Commit ad55fa63 authored by Kong, Fanchen's avatar Kong, Fanchen Committed by Commit Bot

[wasm-simd] [liftoff] Implement div/sqrt on x64 and ia32

Bug: v8:9909
Change-Id: I5afad8ef0c5bf2ccad61d71fcc0c89db48851e24
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2147148
Commit-Queue: Fanchen Kong <fanchen.kong@intel.com>
Reviewed-by: 's avatarZhi An Ng <zhin@chromium.org>
Reviewed-by: 's avatarClemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67146}
parent 9dc37ed0
......@@ -280,6 +280,8 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
AVX_OP2_WITH_TYPE(Movd, movd, Register, XMMRegister)
AVX_OP2_WITH_TYPE(Movd, movd, Operand, XMMRegister)
AVX_OP2_WITH_TYPE(Cvtdq2ps, cvtdq2ps, XMMRegister, Operand)
AVX_OP2_WITH_TYPE(Sqrtps, sqrtps, XMMRegister, XMMRegister)
AVX_OP2_WITH_TYPE(Sqrtpd, sqrtpd, XMMRegister, XMMRegister)
AVX_OP2_WITH_TYPE(Sqrtpd, sqrtpd, XMMRegister, const Operand&)
AVX_OP2_WITH_TYPE(Movaps, movaps, XMMRegister, XMMRegister)
AVX_OP2_WITH_TYPE(Movapd, movapd, XMMRegister, XMMRegister)
......
......@@ -1660,6 +1660,11 @@ void LiftoffAssembler::emit_f64x2_neg(LiftoffRegister dst,
vneg(dst.high_fp(), src.high_fp());
}
void LiftoffAssembler::emit_f64x2_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
bailout(kSimd, "f64x2sqrt");
}
void LiftoffAssembler::emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
vadd(dst.low_fp(), lhs.low_fp(), rhs.low_fp());
......@@ -1678,6 +1683,11 @@ void LiftoffAssembler::emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
vmul(dst.high_fp(), lhs.high_fp(), rhs.high_fp());
}
void LiftoffAssembler::emit_f64x2_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
bailout(kSimd, "f64x2div");
}
void LiftoffAssembler::emit_f32x4_splat(LiftoffRegister dst,
LiftoffRegister src) {
vdup(Neon32, liftoff::GetSimd128Register(dst), src.fp(), 0);
......@@ -1709,6 +1719,11 @@ void LiftoffAssembler::emit_f32x4_neg(LiftoffRegister dst,
vneg(liftoff::GetSimd128Register(dst), liftoff::GetSimd128Register(src));
}
void LiftoffAssembler::emit_f32x4_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
bailout(kSimd, "f32x4sqrt");
}
void LiftoffAssembler::emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
vadd(liftoff::GetSimd128Register(dst), liftoff::GetSimd128Register(lhs),
......@@ -1727,6 +1742,11 @@ void LiftoffAssembler::emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
liftoff::GetSimd128Register(rhs));
}
void LiftoffAssembler::emit_f32x4_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
bailout(kSimd, "f32x4div");
}
void LiftoffAssembler::emit_i64x2_splat(LiftoffRegister dst,
LiftoffRegister src) {
Simd128Register dst_simd = liftoff::GetSimd128Register(dst);
......
