[ARM] Add Neon saturating add and subtract instructions.

test/cctest/test-assembler-arm.cc

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 1281 matching lines @@
     float vcvt_f32_s32[4], vcvt_f32_u32[4];
     uint32_t vdup8[4], vdup16[4], vdup32[4];
     float vabsf[4], vnegf[4];
     uint32_t vabs_s8[4], vabs_s16[4], vabs_s32[4];
     uint32_t vneg_s8[4], vneg_s16[4], vneg_s32[4];
     uint32_t veor[4], vand[4], vorr[4];
     float vdupf[4], vaddf[4], vsubf[4], vmulf[4];
     uint32_t vmin_s8[4], vmin_u16[4], vmin_s32[4];
     uint32_t vmax_s8[4], vmax_u16[4], vmax_s32[4];
     uint32_t vadd8[4], vadd16[4], vadd32[4];
+    uint32_t vqadd_s8[4], vqadd_u16[4], vqadd_s32[4];
     uint32_t vsub8[4], vsub16[4], vsub32[4];
+    uint32_t vqsub_u8[4], vqsub_s16[4], vqsub_u32[4];
     uint32_t vmul8[4], vmul16[4], vmul32[4];
     uint32_t vshl8[4], vshl16[4], vshl32[5];
     uint32_t vshr_s8[4], vshr_u16[4], vshr_s32[5];
     uint32_t vceq[4], vceqf[4], vcgef[4], vcgtf[4];
     uint32_t vcge_s8[4], vcge_u16[4], vcge_s32[4];
     uint32_t vcgt_s8[4], vcgt_u16[4], vcgt_s32[4];
     float vrecpe[4], vrecps[4], vrsqrte[4], vrsqrts[4];
     float vminf[4], vmaxf[4];
     uint32_t vtst[4], vbsl[4];
     uint32_t vext[4];
@@ skipping 313 matching lines @@
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vadd16))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     __ mov(r4, Operand(0x80000001));
     __ vdup(Neon32, q0, r4);
     __ mov(r4, Operand(0x80000002));
     __ vdup(Neon32, q1, r4);
     __ vadd(Neon32, q1, q1, q0);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vadd32))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
 
+    // vqadd.
+    __ mov(r4, Operand(0x81));
+    __ vdup(Neon8, q0, r4);
+    __ mov(r4, Operand(0x82));
+    __ vdup(Neon8, q1, r4);
+    __ vqadd(NeonS8, q1, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vqadd_s8))));
+    __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+    __ mov(r4, Operand(0x8000));
+    __ vdup(Neon16, q0, r4);
+    __ vdup(Neon16, q1, r4);
+    __ vqadd(NeonU16, q1, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vqadd_u16))));
+    __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+    __ mov(r4, Operand(0x80000001));
+    __ vdup(Neon32, q0, r4);
+    __ mov(r4, Operand(0x80000002));
+    __ vdup(Neon32, q1, r4);
+    __ vqadd(NeonS32, q1, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vqadd_s32))));
+    __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+
     // vsub (integer).
     __ mov(r4, Operand(0x01));
     __ vdup(Neon8, q0, r4);
     __ mov(r4, Operand(0x03));
     __ vdup(Neon8, q1, r4);
     __ vsub(Neon8, q1, q0, q1);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vsub8))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     __ mov(r4, Operand(0x0001));
     __ vdup(Neon16, q0, r4);
     __ mov(r4, Operand(0x0003));
     __ vdup(Neon16, q1, r4);
     __ vsub(Neon16, q1, q0, q1);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vsub16))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     __ mov(r4, Operand(0x00000001));
     __ vdup(Neon32, q0, r4);
     __ mov(r4, Operand(0x00000003));
     __ vdup(Neon32, q1, r4);
     __ vsub(Neon32, q1, q0, q1);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vsub32))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
 
+    // vqsub.
+    __ mov(r4, Operand(0x7f));
+    __ vdup(Neon8, q0, r4);
+    __ mov(r4, Operand(0x3f));
+    __ vdup(Neon8, q1, r4);
+    __ vqsub(NeonU8, q1, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vqsub_u8))));
+    __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+    __ mov(r4, Operand(0x8000));
+    __ vdup(Neon16, q0, r4);
+    __ mov(r4, Operand(0x7fff));
+    __ vdup(Neon16, q1, r4);
+    __ vqsub(NeonS16, q1, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vqsub_s16))));
+    __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+    __ mov(r4, Operand(0x80000001));
+    __ vdup(Neon32, q0, r4);
+    __ mov(r4, Operand(0x80000000));
+    __ vdup(Neon32, q1, r4);
+    __ vqsub(NeonU32, q1, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vqsub_u32))));
+    __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+
     // vmul (integer).
     __ mov(r4, Operand(0x02));
     __ vdup(Neon8, q0, r4);
     __ vmul(Neon8, q1, q0, q0);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vmul8))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     __ mov(r4, Operand(0x0002));
     __ vdup(Neon16, q0, r4);
     __ vmul(Neon16, q1, q0, q0);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vmul16))));
@@ skipping 272 matching lines @@
     CHECK_EQ_SPLAT(vmax_s8, 0x03030303u);
     // [0x00ff, 0x00ff, ...] and [0xffff, 0xffff, ...]
     CHECK_EQ_SPLAT(vmin_u16, 0x00ff00ffu);
     CHECK_EQ_SPLAT(vmax_u16, 0xffffffffu);
     // [0x000000ff, 0x000000ff, ...] and [0xffffffff, 0xffffffff, ...]
     CHECK_EQ_SPLAT(vmin_s32, 0xffffffffu);
     CHECK_EQ_SPLAT(vmax_s32, 0xffu);
     CHECK_EQ_SPLAT(vadd8, 0x03030303u);
     CHECK_EQ_SPLAT(vadd16, 0x00030003u);
     CHECK_EQ_SPLAT(vadd32, 0x00000003u);
+    CHECK_EQ_SPLAT(vqadd_s8, 0x80808080u);
+    CHECK_EQ_SPLAT(vqadd_u16, 0xffffffffu);
+    CHECK_EQ_SPLAT(vqadd_s32, 0x80000000u);
+    CHECK_EQ_SPLAT(vqsub_u8, 0x00000000u);
+    CHECK_EQ_SPLAT(vqsub_s16, 0x7fff7fffu);
+    CHECK_EQ_SPLAT(vqsub_u32, 0x00000000u);
     CHECK_EQ_SPLAT(vsub8, 0xfefefefeu);
     CHECK_EQ_SPLAT(vsub16, 0xfffefffeu);
     CHECK_EQ_SPLAT(vsub32, 0xfffffffeu);
     CHECK_EQ_SPLAT(vmul8, 0x04040404u);
     CHECK_EQ_SPLAT(vmul16, 0x00040004u);
     CHECK_EQ_SPLAT(vmul32, 0x00000004u);
     CHECK_EQ_SPLAT(vshl8, 0xaaaaaaaau);
     CHECK_EQ_SPLAT(vshl16, 0xaa00aa00u);
     CHECK_EQ_SPLAT(vshl32, 0xaaaa0000u);
     CHECK_EQ_SPLAT(vshr_s8, 0xc0c0c0c0u);
@@ skipping 1711 matching lines @@
   HandleScope scope(isolate);
 
   Assembler assm(isolate, NULL, 0);
   __ mov(r0, Operand(isolate->factory()->infinity_value()));
   __ BlockConstPoolFor(1019);
   for (int i = 0; i < 1019; ++i) __ nop();
   __ vldr(d0, MemOperand(r0, 0));
 }
 
 #undef __