When rewiring a block to throw, we need to remove the current block from the list
of predecessors for all of our successors, as well as clearing our current successors.
BUG=v8:9684
Change-Id: I0da063b2ef707f07ea27a5f72cabd2ff9a91cc42
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1881154
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64579}
Adds the ability for the GraphAssembler to operate on, and maintain, a
scheduled graph. This will be used by TurboProp to maintain the initial
schedule created before effect-control-linearization, by updating this schedule
during effect-control, select and memory lowering stages rather than doing a
later reschedule.
In order to do this, an internal BlockUpdater is added to GraphAssembler,
which is enabled by passing the schedule to the GraphAssembler. The
GraphAssembler is modified to call into the block updater when nodes are added
and updates the schedule with new basic blocks when new control flow is updated.
BUG=v8:9684
Change-Id: I6d428ad21d869c472bb20f43cc8caf44722f090a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1841355
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64519}
This removes the explicit {kCallWithCallerSavedRegisters} opcode which
is just a regular call node with special handling for saving/restoring
caller saved registers before/after the call. This is now handled via
the {CallDescriptor::kCallerSavedRegisters} flag.
R=neis@chromium.org
BUG=v8:9396
Change-Id: Ie6421085eb2be8a067040222cd5215a9b1013048
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1728611
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#63021}
This allows to return bool values from Torque macros and branch on them
without performance penalty, reconstructing good control flow.
Drive-by cleanup: Delete EnsureDeferredCodeSingleEntryPoint(), since
it's no longer needed. Constructing a graph and then re-inferring
deferred blocks based on branch hints achieves this effect
automatically.
Bug: v8:7793
Change-Id: Idb6802372b407549e4760f290933d5b8f1e9d952
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1681132
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#62979}
This CL is an experiment to get more performance data from the perf-bots
and will likely lead to regressions. The try-bots (see patcheset 9)
indicate some regressions, but it doesn't seem too bad.
Change-Id: Ia173ab20ee2a4904663db0f4ca2ffb196b203c77
Reviewed-on: https://chromium-review.googlesource.com/c/1319763
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57483}
This removes the burden from Torque not to emit unnecessary phis.
This is factored out from the Torque IR CL (https://crrev.com/c/1245766).
Change-Id: I302714250e9ea6367f37613c09caa522d56c151c
Reviewed-on: https://chromium-review.googlesource.com/1254121
Reviewed-by: Daniel Clifford <danno@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56312}
This is a reland of 0909dbe3d6.
Added missing V8_EXPORT_PRIVATE to AndroidLogStream.
TBR=mstarzinger@chromium.org
Original change's description:
> Introduce StdoutStream which prints to Android log or stdout
>
> The often used construct {OFStream(stdout)} does not work on Android.
> This CL introduces an {StdoutStream} which behaves exactly like
> {OFStream(stdout)} on non-android platforms, and redirects to the
> Android log on appropriate systems and configurations.
>
> R=mstarzinger@chromium.org
>
> Bug: v8:7820
> Change-Id: Ia682fdf6d064e37c605c19b032f5a10b96ac825b
> Reviewed-on: https://chromium-review.googlesource.com/1088911
> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
> Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#53692}
Bug: v8:7820
Change-Id: I8164bad78a401dbe4246c9ffcacd050fe511ed58
Reviewed-on: https://chromium-review.googlesource.com/1100636
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53733}
This reverts commit 0909dbe3d6.
Reason for revert: Blocks roll:
https://chromium-review.googlesource.com/c/chromium/src/+/1099143
Original change's description:
> Introduce StdoutStream which prints to Android log or stdout
>
> The often used construct {OFStream(stdout)} does not work on Android.
> This CL introduces an {StdoutStream} which behaves exactly like
> {OFStream(stdout)} on non-android platforms, and redirects to the
> Android log on appropriate systems and configurations.
>
> R=mstarzinger@chromium.org
>
> Bug: v8:7820
> Change-Id: Ia682fdf6d064e37c605c19b032f5a10b96ac825b
> Reviewed-on: https://chromium-review.googlesource.com/1088911
> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
> Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#53692}
TBR=mstarzinger@chromium.org,jarin@chromium.org,jgruber@chromium.org,clemensh@chromium.org,bmeurer@chromium.org
Change-Id: Iadadd9a0df10dca0fad647138a83db50148e864d
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:7820
Reviewed-on: https://chromium-review.googlesource.com/1100635
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53725}
The often used construct {OFStream(stdout)} does not work on Android.
