blob: ec189f9763e7caef0e3807bea757b023a214fd74 [file] [log] [blame]
# Copyright (C) 2011-2021 Apple Inc. All rights reserved.
# Copyright (C) 2021 Igalia S.L. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
# 1. Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# 2. Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
#
# THIS SOFTWARE IS PROVIDED BY APPLE INC. AND ITS CONTRIBUTORS ``AS IS''
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
# THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR ITS CONTRIBUTORS
# BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
# THE POSSIBILITY OF SUCH DAMAGE.
# Constants
const MswOffset = 4
const LswOffset = 0
# Opcodes that should eventually be shared with JS llint
wasmOp(mov, WasmMov, macro(ctx)
mload2i(ctx, m_src, t1, t0)
return2i(ctx, t1, t0)
end)
# Wasm specific bytecodes
macro emitCheckAndPreparePointerAddingOffset(pointer, offset, size)
# This macro updates 'pointer' to target address, and may thrash 'offset'
if ARMv7
# Not enough registers on arm to keep the memory base and size in pinned
# registers, so load them on each access instead. FIXME: improve this.
addps offset, pointer
bcs .throw
addps size - 1, pointer, offset # Use offset as scratch register
bcs .throw
bpb offset, Wasm::Instance::m_cachedBoundsCheckingSize[wasmInstance], .continuation
.throw:
throwException(OutOfBoundsMemoryAccess)
.continuation:
addp Wasm::Instance::m_cachedMemory[wasmInstance], pointer
else
crash()
end
end
macro emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(pointer, offset, size)
# This macro updates 'pointer' to target address, and may thrash 'offset'
if ARMv7
# Not enough registers on arm to keep the memory base and size in pinned
# registers, so load them on each access instead. FIXME: improve this.
addps offset, pointer
bcs .throw
addps size - 1, pointer, offset # Use offset as scratch register
bcs .throw
bpb offset, Wasm::Instance::m_cachedBoundsCheckingSize[wasmInstance], .continuation
.throw:
throwException(OutOfBoundsMemoryAccess)
.continuation:
addp Wasm::Instance::m_cachedMemory[wasmInstance], pointer
btpnz pointer, (size - 1), .throw
else
crash()
end
end
macro wasmLoadOp(name, struct, size, fn)
wasmOp(name, struct, macro(ctx)
mloadi(ctx, m_pointer, t0)
wgetu(ctx, m_offset, t1)
emitCheckAndPreparePointerAddingOffset(t0, t1, size)
fn(t0, t3, t2)
return2i(ctx, t3, t2)
end)
end
wasmLoadOp(load8_u, WasmLoad8U, 1, macro(addr, dstMsw, dstLsw)
loadb LswOffset[addr], dstLsw
move 0, dstMsw
end)
wasmLoadOp(load16_u, WasmLoad16U, 2, macro(addr, dstMsw, dstLsw)
loadh LswOffset[addr], dstLsw
move 0, dstMsw
end)
wasmLoadOp(load32_u, WasmLoad32U, 4, macro(addr, dstMsw, dstLsw)
loadi LswOffset[addr], dstLsw
move 0, dstMsw
end)
wasmLoadOp(load64_u, WasmLoad64U, 8, macro(addr, dstMsw, dstLsw)
# Might be unaligned, so can't use load2ia
loadi LswOffset[addr], dstLsw
loadi MswOffset[addr], dstMsw
end)
