A lightweight WebAssembly runtime that is fast, secure, and standards-compliant
winch(aarch64): v128 shift ops (#13950)
Daniel Macovei9bd335ea41ab
diff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_shl.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_shl.watnew file mode 100644index 0000000000..41358badf4--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_shl.wat@@ -0,0 +1,47 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.shl+ (v128.const i16x8 32768 1 65535 16384 2 3 4 5)+ (i32.const 2)))))+;; 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 #0x7c+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #2+;; ldr q0, #0x80+;; and w0, w0, #0xf+;; dup v31.8h, w0+;; sshl v0.8h, v0.8h, v31.8h+;; 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+;; 7c: udf #0xc11f+;; 80: .byte 0x00, 0x80, 0x01, 0x00+;; 84: .byte 0xff, 0xff, 0x00, 0x40+;; 88: .byte 0x02, 0x00, 0x03, 0x00+;; 8c: .byte 0x04, 0x00, 0x05, 0x00diff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_shr_s.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_shr_s.watnew file mode 100644index 0000000000..0dfbbe1d82--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_shr_s.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.shr_s+ (v128.const i16x8 32768 1 65535 16384 2 3 4 5)+ (i32.const 2)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #2+;; ldr q0, #0x90+;; and w0, w0, #0xf+;; neg x0, x0+;; dup v31.8h, w0+;; sshl v0.8h, v0.8h, v31.8h+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: .byte 0x00, 0x80, 0x01, 0x00+;; 94: .byte 0xff, 0xff, 0x00, 0x40+;; 98: .byte 0x02, 0x00, 0x03, 0x00+;; 9c: .byte 0x04, 0x00, 0x05, 0x00diff --git a/tests/disas/winch/aarch64/v128_ops/i16x8_shr_u.wat b/tests/disas/winch/aarch64/v128_ops/i16x8_shr_u.watnew file mode 100644index 0000000000..e98bffd366--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i16x8_shr_u.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i16x8.shr_u+ (v128.const i16x8 32768 1 65535 16384 2 3 4 5)+ (i32.const 2)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #2+;; ldr q0, #0x90+;; and w0, w0, #0xf+;; neg x0, x0+;; dup v31.8h, w0+;; ushl v0.8h, v0.8h, v31.8h+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: .byte 0x00, 0x80, 0x01, 0x00+;; 94: .byte 0xff, 0xff, 0x00, 0x40+;; 98: .byte 0x02, 0x00, 0x03, 0x00+;; 9c: .byte 0x04, 0x00, 0x05, 0x00diff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_shl.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_shl.watnew file mode 100644index 0000000000..813e321e48--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_shl.wat@@ -0,0 +1,47 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.shl+ (v128.const i32x4 0x80000000 1 0xFFFFFFFF 7)+ (i32.const 3)))))+;; 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 #0x7c+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #3+;; ldr q0, #0x80+;; and w0, w0, #0x1f+;; dup v31.4s, w0+;; sshl v0.4s, v0.4s, v31.4s+;; 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+;; 7c: udf #0xc11f+;; 80: .byte 0x00, 0x00, 0x00, 0x80+;; 84: udf #1+;; 88: .byte 0xff, 0xff, 0xff, 0xff+;; 8c: udf #7diff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_shr_s.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_shr_s.watnew file mode 100644index 0000000000..4fbf275595--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_shr_s.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.shr_s+ (v128.const i32x4 0x80000000 1 0xFFFFFFFF 7)+ (i32.const 3)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #3+;; ldr q0, #0x90+;; and w0, w0, #0x1f+;; neg x0, x0+;; dup v31.4s, w0+;; sshl v0.4s, v0.4s, v31.4s+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: .byte 0x00, 0x00, 0x00, 0x80+;; 94: udf #1+;; 98: .byte 0xff, 0xff, 0xff, 0xff+;; 9c: udf #7diff --git a/tests/disas/winch/aarch64/v128_ops/i32x4_shr_u.wat b/tests/disas/winch/aarch64/v128_ops/i32x4_shr_u.watnew file mode 100644index 0000000000..d2216b16ad--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i32x4_shr_u.