The new frame type is inteneded to represent native C++ stack frames.
JS code may sometimes make calls to helper native functions that do not
provide any special stack layout besides the return address and frame pointer.
Currently the stack iterator bails out when it sees an unknown frame.
The patch allows the iterator to unwind stacks having such frames.
BUG=chromium:768540
Change-Id: I9c273c7015695a6733c0a0c52b522fca7b25de0d
Reviewed-on: https://chromium-review.googlesource.com/794991
Commit-Queue: Alexei Filippov <alph@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50058}
These enums capture the "type" of a load and store in one value, with
accessors to get the value type, memory type, memory size etc.
This will simplify implementing memory operations in Liftoff.
R=titzer@chromium.org
Bug: v8:6600
Change-Id: Ief673e90fe63292853f97e5e95b5c39da1296730
Reviewed-on: https://chromium-review.googlesource.com/822476
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Ben Titzer <titzer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50054}
Turbofan does not always maintain source positions. Since the profiler
is best-effort only, this is ok and we should not assert it.
Source positions for stack traces are computed with an independent
mechanism: the stack walker. It's impossible to use the stack walker
here since it assumes that the frames actually exist.
Bug: chromium:787959
Change-Id: I43c3c4e6de0a9b5f76b05c53af238cb1df6ea1c5
Reviewed-on: https://chromium-review.googlesource.com/822930
Reviewed-by: Alexei Filippov <alph@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50053}
* DisableInlineAllocationSteps was a blunt hammer added to work around
tests that needed to avoid artificially lower limits imposed by
observers. PauseAllocationObserversScope can properly disable step on
a temporary basis.
* Modify tests. Remove DisableInlineAllocationSteps.
This exposed a bug in allocation observers: we were not doing a step
when a fresh page is added.
Fix this by moving the step into UpdateAllocationInfo. We should be
doing a step (and keeping top_on_previous_step_ consistent) whenever
we move move top(). UpdateAllocationInfo is the correct place for this
rather than the callers of UpdateAllocationInfo.
Bug:
Change-Id: I2edc238dc2e73bf9a2e9738c2a9b50efcac5cbf0
Reviewed-on: https://chromium-review.googlesource.com/821052
Commit-Queue: Ali Ijaz Sheikh <ofrobots@google.com>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50052}
If an initializer is a NaryOperation, its position ends up as a start position
of a Scope, and a DCHECK used to fire.
Interestingly, this was not caught by our existing tests.
BUG=chromium:791256
Change-Id: Id47f850c7ad17ca580352f9bd56c9567b485c3b8
Reviewed-on: https://chromium-review.googlesource.com/822093
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50051}
This is a reland of fa18e78dc7.
Mips compile error is fixed.
Original change's description:
> [wasm] [cleanup] Only pass information really needed
>
> Instead of always passing the MachineType, we can often just pass the
> accessed memory size or the MachineRepresentation, which is less
> information to pass and will simplify the upcoming refactoring for
> memory operations in Liftoff.
>
> R=titzer@chromium.org
>
> Bug: v8:6600
> Change-Id: I8748f8e00dcfdbc4082893143fe88bdafde99053
> Reviewed-on: https://chromium-review.googlesource.com/822194
> Reviewed-by: Ben Titzer <titzer@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#50041}
TBR=titzer@chromium.org
Bug: v8:6600
Change-Id: I3dff3072d6ceebd74873ace0c7dce7cccc3055d5
Reviewed-on: https://chromium-review.googlesource.com/822851
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50050}
This reverts commit ee106cb907.
Reason for revert: Still failing on big endian mips.
Original change's description:
> Reland "[wasm] [cleanup] Only pass information really needed"
>
> This is a reland of fa18e78dc7.
>
> Mips compile error is fixed.
>
> Original change's description:
> > [wasm] [cleanup] Only pass information really needed
> >
> > Instead of always passing the MachineType, we can often just pass the
> > accessed memory size or the MachineRepresentation, which is less
> > information to pass and will simplify the upcoming refactoring for
> > memory operations in Liftoff.
