feat(asm): add vpcmp compare, full opmask set, legacy sse integers and bswap
Test / vet (push) Successful in 47s
Test / test (push) Successful in 2m35s
Test / build (push) Successful in 41s

Assisted-by: GLM 5.3
This commit is contained in:
2026-08-27 22:41:03 +02:00
parent 163480e283
commit 19a26e049b
5 changed files with 170 additions and 32 deletions
+37
View File
@@ -319,6 +319,43 @@ func TestEvexExtendedGroundTruth(t *testing.T) {
{"KORTESTD", "KORTESTD", []Operand{vreg(t, "K1"), vreg(t, "K2")}, "c4e1f998d1"},
{"KMOVQ k,k", "KMOVQ", []Operand{vreg(t, "K1"), vreg(t, "K2")}, "c4e1f890d1"},
{"KMOVQ gpr,k", "KMOVQ", []Operand{BX, vreg(t, "K1")}, "c4e1fb92cb"},
// Completed opmask families (ANDN, NOT, OR/XOR word+qword, TEST,
// word-width shifts; byte-exact against go tool asm).
{"KANDNW", "KANDNW", []Operand{vreg(t, "K1"), vreg(t, "K2"), vreg(t, "K3")}, "c5ec42d9"},
{"KANDNB", "KANDNB", []Operand{vreg(t, "K4"), vreg(t, "K5"), vreg(t, "K6")}, "c5d542f4"},
{"KANDND", "KANDND", []Operand{vreg(t, "K1"), vreg(t, "K2"), vreg(t, "K3")}, "c4e1ed42d9"},
{"KANDNQ", "KANDNQ", []Operand{vreg(t, "K4"), vreg(t, "K5"), vreg(t, "K6")}, "c4e1d442f4"},
{"KANDD", "KANDD", []Operand{vreg(t, "K1"), vreg(t, "K2"), vreg(t, "K3")}, "c4e1ed41d9"},
{"KADDD", "KADDD", []Operand{vreg(t, "K4"), vreg(t, "K5"), vreg(t, "K6")}, "c4e1d54af4"},
{"KNOTW", "KNOTW", []Operand{vreg(t, "K1"), vreg(t, "K2")}, "c5f844d1"},
{"KNOTD", "KNOTD", []Operand{vreg(t, "K3"), vreg(t, "K4")}, "c4e1f944e3"},
{"KNOTQ", "KNOTQ", []Operand{vreg(t, "K5"), vreg(t, "K6")}, "c4e1f844f5"},
{"KORW", "KORW", []Operand{vreg(t, "K1"), vreg(t, "K2"), vreg(t, "K3")}, "c5ec45d9"},
{"KORQ", "KORQ", []Operand{vreg(t, "K4"), vreg(t, "K5"), vreg(t, "K6")}, "c4e1d445f4"},
{"KXNORB", "KXNORB", []Operand{vreg(t, "K1"), vreg(t, "K2"), vreg(t, "K3")}, "c5ed46d9"},
{"KXORW", "KXORW", []Operand{vreg(t, "K1"), vreg(t, "K2"), vreg(t, "K3")}, "c5ec47d9"},
{"KXORQ", "KXORQ", []Operand{vreg(t, "K4"), vreg(t, "K5"), vreg(t, "K6")}, "c4e1d447f4"},
{"KORTESTW", "KORTESTW", []Operand{vreg(t, "K1"), vreg(t, "K2")}, "c5f898d1"},
{"KORTESTB", "KORTESTB", []Operand{vreg(t, "K3"), vreg(t, "K4")}, "c5f998e3"},
{"KORTESTQ", "KORTESTQ", []Operand{vreg(t, "K5"), vreg(t, "K6")}, "c4e1f898f5"},
{"KTESTW", "KTESTW", []Operand{vreg(t, "K1"), vreg(t, "K2")}, "c5f899d1"},
{"KTESTD", "KTESTD", []Operand{vreg(t, "K3"), vreg(t, "K4")}, "c4e1f999e3"},
{"KSHIFTLB", "KSHIFTLB", []Operand{Imm(1), vreg(t, "K1"), vreg(t, "K2")}, "c4e37932d101"},
{"KSHIFTLD", "KSHIFTLD", []Operand{Imm(2), vreg(t, "K3"), vreg(t, "K4")}, "c4e37933e302"},
{"KSHIFTLQ", "KSHIFTLQ", []Operand{Imm(3), vreg(t, "K5"), vreg(t, "K6")}, "c4e3f933f503"},
{"KSHIFTRB", "KSHIFTRB", []Operand{Imm(4), vreg(t, "K1"), vreg(t, "K2")}, "c4e37930d104"},
{"KSHIFTRW", "KSHIFTRW", []Operand{Imm(5), vreg(t, "K3"), vreg(t, "K4")}, "c4e3f930e305"},
{"KSHIFTRQ", "KSHIFTRQ", []Operand{Imm(6), vreg(t, "K5"), vreg(t, "K6")}, "c4e3f931f506"},
// Integer compares with an opmask destination (0F3A map, the
// go-bzip2 partition kernel's classify instructions).
{"VPCMPUB", "VPCMPUB", []Operand{Imm(1), vreg(t, "X1"), vreg(t, "X0"), vreg(t, "K1")}, "62f37d083ec901"},
{"VPCMPB", "VPCMPB", []Operand{Imm(2), vreg(t, "Y2"), vreg(t, "Y3"), vreg(t, "K2")}, "62f365283fd202"},
{"VPCMPUW", "VPCMPUW", []Operand{Imm(5), vreg(t, "Z1"), vreg(t, "Z2"), vreg(t, "K3")}, "62f3ed483ed905"},
{"VPCMPW", "VPCMPW", []Operand{Imm(6), vreg(t, "X3"), vreg(t, "X4"), vreg(t, "K4")}, "62f3dd083fe306"},
{"VPCMPD", "VPCMPD", []Operand{Imm(0), vreg(t, "Y1"), vreg(t, "Y2"), vreg(t, "K1")}, "62f36d281fc900"},
{"VPCMPUD", "VPCMPUD", []Operand{Imm(1), vreg(t, "Z2"), vreg(t, "Z3"), vreg(t, "K2")}, "62f365481ed201"},
{"VPCMPQ", "VPCMPQ", []Operand{Imm(2), vreg(t, "X1"), vreg(t, "X2"), vreg(t, "K3")}, "62f3ed081fd902"},
{"VPCMPUQ", "VPCMPUQ", []Operand{Imm(3), vreg(t, "Y3"), vreg(t, "Y4"), vreg(t, "K4")}, "62f3dd281ee303"},
// Lane extract / insert.
{"VEXTRACTF32X4", "VEXTRACTF32X4", []Operand{Imm(1), vreg(t, "Y1"), vreg(t, "X2")}, "62f37d2819ca01"},
{"VEXTRACTI64X2", "VEXTRACTI64X2", []Operand{Imm(1), vreg(t, "Y1"), vreg(t, "X2")}, "62f3fd2839ca01"},