feat(asm): add EVEX masking, zeroing and the AVX-512 F/BW integer set
Assisted-by: Qwen 3.8 Max Preview
This commit is contained in:
+93
-1
@@ -65,6 +65,23 @@ func TestEvexGroundTruth(t *testing.T) {
|
||||
{"VMOVDQU64 (SI)(R15*4),Z3", "VMOVDQU64", []Operand{Idx(SI, vreg(t, "R15"), 4, 0, 64), vreg(t, "Z3")}, "62b1fe486f1cbe"},
|
||||
{"VMOVDQU64 Z0,4(SI)(AX*1)", "VMOVDQU64", []Operand{vreg(t, "Z0"), Idx(SI, AX, 1, 4, 64)}, "62f1fe487f840604000000"},
|
||||
{"VMOVDQU64 Z1,Z2", "VMOVDQU64", []Operand{vreg(t, "Z1"), vreg(t, "Z2")}, "62f1fe487fca"},
|
||||
// The wider AVX-512 F/BW integer set.
|
||||
{"VPADDB Z1,Z2,Z3", "VPADDB", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f16d48fcd9"},
|
||||
{"VPSUBW Z1,Z2,Z3", "VPSUBW", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f16d48f9d9"},
|
||||
{"VPANDQ Z1,Z2,Z3", "VPANDQ", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f1ed48dbd9"},
|
||||
{"VPANDND Z1,Z2,Z3", "VPANDND", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f16d48dfd9"},
|
||||
{"VPMULLW Z1,Z2,Z3", "VPMULLW", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f16d48d5d9"},
|
||||
{"VPMINUB Z1,Z2,Z3", "VPMINUB", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f16d48dad9"},
|
||||
{"VPMAXUQ Z1,Z2,Z3", "VPMAXUQ", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f2ed483fd9"},
|
||||
{"VPAVGW Z1,Z2,Z3", "VPAVGW", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f16d48e3d9"},
|
||||
{"VPSLLVQ Z3,Z1,Z2", "VPSLLVQ", []Operand{vreg(t, "Z3"), vreg(t, "Z1"), vreg(t, "Z2")}, "62f2f54847d3"},
|
||||
{"VPSRAVQ Z3,Z1,Z2", "VPSRAVQ", []Operand{vreg(t, "Z3"), vreg(t, "Z1"), vreg(t, "Z2")}, "62f2f54846d3"},
|
||||
{"VPSHUFD $0x1B,Z1,Z2", "VPSHUFD", []Operand{Imm(0x1B), vreg(t, "Z1"), vreg(t, "Z2")}, "62f17d4870d11b"},
|
||||
{"VPSHUFB Z1,Z2,Z3", "VPSHUFB", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}, "62f26d4800d9"},
|
||||
{"VMOVDQU8 Z1,Z2", "VMOVDQU8", []Operand{vreg(t, "Z1"), vreg(t, "Z2")}, "62f17f487fca"},
|
||||
{"VMOVDQU16 Z1,Z2", "VMOVDQU16", []Operand{vreg(t, "Z1"), vreg(t, "Z2")}, "62f1ff487fca"},
|
||||
// Indices 16–31: rm[4] rides in X̄ for register operands.
|
||||
{"VPSHUFD $1,X16,X17", "VPSHUFD", []Operand{Imm(1), vreg(t, "X16"), vreg(t, "X17")}, "62a17d0870c801"},
|
||||
{"VMOVUPD (DI),Z14", "VMOVUPD", []Operand{Ptr(DI, 0, 64), vreg(t, "Z14")}, "6271fd481037"},
|
||||
{"VMOVUPD 64(DI),Z14", "VMOVUPD", []Operand{Ptr(DI, 64, 64), vreg(t, "Z14")}, "6271fd48107701"},
|
||||
// Conversions and narrowing stores (reg = wide source).
|
||||
@@ -110,6 +127,81 @@ func TestEvexGroundTruth(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
// TestEvexMasking checks the AVX-512 mask operand (K1–K7, placed freely among
|
||||
// the operands) and the .Z zeroing suffix, byte for byte against the Go
|
||||
// assembler.
|
||||
func TestEvexMasking(t *testing.T) {
|
||||
cases := []struct {
|
||||
name string
|
||||
mnem string
|
||||
ops []Operand
|
||||
want string
|
||||
}{
|
||||
// Masked arithmetic: K anywhere among the operands; .Z sets the z bit.
|
||||
{"VPADDD.Z merging+zeroing", "VPADDD.Z", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K2"), vreg(t, "Z3")}, "62f16dcafed9"},
|
||||
{"VPADDD merging", "VPADDD", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K1"), vreg(t, "Z3")}, "62f16d49fed9"},
|
||||
{"VADDPD.Z", "VADDPD.Z", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K2"), vreg(t, "Z3")}, "62f1edca58d9"},
|
||||
{"VPMINSD.Z", "VPMINSD.Z", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K5"), vreg(t, "Z3")}, "62f26dcd39d9"},
|
||||
{"VPMINSQ.Z", "VPMINSQ.Z", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K5"), vreg(t, "Z3")}, "62f2edcd39d9"},
|
||||
// Masked immediate shift (K before the destination).
|
||||
{"VPSRAD.Z", "VPSRAD.Z", []Operand{Imm(1), vreg(t, "Z2"), vreg(t, "K1"), vreg(t, "Z3")}, "62f165c972e201"},
|
||||
{"VPSLLD merge", "VPSLLD", []Operand{Imm(4), vreg(t, "Z1"), vreg(t, "K2"), vreg(t, "Z3")}, "62f1654a72f104"},
|
||||
// Masked align.
|
||||
{"VALIGND", "VALIGND", []Operand{Imm(12), vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K3"), vreg(t, "Z4")}, "62f36d4b03e10c"},
|
||||
// Masked conversion and extract.
|
||||
{"VCVTQQ2PD.Z", "VCVTQQ2PD.Z", []Operand{vreg(t, "Z1"), vreg(t, "K2"), vreg(t, "Z3")}, "62f1fecae6d9"},
|
||||
{"VEXTRACTI64X4", "VEXTRACTI64X4", []Operand{Imm(1), vreg(t, "Z1"), vreg(t, "K2"), vreg(t, "Y3")}, "62f3fd4a3bcb01"},
|
||||
// Masked moves: K sits between the register and memory operands.
|
||||
{"VMOVDQU8 store", "VMOVDQU8", []Operand{vreg(t, "Z1"), vreg(t, "K3"), Ptr(SI, 0, 64)}, "62f17f4b7f0e"},
|
||||
{"VMOVDQU32 load", "VMOVDQU32", []Operand{Ptr(SI, 0, 64), vreg(t, "K4"), vreg(t, "Z1")}, "62f17e4c6f0e"},
|
||||
{"VMOVDQU32 store", "VMOVDQU32", []Operand{vreg(t, "Z1"), vreg(t, "K4"), Ptr(DI, 0, 64)}, "62f17e4c7f0f"},
|
||||
// Masked comparison with a K destination: dst K1, mask K2.
|
||||
{"VPCMPEQD k-dst+mask", "VPCMPEQD", []Operand{vreg(t, "Z0"), vreg(t, "Z3"), vreg(t, "K2"), vreg(t, "K1")}, "62f1654a76c8"},
|
||||
}
|
||||
for _, c := range cases {
|
||||
code, err := Encode(c.mnem, c.ops...)
|
||||
if err != nil {
|
||||
t.Errorf("%s: Encode: %v", c.name, err)
|
||||
continue
|
||||
}
|
||||
if got := hexCompact(code); got != c.want {
|
||||
t.Errorf("%s: bytes %s, want %s", c.name, got, c.want)
|
||||
continue
|
||||
}
|
||||
inst, err := x86asm.Decode(code, 64)
|
||||
if err != nil {
|
||||
t.Errorf("%s: Decode(%x): %v", c.name, code, err)
|
||||
continue
|
||||
}
|
||||
want := c.mnem
|
||||
if i := len(want) - 2; i > 0 && want[i:] == ".Z" {
|
||||
want = want[:i]
|
||||
}
|
||||
if inst.Op.String() != want {
|
||||
t.Errorf("%s: decoded as %s", c.name, inst.Op.String())
|
||||
}
|
||||
}
|
||||
|
||||
// Error cases.
|
||||
bad := []struct {
|
||||
name string
|
||||
mnem string
|
||||
ops []Operand
|
||||
}{
|
||||
{"zeroing without mask", "VPADDD.Z", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}},
|
||||
{"K0 mask", "VPADDD", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K0"), vreg(t, "Z3")}},
|
||||
{"two masks", "VPADDD", []Operand{vreg(t, "Z1"), vreg(t, "K1"), vreg(t, "K2"), vreg(t, "Z3")}},
|
||||
{".Z on VEX-only", "VPSHUFD.Z", []Operand{Imm(1), vreg(t, "X0"), vreg(t, "X1")}},
|
||||
{"unsupported suffix", "VPADDD.BCST", []Operand{vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "Z3")}},
|
||||
{"KMOVW.Z", "KMOVW.Z", []Operand{vreg(t, "K1"), vreg(t, "K2")}},
|
||||
}
|
||||
for _, c := range bad {
|
||||
if _, err := Encode(c.mnem, c.ops...); err == nil {
|
||||
t.Errorf("%s: expected an error, got none", c.name)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// TestEvexErrors checks the EVEX-specific error paths.
|
||||
func TestEvexErrors(t *testing.T) {
|
||||
cases := []struct {
|
||||
@@ -125,7 +217,7 @@ func TestEvexErrors(t *testing.T) {
|
||||
{"VPMOVDW src", "VPMOVDW", []Operand{AX, vreg(t, "Y0")}},
|
||||
{"align arity", "VALIGND", []Operand{Imm(1), vreg(t, "Z0"), vreg(t, "Z1")}},
|
||||
// VEX-only mnemonics reject registers only EVEX can encode.
|
||||
{"VPSHUFD X16", "VPSHUFD", []Operand{Imm(1), vreg(t, "X16"), vreg(t, "X17")}},
|
||||
{"VMOVMSKPS X16", "VMOVMSKPS", []Operand{vreg(t, "X16"), AX}},
|
||||
}
|
||||
for _, c := range cases {
|
||||
if _, err := Encode(c.mnem, c.ops...); err == nil {
|
||||
|
||||
Reference in New Issue
Block a user