This CL introduces an {StdoutStream} which behaves exactly like
{OFStream(stdout)} on non-android platforms, and redirects to the
Android log on appropriate systems and configurations.
R=mstarzinger@chromium.org
Bug: v8:7820
Change-Id: Ia682fdf6d064e37c605c19b032f5a10b96ac825b
Reviewed-on: https://chromium-review.googlesource.com/1088911
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53692}
This removes Type::operator-> which was used to split the change that
removed undefined misuse of Type* to represent integers.
Bug: v8:3770
Change-Id: I9a5bce5ccdc75461a7b939b4070cb58fe6040d99
Reviewed-on: https://chromium-review.googlesource.com/1033736
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52878}
This is part of the effort to decrease the amount of undefined behavior.
that v8 relies on.
The main change here is to represent types with class Type rather than
with pointer Type*. To make the CL smaller, I used an operator overload
hack to separate the change from `->` to `.`. I am working on a CL that
will remove the operator and change all those arrows to dots.
Bug: v8:3770
Change-Id: I71a197cb739a1467937bc95c2a757fab0469aa22
Reviewed-on: https://chromium-review.googlesource.com/1032551
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52872}
In general, TurboFan doesn't encounter phi nodes with only a single
data input in the backend. However, CSA-based builtins (especially
auto-generated ones, e.g. from Torque), may contain single-input phi nodes,
although outside the auto-generated case this doesn't happen much in practice.
Single input phi nodes (i.e. phis in blocks with one predecessor) don't have
any side effects and are essentially useless and harmless, but to avoid problems
in the backend of TurboFan (whose SSA deconstruction disallows control flow
splits that continue to blocks with phis), this CL tweaks the existing
CSA-only control flow and graph sanitization in the CSA path to ensure
no no-op phis.
Change-Id: I109f4dc6cde5ad1794585a09609a230b1848e0d5
Reviewed-on: https://chromium-review.googlesource.com/963711
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Daniel Clifford <danno@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52056}
Use the (D)CHECK_{EQ,NE,GT,...} macros instead of (D)CHECK with an
embedded comparison. This gives better error messages and also does the
right comparison for signed/unsigned mismatches.
This will allow us to reenable the readability/check cpplint check.
R=jarin@chromium.org
Bug: v8:6837
Change-Id: I712580c2a4326e06ee3d6d0eb4ff8c7d24f5fdb9
Reviewed-on: https://chromium-review.googlesource.com/671227
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48135}
The new node behave the same as its counterpart, CallCFunction, with the
additional saving and restoring caller saved registers before and after
the function call.
Bug: chromium:749486
Change-Id: I0a1dfb2e4e55f7720541a00e6d16fd20220f39ed
Reviewed-on: https://chromium-review.googlesource.com/620709
Commit-Queue: Albert Mingkun Yang <albertnetymk@google.com>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47479}
We use Schedule::EnsureDeferredCodeSingleEntryPoint as a helper for
hand-crafted builtin code, to ensure deferred code isn't entered from a
mix of deferred and non-deferred code (invariant required for hot/cold
allocation, or "splintering").
When we create a "merger" block, it may be the case that the original
block had a few phi operands. Those need to be moved as well.
This bug was uncovered by both v8:6390, and, earlier, by v8:5998. We
fixed the earlier one by authoring a the builtin to avoid the need for
EnsureDeferredCodeSingleEntryPoint. I proposed earlier an alternative
where we'd replace the Ensure... method with a Verify, and throw early
when the builtin is assembled, however, we may want to maintain the
slightly higher level DSL for authoring builtins, and perform such
graph adjustments for the lower level constraints afterwards, hence
this current CL.
Bug: v8:5998 v8:6390
Change-Id: Ia3143f7a66904fe480d8edb5b52bf915b8d185dc
Reviewed-on: https://chromium-review.googlesource.com/505264
Commit-Queue: Mircea Trofin <mtrofin@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45387}
Before this patch, loops in deferred code would defeat the propagation of the
deferred flag, since back edges would usually not come from deferred blocks,
thus stoping the forward propagation of the deferred flag at loop headers. This
patch ensures that back edges are ignored in the deferred propations, properly
placing loops dominated by deferred labels and the code that follows them into
deferred code.
R=epertoso@chromium.org
LOG=N
Review-Url: https://codereview.chromium.org/2606923002
Cr-Commit-Position: refs/heads/master@{#41976}
Previously, split-edge blocks inserted into CodeStubAssembler schedules
inherited the "deferredness" of their predecessor block. This tended to inline
newly inserted blocks that preceeded deferred code inline with non-deferred
code. The stack frame elider sometimes inserted stack construction into these
blocks, code which was actually was only needed in the deferred case.
