v8/test/cctest/interpreter/bytecode_expectations/ForAwaitOf.golden

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

676 lines
24 KiB
Plaintext
Raw Normal View History

#
# Autogenerated by generate-bytecode-expectations.
#
---
wrap: no
test function name: f
async iteration: yes
---
snippet: "
async function f() {
for await (let x of [1, 2, 3]) {}
}
f();
"
frame size: 21
parameter count: 1
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
bytecode array length: 325
bytecodes: [
B(SwitchOnGeneratorState), R(0), U8(0), U8(2),
B(Mov), R(closure), R(4),
B(Mov), R(this), R(5),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionEnter), R(4), U8(2),
B(Star), R(0),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
/* 16 E> */ B(StackCheck),
B(Mov), R(context), R(4),
/* 43 S> */ B(CreateArrayLiteral), U8(2), U8(0), U8(37),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(7),
B(LdaNamedProperty), R(7), U8(3), U8(1),
B(JumpIfUndefined), U8(17),
B(JumpIfNull), U8(15),
B(Star), R(8),
B(CallProperty0), R(8), R(7), U8(3),
B(JumpIfJSReceiver), U8(23),
B(CallRuntime), U16(Runtime::kThrowSymbolAsyncIteratorInvalid), R(0), U8(0),
B(LdaNamedProperty), R(7), U8(4), U8(5),
B(Star), R(8),
B(CallProperty0), R(8), R(7), U8(7),
B(Star), R(8),
B(InvokeIntrinsic), U8(Runtime::k_CreateAsyncFromSyncIterator), R(8), U8(1),
B(Star), R(6),
B(LdaNamedProperty), R(6), U8(5), U8(9),
B(Star), R(5),
B(LdaFalse),
B(Star), R(9),
B(Mov), R(context), R(12),
B(LdaTrue),
B(Star), R(9),
/* 38 S> */ B(CallProperty0), R(5), R(6), U8(11),
B(Star), R(15),
B(Mov), R(0), R(14),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionAwaitUncaught), R(14), U8(2),
B(SuspendGenerator), R(0), R(0), U8(14), U8(0),
B(ResumeGenerator), R(0), R(0), U8(14),
B(Star), R(14),
B(InvokeIntrinsic), U8(Runtime::k_GeneratorGetResumeMode), R(0), U8(1),
B(Star), R(15),
B(LdaZero),
B(TestReferenceEqual), R(15),
B(JumpIfTrue), U8(5),
B(Ldar), R(14),
B(ReThrow),
B(Ldar), R(14),
B(Mov), R(14), R(13),
B(JumpIfJSReceiver), U8(7),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(13), U8(1),
B(LdaNamedProperty), R(13), U8(6), U8(13),
B(JumpIfToBooleanTrue), U8(23),
B(LdaNamedProperty), R(13), U8(7), U8(15),
B(Star), R(13),
B(LdaFalse),
B(Star), R(9),
B(Mov), R(13), R(1),
/* 23 E> */ B(StackCheck),
/* 38 S> */ B(Mov), R(1), R(3),
B(Ldar), R(13),
B(JumpLoop), U8(77), I8(0),
B(LdaSmi), I8(-1),
B(Star), R(11),
B(Star), R(10),
B(Jump), U8(7),
B(Star), R(11),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(LdaZero),
B(Star), R(10),
B(LdaTheHole),
/* 38 E> */ B(SetPendingMessage),
B(Star), R(12),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Ldar), R(9),
B(JumpIfToBooleanTrue), U8(96),
B(LdaNamedProperty), R(6), U8(8), U8(17),
B(Star), R(16),
B(JumpIfUndefined), U8(88),
B(JumpIfNull), U8(86),
B(TestTypeOf), U8(6),
B(JumpIfTrue), U8(18),
B(Wide), B(LdaSmi), I16(155),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(17),
B(LdaConstant), U8(9),
B(Star), R(18),
B(CallRuntime), U16(Runtime::kNewTypeError), R(17), U8(2),
B(Throw),
B(Mov), R(context), R(17),
B(CallProperty0), R(16), R(6), U8(19),
B(Star), R(19),
B(Mov), R(0), R(18),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionAwaitUncaught), R(18), U8(2),
B(SuspendGenerator), R(0), R(0), U8(18), U8(1),
B(ResumeGenerator), R(0), R(0), U8(18),
B(Star), R(18),
B(InvokeIntrinsic), U8(Runtime::k_GeneratorGetResumeMode), R(0), U8(1),
B(Star), R(19),
B(LdaZero),
B(TestReferenceEqual), R(19),
B(JumpIfTrue), U8(5),
B(Ldar), R(18),
B(ReThrow),
B(Ldar), R(18),
B(JumpIfJSReceiver), U8(21),
B(Star), R(20),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(20), U8(1),
B(Jump), U8(12),
B(Star), R(17),
B(LdaZero),
B(TestReferenceEqual), R(10),
B(JumpIfTrue), U8(5),
B(Ldar), R(17),
B(ReThrow),
B(Ldar), R(12),
B(SetPendingMessage),
B(LdaZero),
B(TestReferenceEqual), R(10),
B(JumpIfFalse), U8(5),
B(Ldar), R(11),
B(ReThrow),
B(LdaUndefined),
B(Star), R(6),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaTrue),
B(Star), R(7),
B(Mov), R(0), R(5),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionResolve), R(5), U8(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
/* 57 S> */ B(Return),
B(Star), R(5),
B(CreateCatchContext), R(5), U8(10),
B(Star), R(4),
B(LdaTheHole),
B(SetPendingMessage),
B(Ldar), R(4),
B(PushContext), R(5),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaImmutableCurrentContextSlot), U8(4),
B(Star), R(7),
B(LdaTrue),
B(Star), R(8),
B(Mov), R(0), R(6),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionReject), R(6), U8(3),
/* 57 S> */ B(Return),
]
constant pool: [
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
Smi [98],
Smi [229],
ARRAY_BOILERPLATE_DESCRIPTION_TYPE,
SYMBOL_TYPE,
SYMBOL_TYPE,
ONE_BYTE_INTERNALIZED_STRING_TYPE ["next"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["done"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["value"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["return"],