......@@ -1115,6 +1115,11 @@ void LiftoffAssembler::emit_f64x2_neg(LiftoffRegister dst,
Fneg(dst.fp().V2D(), src.fp().V2D());
}
void LiftoffAssembler::emit_f64x2_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
bailout(kSimd, "f64x2sqrt");
}
void LiftoffAssembler::emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
Fadd(dst.fp().V2D(), lhs.fp().V2D(), rhs.fp().V2D());
......@@ -1130,6 +1135,11 @@ void LiftoffAssembler::emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
Fmul(dst.fp().V2D(), lhs.fp().V2D(), rhs.fp().V2D());
}
void LiftoffAssembler::emit_f64x2_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
bailout(kSimd, "f64x2div");
}
void LiftoffAssembler::emit_f32x4_splat(LiftoffRegister dst,
LiftoffRegister src) {
Dup(dst.fp().V4S(), src.fp().S(), 0);
......@@ -1161,6 +1171,11 @@ void LiftoffAssembler::emit_f32x4_neg(LiftoffRegister dst,
Fneg(dst.fp().V4S(), src.fp().V4S());
}
void LiftoffAssembler::emit_f32x4_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
bailout(kSimd, "f32x4sqrt");
}
void LiftoffAssembler::emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
Fadd(dst.fp().V4S(), lhs.fp().V4S(), rhs.fp().V4S());
......@@ -1176,6 +1191,11 @@ void LiftoffAssembler::emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
Fmul(dst.fp().V4S(), lhs.fp().V4S(), rhs.fp().V4S());
}
void LiftoffAssembler::emit_f32x4_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
bailout(kSimd, "f32x4div");
}
void LiftoffAssembler::emit_i64x2_splat(LiftoffRegister dst,
LiftoffRegister src) {
Dup(dst.fp().V2D(), src.gp().X());
......
......@@ -1950,13 +1950,11 @@ void EmitSimdCommutativeBinOp(
template <void (Assembler::*avx_op)(XMMRegister, XMMRegister, XMMRegister),
void (Assembler::*sse_op)(XMMRegister, XMMRegister)>
void EmitSimdSub(LiftoffAssembler* assm, LiftoffRegister dst,
void EmitSimdNonCommutativeBinOp(LiftoffAssembler* assm, LiftoffRegister dst,
LiftoffRegister lhs, LiftoffRegister rhs) {
if (CpuFeatures::IsSupported(AVX)) {
CpuFeatureScope scope(assm, AVX);
(assm->*avx_op)(dst.fp(), lhs.fp(), rhs.fp());
} else if (lhs.fp() == rhs.fp()) {
assm->pxor(dst.fp(), dst.fp());
} else if (dst.fp() == rhs.fp()) {
assm->movaps(kScratchDoubleReg, rhs.fp());
assm->movaps(dst.fp(), lhs.fp());
......@@ -2076,22 +2074,23 @@ void LiftoffAssembler::emit_i8x16_add_saturate_u(LiftoffRegister dst,
void LiftoffAssembler::emit_i8x16_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubb, &Assembler::psubb>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubb, &Assembler::psubb>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i8x16_sub_saturate_s(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubsb, &Assembler::psubsb>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubsb, &Assembler::psubsb>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i8x16_sub_saturate_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubusb, &Assembler::psubusb>(this, dst,
lhs, rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubusb,
&Assembler::psubusb>(this, dst, lhs,
rhs);
}
void LiftoffAssembler::emit_i8x16_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2217,22 +2216,23 @@ void LiftoffAssembler::emit_i16x8_add_saturate_u(LiftoffRegister dst,
void LiftoffAssembler::emit_i16x8_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubw, &Assembler::psubw>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubw, &Assembler::psubw>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i16x8_sub_saturate_s(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubsw, &Assembler::psubsw>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubsw, &Assembler::psubsw>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i16x8_sub_saturate_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubusw, &Assembler::psubusw>(this, dst,
lhs, rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubusw,
&Assembler::psubusw>(this, dst, lhs,
rhs);
}
void LiftoffAssembler::emit_i16x8_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2288,8 +2288,8 @@ void LiftoffAssembler::emit_i32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_i32x4_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubd, &Assembler::psubd>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubd, &Assembler::psubd>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2348,8 +2348,8 @@ void LiftoffAssembler::emit_i64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_i64x2_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubq, &Assembler::psubq>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubq, &Assembler::psubq>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2405,6 +2405,11 @@ void LiftoffAssembler::emit_f32x4_neg(LiftoffRegister dst,
