A lightweight WebAssembly runtime that is fast, secure, and standards-compliant
winch(aarch64): v128 extend and narrow ops (#13945)
Daniel Macovei2dd7686a7689
diff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_extend_high_i8x16_s.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_high_i8x16_s.watnew file mode 100644index 0000000000..099796b361--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_high_i8x16_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.extend_high_i8x16_s (v128.const i8x16 255 1 128 3 4 5 6 7 8 9 10 11 12 13 14 250)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; sshll2 v0.8h, v0.16b, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0xff, 0x01, 0x80, 0x03+;; 84: .byte 0x04, 0x05, 0x06, 0x07+;; 88: add w8, w8, w10, lsl #2+;; 8c: .byte 0x0c, 0x0d, 0x0e, 0xfadiff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_extend_high_i8x16_u.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_high_i8x16_u.watnew file mode 100644index 0000000000..9109ebecc1--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_high_i8x16_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.extend_high_i8x16_u (v128.const i8x16 255 1 128 3 4 5 6 7 8 9 10 11 12 13 14 250)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ushll2 v0.8h, v0.16b, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0xff, 0x01, 0x80, 0x03+;; 84: .byte 0x04, 0x05, 0x06, 0x07+;; 88: add w8, w8, w10, lsl #2+;; 8c: .byte 0x0c, 0x0d, 0x0e, 0xfadiff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_extend_low_i8x16_s.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_low_i8x16_s.watnew file mode 100644index 0000000000..e17a2d15f4--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_low_i8x16_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.extend_low_i8x16_s (v128.const i8x16 255 1 128 3 4 5 6 7 8 9 10 11 12 13 14 250)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; sshll v0.8h, v0.8b, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0xff, 0x01, 0x80, 0x03+;; 84: .byte 0x04, 0x05, 0x06, 0x07+;; 88: add w8, w8, w10, lsl #2+;; 8c: .byte 0x0c, 0x0d, 0x0e, 0xfadiff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_extend_low_i8x16_u.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_low_i8x16_u.watnew file mode 100644index 0000000000..ed21799874--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_extend_low_i8x16_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.extend_low_i8x16_u (v128.const i8x16 255 1 128 3 4 5 6 7 8 9 10 11 12 13 14 250)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ushll v0.8h, v0.8b, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0xff, 0x01, 0x80, 0x03+;; 84: .byte 0x04, 0x05, 0x06, 0x07+;; 88: add w8, w8, w10, lsl #2+;; 8c: .byte 0x0c, 0x0d, 0x0e, 0xfadiff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_narrow_i32x4_s.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_narrow_i32x4_s.watnew file mode 100644index 0000000000..e41b2aee56--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_narrow_i32x4_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.narrow_i32x4_s (v128.const i32x4 0x80000000 -1 40000 -40000) (v128.const i32x4 0x80000000 -1 40000 -40000)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x78+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ldr q1, #0x80+;; sqxtn v1.4h, v1.4s+;; sqxtn2 v1.8h, v0.4s+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q1, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 78: udf #0xc11f+;; 7c: udf #0+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: .byte 0xff, 0xff, 0xff, 0xff+;; 88: udf #0x9c40+;; 8c: .byte 0xc0, 0x63, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_narrow_i32x4_u.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_narrow_i32x4_u.watnew file mode 100644index 0000000000..ad1078a705--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_narrow_i32x4_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.narrow_i32x4_u (v128.const i32x4 0x80000000 -1 40000 -40000) (v128.const i32x4 0x80000000 -1 40000 -40000)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x78+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ldr q1, #0x80+;; sqxtun v1.4h, v1.4s+;; sqxtun2 v1.8h, v0.4s+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q1, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 78: udf #0xc11f+;; 7c: udf #0+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: .byte 0xff, 0xff, 0xff, 0xff+;; 88: udf #0x9c40+;; 8c: .byte 0xc0, 0x63, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_extend_high_i16x8_s.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_high_i16x8_s.watnew file mode 100644index 0000000000..fe48d557de--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_high_i16x8_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.extend_high_i16x8_s (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; sshll2 v0.4s, v0.8h, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x80, 0xff, 0xff+;; 84: .byte 0x2c, 0x01, 0xd4, 0xfe+;; 88: .byte 0x7f, 0x00, 0x80, 0x00+;; 8c: udf #0x7fffdiff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_extend_high_i16x8_u.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_high_i16x8_u.watnew file mode 100644index 0000000000..624fa5910a--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_high_i16x8_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.extend_high_i16x8_u (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ushll2 v0.4s, v0.8h, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x80, 0xff, 0xff+;; 84: .byte 0x2c, 0x01, 0xd4, 0xfe+;; 88: .byte 0x7f, 0x00, 0x80, 0x00+;; 8c: udf #0x7fffdiff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_extend_low_i16x8_s.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_low_i16x8_s.watnew file mode 100644index 0000000000..048cb58f68--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_low_i16x8_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.extend_low_i16x8_s (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; sshll v0.4s, v0.4h, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x80, 0xff, 0xff+;; 84: .byte 0x2c, 0x01, 0xd4, 0xfe+;; 88: .byte 0x7f, 0x00, 0x80, 0x00+;; 8c: udf #0x7fffdiff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_extend_low_i16x8_u.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_low_i16x8_u.watnew file mode 100644index 0000000000..b7a5ee6d0f--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_extend_low_i16x8_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.extend_low_i16x8_u (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ushll v0.4s, v0.4h, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x80, 0xff, 0xff+;; 84: .byte 0x2c, 0x01, 0xd4, 0xfe+;; 88: .byte 0x7f, 0x00, 0x80, 0x00+;; 8c: udf #0x7fffdiff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_extend_high_i32x4_s.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_high_i32x4_s.watnew file mode 100644index 0000000000..e4e4d210cd--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_high_i32x4_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.extend_high_i32x4_s (v128.const i32x4 0x80000000 -1 40000 -40000)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; sshll2 v0.2d, v0.4s, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: .byte 0xff, 0xff, 0xff, 0xff+;; 88: udf #0x9c40+;; 8c: .byte 0xc0, 0x63, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_extend_high_i32x4_u.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_high_i32x4_u.watnew file mode 100644index 0000000000..21ff4054b4--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_high_i32x4_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.extend_high_i32x4_u (v128.const i32x4 0x80000000 -1 40000 -40000)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ushll2 v0.2d, v0.4s, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: .byte 0xff, 0xff, 0xff, 0xff+;; 88: udf #0x9c40+;; 8c: .byte 0xc0, 0x63, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_extend_low_i32x4_s.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_low_i32x4_s.watnew file mode 100644index 0000000000..7a68d6f515--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_low_i32x4_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.extend_low_i32x4_s (v128.const i32x4 0x80000000 -1 40000 -40000)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; sshll v0.2d, v0.2s, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: .byte 0xff, 0xff, 0xff, 0xff+;; 88: udf #0x9c40+;; 8c: .byte 0xc0, 0x63, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_extend_low_i32x4_u.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_low_i32x4_u.watnew file mode 100644index 0000000000..2b93a3e920--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_extend_low_i32x4_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.extend_low_i32x4_u (v128.const i32x4 0x80000000 -1 40000 -40000)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x70+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ushll v0.2d, v0.2s, #0+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q0, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 70: udf #0xc11f+;; 74: udf #0+;; 78: udf #0+;; 7c: udf #0+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: .byte 0xff, 0xff, 0xff, 0xff+;; 88: udf #0x9c40+;; 8c: .byte 0xc0, 0x63, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i8x16_narrow_i16x8_s.wat b/tests/disas/winch/aarch64/v128_ops/i8x16_narrow_i16x8_s.watnew file mode 100644index 0000000000..9c85dad0ca--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i8x16_narrow_i16x8_s.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i8x16.narrow_i16x8_s (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0) (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x78+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ldr q1, #0x80+;; sqxtn v1.8b, v1.8h+;; sqxtn2 v1.16b, v0.8h+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q1, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 78: udf #0xc11f+;; 7c: udf #0+;; 80: .byte 0x00, 0x80, 0xff, 0xff+;; 84: .byte 0x2c, 0x01, 0xd4, 0xfe+;; 88: .byte 0x7f, 0x00, 0x80, 0x00+;; 8c: udf #0x7fffdiff --git a/tests/disas/winch/aarch64/v128_ops/i8x16_narrow_i16x8_u.wat b/tests/disas/winch/aarch64/v128_ops/i8x16_narrow_i16x8_u.watnew file mode 100644index 0000000000..fdfa829775--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i8x16_narrow_i16x8_u.wat@@ -0,0 +1,45 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i8x16.narrow_i16x8_u (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0) (v128.const i16x8 0x8000 -1 300 -300 127 128 0x7FFF 0)))))+;; wasm[0]::function[0]:+;; stp x29, x30, [sp, #-0x10]!+;; mov x29, sp+;; str x28, [sp, #-0x10]!+;; mov x28, sp+;; ldur x16, [x0, #8]+;; ldur x16, [x16, #0x18]+;; mov x17, #0+;; movk x17, #0x10+;; add x16, x16, x17+;; cmp sp, x16+;; b.lo #0x78+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; ldr q0, #0x80+;; ldr q1, #0x80+;; sqxtun v1.8b, v1.8h+;; sqxtun2 v1.16b, v0.8h+;; mov x0, #0+;; ldur x1, [x9, #0x38]+;; add x1, x1, w0, uxtw+;; stur q1, [x1]+;; add x28, x28, #0x10+;; mov sp, x28+;; mov sp, x28+;; ldr x28, [sp], #0x10+;; ldp x29, x30, [sp], #0x10+;; ret+;; 78: udf #0xc11f+;; 7c: udf #0+;; 80: .byte 0x00, 0x80, 0xff, 0xff+;; 84: .byte 0x2c, 0x01, 0xd4, 0xfe+;; 88: .byte 0x7f, 0x00, 0x80, 0x00+;; 8c: udf #0x7fffdiff --git a/winch/codegen/src/isa/aarch64/asm.rs b/winch/codegen/src/isa/aarch64/asm.rsindex d461456de6..43459e4c54 100644--- a/winch/codegen/src/isa/aarch64/asm.rs+++ b/winch/codegen/src/isa/aarch64/asm.rs@@ -15,7 +15,7 @@ use crate::{use cranelift_codegen::isa::aarch64;use cranelift_codegen::isa::aarch64::inst::emit::{enc_arith_rrr, enc_move_wide, enc_movk};use cranelift_codegen::isa::aarch64::inst::{- ASIMDFPModImm, FpuToIntOp, MoveWideConst, NZCV, UImm5, VecALUModOp,+ ASIMDFPModImm, FpuToIntOp, MoveWideConst, NZCV, UImm5, VecALUModOp, VecExtendOp, VecRRNarrowOp,};use cranelift_codegen::{Final, MachBuffer, MachBufferFinalized, MachInst, MachInstEmit, MachInstEmitState, MachLabel,@@ -423,6 +423,51 @@ impl Assembler {self.alu_rrr(ALUOp::Adc, rm, rn, rd, size);}+ /// Vector extend: widen the low or high half's lanes to `lane_size`.