> >
> > R=titzer@chromium.org
> >
> > Bug: v8:6600
> > Change-Id: I8748f8e00dcfdbc4082893143fe88bdafde99053
> > Reviewed-on: https://chromium-review.googlesource.com/822194
> > Reviewed-by: Ben Titzer <titzer@chromium.org>
> > Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#50041}
>
> TBR=titzer@chromium.org
>
> Bug: v8:6600
> Change-Id: Ia2c47464277a7c7a70099e7f7d6e32003e34d87e
> Reviewed-on: https://chromium-review.googlesource.com/822342
> Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#50047}
TBR=titzer@chromium.org,clemensh@chromium.org
Change-Id: I448e3ad433451fb468c2db66b4b297e8c72a8c69
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:6600
Reviewed-on: https://chromium-review.googlesource.com/822696
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50048}
This is a reland of fa18e78dc7.
Mips compile error is fixed.
Original change's description:
> [wasm] [cleanup] Only pass information really needed
>
> Instead of always passing the MachineType, we can often just pass the
> accessed memory size or the MachineRepresentation, which is less
> information to pass and will simplify the upcoming refactoring for
> memory operations in Liftoff.
>
> R=titzer@chromium.org
>
> Bug: v8:6600
> Change-Id: I8748f8e00dcfdbc4082893143fe88bdafde99053
> Reviewed-on: https://chromium-review.googlesource.com/822194
> Reviewed-by: Ben Titzer <titzer@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#50041}
TBR=titzer@chromium.org
Bug: v8:6600
Change-Id: Ia2c47464277a7c7a70099e7f7d6e32003e34d87e
Reviewed-on: https://chromium-review.googlesource.com/822342
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50047}
This operation was used to implement asm.js stores, but is obsolete
with asm.js stores now being lowered to normal graph nodes.
R=mstarzinger@chromium.orgCC=jarin@chromium.org
Bug:
Change-Id: Iea90b1a62be2e273c0562058642adc5b63ae2cf8
Reviewed-on: https://chromium-review.googlesource.com/822570
Commit-Queue: Ben Titzer <titzer@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50046}
This is a reland of c3bd741efd
Original change's description:
> Fix "this" value in lazily-parsed module functions.
>
> When preparsing top-level functions in a module, we didn't track
> unresolved variables. Consequently, "this" ended up referencing
> the global "this", which has the wrong value (in a module "this"
> is supposed to be the undefined value).
>
> This patch fixes that. This also lets us stop forcing context
> allocation of all variables in module scopes, which the patch
> takes care of as well.
>
> Bug: chromium:791334
> Change-Id: Ifac1f1adc033f3facfb3d29dd4bca32ee27bffcf
> Reviewed-on: https://chromium-review.googlesource.com/808938
> Reviewed-by: Marja Hölttä <marja@chromium.org>
> Reviewed-by: Adam Klein <adamk@chromium.org>
> Reviewed-by: Aleksey Kozyatinskiy <kozyatinskiy@chromium.org>
> Commit-Queue: Georg Neis <neis@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#50025}
TBR=adamk@chromium.orgTBR=kozyatinskiy@chromium.org
Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel
Bug: chromium:791334
Change-Id: I57acc7b84a345565b36cbb55924fa2ff9b449eec
Reviewed-on: https://chromium-review.googlesource.com/822341
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50045}
The CheckedLoad and CheckedStore machine operators are a form of
macro node used in the backend of TurboFan. They express bounds-checked
accesses to typed arrays in the asm.js style. While their semantics
can be expressed as normal machine operations and branches in the
graph, they were originally used to save compile time when graphs
were very big. ASM->WASM->TF graphs are smaller.
This CL implements asm.js bounds checks for stores using normal
branch nodes and removes the need for CheckedStore, improving
maintainability at some small cost to compilation time.
R=mstarzinger@chromium.org
CC=bradnelson@chromium.org
Bug:
Change-Id: I928689cfe33acac5a422f32cb014850ba09196f1
Reviewed-on: https://chromium-review.googlesource.com/822471
Commit-Queue: Ben Titzer <titzer@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50044}
This reverts commit fa18e78dc7.
Reason for revert:
https://build.chromium.org/p/client.v8.ports/builders/V8%20Mips%20-%20builder/builds/14055
Original change's description:
> [wasm] [cleanup] Only pass information really needed
>
> Instead of always passing the MachineType, we can often just pass the
> accessed memory size or the MachineRepresentation, which is less
> information to pass and will simplify the upcoming refactoring for
> memory operations in Liftoff.
>
> R=titzer@chromium.org
>
> Bug: v8:6600
> Change-Id: I8748f8e00dcfdbc4082893143fe88bdafde99053
> Reviewed-on: https://chromium-review.googlesource.com/822194
> Reviewed-by: Ben Titzer <titzer@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#50041}
TBR=titzer@chromium.org,clemensh@chromium.org
Change-Id: I3c85e3bc9b37c97b373b5e51a094be9561767f86
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:6600
Reviewed-on: https://chromium-review.googlesource.com/822850
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50043}
Instead of always passing the MachineType, we can often just pass the
accessed memory size or the MachineRepresentation, which is less
information to pass and will simplify the upcoming refactoring for
memory operations in Liftoff.
R=titzer@chromium.org
Bug: v8:6600
Change-Id: I8748f8e00dcfdbc4082893143fe88bdafde99053
Reviewed-on: https://chromium-review.googlesource.com/822194
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50041}
Currently Page::ShrinkToHighWaterMark checks that there is only one
filler in the to-be-freed area at the end of the page. This does not
hold if an allocation observer is active.
We should instead check that the to-be-freed area does not contain
allocated objects and will not contain allocated objects:
1) Following chain of fillers we arrive at the end of the page.
2) The free list of the page is empty.
This patch also changes PagedSpace::ResetFreeList to evict free list
entries of each page, instead of just reseting the global free list.
It also removes invalidation of free list categories.
Now FreeList::EvictFreeListItems simply evicts free list entries without
invalidating free list categories.
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I6051578ec2c91c9614d14c7a6ce188d2db5ace3b
Reviewed-on: https://chromium-review.googlesource.com/822571
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50040}
Instead of specializing {DecodeLoadMem} and {DecodeStoreMem} for
prefixed opcodes, just pass the offset as parameter.
R=titzer@chromium.org
Change-Id: I0581f6e9cb562885709aaa0317bb550b175ad160
Reviewed-on: https://chromium-review.googlesource.com/822391
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50039}
Trampoline ICs differ from normal ICs only in that they
load the feedback vector from the stack instead of accepting it
as an argument. We created them for FullCodeGen, where you always
have the feedback vector on the stack, and can reduce the amount
of code required to invoke an ICs by hiding the stack load down
in the trampoline IC.
Now the concept isn't really necessary, and we can achieve a
nice cleanup of our idea-space by removing it. The first step is
to remove their usage in TurboFan.
Bug: v8:7199
Change-Id: I08aefa5de199b1f19adea261b7086a84338c0805
Reviewed-on: https://chromium-review.googlesource.com/822193
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Michael Stanton <mvstanton@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50038}
Corrected mips code generation for WASM_CALL and JS_TO_WASM_CALL.
The logic for fetching raw call sites needed changing, too, in light
of mips' 2-instruction calls, where using target_address_address is
incorrect. The CL adds platform-specific accessors.
Bug: chromium:793292 chromium:793282
Change-Id: I879ea6bffdad60791d88a6f5ea15087cdcd3f2e9
Reviewed-on: https://chromium-review.googlesource.com/818460
Reviewed-by: Ben Titzer <titzer@chromium.org>
Reviewed-by: Ivica Bogosavljevic <ivica.bogosavljevic@mips.com>
Commit-Queue: Mircea Trofin <mtrofin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50037}
This reverts commit c3bd741efd.
Reason for revert: Breaks layout tests:
https://build.chromium.org/p/client.v8.fyi/builders/V8-Blink%20Linux%2064/builds/20384
Original change's description:
> Fix "this" value in lazily-parsed module functions.
>
> When preparsing top-level functions in a module, we didn't track
> unresolved variables. Consequently, "this" ended up referencing
> the global "this", which has the wrong value (in a module "this"
> is supposed to be the undefined value).
>
> This patch fixes that. This also lets us stop forcing context
> allocation of all variables in module scopes, which the patch
> takes care of as well.
>
> Bug: chromium:791334
> Change-Id: Ifac1f1adc033f3facfb3d29dd4bca32ee27bffcf
> Reviewed-on: https://chromium-review.googlesource.com/808938
> Reviewed-by: Marja Hölttä <marja@chromium.org>
> Reviewed-by: Adam Klein <adamk@chromium.org>
> Reviewed-by: Aleksey Kozyatinskiy <kozyatinskiy@chromium.org>
> Commit-Queue: Georg Neis <neis@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#50025}
TBR=adamk@chromium.org,marja@chromium.org,neis@chromium.org,kozyatinskiy@chromium.org
Change-Id: I81f69334ed2ce104c00e6205d50001e4bdf07d15
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:791334
Reviewed-on: https://chromium-review.googlesource.com/822258
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50036}
This makes the code dealing with type feedback more concise and uniform
(at the cost of a few redundant comparisons).
Bug:
Change-Id: If6b98bd1f0dddd392d7b00d65b600127bd30ff7e
Reviewed-on: https://chromium-review.googlesource.com/818984
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50035}
The original CL introduced a test which uses a random number generator.
I disable the test for now, which is okay because this CL adds to a
work-in-progress feature anyways, and I will fix the problem in another
CL.
Original description:
Add the ability to return (multiple) return values on the stack:
- Extend stack frames with a new buffer region for return slots.
This region is located at the end of a caller's frame such that
its slots can be indexed as caller frame slots in a callee
(located beyond its parameters) and assigned return values.
- Adjust stack frame constructon and deconstruction accordingly.
- Extend linkage computation to support register plus stack returns.
- Reserve return slots in caller frame when respective calls occur.
- Introduce and generate architecture instructions ('peek') for
reading back results from return slots in the caller.
- Aggressive tests.
- Some minor clean-up.
So far, only ia32 and x64 are implemented.
Change-Id: I8b03fc4e53946daaa0e14a34603f4824a04fad7e
Reviewed-on: https://chromium-review.googlesource.com/819557
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50031}
I tested it with 30000 tests in the wasm spec test suite, and they all
pass.
Change-Id: Ice0fb64f56685f67521431213e3fc8fbc263750e
Reviewed-on: https://chromium-review.googlesource.com/822111
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50026}
When preparsing top-level functions in a module, we didn't track
unresolved variables. Consequently, "this" ended up referencing
the global "this", which has the wrong value (in a module "this"
is supposed to be the undefined value).
This patch fixes that. This also lets us stop forcing context
allocation of all variables in module scopes, which the patch
takes care of as well.
Bug: chromium:791334
Change-Id: Ifac1f1adc033f3facfb3d29dd4bca32ee27bffcf
Reviewed-on: https://chromium-review.googlesource.com/808938
Reviewed-by: Marja Hölttä <marja@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Reviewed-by: Aleksey Kozyatinskiy <kozyatinskiy@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50025}
Use large zone segments for CSA builin generation when serializer is
active. Turbofan backend uses pointer comparison of zone allocated
objects. Large zone segments ensure determistic order of objects.
Bug: v8:7188
Change-Id: I18e1e18fa79cded561563de1329bc3d9a8c364fb
Reviewed-on: https://chromium-review.googlesource.com/817601
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50023}
Some refactor moved from
https://chromium-review.googlesource.com/c/v8/v8/+/798331.
Bug: v8:6917
Change-Id: I8cae6cfca7a0d7d8e234052c0ab0bfe252355e60
Reviewed-on: https://chromium-review.googlesource.com/819550
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Sergiy Byelozyorov <sergiyb@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50020}
If we have several scripts with the same url (see many <script> tags in
one page), then we try to set breakpoint only in script with given
lineNumber inside and ignore all other scripts. We should follow the
same logic when we capture hint for later breakpoint restore.
R=yangguo@chromium.org
Bug: none
Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel
Change-Id: I44a332ac64f62ec9a0d24d5fe4688f8ced125e39
Reviewed-on: https://chromium-review.googlesource.com/821053
Commit-Queue: Yang Guo <yangguo@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50019}
Port 52ff3ae48b
Original Commit Message:
- Implement RunMicrotasks in CSA to prevent a potentially large number
of jumps between C++ and JS code while consuming te queue. Appears to
provide a ~60% speedup in microtask-heavy code, which from limited
testing appears to scale linearly.
The code-stub microtask pump bails out to the old C++ microtask pump
if it encounters a CallHandlerInfo microtask, and remains in C++ for
the remainder of the queue (returning to the JS/stub implementation
after the bailed out queue is exhausted).
- Add a variation of JSEntryStub which enters the new RunMicrotasks code
stub.
- Add a new RunMicrotasks helper to Execution, which uses the
RunMicrotasks entry stub.
R=caitp@igalia.com, joransiu@ca.ibm.com, jbarboza@ca.ibm.com
BUG=
LOG=N
Change-Id: Ifa15ca19312bb92758e82d19c3e3fc0a8b908d82
Reviewed-on: https://chromium-review.googlesource.com/820197
Reviewed-by: Joran Siu <joransiu@ca.ibm.com>
Commit-Queue: Junliang Yan <jyan@ca.ibm.com>
Cr-Commit-Position: refs/heads/master@{#50018}
This is a separation of the DFA Unicode Decoder from
https://chromium-review.googlesource.com/c/v8/v8/+/789560
I attempted to make the DFA's table a bit more explicit in this CL. Still, the
linter prevents me from letting me present the array as a "table" in source
code. For a better representation, please refer to
https://docs.google.com/spreadsheets/d/1L9STtkmWs-A7HdK5ZmZ-wPZ_VBjQ3-Jj_xN9c6_hLKA
- - - - -
Now for a big copy-paste from 789560:
Essentially, reworks a standard FSM (imagine an
array of structs) and flattens it out into a single-dimension array.
Using Table 3-7 of the Unicode 10.0.0 standard (page 126 of
http://www.unicode.org/versions/Unicode10.0.0/ch03.pdf), we can nicely
map all bytes into one of 12 character classes:
00. 0x00-0x7F
01. 0x80-0x8F (split from general continuation because this range is not
valid after a 0xF0 leading byte)
02. 0x90-0x9F (split from general continuation because this range is not
valid after a 0xE0 nor a 0xF4 leading byte)
03. 0xA0-0xBF (the rest of the continuation range)
04. 0xC0-0xC1, 0xF5-0xFF (the joined range of invalid bytes, notice this
includes 255 which we use as a known bad byte during hex-to-int
decoding)
05. 0xC2-0xDF (leading bytes which require any continuation byte
afterwards)
06. 0xE0 (leading byte which requires a 0xA0-0xBF afterwards then any
continuation byte after that)
07. 0xE1-0xEC, 0xEE-0xEF (leading bytes which requires any continuation
afterwards then any continuation byte after that)
08. 0xED (leading byte which requires a 0x80-0x9F afterwards then any
continuation byte after that)
09. 0xF1-F3 (leading bytes which requires any continuation byte
afterwards then any continuation byte then any continuation byte)
10. 0xF0 (leading bytes which requires a 0x90-0xBF afterwards then any
continuation byte then any continuation byte)
11. 0xF4 (leading bytes which requires a 0x80-0x8F afterwards then any
continuation byte then any continuation byte)
Note that 0xF0 and 0xF1-0xF3 were swapped so that fewer bytes were
needed to represent the transition state ("9, 10, 10, 10" vs.
"10, 9, 9, 9").
Using these 12 classes as "transitions", we can map from one state to
the next. Each state is defined as some multiple of 12, so that we're
always starting at the 0th column of each row of the FSM. From each
state, we add the transition and get a index of the new row the FSM is
entering.
If at any point we encounter a bad byte, the state + bad-byte-transition
is guaranteed to map us into the first row of the FSM (which contains no
valid exiting transitions).
The key differences from Björn's original (or his self-modified) DFA is
the "bad" state is now mapped to 0 (or the first row of the FSM) instead
of 12 (the second row). This saves ~50 bytes when gzipping, and also
speeds up determining if a string is properly encoded (see his sample
code at http://bjoern.hoehrmann.de/utf-8/decoder/dfa/#performance).
Finally, I've replace his ternary check with an array access, to make
the algorithm branchless. This places a requirement on the caller to 0
out the code point between successful decodings, which it could always
have done because it's already branching.
R=marja@google.com
Bug:
Change-Id: I574f208a84dc5d06caba17127b0d41f7ce1a3395
Reviewed-on: https://chromium-review.googlesource.com/805357
Commit-Queue: Justin Ridgewell <jridgewell@google.com>
Reviewed-by: Marja Hölttä <marja@chromium.org>
Reviewed-by: Mathias Bynens <mathias@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50012}