}
}
void LiftoffAssembler::emit_f32x4_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
Sqrtps(dst.fp(), src.fp());
}
void LiftoffAssembler::emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdCommutativeBinOp<&Assembler::vaddps, &Assembler::addps>(
......@@ -2413,8 +2418,8 @@ void LiftoffAssembler::emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_f32x4_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vsubps, &Assembler::subps>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vsubps, &Assembler::subps>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2423,6 +2428,12 @@ void LiftoffAssembler::emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f32x4_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vdivps, &Assembler::divps>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f64x2_abs(LiftoffRegister dst,
LiftoffRegister src) {
if (dst.fp() == src.fp()) {
......@@ -2449,6 +2460,11 @@ void LiftoffAssembler::emit_f64x2_neg(LiftoffRegister dst,
}
}
void LiftoffAssembler::emit_f64x2_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
Sqrtpd(dst.fp(), src.fp());
}
void LiftoffAssembler::emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdCommutativeBinOp<&Assembler::vaddpd, &Assembler::addpd>(
......@@ -2457,8 +2473,8 @@ void LiftoffAssembler::emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_f64x2_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vsubpd, &Assembler::subpd>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vsubpd, &Assembler::subpd>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2467,6 +2483,12 @@ void LiftoffAssembler::emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f64x2_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vdivpd, &Assembler::divpd>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
......
......@@ -817,20 +817,26 @@ class LiftoffAssembler : public TurboAssembler {
LiftoffRegister rhs);
inline void emit_f32x4_abs(LiftoffRegister dst, LiftoffRegister src);
inline void emit_f32x4_neg(LiftoffRegister dst, LiftoffRegister src);
inline void emit_f32x4_sqrt(LiftoffRegister dst, LiftoffRegister src);
inline void emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f32x4_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f32x4_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f64x2_abs(LiftoffRegister dst, LiftoffRegister src);
inline void emit_f64x2_neg(LiftoffRegister dst, LiftoffRegister src);
inline void emit_f64x2_sqrt(LiftoffRegister dst, LiftoffRegister src);
inline void emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f64x2_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_f64x2_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs);
inline void emit_i8x16_rounding_average_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs);
......
......@@ -2438,22 +2438,30 @@ class LiftoffCompiler {
return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_abs);
case wasm::kExprF32x4Neg:
return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_neg);
case wasm::kExprF32x4Sqrt:
return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_sqrt);
case wasm::kExprF32x4Add:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_add);
case wasm::kExprF32x4Sub:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_sub);
case wasm::kExprF32x4Mul:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_mul);
case wasm::kExprF32x4Div:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_div);
case wasm::kExprF64x2Abs:
return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_abs);
case wasm::kExprF64x2Neg:
return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_neg);
case wasm::kExprF64x2Sqrt:
return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_sqrt);
case wasm::kExprF64x2Add:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_add);
case wasm::kExprF64x2Sub:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_sub);
case wasm::kExprF64x2Mul:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_mul);
case wasm::kExprF64x2Div:
return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_div);
case wasm::kExprI8x16RoundingAverageU:
return EmitBinOp<kS128, kS128>(
&LiftoffAssembler::emit_i8x16_rounding_average_u);
......
