[ARM] Add vcge, vcgt instructions to assembler.

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 1271 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];
     float vdupf[4], vaddf[4], vsubf[4], vmulf[4];
     uint32_t vadd8[4], vadd16[4], vadd32[4];
     uint32_t vsub8[4], vsub16[4], vsub32[4];
     uint32_t vmul8[4], vmul16[4], vmul32[4];
+    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];
-    uint32_t vtst[4], vceq[4], vceqf[4], vbsl[4];
+    uint32_t vtst[4], vbsl[4];
     uint32_t vext[4];
     uint32_t vzip8a[4], vzip8b[4], vzip16a[4], vzip16b[4], vzip32a[4],
         vzip32b[4];
     uint32_t vrev64_32[4], vrev64_16[4], vrev64_8[4];
     uint32_t vrev32_16[4], vrev32_8[4];
     uint32_t vrev16_8[4];
     uint32_t vtbl[2], vtbx[2];
   } T;
   T t;
 
@@ skipping 207 matching lines @@
     __ vrsqrts(q1, q0, q1);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vrsqrts))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     // vceq (float).
     __ vmov(s4, 1.0);
     __ vdup(q0, s4);
     __ vdup(q1, s4);
     __ vceq(q1, q1, q0);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vceqf))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
+    // vcge (float).
+    __ vmov(s0, 1.0);
+    __ vmov(s1, -1.0);
+    __ vmov(s2, -0.0);
+    __ vmov(s3, 0.0);
+    __ vdup(q1, s3);
+    __ vcge(q2, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcgef))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+    __ vcgt(q2, q1, q0);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcgtf))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
 
     // vadd (integer).
     __ mov(r4, Operand(0x81));
     __ vdup(Neon8, q0, r4);
     __ mov(r4, Operand(0x82));
     __ vdup(Neon8, q1, r4);
     __ vadd(Neon8, q1, q1, q0);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vadd8))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     __ mov(r4, Operand(0x8001));
@@ skipping 47 matching lines @@
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
     __ mov(r4, Operand(0x00000002));
     __ vdup(Neon32, q0, r4);
     __ vmul(Neon32, q1, q0, q0);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vmul32))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
 
     // vceq.
     __ mov(r4, Operand(0x03));
     __ vdup(Neon8, q0, r4);
-    __ mov(r4, Operand(0x03));
     __ vdup(Neon16, q1, r4);
     __ vceq(Neon8, q1, q0, q1);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vceq))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
 
+    // vcge/vcgt.
+    __ mov(r4, Operand(0x03));
+    __ vdup(Neon16, q0, r4);
+    __ vdup(Neon8, q1, r4);
+    __ vcge(NeonS8, q2, q0, q1);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcge_s8))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+    __ vcgt(NeonS8, q2, q0, q1);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcgt_s8))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+    __ mov(r4, Operand(0xff));
+    __ vdup(Neon16, q0, r4);
+    __ vdup(Neon8, q1, r4);
+    __ vcge(NeonU16, q2, q0, q1);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcge_u16))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+    __ vcgt(NeonU16, q2, q0, q1);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcgt_u16))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+    __ mov(r4, Operand(0xff));
+    __ vdup(Neon32, q0, r4);
+    __ vdup(Neon8, q1, r4);
+    __ vcge(NeonS32, q2, q0, q1);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcge_s32))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+    __ vcgt(NeonS32, q2, q0, q1);
+    __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vcgt_s32))));
+    __ vst1(Neon8, NeonListOperand(q2), NeonMemOperand(r4));
+
     // vtst.
     __ mov(r4, Operand(0x03));
     __ vdup(Neon8, q0, r4);
     __ mov(r4, Operand(0x02));
     __ vdup(Neon16, q1, r4);
     __ vtst(Neon8, q1, q0, q1);
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, vtst))));
     __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
 
     // vbsl.
@@ skipping 173 matching lines @@
 
     CHECK_EQ_SPLAT(veor, 0x00ff00ffu);
     CHECK_EQ_SPLAT(vaddf, 2.0);
     CHECK_EQ_SPLAT(vsubf, -1.0);
     CHECK_EQ_SPLAT(vmulf, 4.0);
     CHECK_EQ_SPLAT(vrecpe, 0.5f);    // 1 / 2
     CHECK_EQ_SPLAT(vrecps, -1.0f);   // 2 - (2 * 1.5)
     CHECK_EQ_SPLAT(vrsqrte, 0.5f);   // 1 / sqrt(4)
     CHECK_EQ_SPLAT(vrsqrts, -1.0f);  // (3 - (2 * 2.5)) / 2
     CHECK_EQ_SPLAT(vceqf, 0xffffffffu);
+    // [0] >= [-1, 1, -0, 0]
+    CHECK_EQ_32X4(vcgef, 0u, 0xffffffffu, 0xffffffffu, 0xffffffffu);
+    CHECK_EQ_32X4(vcgtf, 0u, 0xffffffffu, 0u, 0u);
     CHECK_EQ_SPLAT(vadd8, 0x03030303u);
     CHECK_EQ_SPLAT(vadd16, 0x00030003u);
     CHECK_EQ_SPLAT(vadd32, 0x00000003u);
     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(vceq, 0x00ff00ffu);
+    // [0, 3, 0, 3, ...] >= [3, 3, 3, 3, ...]
+    CHECK_EQ_SPLAT(vcge_s8, 0x00ff00ffu);
+    CHECK_EQ_SPLAT(vcgt_s8, 0u);
+    // [0x00ff, 0x00ff, ...] >= [0xffff, 0xffff, ...]
+    CHECK_EQ_SPLAT(vcge_u16, 0u);
+    CHECK_EQ_SPLAT(vcgt_u16, 0u);
+    // [0x000000ff, 0x000000ff, ...] >= [0xffffffff, 0xffffffff, ...]
+    CHECK_EQ_SPLAT(vcge_s32, 0xffffffffu);
+    CHECK_EQ_SPLAT(vcgt_s32, 0xffffffffu);
     CHECK_EQ_SPLAT(vtst, 0x00ff00ffu);
     CHECK_EQ_SPLAT(vbsl, 0x02010201u);
 
     CHECK_EQ_32X4(vext, 0x06050403u, 0x0a090807u, 0x0e0d0c0bu, 0x0201000fu);
 
     CHECK_EQ_32X4(vzip8a, 0x01010000u, 0x03030202u, 0x05050404u, 0x07070606u);
     CHECK_EQ_32X4(vzip8b, 0x09090808u, 0x0b0b0a0au, 0x0d0d0c0cu, 0x0f0f0e0eu);
     CHECK_EQ_32X4(vzip16a, 0x01000100u, 0x03020302u, 0x05040504u, 0x07060706u);
     CHECK_EQ_32X4(vzip16b, 0x09080908u, 0x0b0a0b0au, 0x0d0c0d0cu, 0x0f0e0f0eu);
     CHECK_EQ_32X4(vzip32a, 0x03020100u, 0x03020100u, 0x07060504u, 0x07060504u);
@@ skipping 1689 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 __