wasmLoadOp(i32_load8_s, WasmI32Load8S, 1, macro(addr, dstMsw, dstLsw)
loadbsi LswOffset[addr], dstLsw
move 0, dstMsw
end)
wasmLoadOp(i32_load16_s, WasmI32Load16S, 2, macro(addr, dstMsw, dstLsw)
loadhsi LswOffset[addr], dstLsw
move 0, dstMsw
end)
wasmLoadOp(i64_load8_s, WasmI64Load8S, 1, macro(addr, dstMsw, dstLsw)
loadbsi LswOffset[addr], dstLsw
rshifti dstLsw, 31, dstMsw
end)
wasmLoadOp(i64_load16_s, WasmI64Load16S, 2, macro(addr, dstMsw, dstLsw)
loadhsi LswOffset[addr], dstLsw
rshifti dstLsw, 31, dstMsw
end)
wasmLoadOp(i64_load32_s, WasmI64Load32S, 4, macro(addr, dstMsw, dstLsw)
loadi LswOffset[addr], dstLsw
rshifti dstLsw, 31, dstMsw
end)
macro wasmStoreOp(name, struct, size, fn)
wasmOp(name, struct, macro(ctx)
mloadi(ctx, m_pointer, t0)
wgetu(ctx, m_offset, t1)
emitCheckAndPreparePointerAddingOffset(t0, t1, size)
mload2i(ctx, m_value, t3, t2)
fn(t3, t2, t0)
dispatch(ctx)
end)
end
wasmStoreOp(store8, WasmStore8, 1, macro(srcMsw, srcLsw, addr)
storeb srcLsw, LswOffset[addr]
end)
wasmStoreOp(store16, WasmStore16, 2, macro(srcMsw, srcLsw, addr)
storeh srcLsw, LswOffset[addr]
end)
wasmStoreOp(store32, WasmStore32, 4, macro(srcMsw, srcLsw, addr)
storei srcLsw, LswOffset[addr]
end)
wasmStoreOp(store64, WasmStore64, 8, macro(srcMsw, srcLsw, addr)
# Might be unaligned, so can't use store2ia
storei srcLsw, LswOffset[addr]
storei srcMsw, MswOffset[addr]
end)
wasmOp(ref_is_null, WasmRefIsNull, macro(ctx)
mload2i(ctx, m_ref, t1, t0) # t0 is unused
cieq t1, NullTag, t0
returni(ctx, t0)
end)
wasmOp(get_global, WasmGetGlobal, macro(ctx)
loadp Wasm::Instance::m_globals[wasmInstance], t0
wgetu(ctx, m_globalIndex, t1)
load2ia [t0, t1, 8], t0, t1
return2i(ctx, t1, t0)
end)
wasmOp(set_global, WasmSetGlobal, macro(ctx)
loadp Wasm::Instance::m_globals[wasmInstance], t0
wgetu(ctx, m_globalIndex, t1)
mload2i(ctx, m_value, t3, t2)
store2ia t2, t3, [t0, t1, 8]
dispatch(ctx)
end)
wasmOp(get_global_portable_binding, WasmGetGlobalPortableBinding, macro(ctx)
loadp Wasm::Instance::m_globals[wasmInstance], t0
wgetu(ctx, m_globalIndex, t1)
loadp [t0, t1, 8], t0
load2ia [t0], t0, t1
return2i(ctx, t1, t0)
end)
wasmOp(set_global_portable_binding, WasmSetGlobalPortableBinding, macro(ctx)
loadp Wasm::Instance::m_globals[wasmInstance], t0
wgetu(ctx, m_globalIndex, t1)
mload2i(ctx, m_value, t3, t2)
loadp [t0, t1, 8], t0
store2ia t2, t3, [t0]
dispatch(ctx)
end)
# Opcodes that don't have the `b3op` entry in wasm.json. This should be kept in sync
macro callDivRem(fn)
if ARMv7
subp StackAlignment, sp
storep PC, [sp]
cCall4(fn)
loadp [sp], PC
addp StackAlignment, sp
else
error
end
end
# i32 binary ops
wasmOp(i32_div_s, WasmI32DivS, macro (ctx)
mloadi(ctx, m_lhs, a0)
mloadi(ctx, m_rhs, a1)
btiz a1, .throwDivisionByZero
bineq a1, -1, .safe
bieq a0, constexpr INT32_MIN, .throwIntegerOverflow
.safe:
callDivRem(_slow_path_wasm_i32_div_s)
returni(ctx, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
.throwIntegerOverflow:
throwException(IntegerOverflow)