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i32x4.shr_u+ (v128.const i32x4 0x80000000 1 0xFFFFFFFF 7)+ (i32.const 3)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #3+;; ldr q0, #0x90+;; and w0, w0, #0x1f+;; neg x0, x0+;; dup v31.4s, w0+;; ushl v0.4s, v0.4s, v31.4s+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: .byte 0x00, 0x00, 0x00, 0x80+;; 94: udf #1+;; 98: .byte 0xff, 0xff, 0xff, 0xff+;; 9c: udf #7diff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_shl.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_shl.watnew file mode 100644index 0000000000..cdd0f7a1ea--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_shl.wat@@ -0,0 +1,47 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.shl+ (v128.const i64x2 0x8000000000000000 0xFFFFFFFFFFFFFFFF)+ (i32.const 2)))))+;; 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 #0x7c+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #2+;; ldr q0, #0x80+;; and w0, w0, #0x3f+;; dup v31.2d, x0+;; sshl v0.2d, v0.2d, v31.2d+;; 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+;; 7c: udf #0xc11f+;; 80: udf #0+;; 84: .byte 0x00, 0x00, 0x00, 0x80+;; 88: .byte 0xff, 0xff, 0xff, 0xff+;; 8c: .byte 0xff, 0xff, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_shr_s.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_shr_s.watnew file mode 100644index 0000000000..eaebafd797--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_shr_s.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.shr_s+ (v128.const i64x2 0x8000000000000000 0xFFFFFFFFFFFFFFFF)+ (i32.const 2)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #2+;; ldr q0, #0x90+;; and w0, w0, #0x3f+;; neg x0, x0+;; dup v31.2d, x0+;; sshl v0.2d, v0.2d, v31.2d+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: udf #0+;; 94: .byte 0x00, 0x00, 0x00, 0x80+;; 98: .byte 0xff, 0xff, 0xff, 0xff+;; 9c: .byte 0xff, 0xff, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i64x2_shr_u.wat b/tests/disas/winch/aarch64/v128_ops/i64x2_shr_u.watnew file mode 100644index 0000000000..4eacd62235--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i64x2_shr_u.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i64x2.shr_u+ (v128.const i64x2 0x8000000000000000 0xFFFFFFFFFFFFFFFF)+ (i32.const 2)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #2+;; ldr q0, #0x90+;; and w0, w0, #0x3f+;; neg x0, x0+;; dup v31.2d, x0+;; ushl v0.2d, v0.2d, v31.2d+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: udf #0+;; 94: .byte 0x00, 0x00, 0x00, 0x80+;; 98: .byte 0xff, 0xff, 0xff, 0xff+;; 9c: .byte 0xff, 0xff, 0xff, 0xffdiff --git a/tests/disas/winch/aarch64/v128_ops/i8x16_shl.wat b/tests/disas/winch/aarch64/v128_ops/i8x16_shl.watnew file mode 100644index 0000000000..696532a9c1--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i8x16_shl.wat@@ -0,0 +1,47 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i8x16.shl+ (v128.const i8x16 128 1 255 64 2 250 7 8 9 10 11 12 13 14 15 16)+ (i32.const 3)))))+;; 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 #0x7c+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #3+;; ldr q0, #0x80+;; and w0, w0, #7+;; dup v31.16b, w0+;; sshl v0.16b, v0.16b, v31.16b+;; 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+;; 7c: udf #0xc11f+;; 80: .byte 0x80, 0x01, 0xff, 0x40+;; 84: stlxrb w7, w2, [x16]+;; 88: .byte 0x09, 0x0a, 0x0b, 0x0c+;; 8c: adr x13, #0x1e24cdiff --git a/tests/disas/winch/aarch64/v128_ops/i8x16_shr_s.wat b/tests/disas/winch/aarch64/v128_ops/i8x16_shr_s.watnew file mode 100644index 0000000000..81227e98b6--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i8x16_shr_s.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i8x16.shr_s+ (v128.const i8x16 128 1 255 64 2 250 7 8 9 10 11 12 13 14 15 16)+ (i32.const 3)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #3+;; ldr q0, #0x90+;; and w0, w0, #7+;; neg x0, x0+;; dup v31.16b, w0+;; sshl v0.16b, v0.16b, v31.16b+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: .byte 0x80, 0x01, 0xff, 0x40+;; 94: stlxrb w7, w2, [x16]+;; 98: .byte 0x09, 0x0a, 0x0b, 0x0c+;; 9c: adr x13, #0x1e25cdiff --git a/tests/disas/winch/aarch64/v128_ops/i8x16_shr_u.wat b/tests/disas/winch/aarch64/v128_ops/i8x16_shr_u.watnew file mode 100644index 0000000000..cf2e76a5aa--- /dev/null+++ b/tests/disas/winch/aarch64/v128_ops/i8x16_shr_u.wat@@ -0,0 +1,51 @@+;;! target = "aarch64"+;;! test = "winch"++(module+ (memory 1)+ (func (export "run")+ (v128.store (i32.const 0)+ (i8x16.shr_u+ (v128.const i8x16 128 1 255 64 2 250 7 8 9 10 11 12 13 14 15 16)+ (i32.const 3)))))+;; 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 #0x80+;; 2c: mov x9, x0+;; sub x28, x28, #0x10+;; mov sp, x28+;; stur x0, [x28, #8]+;; stur x1, [x28]+;; mov x0, #3+;; ldr q0, #0x90+;; and w0, w0, #7+;; neg x0, x0+;; dup v31.16b, w0+;; ushl v0.16b, v0.16b, v31.16b+;; 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+;; 80: udf #0xc11f+;; 84: udf #0+;; 88: udf #0+;; 8c: udf #0+;; 90: .byte 0x80, 0x01, 0xff, 0x40+;; 94: stlxrb w7, w2, [x16]+;; 98: .byte 0x09, 0x0a, 0x0b, 0x0c+;; 9c: adr x13, #0x1e25cdiff --git a/winch/codegen/src/isa/aarch64/asm.rs b/winch/codegen/src/isa/aarch64/asm.rsindex 2a8d485f52..ec9cb39766 100644--- a/winch/codegen/src/isa/aarch64/asm.rs+++ b/winch/codegen/src/isa/aarch64/asm.rs@@ -660,6 +660,11 @@ impl Assembler {self.alu_rrr_extend(ALUOp::Sub, rm, rn, rd, size, ExtendOp::UXTX);}+ /// Negate a general-purpose register.+ pub fn neg_rr(&mut self, rn: Reg, rd: WritableReg, size: OperandSize) {+ self.alu_rrr(ALUOp::Sub, rn, zero(), rd, size);+ }+/// Subtract with three registers, setting flags.pub fn subs_rrr(&mut self, rm: Reg, rn: Reg, rd: WritableReg, size: OperandSize) {self.alu_rrr_extend(ALUOp::SubS, rm, rn, rd, size, ExtendOp::UXTX);diff --git a/winch/codegen/src/isa/aarch64/masm.rs b/winch/codegen/src/isa/aarch64/masm.rsindex f66873b92a..d747f8ae72 100644--- a/winch/codegen/src/isa/aarch64/masm.rs+++ b/winch/codegen/src/isa/aarch64/masm.rs@@ -1868,11 +1868,51 @@ impl Masm for MacroAssembler {fn v128_shift(&mut self,- _context: &mut CodeGenContext<Emission>,- _lane_width: OperandSize,- _shift_kind: ShiftKind,+ context: &mut CodeGenContext<Emission>,+ lane_width: OperandSize,+ shift_kind: ShiftKind,) -> Result<()> {- Err(format_err!(CodeGenError::unimplemented_masm_instruction()))+ let shift_amount = context.pop_to_reg(self, None)?.reg;+ let operand = context.pop_to_reg(self, None)?.reg;+ let amount_mask = lane_width.num_bits() - 1;+ self.and(+ writable!(shift_amount),+ shift_amount,+ RegImm::i32(amount_mask as i32),+ OperandSize::S32,+ )?;++ let size = match lane_width {+ OperandSize::S8 => VectorSize::Size8x16,+ OperandSize::S16 => VectorSize::Size16x8,+ OperandSize::S32 => VectorSize::Size32x4,+ OperandSize::S64 => VectorSize::Size64x2,+ _ => bail!(CodeGenError::unexpected_operand_size()),+ };++ let (op, negate) = match shift_kind {+ ShiftKind::Shl => (VecALUOp::Sshl, false),+ ShiftKind::ShrS => (VecALUOp::Sshl, true),+ ShiftKind::ShrU => (VecALUOp::Ushl, true),+ ShiftKind::Rotl | ShiftKind::Rotr => {+ bail!(CodeGenError::unimplemented_masm_instruction())+ }+ };++ if negate {+ self.asm+ .neg_rr(shift_amount, writable!(shift_amount), OperandSize::S64);+ }++ self.with_scratch::<FloatScratch, _>(|masm, tmp| {+ masm.asm.vec_dup(shift_amount, tmp.writable(), size);+ masm.asm+ .vec_rrr(op, operand, tmp.inner(), writable!(operand), size);+ });++ context.free_reg(shift_amount);+ context.stack.push(TypedReg::v128(operand).into());+ Ok(())}fn v128_q15mulr_sat_s(