Skip to content

fix(jsc): detect usable AVX on Darwin x64 #36

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
steipete merged 1 commit into from
Oct 25, 2026
Merged

fix(jsc): detect usable AVX on Darwin x64 #36

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 1 addition & 37 deletions Source/JavaScriptCore/assembler/MacroAssemblerX86_64.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,18 +32,12 @@
#include "ProbeContext.h"
#include <wtf/InlineASM.h>

#if OS(DARWIN)
#include <sys/sysctl.h>
#endif

namespace JSC {

JSC_DECLARE_NOEXCEPT_JIT_OPERATION(ctiMasmProbeTrampoline, void, ());
JSC_ANNOTATE_JIT_OPERATION_PROBE(ctiMasmProbeTrampoline);
#if !OS(DARWIN)
JSC_DECLARE_NOEXCEPT_JIT_OPERATION(ctiMasmProbeTrampolineAVX, void, ());
JSC_ANNOTATE_JIT_OPERATION_PROBE(ctiMasmProbeTrampolineAVX);
#endif

// The following are offsets for Probe::State fields accessed by the ctiMasmProbeTrampoline stub.

Expand Down Expand Up @@ -193,7 +187,6 @@ static_assert(sizeof(Probe::State) == PROBE_SIZE, "Probe::State::size's matches
"vmovaps " STRINGIZE_VALUE_OF(PROBE_CPU_XMM14_VECTOR_OFFSET) "(%rbp), %xmm14" "\n" \
"vmovaps " STRINGIZE_VALUE_OF(PROBE_CPU_XMM15_VECTOR_OFFSET) "(%rbp), %xmm15" "\n"

#if !OS(DARWIN)
#define PROBE_XMM_SAVE_MOVAPS \
"movaps %xmm0, " STRINGIZE_VALUE_OF(PROBE_CPU_XMM0_VECTOR_OFFSET) "(%rsp)" "\n" \
"movaps %xmm1, " STRINGIZE_VALUE_OF(PROBE_CPU_XMM1_VECTOR_OFFSET) "(%rsp)" "\n" \
Expand Down Expand Up @@ -229,7 +222,6 @@ static_assert(sizeof(Probe::State) == PROBE_SIZE, "Probe::State::size's matches
"movaps " STRINGIZE_VALUE_OF(PROBE_CPU_XMM13_VECTOR_OFFSET) "(%rbp), %xmm13" "\n" \
"movaps " STRINGIZE_VALUE_OF(PROBE_CPU_XMM14_VECTOR_OFFSET) "(%rbp), %xmm14" "\n" \
"movaps " STRINGIZE_VALUE_OF(PROBE_CPU_XMM15_VECTOR_OFFSET) "(%rbp), %xmm15" "\n"
#endif // !OS(DARWIN)

#if COMPILER(MSVC)
#define ASM_PREVIOUS_SECTION
Expand Down Expand Up @@ -404,19 +396,13 @@ __asm__( \
ASM_PREVIOUS_SECTION \
);

#if OS(DARWIN)
// On macOS, all x86_64 CPUs support AVX. Use vmovaps unconditionally.
DEFINE_PROBE_TRAMPOLINE(ctiMasmProbeTrampoline, PROBE_XMM_SAVE_VMOVAPS, PROBE_XMM_RESTORE_VMOVAPS,
ctiMasmProbeTrampolineCopyLoop, ctiMasmProbeTrampolineProbeStateIsSafe, ctiMasmProbeTrampolineRestoreRegisters)
#else
// On Linux/Windows, AVX may not be available. Provide two trampolines:
// AVX may not be available, including under Rosetta. Provide two trampolines:
// - ctiMasmProbeTrampoline: uses movaps (SSE, always available on x86_64)
// - ctiMasmProbeTrampolineAVX: uses vmovaps (requires AVX)
DEFINE_PROBE_TRAMPOLINE(ctiMasmProbeTrampoline, PROBE_XMM_SAVE_MOVAPS, PROBE_XMM_RESTORE_MOVAPS,
ctiMasmProbeTrampolineCopyLoop, ctiMasmProbeTrampolineProbeStateIsSafe, ctiMasmProbeTrampolineRestoreRegisters)
DEFINE_PROBE_TRAMPOLINE(ctiMasmProbeTrampolineAVX, PROBE_XMM_SAVE_VMOVAPS, PROBE_XMM_RESTORE_VMOVAPS,
ctiMasmProbeTrampolineAVXCopyLoop, ctiMasmProbeTrampolineAVXProbeStateIsSafe, ctiMasmProbeTrampolineAVXRestoreRegisters)
#endif