end)
wasmOp(i32_div_u, WasmI32DivU, macro (ctx)
mloadi(ctx, m_lhs, a0)
mloadi(ctx, m_rhs, a1)
btiz a1, .throwDivisionByZero
callDivRem(_slow_path_wasm_i32_div_u)
returni(ctx, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
end)
wasmOp(i32_rem_s, WasmI32RemS, macro (ctx)
mloadi(ctx, m_lhs, a0)
mloadi(ctx, m_rhs, a1)
btiz a1, .throwDivisionByZero
bineq a1, -1, .safe
bineq a0, constexpr INT32_MIN, .safe
move 0, r0
jmp .return
.safe:
callDivRem(_slow_path_wasm_i32_rem_s)
.return:
returni(ctx, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
end)
wasmOp(i32_rem_u, WasmI32RemU, macro (ctx)
mloadi(ctx, m_lhs, a0)
mloadi(ctx, m_rhs, a1)
btiz a1, .throwDivisionByZero
callDivRem(_slow_path_wasm_i32_rem_u)
returni(ctx, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
end)
# i64 binary ops
wasmOp(i64_add, WasmI64Add, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
addis t2, t0
adci t3, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_sub, WasmI64Sub, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
subis t2, t0
sbci t3, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_mul, WasmI64Mul, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
muli t2, t1
muli t0, t3
umulli t0, t2, t0, t2
addi t1, t2
addi t3, t2
return2i(ctx, t2, t0)
end)
wasmOp(i64_div_s, WasmI64DivS, macro (ctx)
mload2i(ctx, m_lhs, a1, a0)
mload2i(ctx, m_rhs, a3, a2)
btinz a3, .nonZeroDivisor
btiz a2, .throwDivisionByZero
.nonZeroDivisor:
bineq a3, -1, .safe
bineq a2, -1, .safe
bineq a1, constexpr INT32_MIN, .safe
btiz a0, .throwIntegerOverflow
.safe:
callDivRem(_slow_path_wasm_i64_div_s)
return2i(ctx, r1, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
.throwIntegerOverflow:
throwException(IntegerOverflow)
end)
wasmOp(i64_div_u, WasmI64DivU, macro (ctx)
mload2i(ctx, m_lhs, a1, a0)
mload2i(ctx, m_rhs, a3, a2)
btinz a3, .nonZeroDivisor
btiz a2, .throwDivisionByZero
.nonZeroDivisor:
callDivRem(_slow_path_wasm_i64_div_u)
return2i(ctx, r1, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
end)
wasmOp(i64_rem_s, WasmI64RemS, macro (ctx)
mload2i(ctx, m_lhs, a1, a0)
mload2i(ctx, m_rhs, a3, a2)
btinz a3, .nonZeroDivisor
btiz a2, .throwDivisionByZero
.nonZeroDivisor:
bineq a3, -1, .safe
bineq a2, -1, .safe
bineq a1, constexpr INT32_MIN, .safe
btinz a0, .safe
move 0, r1
move 0, r0
jmp .return
.safe:
callDivRem(_slow_path_wasm_i64_rem_s)
.return:
return2i(ctx, r1, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
end)
wasmOp(i64_rem_u, WasmI64RemU, macro (ctx)
mload2i(ctx, m_lhs, a1, a0)
mload2i(ctx, m_rhs, a3, a2)
btinz a3, .nonZeroDivisor
btiz a2, .throwDivisionByZero
.nonZeroDivisor:
callDivRem(_slow_path_wasm_i64_rem_u)
return2i(ctx, r1, r0)
.throwDivisionByZero:
throwException(DivisionByZero)
end)
wasmOp(i64_and, WasmI64And, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
andi t2, t0
andi t3, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_or, WasmI64Or, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
ori t2, t0
ori t3, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_xor, WasmI64Xor, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
xori t2, t0
xori t3, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_shl, WasmI64Shl, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mloadi(ctx, m_rhs, t2)
andi 0x3f, t2
btiz t2, .return
bib t2, 32, .lessThan32
subi 32, t2
lshifti t0, t2, t1
move 0, t0
jmp .return
.lessThan32:
lshifti t2, t1
move 32, t3
subi t2, t3
urshifti t0, t3, t3
ori t3, t1
lshifti t2, t0
.return:
return2i(ctx, t1, t0)
end)
wasmOp(i64_shr_u, WasmI64ShrU, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mloadi(ctx, m_rhs, t2)
andi 0x3f, t2
btiz t2, .return
bib t2, 32, .lessThan32
subi 32, t2
urshifti t1, t2, t0
move 0, t1
jmp .return
.lessThan32:
urshifti t2, t0
move 32, t3
subi t2, t3
lshifti t1, t3, t3
ori t3, t0
urshifti t2, t1
.return:
return2i(ctx, t1, t0)
end)
wasmOp(i64_shr_s, WasmI64ShrS, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mloadi(ctx, m_rhs, t2)
andi 0x3f, t2
btiz t2, .return
bib t2, 32, .lessThan32
subi 32, t2
rshifti t1, t2, t0
rshifti 31, t1
jmp .return
.lessThan32:
urshifti t2, t0
move 32, t3
subi t2, t3
lshifti t1, t3, t3
ori t3, t0
rshifti t2, t1
.return:
return2i(ctx, t1, t0)
end)
wasmOp(i64_rotr, WasmI64Rotr, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mloadi(ctx, m_rhs, t2)
andi t2, 0x20, t3
btiz t3, .noSwap
move t0, t3
move t1, t0
move t3, t1
.noSwap:
andi 0x1f, t2
btiz t2, .return
move 32, t5
subi t2, t5
lshifti t0, t5, t3
lshifti t1, t5, t5
urshifti t2, t0
urshifti t2, t1
ori t5, t0
ori t3, t1
.return:
return2i(ctx, t1, t0)
end)
wasmOp(i64_rotl, WasmI64Rotl, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mloadi(ctx, m_rhs, t2)
andi t2, 0x20, t3
btiz t3, .noSwap
move t0, t3
move t1, t0
move t3, t1
.noSwap:
andi 0x1f, t2
btiz t2, .return
move 32, t5
subi t2, t5
urshifti t0, t5, t3
urshifti t1, t5, t5
lshifti t2, t0
lshifti t2, t1
ori t5, t0
ori t3, t1
.return:
return2i(ctx, t1, t0)
end)
wasmOp(i64_eq, WasmI64Eq, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 0, t6
bineq t1, t3, .return
cieq t0, t2, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_ne, WasmI64Ne, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bineq t1, t3, .return
cineq t0, t2, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_lt_s, WasmI64LtS, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bilt t1, t3, .return
move 0, t6
bigt t1, t3, .return
cib t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_le_s, WasmI64LeS, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bilt t1, t3, .return
move 0, t6
bigt t1, t3, .return
cibeq t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_lt_u, WasmI64LtU, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bib t1, t3, .return
move 0, t6
bia t1, t3, .return
cib t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_le_u, WasmI64LeU, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bib t1, t3, .return
move 0, t6
bia t1, t3, .return
cibeq t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_gt_s, WasmI64GtS, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bigt t1, t3, .return
move 0, t6
bilt t1, t3, .return
cia t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_ge_s, WasmI64GeS, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bigt t1, t3, .return
move 0, t6
bilt t1, t3, .return
ciaeq t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_gt_u, WasmI64GtU, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 1, t6
bia t1, t3, .return
move 0, t6
bib t1, t3, .return
cia t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
wasmOp(i64_ge_u, WasmI64GeU, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2)
move 0, t6
bib t1, t3, .return
move 1, t6
bia t1, t3, .return
ciaeq t0, t2, t6
andi 1, t6
.return:
returni(ctx, t6)
end)
# i64 unary ops
wasmOp(i64_ctz, WasmI64Ctz, macro (ctx)
mload2i(ctx, m_operand, t1, t0)
btiz t0, .top
tzcnti t0, t0
jmp .return
.top:
tzcnti t1, t0
addi 32, t0
.return:
return2i(ctx, 0, t0)
end)
wasmOp(i64_clz, WasmI64Clz, macro(ctx)
mload2i(ctx, m_operand, t1, t0)
btiz t1, .bottom
lzcnti t1, t0
jmp .return
.bottom:
lzcnti t0, t0
addi 32, t0
.return:
return2i(ctx, 0, t0)
end)
wasmOp(i64_popcnt, WasmI64Popcnt, macro (ctx)
mload2i(ctx, m_operand, a3, a2)
prepareStateForCCall()
move PC, a0
cCall2(_slow_path_wasm_popcountll)
restoreStateAfterCCall()
return2i(ctx, 0, r1)
end)
wasmOp(i64_eqz, WasmI64Eqz, macro(ctx)
mload2i(ctx, m_operand, t1, t0)
move 0, t2
btinz t1, .return
cieq t0, 0, t2
.return:
returni(ctx, t2)
end)
# f64 binary ops
wasmOp(f64_copysign, WasmF64Copysign, macro(ctx)
mload2i(ctx, m_lhs, t1, t0)
mload2i(ctx, m_rhs, t3, t2) # t2 unused
andi 0x7fffffff, t1
andi 0x80000000, t3
ori t3, t1
return2i(ctx, t1, t0)
end)
# Type conversion ops
wasmOp(i64_extend_s_i32, WasmI64ExtendSI32, macro(ctx)
mloadi(ctx, m_operand, t0)
rshifti t0, 31, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_extend_u_i32, WasmI64ExtendUI32, macro(ctx)
mloadi(ctx, m_operand, t0)
return2i(ctx, 0, t0)
end)
wasmOp(f64_reinterpret_i64, WasmF64ReinterpretI64, macro(ctx)
mload2i(ctx, m_operand, t1, t0)
fii2d t0, t1, ft0
returnd(ctx, ft0)
end)
wasmOp(i64_reinterpret_f64, WasmI64ReinterpretF64, macro(ctx)
mloadd(ctx, m_operand, ft0)
fd2ii ft0, t0, t1
return2i(ctx, t1, t0)
end)
wasmOp(i32_trunc_s_f64, WasmI32TruncSF64, macro (ctx)
mloadd(ctx, m_operand, ft0)
moveii 0xc1e0000000200000, t1, t0 # INT32_MIN - 1.0
fii2d t0, t1, ft1
bdltequn ft0, ft1, .outOfBoundsTrunc
moveii 0x41e0000000000000, t1, t0 # -INT32_MIN
fii2d t0, t1, ft1
bdgtequn ft0, ft1, .outOfBoundsTrunc
truncated2is ft0, t0
returni(ctx, t0)
.outOfBoundsTrunc:
throwException(OutOfBoundsTrunc)
end)
wasmOp(i32_trunc_u_f64, WasmI32TruncUF64, macro (ctx)
mloadd(ctx, m_operand, ft0)
moveii 0xbff0000000000000, t1, t0 # -1.0
fii2d t0, t1, ft1