[turbofan] Significantly improve ConsString creation performance. This change significantly improves the performance of string concatenation in optimized code for the case where the resulting string is represented as a ConsString. On the relevant test cases we go from serializeNaive: 10762 ms. serializeClever: 7813 ms. serializeConcat: 10271 ms. to serializeNaive: 10278 ms. serializeClever: 5533 ms. serializeConcat: 10310 ms. which represents a 30% improvement on the "clever" benchmark, which tests specifically the ConsString creation performance. This was accomplished via a couple of different steps, which are briefly outlined here: 1. The empty_string gets its own map, so that we can easily recognize and handle it appropriately in the TurboFan type system. This allows us to express (and assert) that the inputs to NewConsString are non-empty strings, making sure that TurboFan no longer creates "crippled ConsStrings" with empty left or right hand sides. 2. Further split the existing String types in TurboFan to be able to distinguish between OneByte and TwoByte strings on the type system level. This allows us to avoid having to dynamically lookup the resulting ConsString map in case of ConsString creation (i.e. when we know that both input strings are OneByte strings or at least one of the input strings is TwoByte). 3. We also introduced more finegrained feedback for the Add bytecode in the interpreter, having it collect feedback about ConsStrings, specifically ConsOneByteString and ConsTwoByteString. This feedback can be used by TurboFan to only inline the relevant code for what was seen so far. This allows us to remove the Octane/Splay specific magic in JSTypedLowering to detect ConsString creation, and instead purely rely on the feedback of what was seen so far (also making it possible to change the semantics of NewConsString to be a low-level operator, which is only introduced in SimplifiedLowering by looking at the input types of StringConcat). 4. On top of the before mentioned type and interpreter changes we added new operators CheckNonEmptyString, CheckNonEmptyOneByteString, and CheckNonEmptyTwoByteString, which perform the appropriate (dynamic) checks. There are several more improvements that are possible based on this, but since the change was already quite big, we decided not to put everything into the first change, but do some follow up tweaks to the type system, and builtin optimizations later. Tbr: mstarzinger@chromium.org Bug: v8:8834, v8:8931, v8:8939, v8:8951 Change-Id: Ia24e17c6048bf2b04df966d3cd441f0edda05c93 Cq-Include-Trybots: luci.chromium.try:linux-blink-rel Doc: https://bit.ly/fast-string-concatenation-in-javascript Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1499497 Commit-Queue: Michael Achenbach <machenbach@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Cr-Commit-Position: refs/heads/master@{#60318}
2019-03-18 19:33:46 +00:00
EMPTY_STRING_TYPE [""],
SCOPE_INFO_TYPE,
]
handlers: [
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
[20, 297, 297],
[77, 157, 165],
[211, 260, 262],
]
---
snippet: "
async function f() {
for await (let x of [1, 2, 3]) { return x; }
}
f();
"
frame size: 21
parameter count: 1
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
bytecode array length: 346
bytecodes: [
B(SwitchOnGeneratorState), R(0), U8(0), U8(2),
B(Mov), R(closure), R(4),
B(Mov), R(this), R(5),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionEnter), R(4), U8(2),
B(Star), R(0),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
/* 16 E> */ B(StackCheck),
B(Mov), R(context), R(4),
/* 43 S> */ B(CreateArrayLiteral), U8(2), U8(0), U8(37),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(7),
B(LdaNamedProperty), R(7), U8(3), U8(1),
B(JumpIfUndefined), U8(17),
B(JumpIfNull), U8(15),
B(Star), R(8),
B(CallProperty0), R(8), R(7), U8(3),
B(JumpIfJSReceiver), U8(23),
B(CallRuntime), U16(Runtime::kThrowSymbolAsyncIteratorInvalid), R(0), U8(0),
B(LdaNamedProperty), R(7), U8(4), U8(5),
B(Star), R(8),
B(CallProperty0), R(8), R(7), U8(7),
B(Star), R(8),
B(InvokeIntrinsic), U8(Runtime::k_CreateAsyncFromSyncIterator), R(8), U8(1),
B(Star), R(6),
B(LdaNamedProperty), R(6), U8(5), U8(9),
B(Star), R(5),
B(LdaFalse),
B(Star), R(9),
B(Mov), R(context), R(12),
B(LdaTrue),
B(Star), R(9),
/* 38 S> */ B(CallProperty0), R(5), R(6), U8(11),
B(Star), R(15),
B(Mov), R(0), R(14),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionAwaitUncaught), R(14), U8(2),
B(SuspendGenerator), R(0), R(0), U8(14), U8(0),
B(ResumeGenerator), R(0), R(0), U8(14),
B(Star), R(14),
B(InvokeIntrinsic), U8(Runtime::k_GeneratorGetResumeMode), R(0), U8(1),
B(Star), R(15),
B(LdaZero),
B(TestReferenceEqual), R(15),
B(JumpIfTrue), U8(5),
B(Ldar), R(14),
B(ReThrow),
B(Ldar), R(14),
B(Mov), R(14), R(13),
B(JumpIfJSReceiver), U8(7),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(13), U8(1),
B(LdaNamedProperty), R(13), U8(6), U8(13),
B(JumpIfToBooleanTrue), U8(27),
B(LdaNamedProperty), R(13), U8(7), U8(15),
B(Star), R(13),
B(LdaFalse),
B(Star), R(9),
B(Mov), R(13), R(1),
/* 23 E> */ B(StackCheck),
/* 38 S> */ B(Mov), R(1), R(3),
/* 56 S> */ B(LdaSmi), I8(1),
B(Mov), R(13), R(11),
B(Star), R(10),
B(Jump), U8(15),
B(LdaSmi), I8(-1),
B(Star), R(11),
B(Star), R(10),
B(Jump), U8(7),
B(Star), R(11),
B(LdaZero),
B(Star), R(10),
B(LdaTheHole),
B(SetPendingMessage),
B(Star), R(12),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Ldar), R(9),
B(JumpIfToBooleanTrue), U8(96),
B(LdaNamedProperty), R(6), U8(8), U8(17),
B(Star), R(16),
B(JumpIfUndefined), U8(88),
B(JumpIfNull), U8(86),
B(TestTypeOf), U8(6),
B(JumpIfTrue), U8(18),
B(Wide), B(LdaSmi), I16(155),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(17),
B(LdaConstant), U8(9),
B(Star), R(18),
B(CallRuntime), U16(Runtime::kNewTypeError), R(17), U8(2),
B(Throw),
B(Mov), R(context), R(17),
B(CallProperty0), R(16), R(6), U8(19),
B(Star), R(19),
B(Mov), R(0), R(18),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionAwaitUncaught), R(18), U8(2),
B(SuspendGenerator), R(0), R(0), U8(18), U8(1),
B(ResumeGenerator), R(0), R(0), U8(18),
B(Star), R(18),
B(InvokeIntrinsic), U8(Runtime::k_GeneratorGetResumeMode), R(0), U8(1),
B(Star), R(19),
B(LdaZero),
B(TestReferenceEqual), R(19),
B(JumpIfTrue), U8(5),
B(Ldar), R(18),
B(ReThrow),
B(Ldar), R(18),
B(JumpIfJSReceiver), U8(21),
B(Star), R(20),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(20), U8(1),
B(Jump), U8(12),
B(Star), R(17),
B(LdaZero),
B(TestReferenceEqual), R(10),
B(JumpIfTrue), U8(5),
B(Ldar), R(17),
B(ReThrow),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Ldar), R(12),
B(SetPendingMessage),
B(Ldar), R(10),
B(SwitchOnSmiNoFeedback), U8(10), U8(2), I8(0),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(Jump), U8(19),
B(Ldar), R(11),
B(ReThrow),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaTrue),
B(Star), R(18),
B(Mov), R(0), R(16),
B(Mov), R(11), R(17),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionResolve), R(16), U8(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
/* 68 S> */ B(Return),
B(LdaUndefined),
B(Star), R(6),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaTrue),
B(Star), R(7),
B(Mov), R(0), R(5),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionResolve), R(5), U8(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
/* 68 S> */ B(Return),
B(Star), R(5),
B(CreateCatchContext), R(5), U8(12),
B(Star), R(4),
B(LdaTheHole),
B(SetPendingMessage),
B(Ldar), R(4),
B(PushContext), R(5),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaImmutableCurrentContextSlot), U8(4),
B(Star), R(7),
B(LdaTrue),
B(Star), R(8),
B(Mov), R(0), R(6),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionReject), R(6), U8(3),
/* 68 S> */ B(Return),
]
constant pool: [
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
Smi [98],
Smi [233],
ARRAY_BOILERPLATE_DESCRIPTION_TYPE,
SYMBOL_TYPE,
SYMBOL_TYPE,
ONE_BYTE_INTERNALIZED_STRING_TYPE ["next"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["done"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["value"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["return"],
[turbofan] Significantly improve ConsString creation performance. This change significantly improves the performance of string concatenation in optimized code for the case where the resulting string is represented as a ConsString. On the relevant test cases we go from serializeNaive: 10762 ms. serializeClever: 7813 ms. serializeConcat: 10271 ms. to serializeNaive: 10278 ms. serializeClever: 5533 ms. serializeConcat: 10310 ms. which represents a 30% improvement on the "clever" benchmark, which tests specifically the ConsString creation performance. This was accomplished via a couple of different steps, which are briefly outlined here: 1. The empty_string gets its own map, so that we can easily recognize and handle it appropriately in the TurboFan type system. This allows us to express (and assert) that the inputs to NewConsString are non-empty strings, making sure that TurboFan no longer creates "crippled ConsStrings" with empty left or right hand sides. 2. Further split the existing String types in TurboFan to be able to distinguish between OneByte and TwoByte strings on the type system level. This allows us to avoid having to dynamically lookup the resulting ConsString map in case of ConsString creation (i.e. when we know that both input strings are OneByte strings or at least one of the input strings is TwoByte). 