stack.
Before the Code object is created, the same information can be found in CompilationInfo::inlined_functions().
If SourcePosition::InliningId() is SourcePosition::kNotInlined, it refers to the outer (non-inlined) function.
So every SourcePosition has full information about its inlining stack, as long as the corresponding Code object is known. The internal represenation of a source position is a positive 64bit integer.
All compilers create now appropriate source positions for inlined functions. In the case of Turbofan, this required using AstGraphBuilderWithPositions for inlined functions too. So this class is now moved to a header file.
At the moment, the additional information in source positions is only used in --trace-deopt and --code-comments. The profiler needs to be updated, at the moment it gets the correct script offsets from the deopt info, but the wrong script id from the reconstructed deopt stack, which can lead to wrong outputs. This should be resolved by making the profiler use the new inlining information for deopts.
I activated the inlined deoptimization tests in test-cpu-profiler.cc for Turbofan, changing them to a case where the deopt stack and the inlining position agree. It is currently still broken for other cases.
The following additional changes were necessary:
- The source position table (internal::SourcePositionTableBuilder etc.) supports now 64bit source positions. Encoding source positions in a single 64bit int together with the difference encoding in the source position table results in very little overhead for the inlining id, since only 12% of the source positions in Octane have a changed inlining id.
- The class HPositionInfo was effectively dead code and is now removed.
- SourcePosition has new printing and information facilities, including computing a full inlining stack.
- I had to rename compiler/source-position.{h,cc} to compiler/compiler-source-position-table.{h,cc} to avoid clashes with the new src/source-position.cc file.
- I wrote the new wrapper PodArray for ByteArray. It is a template working with any POD-type. This is used in DeoptimizationInputData::InliningPositions().
- I removed HInlinedFunctionInfo and HGraph::inlined_function_infos, because they were only used for the now obsolete Crankshaft inlining ids.
- Crankshaft managed a list of inlined functions in Lithium: LChunk::inlined_functions. This is an analog structure to CompilationInfo::inlined_functions. So I removed LChunk::inlined_functions and made Crankshaft use CompilationInfo::inlined_functions instead, because this was necessary to register the offsets into the literal array in a uniform way. This is a safe change because LChunk::inlined_functions has no other uses and the functions in CompilationInfo::inlined_functions have a strictly longer lifespan, being created earlier (in Hydrogen already).
BUG=v8:5432
Review-Url: https://codereview.chromium.org/2451853002
Cr-Commit-Position: refs/heads/master@{#40975}
Original commit message:
Also, CodeGenerator::MakeCodeEpilogue now accepts an optional pointer
to a EhFrameWriter and will attach unwinding information to the code
object when passed one.
Reason for reverting:
The STATIC_CONST_MEMBER_DEFINITION in eh-frame-writer-unittest.cc
causes a compiler error on V8 Win64 - clang buildbot.
Removing that bit.
BUG=v8:4899
LOG=N
Review-Url: https://codereview.chromium.org/2023503002
Cr-Commit-Position: refs/heads/master@{#37754}
Original commit message:
Also, CodeGenerator::MakeCodeEpilogue now accepts an optional pointer
to a EhFrameWriter and will attach unwinding information to the code
object when passed one.
Reason for reverting:
The STATIC_CONST_MEMBER_DEFINITION in eh-frame-writer-unittest.cc
causes a compiler error on V8 Win64 - clang buildbot.
Removing that bit.
BUG=v8:4899
LOG=N
Review-Url: https://codereview.chromium.org/2023503002
Cr-Commit-Position: refs/heads/master@{#37707}
Reason for revert:
The STATIC_CONST_MEMBER_DEFINITION in eh-frame-writer-unittest.cc causes a compiler error on V8 Win64 - clang buildbot. Removing that bit should be sufficient.
Original issue's description:
> Implement .eh_frame writer and disassembler.
>
> Also, CodeGenerator::MakeCodeEpilogue now accepts an optional pointer
> to a EhFrameWriter and will attach unwinding information to the code
> object when passed one.
>
> BUG=v8:4899
> LOG=N
>
> Committed: https://crrev.com/27d810e63b744b5b3d9aa28ff21413247773e6c2
> Cr-Commit-Position: refs/heads/master@{#37683}
TBR=rmcilroy@chromium.org,jarin@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4899
Review-Url: https://codereview.chromium.org/2143033002
Cr-Commit-Position: refs/heads/master@{#37688}
Also, CodeGenerator::MakeCodeEpilogue now accepts an optional pointer
to a EhFrameWriter and will attach unwinding information to the code
object when passed one.
BUG=v8:4899
LOG=N
Review-Url: https://codereview.chromium.org/2023503002
Cr-Commit-Position: refs/heads/master@{#37683}
the .eh_frame format as part of the jitdump generated when
FLAG_perf_prof is enabled. The final goal is allowing precise unwinding
of callchains that include JITted code when profiling V8 using perf.
Unwinding information is stored in the body of code objects after the
code itself, prefixed with its length and aligned to a 8-byte boundary.
A boolean flag in the header signals its presence, resulting in zero
memory overhead when the generation of unwinding info is disabled or
no such information was attached to the code object.
A new jitdump record type (with id 4) is introduced for specifying
optional unwinding information for code load records. The EhFrameHdr
struct is also introduced, together with a constructor to initialise it
from the associated code object.
At this stage no unwinding information is written to the jitdump, but
the infrastructure for doing so is ready in place.
BUG=v8:4899
LOG=N
Review-Url: https://codereview.chromium.org/1993653003
Cr-Commit-Position: refs/heads/master@{#37296}
This CL introduces static variables and global lock for writing to the
dump files, so that multiple web workers do not run into trouble.
Review URL: https://codereview.chromium.org/1839133002
Cr-Commit-Position: refs/heads/master@{#35123}
This fixes support for debug info in perf. Thanks to Stephane Eranian for
identifying the problem - debug info event has to be emitted before the
code load event. It also seems that perf does not yet support the shorthand
for repeated source files in the debug info entry ("\xff\0"), so I changed
it to always write the script name.
Review URL: https://codereview.chromium.org/1843563002
Cr-Commit-Position: refs/heads/master@{#35092}
Difference from --perf-basic-prof:
- correctly attributes samples when code space gets reused (when unused code object dies and a new code objects is allocated at the same place).
- outputs compiled machine code for instruction-level profile.
Just like --perf-basic-prof, the file writer is not synchronized (even worse, there is a per-isolate file handle), so we will run into trouble with multiple isolates. However, this patch is still an improvement on --perf-basic-prof, and it should be fine to replace ll-prof.
The patch also introduces experimental support for debug info, but it does not seem to be picked by the perf tool.
Usage:
You need the perf tool from Linux kernel >4.5. Then run:
$ perf record -k mono d8 --perf-prof <your JS file>
$ perf inject -j -i perf.data -o perf.data.jitted
$ perf report -i perf.data.jitted
Some explanations:
The "-k mono" switch from "perf record" tells the perf tool to use the monotonic clock for perf sample timestamping. The "perf inject -j" command injects the collected code events into the perf data file, writing the output into perf.data.jitted. The perf report command then creates the report.
Review URL: https://codereview.chromium.org/1809203007
Cr-Commit-Position: refs/heads/master@{#35091}
This adds timestamps to allow profiling with code space reuse. Also a couple of updates to reflect the changes in the JIT perf interface + a move of the perf-related stuff into separate files.
Unfortunately, the change only works with the latest patch to the perf tool from a Linux perf tool contributor (Stephane Eranian).
BUG=
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/255803003
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@22098 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