bdltequn ft0, ft1, .outOfBoundsTrunc
moveii 0x41f0000000000000, t1, t0 # INT32_MIN * -2.0
fii2d t0, t1, ft1
bdgtequn ft0, ft1, .outOfBoundsTrunc
truncated2i ft0, t0
returni(ctx, t0)
.outOfBoundsTrunc:
throwException(OutOfBoundsTrunc)
end)
slowWasmOp(i64_trunc_s_f32)
slowWasmOp(i64_trunc_u_f32)
slowWasmOp(i64_trunc_s_f64)
slowWasmOp(i64_trunc_u_f64)
wasmOp(i32_trunc_sat_f64_s, WasmI32TruncSatF64S, macro (ctx)
mloadd(ctx, m_operand, ft0)
moveii 0xc1e0000000200000, t1, t0 # INT32_MIN - 1.0
fii2d t0, t1, ft1
bdltequn ft0, ft1, .outOfBoundsTruncSatMinOrNaN
moveii 0x41e0000000000000, t1, t0 # -INT32_MIN
fii2d t0, t1, ft1
bdgtequn ft0, ft1, .outOfBoundsTruncSatMax
truncated2is ft0, t0
returni(ctx, t0)
.outOfBoundsTruncSatMinOrNaN:
bdeq ft0, ft0, .outOfBoundsTruncSatMin
move 0, t0
returni(ctx, t0)
.outOfBoundsTruncSatMax:
move (constexpr INT32_MAX), t0
returni(ctx, t0)
.outOfBoundsTruncSatMin:
move (constexpr INT32_MIN), t0
returni(ctx, t0)
end)
wasmOp(i32_trunc_sat_f64_u, WasmI32TruncSatF64U, macro (ctx)
mloadd(ctx, m_operand, ft0)
moveii 0xbff0000000000000, t1, t0 # -1.0
fii2d t0, t1, ft1
bdltequn ft0, ft1, .outOfBoundsTruncSatMin
moveii 0x41f0000000000000, t1, t0 # INT32_MIN * -2.0
fii2d t0, t1, ft1
bdgtequn ft0, ft1, .outOfBoundsTruncSatMax
truncated2i ft0, t0
returni(ctx, t0)
.outOfBoundsTruncSatMin:
move 0, t0
returni(ctx, t0)
.outOfBoundsTruncSatMax:
move (constexpr UINT32_MAX), t0
returni(ctx, t0)
end)
slowWasmOp(i64_trunc_sat_f32_s)
slowWasmOp(i64_trunc_sat_f32_u)
slowWasmOp(i64_trunc_sat_f64_s)
slowWasmOp(i64_trunc_sat_f64_u)
# Extend ops
wasmOp(i64_extend8_s, WasmI64Extend8S, macro(ctx)
mloadi(ctx, m_operand, t0)
sxb2i t0, t0
rshifti t0, 31, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_extend16_s, WasmI64Extend16S, macro(ctx)
mloadi(ctx, m_operand, t0)
sxh2i t0, t0
rshifti t0, 31, t1
return2i(ctx, t1, t0)
end)
wasmOp(i64_extend32_s, WasmI64Extend16S, macro(ctx)
mloadi(ctx, m_operand, t0)
rshifti t0, 31, t1
return2i(ctx, t1, t0)
end)
# Atomics
macro wasmAtomicBinaryRMWOps(lowerCaseOpcode, upperCaseOpcode, fnb, fnh, fni, fn2i)
wasmOp(i64_atomic_rmw8%lowerCaseOpcode%_u, WasmI64AtomicRmw8%upperCaseOpcode%U, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mloadi(ctx, m_value, t0)
emitCheckAndPreparePointerAddingOffset(t3, t5, 1)
fnb(t0, [t3], t2, t5, t6)
andi 0xff, t6 # FIXME: ZeroExtend8To64
assert(macro(ok) bibeq t6, 0xff, .ok end)
return2i(ctx, 0, t6)
end)
wasmOp(i64_atomic_rmw16%lowerCaseOpcode%_u, WasmI64AtomicRmw16%upperCaseOpcode%U, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mloadi(ctx, m_value, t0)
emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(t3, t5, 2)
fnh(t0, [t3], t2, t5, t6)
andi 0xffff, t6 # FIXME: ZeroExtend16To64
assert(macro(ok) bibeq t6, 0xffff, .ok end)
return2i(ctx, 0, t6)
end)
wasmOp(i64_atomic_rmw32%lowerCaseOpcode%_u, WasmI64AtomicRmw32%upperCaseOpcode%U, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mloadi(ctx, m_value, t0)
emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(t3, t5, 4)
fni(t0, [t3], t2, t5, t6)
return2i(ctx, 0, t6)
end)
wasmOp(i64_atomic_rmw%lowerCaseOpcode%, WasmI64AtomicRmw%upperCaseOpcode%, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mload2i(ctx, m_value, t1, t0)
emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(t3, t5, 8)
push t4 # This is the PB, so need to be saved and restored
fn2i(t1, t0, [t3], t4, t5, t7, t2, t6)
pop t4
return2i(ctx, t2, t6)
end)
end
macro wasmAtomicBinaryRMWOpsWithWeakCAS(lowerCaseOpcode, upperCaseOpcode, fni, fn2i)
wasmAtomicBinaryRMWOps(lowerCaseOpcode, upperCaseOpcode,
macro(t0GPR, mem, t2GPR, t5GPR, t6GPR)
fence
.loop:
loadlinkb mem, t6GPR
fni(t0GPR, t6GPR, t2GPR)
storecondb t5GPR, t2GPR, mem
bineq t5GPR, 0, .loop
fence
end,
macro(t0GPR, mem, t2GPR, t5GPR, t6GPR)
fence
.loop:
loadlinkh mem, t6GPR
fni(t0GPR, t6GPR, t2GPR)
storecondh t5GPR, t2GPR, mem
bineq t5GPR, 0, .loop
fence
end,
macro(t0GPR, mem, t2GPR, t5GPR, t6GPR)
fence
.loop:
loadlinki mem, t6GPR
fni(t0GPR, t6GPR, t2GPR)
storecondi t5GPR, t2GPR, mem
bineq t5GPR, 0, .loop
fence
end,
macro(t1GPR, t0GPR, mem, t4GPR, t5GPR, t7GPR, t2GPR, t6GPR)
fence
.loop:
loadlink2i mem, t6GPR, t2GPR
fn2i(t1GPR, t0GPR, t2GPR, t6GPR, t4GPR, t5GPR)
storecond2i t7GPR, t5GPR, t4GPR, mem
bineq t7GPR, 0, .loop
fence
end)
end
if ARMv7
wasmAtomicBinaryRMWOpsWithWeakCAS(_add, Add,
macro(src0, src1, dst)
addi src0, src1, dst
end,
macro(src0Msw, src0Lsw, src1Msw, src1Lsw, dstMsw, dstLsw)
addis src0Lsw, src1Lsw, dstLsw
adci src0Msw, src1Msw, dstMsw
end)
wasmAtomicBinaryRMWOpsWithWeakCAS(_sub, Sub,
macro(src0, src1, dst)
subi src1, src0, dst
end,
macro(src0Msw, src0Lsw, src1Msw, src1Lsw, dstMsw, dstLsw)
subis src1Lsw, src0Lsw, dstLsw
sbci src1Msw, src0Msw, dstMsw
end)
wasmAtomicBinaryRMWOpsWithWeakCAS(_xchg, Xchg,
macro(src0, src1, dst)
move src0, dst
end,
macro(src0Msw, src0Lsw, src1Msw, src1Lsw, dstMsw, dstLsw)
move src0Lsw, dstLsw
move src0Msw, dstMsw
end)
wasmAtomicBinaryRMWOpsWithWeakCAS(_and, And,
macro(src0, src1, dst)
andi src0, src1, dst
end,
macro(src0Msw, src0Lsw, src1Msw, src1Lsw, dstMsw, dstLsw)
andi src0Lsw, src1Lsw, dstLsw
andi src0Msw, src1Msw, dstMsw
end)
wasmAtomicBinaryRMWOpsWithWeakCAS(_or, Or,
macro(src0, src1, dst)
ori src0, src1, dst
end,
macro(src0Msw, src0Lsw, src1Msw, src1Lsw, dstMsw, dstLsw)
ori src0Lsw, src1Lsw, dstLsw
ori src0Msw, src1Msw, dstMsw
end)
wasmAtomicBinaryRMWOpsWithWeakCAS(_xor, Xor,
macro(src0, src1, dst)
xori src0, src1, dst
end,
macro(src0Msw, src0Lsw, src1Msw, src1Lsw, dstMsw, dstLsw)
xori src0Lsw, src1Lsw, dstLsw
xori src0Msw, src1Msw, dstMsw
end)
end
macro wasmAtomicCompareExchangeOps(lowerCaseOpcode, upperCaseOpcode, fnb, fnh, fni, fn2i)
wasmOp(i64_atomic_rmw8%lowerCaseOpcode%_u, WasmI64AtomicRmw8%upperCaseOpcode%U, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mload2i(ctx, m_expected, t1, t0)
mload2i(ctx, m_value, t6, t2)