3. We also introduced more finegrained feedback for the Add bytecode in the interpreter, having it collect feedback about ConsStrings, specifically ConsOneByteString and ConsTwoByteString. This feedback can be used by TurboFan to only inline the relevant code for what was seen so far. This allows us to remove the Octane/Splay specific magic in JSTypedLowering to detect ConsString creation, and instead purely rely on the feedback of what was seen so far (also making it possible to change the semantics of NewConsString to be a low-level operator, which is only introduced in SimplifiedLowering by looking at the input types of StringConcat). 4. On top of the before mentioned type and interpreter changes we added new operators CheckNonEmptyString, CheckNonEmptyOneByteString, and CheckNonEmptyTwoByteString, which perform the appropriate (dynamic) checks. There are several more improvements that are possible based on this, but since the change was already quite big, we decided not to put everything into the first change, but do some follow up tweaks to the type system, and builtin optimizations later. Tbr: mstarzinger@chromium.org Bug: v8:8834, v8:8931, v8:8939, v8:8951 Change-Id: Ia24e17c6048bf2b04df966d3cd441f0edda05c93 Cq-Include-Trybots: luci.chromium.try:linux-blink-rel Doc: https://bit.ly/fast-string-concatenation-in-javascript Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1499497 Commit-Queue: Michael Achenbach <machenbach@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Cr-Commit-Position: refs/heads/master@{#60318}
2019-03-18 19:33:46 +00:00
EMPTY_STRING_TYPE [""],
Smi [6],
Smi [9],
SCOPE_INFO_TYPE,
]
handlers: [
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
[20, 318, 318],
[77, 161, 169],
[215, 264, 266],
]
---
snippet: "
async function f() {
for await (let x of [10, 20, 30]) {
if (x == 10) continue;
if (x == 20) break;
}
}
f();
"
frame size: 21
parameter count: 1
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
bytecode array length: 341
bytecodes: [
B(SwitchOnGeneratorState), R(0), U8(0), U8(2),
B(Mov), R(closure), R(4),
B(Mov), R(this), R(5),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionEnter), R(4), U8(2),
B(Star), R(0),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
/* 16 E> */ B(StackCheck),
B(Mov), R(context), R(4),
/* 43 S> */ B(CreateArrayLiteral), U8(2), U8(0), U8(37),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(7),
B(LdaNamedProperty), R(7), U8(3), U8(1),
B(JumpIfUndefined), U8(17),
B(JumpIfNull), U8(15),
B(Star), R(8),
B(CallProperty0), R(8), R(7), U8(3),
B(JumpIfJSReceiver), U8(23),
B(CallRuntime), U16(Runtime::kThrowSymbolAsyncIteratorInvalid), R(0), U8(0),
B(LdaNamedProperty), R(7), U8(4), U8(5),
B(Star), R(8),
B(CallProperty0), R(8), R(7), U8(7),
B(Star), R(8),
B(InvokeIntrinsic), U8(Runtime::k_CreateAsyncFromSyncIterator), R(8), U8(1),
B(Star), R(6),
B(LdaNamedProperty), R(6), U8(5), U8(9),
B(Star), R(5),
B(LdaFalse),
B(Star), R(9),
B(Mov), R(context), R(12),
B(LdaTrue),
B(Star), R(9),
/* 38 S> */ B(CallProperty0), R(5), R(6), U8(11),
B(Star), R(15),
B(Mov), R(0), R(14),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionAwaitUncaught), R(14), U8(2),
B(SuspendGenerator), R(0), R(0), U8(14), U8(0),
B(ResumeGenerator), R(0), R(0), U8(14),
B(Star), R(14),
B(InvokeIntrinsic), U8(Runtime::k_GeneratorGetResumeMode), R(0), U8(1),
B(Star), R(15),
B(LdaZero),
B(TestReferenceEqual), R(15),
B(JumpIfTrue), U8(5),
B(Ldar), R(14),
B(ReThrow),
B(Ldar), R(14),
B(Mov), R(14), R(13),
B(JumpIfJSReceiver), U8(7),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(13), U8(1),
B(LdaNamedProperty), R(13), U8(6), U8(13),
B(JumpIfToBooleanTrue), U8(39),
B(LdaNamedProperty), R(13), U8(7), U8(15),
B(Star), R(13),
B(LdaFalse),
B(Star), R(9),
B(Mov), R(13), R(1),
/* 23 E> */ B(StackCheck),
/* 38 S> */ B(Mov), R(1), R(3),
/* 63 S> */ B(LdaSmi), I8(10),
/* 69 E> */ B(TestEqual), R(3), U8(17),
B(JumpIfFalse), U8(4),
/* 76 S> */ B(Jump), U8(11),
/* 90 S> */ B(LdaSmi), I8(20),
/* 96 E> */ B(TestEqual), R(3), U8(18),
B(JumpIfFalse), U8(4),
/* 103 S> */ B(Jump), U8(5),
B(JumpLoop), U8(93), I8(0),
B(LdaSmi), I8(-1),
B(Star), R(11),
B(Star), R(10),
B(Jump), U8(7),
B(Star), R(11),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(LdaZero),
B(Star), R(10),
B(LdaTheHole),
B(SetPendingMessage),
B(Star), R(12),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Ldar), R(9),
B(JumpIfToBooleanTrue), U8(96),
B(LdaNamedProperty), R(6), U8(8), U8(19),
B(Star), R(16),
B(JumpIfUndefined), U8(88),
B(JumpIfNull), U8(86),
B(TestTypeOf), U8(6),
B(JumpIfTrue), U8(18),
B(Wide), B(LdaSmi), I16(155),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(17),
B(LdaConstant), U8(9),
B(Star), R(18),
B(CallRuntime), U16(Runtime::kNewTypeError), R(17), U8(2),
B(Throw),
B(Mov), R(context), R(17),
B(CallProperty0), R(16), R(6), U8(21),
B(Star), R(19),
B(Mov), R(0), R(18),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionAwaitUncaught), R(18), U8(2),
B(SuspendGenerator), R(0), R(0), U8(18), U8(1),
B(ResumeGenerator), R(0), R(0), U8(18),
B(Star), R(18),
