Enable JIT and FTL by default on the Windows Port

[iangrunert]

847f97e

Enable JIT and FTL by default on the Windows Port
https://bugs.webkit.org/show_bug.cgi?id=278878

Reviewed by Yusuke Suzuki.

CodeBlock is 8 bytes larger on Windows - upping the static_assert.

Bumping the stack size for the main process from the Windows default 1MB
up to 8MB to match Mac and Linux, which fixes a stack overflow in test
JSTests/stress/object-create-will-do-transition.js

get-array-length-concurrently-change-mode.js adds an explicit time check
as the watchdog isn't firing. Likely a bug / problem in RunLoopWin if
this works on other platforms.

Making PrintStream printInternal add the 0x prefix to pointers on
Windows, fixing dfg-ai-direct-get-by-id-attribute-change-transition.js

There are still some test failures in JSC stress tests, but with the
exception of wasm/gc/array_new_fixed_long.js the remaining failures are
unrelated to JIT.

* JSTests/stress/get-array-length-concurrently-change-mode.js:
(main):
* Source/JavaScriptCore/bytecode/CodeBlock.h:
* Source/WTF/wtf/PrintStream.cpp:
(WTF::printInternal):
* Source/cmake/OptionsMSVC.cmake:
* Source/cmake/WebKitFeatures.cmake:

Canonical link: https://commits.webkit.org/284574@main

8265788

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🧪 wpe-wk2	✅ 🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 api-wpe
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🛠 wpe-cairo
✅ 🛠 🧪 jsc	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
✅ 🛠 🧪 jsc-arm64	✅ 🛠 vision	✅ 🧪 mac-AS-debug-wk2	✅ 🧪 gtk-wk2
	✅ 🛠 vision-sim	✅ 🧪 mac-wk2-stress	✅ 🧪 api-gtk
	✅ 🧪 vision-wk2	✅ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
✅ 🛠 🧪 unsafe-merge	✅ 🛠 tv		✅ 🧪 jsc-armv7-tests
	✅ 🛠 tv-sim
	✅ 🛠 watch
	✅ 🛠 watch-sim

[webkit-early-warning-system]

EWS run on previous version of this PR (hash f3d58df)

Details

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	⏳ 🛠 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🧪 wpe-wk2	⏳ 🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 api-wpe
	✅ 🧪 ios-wk2-wpt		✅ 🛠 wpe-cairo
✅ 🛠 🧪 jsc	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
✅ 🛠 🧪 jsc-arm64	✅ 🛠 vision	✅ 🧪 mac-AS-debug-wk2	✅ 🧪 gtk-wk2
	✅ 🛠 vision-sim	✅ 🧪 mac-wk2-stress	✅ 🧪 api-gtk
	✅ 🧪 vision-wk2	✅ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv		✅ 🧪 jsc-armv7-tests
	✅ 🛠 tv-sim
	✅ 🛠 watch
	✅ 🛠 watch-sim

[Constellation]

This change looks strange as using return does not make it tail-call (since it is not a strict mode code).

[iangrunert]

Yep - this test was failing because Window's main thread stack size was only 1MB so it'd overflow the call stack well and truly before the other platforms.

Not 100% sure why adding the return caused it to not stack overflow, I'm surprised that would've changed the number of call frames or size of the call frames.

[Constellation]

Can you fix the Windows' output instead of changing the test?

[iangrunert]

Yep - I still have to address this, I didn't notice I hadn't fully reverted my changes to this test file before pushing.

[fujii]

This code was added by me to disable JIT for Windows.
#28269
Can we just remove this code?

[webkit-early-warning-system]

EWS run on previous version of this PR (hash 472cabf)

Details

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	🛠 wpe	🛠 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	🧪 wpe-wk2	🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	🧪 api-wpe
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	🛠 wpe-cairo
🛠 🧪 jsc	⏳ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
✅ 🛠 🧪 jsc-arm64	🛠 vision	⏳ 🧪 mac-AS-debug-wk2	⏳ 🧪 gtk-wk2
	🛠 vision-sim	✅ 🧪 mac-wk2-stress	🧪 api-gtk
	⏳ 🧪 vision-wk2	⏳ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv		⏳ 🧪 jsc-armv7-tests
	🛠 tv-sim
	🛠 watch
	✅ 🛠 watch-sim

[webkit-early-warning-system]

EWS run on current version of this PR (hash 8265788)

Details

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
✅ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 win
	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🧪 wpe-wk2	✅ 🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 api-wpe
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🛠 wpe-cairo
✅ 🛠 🧪 jsc	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
✅ 🛠 🧪 jsc-arm64	✅ 🛠 vision	✅ 🧪 mac-AS-debug-wk2	✅ 🧪 gtk-wk2
	✅ 🛠 vision-sim	✅ 🧪 mac-wk2-stress	✅ 🧪 api-gtk
	✅ 🧪 vision-wk2	✅ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
✅ 🛠 🧪 unsafe-merge	✅ 🛠 tv		✅ 🧪 jsc-armv7-tests
	✅ 🛠 tv-sim
	✅ 🛠 watch
	✅ 🛠 watch-sim

[fujii]

Why does Windows port need such a big 8MB stack? Other port doesn't seem to use such a big stack.

WebKit/Source/WTF/wtf/Threading.cpp

Lines 108 to 141 in 4d33c93

	static std::optional<size_t> stackSize(ThreadType threadType)
	{
	// Return the stack size for the created thread based on its type.
	// If the stack size is not specified, then use the system default. Platforms can tune the values here.
	// Enable STACK_STATS in StackStats.h to create a build that will track the information for tuning.
	#if PLATFORM(PLAYSTATION)
	if (threadType == ThreadType::JavaScript)
	return 512 * KB;
	#elif OS(DARWIN) && (ASAN_ENABLED || ASSERT_ENABLED)
	if (threadType == ThreadType::Compiler)
	return 1 * MB; // ASan / Debug build needs more stack space.
	#elif OS(WINDOWS)
	// WebGL conformance tests need more stack space <https://webkit.org/b/261297>
	if (threadType == ThreadType::Graphics)
	#if defined(NDEBUG)
	return 2 * MB;
	#else
	return 4 * MB;
	#endif
	#else
	UNUSED_PARAM(threadType);
	#endif
	#if defined(DEFAULT_THREAD_STACK_SIZE_IN_KB) && DEFAULT_THREAD_STACK_SIZE_IN_KB > 0
	return DEFAULT_THREAD_STACK_SIZE_IN_KB * 1024;
	#elif OS(LINUX) && !defined(__BIONIC__) && !defined(__GLIBC__)
	// on libcs other than glibc and bionic (e.g. musl) we are either unsure how big
	// the default thread stack is, or we know it's too small - pick a robust default
	return 1 * MB;
	#else
	// Use the platform's default stack size
	return std::nullopt;
	#endif
	}

[iangrunert]

We need a larger stack for the main thread - the JSTests/stress/object-create-will-do-transition.js test was overflowing the call stack. Mac and Linux both have 8MB stack size for the main thread, but with a 1MB stack size on Windows we're overflowing before we get close to the 1.5MB maxPerThreadStackUsage set for jsc stress tests.

We could update Threading.cpp#stackSize to keep a 1MB stack for other threads, we only need the increased stack size on the main thread.

[webkit-commit-queue]

Committed 284574@main (847f97e): https://commits.webkit.org/284574@main

Reviewed commits have been landed. Closing PR #32972 and removing active labels.