+ pub fn vec_extend(+ &mut self,+ t: VecExtendOp,+ rn: Reg,+ rd: WritableReg,+ high_half: bool,+ lane_size: ScalarSize,+ ) {+ self.emit(Inst::VecExtend {+ t,+ rd: rd.map(Into::into),+ rn: rn.into(),+ high_half,+ lane_size,+ });+ }++ /// Vector narrow: saturate `rn`'s lanes to `lane_size` into the low or+ /// high half of `rd`; the high-half form preserves `rd`'s other lanes.+ pub fn vec_narrow(+ &mut self,+ op: VecRRNarrowOp,+ rn: Reg,+ rd: WritableReg,+ high_half: bool,+ lane_size: ScalarSize,+ ) {+ if high_half {+ self.emit(Inst::VecRRNarrowHigh {+ op,+ rd: rd.map(Into::into),+ ri: rd.to_reg().into(),+ rn: rn.into(),+ lane_size,+ });+ } else {+ self.emit(Inst::VecRRNarrowLow {+ op,+ rd: rd.map(Into::into),+ rn: rn.into(),+ lane_size,+ });+ }+ }/// Vector three-register ALU operation whose destination is also an inputpub fn vec_rrr_mod(&mut self,diff --git a/winch/codegen/src/isa/aarch64/masm.rs b/winch/codegen/src/isa/aarch64/masm.rsindex bfc8b72e8a..c4779f9f2b 100644--- a/winch/codegen/src/isa/aarch64/masm.rs+++ b/winch/codegen/src/isa/aarch64/masm.rs@@ -33,8 +33,8 @@ use cranelift_codegen::{ir::{MemFlagsData, RelSourceLoc, SourceLoc, types},isa::aarch64,isa::aarch64::inst::{- self, Cond, ExtendOp, Imm12, ImmLogic, ImmShift, SImm7Scaled, SImm9, VecALUModOp, VecALUOp,- VecMisc2, VectorSize,+ self, Cond, ExtendOp, Imm12, ImmLogic, ImmShift, SImm7Scaled, SImm9, ScalarSize,+ VecALUModOp, VecALUOp, VecExtendOp, VecMisc2, VecRRNarrowOp, VectorSize,},settings,};@@ -1615,12 +1615,21 @@ impl Masm for MacroAssembler {fn v128_narrow(&mut self,- _src1: Reg,- _src2: Reg,- _dst: WritableReg,- _kind: V128NarrowKind,+ src1: Reg,+ src2: Reg,+ dst: WritableReg,+ kind: V128NarrowKind,) -> Result<()> {- bail!(CodeGenError::unimplemented_masm_instruction())+ let (op, lane_size) = match kind {+ V128NarrowKind::I16x8S => (VecRRNarrowOp::Sqxtn, ScalarSize::Size8),+ V128NarrowKind::I16x8U => (VecRRNarrowOp::Sqxtun, ScalarSize::Size8),+ V128NarrowKind::I32x4S => (VecRRNarrowOp::Sqxtn, ScalarSize::Size16),+ V128NarrowKind::I32x4U => (VecRRNarrowOp::Sqxtun, ScalarSize::Size16),+ };+ debug_assert!(dst.to_reg() != src2);+ self.asm.vec_narrow(op, src1, dst, false, lane_size);+ self.asm.vec_narrow(op, src2, dst, true, lane_size);+ Ok(())}fn v128_demote(&mut self, _src: Reg, _dst: WritableReg) -> Result<()> {@@ -1631,8 +1640,24 @@ impl Masm for MacroAssembler {bail!(CodeGenError::unimplemented_masm_instruction())}- fn v128_extend(&mut self, _src: Reg, _dst: WritableReg, _kind: V128ExtendKind) -> Result<()> {- bail!(CodeGenError::unimplemented_masm_instruction())+ fn v128_extend(&mut self, src: Reg, dst: WritableReg, kind: V128ExtendKind) -> Result<()> {+ use VecExtendOp::{Sxtl, Uxtl};+ let (op, high_half, lane_size) = match kind {+ V128ExtendKind::LowI8x16S => (Sxtl, false, ScalarSize::Size16),+ V128ExtendKind::HighI8x16S => (Sxtl, true, ScalarSize::Size16),+ V128ExtendKind::LowI8x16U => (Uxtl, false, ScalarSize::Size16),+ V128ExtendKind::HighI8x16U => (Uxtl, true, ScalarSize::Size16),+ V128ExtendKind::LowI16x8S => (Sxtl, false, ScalarSize::Size32),+ V128ExtendKind::HighI16x8S => (Sxtl, true, ScalarSize::Size32),+ V128ExtendKind::LowI16x8U => (Uxtl, false, ScalarSize::Size32),+ V128ExtendKind::HighI16x8U => (Uxtl, true, ScalarSize::Size32),+ V128ExtendKind::LowI32x4S => (Sxtl, false, ScalarSize::Size64),+ V128ExtendKind::HighI32x4S => (Sxtl, true, ScalarSize::Size64),+ V128ExtendKind::LowI32x4U => (Uxtl, false, ScalarSize::Size64),+ V128ExtendKind::HighI32x4U => (Uxtl, true, ScalarSize::Size64),+ };+ self.asm.vec_extend(op, src, dst, high_half, lane_size);+ Ok(())}fn v128_add(&mut self, lhs: Reg, rhs: Reg, dst: WritableReg, kind: V128AddKind) -> Result<()> {