B(InvokeIntrinsic), U8(Runtime::k_GeneratorGetResumeMode), R(0), U8(1),
B(Star), R(19),
B(LdaZero),
B(TestReferenceEqual), R(19),
B(JumpIfTrue), U8(5),
B(Ldar), R(18),
B(ReThrow),
B(Ldar), R(18),
B(JumpIfJSReceiver), U8(21),
B(Star), R(20),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(20), U8(1),
B(Jump), U8(12),
B(Star), R(17),
B(LdaZero),
B(TestReferenceEqual), R(10),
B(JumpIfTrue), U8(5),
B(Ldar), R(17),
B(ReThrow),
B(Ldar), R(12),
B(SetPendingMessage),
B(LdaZero),
B(TestReferenceEqual), R(10),
B(JumpIfFalse), U8(5),
B(Ldar), R(11),
B(ReThrow),
B(LdaUndefined),
B(Star), R(6),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaTrue),
B(Star), R(7),
B(Mov), R(0), R(5),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionResolve), R(5), U8(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
/* 114 S> */ B(Return),
B(Star), R(5),
B(CreateCatchContext), R(5), U8(10),
B(Star), R(4),
B(LdaTheHole),
B(SetPendingMessage),
B(Ldar), R(4),
B(PushContext), R(5),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaImmutableCurrentContextSlot), U8(4),
B(Star), R(7),
B(LdaTrue),
B(Star), R(8),
B(Mov), R(0), R(6),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionReject), R(6), U8(3),
/* 114 S> */ B(Return),
]
constant pool: [
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
Smi [98],
Smi [245],
ARRAY_BOILERPLATE_DESCRIPTION_TYPE,
SYMBOL_TYPE,
SYMBOL_TYPE,
ONE_BYTE_INTERNALIZED_STRING_TYPE ["next"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["done"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["value"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["return"],
[turbofan] Significantly improve ConsString creation performance. This change significantly improves the performance of string concatenation in optimized code for the case where the resulting string is represented as a ConsString. On the relevant test cases we go from serializeNaive: 10762 ms. serializeClever: 7813 ms. serializeConcat: 10271 ms. to serializeNaive: 10278 ms. serializeClever: 5533 ms. serializeConcat: 10310 ms. which represents a 30% improvement on the "clever" benchmark, which tests specifically the ConsString creation performance. This was accomplished via a couple of different steps, which are briefly outlined here: 1. The empty_string gets its own map, so that we can easily recognize and handle it appropriately in the TurboFan type system. This allows us to express (and assert) that the inputs to NewConsString are non-empty strings, making sure that TurboFan no longer creates "crippled ConsStrings" with empty left or right hand sides. 2. Further split the existing String types in TurboFan to be able to distinguish between OneByte and TwoByte strings on the type system level. This allows us to avoid having to dynamically lookup the resulting ConsString map in case of ConsString creation (i.e. when we know that both input strings are OneByte strings or at least one of the input strings is TwoByte). 3. We also introduced more finegrained feedback for the Add bytecode in the interpreter, having it collect feedback about ConsStrings, specifically ConsOneByteString and ConsTwoByteString. This feedback can be used by TurboFan to only inline the relevant code for what was seen so far. This allows us to remove the Octane/Splay specific magic in JSTypedLowering to detect ConsString creation, and instead purely rely on the feedback of what was seen so far (also making it possible to change the semantics of NewConsString to be a low-level operator, which is only introduced in SimplifiedLowering by looking at the input types of StringConcat). 4. On top of the before mentioned type and interpreter changes we added new operators CheckNonEmptyString, CheckNonEmptyOneByteString, and CheckNonEmptyTwoByteString, which perform the appropriate (dynamic) checks. There are several more improvements that are possible based on this, but since the change was already quite big, we decided not to put everything into the first change, but do some follow up tweaks to the type system, and builtin optimizations later. Tbr: mstarzinger@chromium.org Bug: v8:8834, v8:8931, v8:8939, v8:8951 Change-Id: Ia24e17c6048bf2b04df966d3cd441f0edda05c93 Cq-Include-Trybots: luci.chromium.try:linux-blink-rel Doc: https://bit.ly/fast-string-concatenation-in-javascript Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1499497 Commit-Queue: Michael Achenbach <machenbach@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Cr-Commit-Position: refs/heads/master@{#60318}
2019-03-18 19:33:46 +00:00
EMPTY_STRING_TYPE [""],
SCOPE_INFO_TYPE,
]
handlers: [
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
[20, 313, 313],
[77, 173, 181],
[227, 276, 278],
]
---
snippet: "
async function f() {
var x = { 'a': 1, 'b': 2 };
for (x['a'] of [1,2,3]) { return x['a']; }
}
f();
"
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