......@@ -1880,13 +1880,11 @@ void EmitSimdCommutativeBinOp(
template <void (Assembler::*avx_op)(XMMRegister, XMMRegister, XMMRegister),
void (Assembler::*sse_op)(XMMRegister, XMMRegister)>
void EmitSimdSub(LiftoffAssembler* assm, LiftoffRegister dst,
void EmitSimdNonCommutativeBinOp(LiftoffAssembler* assm, LiftoffRegister dst,
LiftoffRegister lhs, LiftoffRegister rhs) {
if (CpuFeatures::IsSupported(AVX)) {
CpuFeatureScope scope(assm, AVX);
(assm->*avx_op)(dst.fp(), lhs.fp(), rhs.fp());
} else if (lhs.fp() == rhs.fp()) {
assm->pxor(dst.fp(), dst.fp());
} else if (dst.fp() == rhs.fp()) {
assm->movaps(kScratchDoubleReg, rhs.fp());
assm->movaps(dst.fp(), lhs.fp());
......@@ -2000,22 +1998,23 @@ void LiftoffAssembler::emit_i8x16_add_saturate_u(LiftoffRegister dst,
void LiftoffAssembler::emit_i8x16_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubb, &Assembler::psubb>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubb, &Assembler::psubb>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i8x16_sub_saturate_s(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubsb, &Assembler::psubsb>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubsb, &Assembler::psubsb>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i8x16_sub_saturate_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubusb, &Assembler::psubusb>(this, dst,
lhs, rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubusb,
&Assembler::psubusb>(this, dst, lhs,
rhs);
}
void LiftoffAssembler::emit_i8x16_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2141,22 +2140,23 @@ void LiftoffAssembler::emit_i16x8_add_saturate_u(LiftoffRegister dst,
void LiftoffAssembler::emit_i16x8_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubw, &Assembler::psubw>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubw, &Assembler::psubw>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i16x8_sub_saturate_s(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubsw, &Assembler::psubsw>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubsw, &Assembler::psubsw>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i16x8_sub_saturate_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubusw, &Assembler::psubusw>(this, dst,
lhs, rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubusw,
&Assembler::psubusw>(this, dst, lhs,
rhs);
}
void LiftoffAssembler::emit_i16x8_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2212,8 +2212,8 @@ void LiftoffAssembler::emit_i32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_i32x4_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubd, &Assembler::psubd>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubd, &Assembler::psubd>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2271,8 +2271,8 @@ void LiftoffAssembler::emit_i64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_i64x2_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vpsubq, &Assembler::psubq>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpsubq, &Assembler::psubq>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2328,6 +2328,11 @@ void LiftoffAssembler::emit_f32x4_neg(LiftoffRegister dst,
}
}
void LiftoffAssembler::emit_f32x4_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
Sqrtps(dst.fp(), src.fp());
}
void LiftoffAssembler::emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdCommutativeBinOp<&Assembler::vaddps, &Assembler::addps>(
......@@ -2336,8 +2341,8 @@ void LiftoffAssembler::emit_f32x4_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_f32x4_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vsubps, &Assembler::subps>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vsubps, &Assembler::subps>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2346,6 +2351,12 @@ void LiftoffAssembler::emit_f32x4_mul(LiftoffRegister dst, LiftoffRegister lhs,
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f32x4_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vdivps, &Assembler::divps>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f64x2_abs(LiftoffRegister dst,
LiftoffRegister src) {
if (dst.fp() == src.fp()) {
......@@ -2372,6 +2383,11 @@ void LiftoffAssembler::emit_f64x2_neg(LiftoffRegister dst,
}
}
void LiftoffAssembler::emit_f64x2_sqrt(LiftoffRegister dst,
LiftoffRegister src) {
Sqrtpd(dst.fp(), src.fp());
}
void LiftoffAssembler::emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdCommutativeBinOp<&Assembler::vaddpd, &Assembler::addpd>(
......@@ -2380,8 +2396,8 @@ void LiftoffAssembler::emit_f64x2_add(LiftoffRegister dst, LiftoffRegister lhs,
void LiftoffAssembler::emit_f64x2_sub(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdSub<&Assembler::vsubpd, &Assembler::subpd>(this, dst, lhs,
rhs);
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vsubpd, &Assembler::subpd>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
......@@ -2390,6 +2406,12 @@ void LiftoffAssembler::emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_f64x2_div(LiftoffRegister dst, LiftoffRegister lhs,
LiftoffRegister rhs) {
liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vdivpd, &Assembler::divpd>(
this, dst, lhs, rhs);
}
void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
LiftoffRegister lhs,
LiftoffRegister rhs) {
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment