[wasm-simd][liftoff][arm][arm64] Implement all_true

Implement all_true for arm and arm64. Instruction sequence is the same
as TurboFan.

Bug: v8:9909
Change-Id: Ibe57c6ae6f700dfe5bd23a91a243778b6481c5a0
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2222606Reviewed-by: Clemens Backes <clemensb@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#68089}

src/wasm/baseline/arm/liftoff-assembler-arm.h

@@ -2564,7 +2564,13 @@ void LiftoffAssembler::emit_v32x4_anytrue(LiftoffRegister dst,
 
 void LiftoffAssembler::emit_v32x4_alltrue(LiftoffRegister dst,
                                           LiftoffRegister src) {
-  bailout(kSimd, "v32x4_alltrue");
+  UseScratchRegisterScope temps(this);
+  DwVfpRegister scratch = temps.AcquireD();
+  vpmin(NeonU32, scratch, src.low_fp(), src.high_fp());
+  vpmin(NeonU32, scratch, scratch, scratch);
+  ExtractLane(dst.gp(), scratch, NeonS32, 0);
+  cmp(dst.gp(), Operand(0));
+  mov(dst.gp(), Operand(1), LeaveCC, ne);
 }
 
 void LiftoffAssembler::emit_i32x4_shl(LiftoffRegister dst, LiftoffRegister lhs,
@@ -2666,7 +2672,14 @@ void LiftoffAssembler::emit_v16x8_anytrue(LiftoffRegister dst,
 
 void LiftoffAssembler::emit_v16x8_alltrue(LiftoffRegister dst,
                                           LiftoffRegister src) {
-  bailout(kSimd, "v16x8_alltrue");
+  UseScratchRegisterScope temps(this);
+  DwVfpRegister scratch = temps.AcquireD();
+  vpmin(NeonU16, scratch, src.low_fp(), src.high_fp());
+  vpmin(NeonU16, scratch, scratch, scratch);
+  vpmin(NeonU16, scratch, scratch, scratch);
+  ExtractLane(dst.gp(), scratch, NeonS16, 0);
+  cmp(dst.gp(), Operand(0));
+  mov(dst.gp(), Operand(1), LeaveCC, ne);
 }
 
 void LiftoffAssembler::emit_i16x8_shl(LiftoffRegister dst, LiftoffRegister lhs,
@@ -2840,7 +2853,15 @@ void LiftoffAssembler::emit_v8x16_anytrue(LiftoffRegister dst,
 
 void LiftoffAssembler::emit_v8x16_alltrue(LiftoffRegister dst,
                                           LiftoffRegister src) {
-  bailout(kSimd, "v8x16_alltrue");
+  UseScratchRegisterScope temps(this);
+  DwVfpRegister scratch = temps.AcquireD();
+  vpmin(NeonU8, scratch, src.low_fp(), src.high_fp());
+  vpmin(NeonU8, scratch, scratch, scratch);
+  vpmin(NeonU8, scratch, scratch, scratch);
+  vpmin(NeonU8, scratch, scratch, scratch);
+  ExtractLane(dst.gp(), scratch, NeonS8, 0);
+  cmp(dst.gp(), Operand(0));
+  mov(dst.gp(), Operand(1), LeaveCC, ne);
 }
 
 void LiftoffAssembler::emit_i8x16_shl(LiftoffRegister dst, LiftoffRegister lhs,

src/wasm/baseline/arm64/liftoff-assembler-arm64.h

@@ -164,6 +164,16 @@ inline void EmitAnyTrue(LiftoffAssembler* assm, LiftoffRegister dst,
   assm->Cset(dst.gp().W(), ne);
 }
 
+inline void EmitAllTrue(LiftoffAssembler* assm, LiftoffRegister dst,
+                        LiftoffRegister src, VectorFormat format) {
+  UseScratchRegisterScope scope(assm);
+  VRegister temp = scope.AcquireV(ScalarFormatFromFormat(format));
+  assm->Uminv(temp, VRegister::Create(src.fp().code(), format));
+  assm->Umov(dst.gp().W(), temp, 0);
+  assm->Cmp(dst.gp().W(), 0);
+  assm->Cset(dst.gp().W(), ne);
+}
+
 }  // namespace liftoff
 
 int LiftoffAssembler::PrepareStackFrame() {
@@ -1500,7 +1510,7 @@ void LiftoffAssembler::emit_v32x4_anytrue(LiftoffRegister dst,
 
 void LiftoffAssembler::emit_v32x4_alltrue(LiftoffRegister dst,
                                           LiftoffRegister src) {
-  bailout(kSimd, "v32x4_alltrue");
+  liftoff::EmitAllTrue(this, dst, src, kFormat4S);
 }
 
 void LiftoffAssembler::emit_i32x4_shl(LiftoffRegister dst, LiftoffRegister lhs,
@@ -1621,7 +1631,7 @@ void LiftoffAssembler::emit_v16x8_anytrue(LiftoffRegister dst,
 
 void LiftoffAssembler::emit_v16x8_alltrue(LiftoffRegister dst,
                                           LiftoffRegister src) {
-  bailout(kSimd, "v16x8_alltrue");
+  liftoff::EmitAllTrue(this, dst, src, kFormat8H);
 }
 
 void LiftoffAssembler::emit_i16x8_shl(LiftoffRegister dst, LiftoffRegister lhs,
@@ -1766,7 +1776,7 @@ void LiftoffAssembler::emit_v8x16_anytrue(LiftoffRegister dst,
 
 void LiftoffAssembler::emit_v8x16_alltrue(LiftoffRegister dst,
                                           LiftoffRegister src) {
-  bailout(kSimd, "v8x16_alltrue");
+  liftoff::EmitAllTrue(this, dst, src, kFormat16B);
 }
 
 void LiftoffAssembler::emit_i8x16_shl(LiftoffRegister dst, LiftoffRegister lhs,