frame size: 16
parameter count: 1
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
bytecode array length: 261
bytecodes: [
B(Mov), R(closure), R(2),
B(Mov), R(this), R(3),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionEnter), R(2), U8(2),
[async] First prototype of zero-cost async stack traces. This introduces a new flag --async-stack-traces, which enables zero-cost async stack traces. This enriches the non-standard Error.stack property with async stack frames computed from walking up the promise chains and collecting all the await suspension points along the way. In Error.stack these async frames are marked with "async" to make it possible to distinguish them from regular frames, for example: ``` Error: Some error message at bar (<anonymous>) at async foo (<anonymous>) ``` It's zero-cost because no additional information is collected during the execution of the program, but only the information already present in the promise chains is used to reconstruct an approximation of the async stack in case of an exception. But this approximation is limited to suspension points at await's in async functions. This depends on a recent ECMAScript specification change, flagged behind --harmony-await-optimization and implied the --async-stack-traces flag. Without this change there's no way to get from the outer promise of an async function to the rest of the promise chain, since the link is broken by the indirection introduced by await. For async functions the special outer promise, named .promise in the Parser desugaring, is now forcible allocated to stack slot 0 during scope resolution, to make it accessible to the stack frame construction logic. Note that this first prototype doesn't yet work fully support async generators and might have other limitations. Bug: v8:7522 Ref: nodejs/node#11865 Change-Id: I0cc8e3cdfe45dab56d3d506be2d25907409b01a9 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1256762 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Adam Klein <adamk@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56363}
2018-10-04 07:19:45 +00:00
B(Star), R(0),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
/* 16 E> */ B(StackCheck),
B(Mov), R(context), R(2),
/* 31 S> */ B(CreateObjectLiteral), U8(0), U8(0), U8(41),
B(Star), R(1),
/* 68 S> */ B(CreateArrayLiteral), U8(1), U8(1), U8(37),
B(Star), R(5),
B(LdaNamedProperty), R(5), U8(2), U8(2),
B(Star), R(6),
B(CallProperty0), R(6), R(5), U8(4),
B(JumpIfJSReceiver), U8(7),
B(CallRuntime), U16(Runtime::kThrowSymbolIteratorInvalid), R(0), U8(0),
B(Star), R(4),
B(LdaNamedProperty), R(4), U8(3), U8(6),
B(Star), R(3),
B(LdaFalse),
B(Star), R(7),
B(Mov), R(context), R(10),
B(LdaTrue),
B(Star), R(7),
/* 59 S> */ B(CallProperty0), R(3), R(4), U8(8),
B(Star), R(11),
B(JumpIfJSReceiver), U8(7),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(11), U8(1),
B(LdaNamedProperty), R(11), U8(4), U8(10),
B(JumpIfToBooleanTrue), U8(33),
B(LdaNamedProperty), R(11), U8(5), U8(12),
B(Star), R(11),
B(LdaFalse),
B(Star), R(7),
B(Ldar), R(11),
/* 58 E> */ B(StaNamedProperty), R(1), U8(6), U8(14),
/* 53 E> */ B(StackCheck),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
/* 87 S> */ B(LdaNamedProperty), R(1), U8(6), U8(16),
B(Star), R(9),
B(LdaSmi), I8(1),
B(Star), R(8),
B(Mov), R(1), R(12),
B(Jump), U8(15),
B(LdaSmi), I8(-1),
B(Star), R(9),
B(Star), R(8),
B(Jump), U8(7),
B(Star), R(9),
B(LdaZero),
B(Star), R(8),
B(LdaTheHole),
B(SetPendingMessage),
B(Star), R(10),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Ldar), R(7),
B(JumpIfToBooleanTrue), U8(60),
B(LdaNamedProperty), R(4), U8(7), U8(18),
B(Star), R(13),
B(JumpIfUndefined), U8(52),
B(JumpIfNull), U8(50),
B(TestTypeOf), U8(6),
B(JumpIfTrue), U8(18),
B(Wide), B(LdaSmi), I16(155),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(14),
B(LdaConstant), U8(8),
B(Star), R(15),
B(CallRuntime), U16(Runtime::kNewTypeError), R(14), U8(2),
B(Throw),
B(Mov), R(context), R(14),
B(CallProperty0), R(13), R(4), U8(20),
B(JumpIfJSReceiver), U8(21),
B(Star), R(15),
B(CallRuntime), U16(Runtime::kThrowIteratorResultNotAnObject), R(15), U8(1),
B(Jump), U8(12),
B(Star), R(14),
B(LdaZero),
B(TestReferenceEqual), R(8),
B(JumpIfTrue), U8(5),
B(Ldar), R(14),
B(ReThrow),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Ldar), R(10),
B(SetPendingMessage),
B(Ldar), R(8),
B(SwitchOnSmiNoFeedback), U8(9), U8(2), I8(0),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(Jump), U8(19),
B(Ldar), R(9),
B(ReThrow),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaFalse),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(Star), R(15),
B(Mov), R(0), R(13),
B(Mov), R(9), R(14),
[async] Introduce dedicated JSAsyncFunctionObject. This JSAsyncFunctionObject represents the implicit generator object inside of async functions, and also holds the outer promise for the async functions. This in turn allows us to get rid of the .promise in the Parser / BytecodeGenerator completely, and will make it possible to build zero-cost async stack traces independent of the concrete synchronous part of the stack frame (which currently breaks in Node.js). In the bytecode all the async function operations now take this new JSAsyncFunctionObject instead of passing both the .generator_object and the .promise, which further simplifies and shrinks the bytecode. It also reduces the size of async function frames, potentially making the suspend/resume cheaper. This also changes `await` to use intrinsics instead of calling to special JSFunctions on the native context, and thus reduces the size of the native contexts. Drive-by-fix: Introduce a dedicated JSCreateAsyncFunctionObject operator to TurboFan. Bug: v8:7253, v8:7522 Change-Id: I2305302285156aa1f71328ecac70377abdd92c80 Ref: nodejs/node#11865 Design-Document: http://bit.ly/v8-zero-cost-async-stack-traces Reviewed-on: https://chromium-review.googlesource.com/c/1273049 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#56554}
2018-10-11 08:35:56 +00:00
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionResolve), R(13), U8(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
/* 96 S> */ B(Return),
B(LdaUndefined),
B(Star), R(4),
B(LdaFalse),
B(Star), R(5),
B(Mov), R(0), R(3),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionResolve), R(3), U8(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
/* 96 S> */ B(Return),
B(Star), R(3),
B(CreateCatchContext), R(3), U8(11),
B(Star), R(2),
B(LdaTheHole),
B(SetPendingMessage),
B(Ldar), R(2),
B(PushContext), R(3),
[async] Improve async function handling. This change introduces new intrinsics used to desugar async functions in the Parser and the BytecodeGenerator, namely we introduce a new %_AsyncFunctionEnter intrinsic that constructs the generator object for the async function (and in the future will also create the outer promise for the async function). This generator object is internal and never escapes to user code, plus since async functions don't have a "prototype" property, we can just a single map here instead of tracking the prototype/initial_map on every async function. This saves one word per async function plus one initial_map per async function that was invoked at least once. We also introduce two new intrinsics %_AsyncFunctionReject, which rejects the outer promise with the caught exception, and another %_AsyncFunctionResolve, which resolves the outer promise with the right hand side of the `return` statement. These functions also perform the DevTools part of the job (aka popping from the promise stack and sending the debug event). This allows us to get rid of the implicit try-finally from async functions completely; because the finally block only called to the %AsyncFunctionPromiseRelease builtin, which was used to inform DevTools. In essence we now turn an async function like ```js async function f(x) { return await bar(x); } ``` into something like this (in Parser and BytecodeGenerator respectively): ``` function f(x) { .generator_object = %_AsyncFunctionEnter(.closure, this); .promise = %AsyncFunctionCreatePromise(); try { .tmp = await bar(x); return %_AsyncFunctionResolve(.promise, .tmp); } catch (e) { return %_AsyncFunctionReject(.promise, e); } } ``` Overall the bytecode for async functions gets significantly shorter already (and will get even shorter once we put the outer promise into the async function generator object). For example the bytecode for a simple async function ```js async function f(x) { return await x; } ``` goes from 175 bytes to 110 bytes (a ~38% reduction in size), which is in particular due to the simplification around the try-finally removal. Overall this seems to improve the doxbee-async-es2017-native test by around 2-3%. On the test case mentioned in v8:8276 we go from 1124ms to 441ms, which corresponds to a 60% reduction in total execution time! Tbr: marja@chromium.org Bug: v8:7253, v8:7522, v8:8276 Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4 Reviewed-on: https://chromium-review.googlesource.com/c/1269041 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Maya Lekova <mslekova@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Cr-Commit-Position: refs/heads/master@{#56502}
2018-10-10 05:54:39 +00:00
B(LdaImmutableCurrentContextSlot), U8(4),
B(Star), R(5),
B(LdaFalse),
B(Star), R(6),
B(Mov), R(0), R(4),
B(InvokeIntrinsic), U8(Runtime::k_AsyncFunctionReject), R(4), U8(3),
/* 96 S> */ B(Return),
]
constant pool: [
OBJECT_BOILERPLATE_DESCRIPTION_TYPE,
ARRAY_BOILERPLATE_DESCRIPTION_TYPE,
SYMBOL_TYPE,
ONE_BYTE_INTERNALIZED_STRING_TYPE ["next"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["done"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["value"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["a"],
ONE_BYTE_INTERNALIZED_STRING_TYPE ["return"],
[turbofan] Significantly improve ConsString creation performance. This change significantly improves the performance of string concatenation in optimized code for the case where the resulting string is represented as a ConsString. On the relevant test cases we go from serializeNaive: 10762 ms. serializeClever: 7813 ms. serializeConcat: 10271 ms. to serializeNaive: 10278 ms. serializeClever: 5533 ms. serializeConcat: 10310 ms. which represents a 30% improvement on the "clever" benchmark, which tests specifically the ConsString creation performance. This was accomplished via a couple of different steps, which are briefly outlined here: 1. The empty_string gets its own map, so that we can easily recognize and handle it appropriately in the TurboFan type system. This allows us to express (and assert) that the inputs to NewConsString are non-empty strings, making sure that TurboFan no longer creates "crippled ConsStrings" with empty left or right hand sides. 2. Further split the existing String types in TurboFan to be able to distinguish between OneByte and TwoByte strings on the type system level. This allows us to avoid having to dynamically lookup the resulting ConsString map in case of ConsString creation (i.e. when we know that both input strings are OneByte strings or at least one of the input strings is TwoByte). 3. We also introduced more finegrained feedback for the Add bytecode in the interpreter, having it collect feedback about ConsStrings, specifically ConsOneByteString and ConsTwoByteString. This feedback can be used by TurboFan to only inline the relevant code for what was seen so far. This allows us to remove the Octane/Splay specific magic in JSTypedLowering to detect ConsString creation, and instead purely rely on the feedback of what was seen so far (also making it possible to change the semantics of NewConsString to be a low-level operator, which is only introduced in SimplifiedLowering by looking at the input types of StringConcat). 4. On top of the before mentioned type and interpreter changes we added new operators CheckNonEmptyString, CheckNonEmptyOneByteString, and CheckNonEmptyTwoByteString, which perform the appropriate (dynamic) checks. There are several more improvements that are possible based on this, but since the change was already quite big, we decided not to put everything into the first change, but do some follow up tweaks to the type system, and builtin optimizations later. Tbr: mstarzinger@chromium.org Bug: v8:8834, v8:8931, v8:8939, v8:8951 Change-Id: Ia24e17c6048bf2b04df966d3cd441f0edda05c93 Cq-Include-Trybots: luci.chromium.try:linux-blink-rel Doc: https://bit.ly/fast-string-concatenation-in-javascript Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1499497 Commit-Queue: Michael Achenbach <machenbach@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org> Cr-Commit-Position: refs/heads/master@{#60318}
2019-03-18 19:33:46 +00:00
EMPTY_STRING_TYPE [""],
Smi [6],
Smi [9],
SCOPE_INFO_TYPE,
]
handlers: [
Reland "[ignition] Skip binding dead labels" This is a reland of 35269f77f8a7fb5360717e2cb470a6ef1637944e Switches on an expression that unconditionally throws would have all their case statements dead, causing a DCHECK error in the SwitchBuilder. This fixes up the DCHECK to allow dead labels. Original change's description: > [ignition] Skip binding dead labels > > BytecodeLabels for forward jumps may create a dead basic block if their > corresponding jump was elided (due to it dead code elimination). We can > avoid generating such dead basic blocks by skipping the label bind when > no corresponding jump has been observed. This works because all jumps > except JumpLoop are forward jumps, so we only have to special case one > Bind for loop headers to bind unconditionally. > > Since Binds are now conditional on a jump existing, we can no longer rely > on using Bind to get the current offset (e.g. at the beginning of a try > block). Instead, we now expose the current offset in the bytecode array > writer. Conveniently, this means that we can be a bit smarter about basic > blocks around these statements. > > As a drive-by, remove the unused Bind(target,label) function. > > Bug: chromium:934166 > Change-Id: I532aa452fb083560d07b90da99caca0b1d082aa3 > Reviewed-on: https://chromium-review.googlesource.com/c/1488763 > Commit-Queue: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > Cr-Commit-Position: refs/heads/master@{#59942} TBR=rmcilroy@chromium.org Bug: chromium:934166 Change-Id: If6eab4162106717ce64a2dc477000c6a76354cb4 Reviewed-on: https://chromium-review.googlesource.com/c/1494535 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Commit-Queue: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#59948}
2019-02-28 13:34:26 +00:00
[16, 233, 233],
[59, 112, 120],
[166, 179, 181],
]