// What code is emitted for the probe?
// ==================================
Expand Down Expand Up @@ -464,14 +450,10 @@ void MacroAssembler::probe(Probe::Function function, void* arg)
push(RegisterID::eax);
push(RegisterID::eax);

#if OS(DARWIN)
move(TrustedImmPtr(reinterpret_cast<void*>(ctiMasmProbeTrampoline)), RegisterID::eax);
#else
if (supportsAVX())
move(TrustedImmPtr(reinterpret_cast<void*>(ctiMasmProbeTrampolineAVX)), RegisterID::eax);
else
move(TrustedImmPtr(reinterpret_cast<void*>(ctiMasmProbeTrampoline)), RegisterID::eax);
#endif

push(RegisterID::edx);
move(TrustedImmPtr(reinterpret_cast<void*>(function)), RegisterID::edx);
Expand All @@ -496,14 +478,12 @@ MacroAssemblerX86_64::CPUID MacroAssemblerX86_64::getCPUIDEx(unsigned level, uns
return result;
}

#if !OS(DARWIN)
static inline uint64_t readXCR0()
{
uint32_t lo, hi;
__asm__("xgetbv" : "=a"(lo), "=d"(hi) : "c"(0));
return static_cast<uint64_t>(lo) | (static_cast<uint64_t>(hi) << 32);
}
#endif

void MacroAssemblerX86_64::collectCPUFeatures()
{
Expand All @@ -516,9 +496,6 @@ void MacroAssemblerX86_64::collectCPUFeatures()
s_sse4_1CheckState = (cpuid[2] & (1 << 19)) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;
s_sse4_2CheckState = (cpuid[2] & (1 << 20)) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;
s_popcntCheckState = (cpuid[2] & (1 << 23)) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;
#if OS(DARWIN)
s_avxCheckState = CPUIDCheckState::Set;
#else
// Per the Intel SDM, AVX is usable only when:
// 1. CPUID.1:ECX.AVX[bit 28] = 1 (CPU supports AVX)
// 2. CPUID.1:ECX.OSXSAVE[bit 27] = 1 (OS has enabled XGETBV)
Expand All @@ -533,19 +510,7 @@ void MacroAssemblerX86_64::collectCPUFeatures()
if ((readXCR0() & ymmStateMask) == ymmStateMask)
s_avxCheckState = CPUIDCheckState::Set;
}
#endif
}
#if OS(DARWIN)
{
uint32_t val = 0;
size_t valSize = sizeof(val);
int rc = sysctlbyname("hw.optional.bmi1", &val, &valSize, nullptr, 0);
s_bmi1CheckState = (rc >= 0 && val) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;

rc = sysctlbyname("hw.optional.avx2_0", &val, &valSize, nullptr, 0);
s_avx2CheckState = (rc >= 0 && val) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;
}
#else
{
// CPUID leaf 7 sub-leaf 0: BMI1 = EBX[3], AVX2 = EBX[5].
CPUID cpuid = getCPUID(0x7);
Expand All @@ -555,7 +520,6 @@ void MacroAssemblerX86_64::collectCPUFeatures()
bool hasAVX2Feature = cpuid[1] & (1 << 5);
s_avx2CheckState = (hasAVX2Feature && s_avxCheckState == CPUIDCheckState::Set) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;
}
#endif
{
CPUID cpuid = getCPUID(0x80000001);
s_lzcntCheckState = (cpuid[2] & (1 << 5)) ? CPUIDCheckState::Set : CPUIDCheckState::Clear;
Expand Down