This CL marks inserted split edge blocks with the deferredness of their
successors, e.g. they are defererred if their predecessor is deferrred. This
moves stack building code that logically belongs to deferred code into the
deferred code area rather than inlining it in non-deferred code.
Review-Url: https://codereview.chromium.org/2003663002
Cr-Commit-Position: refs/heads/master@{#36446}
If a deferred block has multiple predecessors, they have to be
all deferred. Otherwise, we can run into a situation where if a range
that spills only in deferred blocks inserts its spill in the block, and
other ranges need moves inserted by ResolveControlFlow in the predecessors,
the register of the range spilled in the deferred block may be clobbered.
To avoid that, when a deferred block has multiple predecessors, and some
are not deferred, we add a non-deferred block to collect all such edges.
This CL addresses the validator assertion failure the referenced issue, as well
as the greedy allocator failure - which was caused by the situation described
above.
BUG=v8:4940
LOG=n
Review URL: https://codereview.chromium.org/1912093005
Cr-Commit-Position: refs/heads/master@{#35742}
This introduces a compiler pass that schedules the graph and re-wires effect chain according to the schedule. It also connects allocating representation changes to the effect chain, and removes the BeginRegion and EndRegion nodes - they should not be needed anymore because all effectful nodes should be already wired-in.
This is an intermediate CL - the next step is to move lowering of the Change*ToTaggedEffect nodes to StateEffectIntroduction so that we do not have to introduce the effectful versions of nodes.
Review URL: https://codereview.chromium.org/1849603002
Cr-Commit-Position: refs/heads/master@{#35565}
We use the pattern "B23" to refer to RPO numbers and "id:42" when
referring to unordered block ids. This unifies printing functions to
follow that pattern throughout the system.
R=danno@chromium.org
Review URL: https://codereview.chromium.org/1824313002
Cr-Commit-Position: refs/heads/master@{#35022}
This CL adds an extra pass before calculating the special RPO
order in the custom RawMachineAssembler pipeline that
walks through the schedule and inserts extra blocks to
guarantee that the control flow graph is in split edge form. It
also propagates deferred block marks forward to these new
blocks if appropriate.
Review URL: https://codereview.chromium.org/1811333002
Cr-Commit-Position: refs/heads/master@{#35014}
Up until now we used int32_t for NodeId, but that was not ideal because
negative values are invalid for NodeId and we use it as an array index
for example in the NodeMarker class, where C++ compilers on x64 have to
generate code that does proper sign extension for the indices, which is
completely unnecessary.
R=svenpanne@chromium.org
Review URL: https://codereview.chromium.org/1178403004
Cr-Commit-Position: refs/heads/master@{#28997}
Tail calls are matched on the graph, with a dedicated tail call
optimization that is actually testable. The instruction selection can
still fall back to a regular if the platform constraints don't allow to
emit a tail call (i.e. the return locations of caller and callee differ
or the callee takes non-register parameters, which is a restriction that
will be removed in the future).
Also explicitly limit tail call optimization to stubs for now and drop
the global flag.
BUG=v8:4076
LOG=n
Review URL: https://codereview.chromium.org/1114163005
Cr-Commit-Position: refs/heads/master@{#28219}
Adds Switch and Case operators to TurboFan and handles them
appropriately in instruction selection and code generation.
BUG=v8:3872
LOG=n
Review URL: https://codereview.chromium.org/892513003
Cr-Commit-Position: refs/heads/master@{#26515}
This contains the following changes squashed together:
- Switch BasicBlock::loop_end to be a basic block instead of an RPO.
- Switch ScheduleLate to use dominator depth instead of RPO.
- Switch ScheduleEarly to use dominator depth instead of RPO.
- Push out absolute RPO ordering everywhere else in the scheduler.
- Keep linked list of blocks in RPO order while scheduling.
- Switch from RPO number to depth for dominator calculation.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/696363002
Cr-Commit-Position: refs/heads/master@{#25138}
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@25138 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Branch can now have an optional hint, when the condition is
likely true or false, and if such a hint is present the other
basic block will be marked as deferred and placed at the end
of the function.
We currently use this feature for tagging int32/uint32 in
change lowering, and for load/store bounds checks in simplified
lowering.
TEST=cctest,unittests
R=dcarney@chromium.org
Review URL: https://codereview.chromium.org/642883003
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@24802 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
- 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