emitCheckAndPreparePointerAddingOffset(t3, t5, 1)
fnb(t1, t0, t6, t2, [t3], t7, t5)
andi 0xff, t0 # FIXME: ZeroExtend8To64
assert(macro(ok) bibeq t0, 0xff, .ok end)
assert(macro(ok) bieq t1, 0, .ok end)
return2i(ctx, t1, t0)
end)
wasmOp(i64_atomic_rmw16%lowerCaseOpcode%_u, WasmI64AtomicRmw16%upperCaseOpcode%U, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mload2i(ctx, m_expected, t1, t0)
mload2i(ctx, m_value, t6, t2)
emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(t3, t5, 2)
fnh(t1, t0, t6, t2, [t3], t7, t5)
andi 0xffff, t0 # FIXME: ZeroExtend16To64
assert(macro(ok) bibeq t0, 0xffff, .ok end)
assert(macro(ok) bieq t1, 0, .ok end)
return2i(ctx, t1, t0)
end)
wasmOp(i64_atomic_rmw32%lowerCaseOpcode%_u, WasmI64AtomicRmw32%upperCaseOpcode%U, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mload2i(ctx, m_expected, t1, t0)
mload2i(ctx, m_value, t6, t2)
emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(t3, t5, 8)
fni(t1, t0, t6, t2, [t3], t7, t5)
assert(macro(ok) bieq t1, 0, .ok end)
return2i(ctx, t1, t0)
end)
wasmOp(i64_atomic_rmw%lowerCaseOpcode%, WasmI64AtomicRmw%upperCaseOpcode%, macro(ctx)
mloadi(ctx, m_pointer, t3)
wgetu(ctx, m_offset, t5)
mload2i(ctx, m_expected, t1, t0)
mload2i(ctx, m_value, t6, t2)
emitCheckAndPreparePointerAddingOffsetWithAlignmentCheck(t3, t5, 8)
push t4 # This is the PB, so need to be saved and restored
fn2i(t1, t0, t6, t2, [t3], t7, t5, t4)
pop t4
return2i(ctx, t1, t0)
end)
end
if ARMv7
// exp: "expected", val: "value", res: "result"
wasmAtomicCompareExchangeOps(_cmpxchg, Cmpxchg,
macro(expMsw, expLsw, valMsw, valLsw, mem, scratch, resLsw)
fence
.loop:
loadlinkb mem, resLsw
bineq expLsw, resLsw, .fail
bineq expMsw, 0, .fail
storecondb scratch, valLsw, mem
bieq scratch, 0, .done
jmp .loop
.fail:
storecondb scratch, resLsw, mem
bieq scratch, 0, .done
jmp .loop
.done:
fence
move resLsw, expLsw
move 0, expMsw
end,
macro(expMsw, expLsw, valMsw, valLsw, mem, scratch, resLsw)
fence
.loop:
loadlinkh mem, resLsw
bineq expLsw, resLsw, .fail
bineq expMsw, 0, .fail
storecondh scratch, valLsw, mem
bieq scratch, 0, .done
jmp .loop
.fail:
storecondh scratch, resLsw, mem
bieq scratch, 0, .done
jmp .loop
.done:
fence
move resLsw, expLsw
move 0, expMsw
end,
macro(expMsw, expLsw, valMsw, valLsw, mem, scratch, resLsw)
fence
.loop:
loadlinki mem, resLsw
bineq expLsw, resLsw, .fail
bineq expMsw, 0, .fail
storecondi scratch, valLsw, mem
bieq scratch, 0, .done
jmp .loop
.fail:
storecondi scratch, resLsw, mem
bieq scratch, 0, .done
jmp .loop
.done:
fence
move resLsw, expLsw
move 0, expMsw
end,
macro(expMsw, expLsw, valMsw, valLsw, mem, scratch, resMsw, resLsw)
fence
.loop:
loadlink2i mem, resLsw, resMsw
bineq expLsw, resLsw, .fail
bineq expMsw, resMsw, .fail
storecond2i scratch, valLsw, valMsw, mem
bieq scratch, 0, .done
jmp .loop
.fail:
storecond2i scratch, resLsw, resMsw, mem
bieq scratch, 0, .done
jmp .loop
.done:
fence
move resLsw, expLsw
move resMsw, expMsw
end)
end