PPC/S390 [simd]: optimize GeS ops on ppc and s390

Doing a `!(B > A)` which is equal to `A >= B`. This way
we use one less instruction.

Change-Id: I49d50f11096e2d542eaabab82c17225c83e89b63
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2846980Reviewed-by: Junliang Yan <junyan@redhat.com>
Commit-Queue: Milad Fa <mfarazma@redhat.com>
Cr-Commit-Position: refs/heads/master@{#74125}

src/compiler/backend/ppc/code-generator-ppc.cc

@@ -2642,21 +2642,17 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kPPC_I64x2GeS: {
-      __ vcmpequd(kScratchSimd128Reg, i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vcmpgtsd(i.OutputSimd128Register(), i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vor(i.OutputSimd128Register(), i.OutputSimd128Register(),
-             kScratchSimd128Reg);
+      __ vcmpgtsd(kScratchSimd128Reg, i.InputSimd128Register(1),
+                  i.InputSimd128Register(0));
+      __ vnor(i.OutputSimd128Register(), kScratchSimd128Reg,
+              kScratchSimd128Reg);
       break;
     }
     case kPPC_I32x4GeS: {
-      __ vcmpequw(kScratchSimd128Reg, i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vcmpgtsw(i.OutputSimd128Register(), i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vor(i.OutputSimd128Register(), i.OutputSimd128Register(),
-             kScratchSimd128Reg);
+      __ vcmpgtsw(kScratchSimd128Reg, i.InputSimd128Register(1),
+                  i.InputSimd128Register(0));
+      __ vnor(i.OutputSimd128Register(), kScratchSimd128Reg,
+              kScratchSimd128Reg);
       break;
     }
     case kPPC_I32x4GtU: {
@@ -2680,12 +2676,10 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kPPC_I16x8GeS: {
-      __ vcmpequh(kScratchSimd128Reg, i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vcmpgtsh(i.OutputSimd128Register(), i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vor(i.OutputSimd128Register(), i.OutputSimd128Register(),
-             kScratchSimd128Reg);
+      __ vcmpgtsh(kScratchSimd128Reg, i.InputSimd128Register(1),
+                  i.InputSimd128Register(0));
+      __ vnor(i.OutputSimd128Register(), kScratchSimd128Reg,
+              kScratchSimd128Reg);
       break;
     }
     case kPPC_I16x8GtU: {
@@ -2708,12 +2702,10 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kPPC_I8x16GeS: {
-      __ vcmpequb(kScratchSimd128Reg, i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vcmpgtsb(i.OutputSimd128Register(), i.InputSimd128Register(0),
-                  i.InputSimd128Register(1));
-      __ vor(i.OutputSimd128Register(), i.OutputSimd128Register(),
-             kScratchSimd128Reg);
+      __ vcmpgtsb(kScratchSimd128Reg, i.InputSimd128Register(1),
+                  i.InputSimd128Register(0));
+      __ vnor(i.OutputSimd128Register(), kScratchSimd128Reg,
+              kScratchSimd128Reg);
       break;
     }
     case kPPC_I8x16GtU: {

src/compiler/backend/s390/code-generator-s390.cc

@@ -2986,12 +2986,10 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kS390_I32x4GeS: {
-      __ vceq(kScratchDoubleReg, i.InputSimd128Register(0),
-              i.InputSimd128Register(1), Condition(0), Condition(2));
-      __ vch(i.OutputSimd128Register(), i.InputSimd128Register(0),
-             i.InputSimd128Register(1), Condition(0), Condition(2));
-      __ vo(i.OutputSimd128Register(), i.OutputSimd128Register(),
-            kScratchDoubleReg, Condition(0), Condition(0), Condition(2));
+      __ vch(kScratchDoubleReg, i.InputSimd128Register(1),
+             i.InputSimd128Register(0), Condition(0), Condition(2));
+      __ vno(i.OutputSimd128Register(), kScratchDoubleReg, kScratchDoubleReg,
+             Condition(0), Condition(0), Condition(2));
       break;
     }
     case kS390_I32x4GtU: {
@@ -3014,12 +3012,10 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kS390_I16x8GeS: {
-      __ vceq(kScratchDoubleReg, i.InputSimd128Register(0),
-              i.InputSimd128Register(1), Condition(0), Condition(1));
-      __ vch(i.OutputSimd128Register(), i.InputSimd128Register(0),
-             i.InputSimd128Register(1), Condition(0), Condition(1));
-      __ vo(i.OutputSimd128Register(), i.OutputSimd128Register(),
-            kScratchDoubleReg, Condition(0), Condition(0), Condition(1));
+      __ vch(kScratchDoubleReg, i.InputSimd128Register(1),
+             i.InputSimd128Register(0), Condition(0), Condition(1));
+      __ vno(i.OutputSimd128Register(), kScratchDoubleReg, kScratchDoubleReg,
+             Condition(0), Condition(0), Condition(1));
       break;
     }
     case kS390_I16x8GtU: {
@@ -3042,12 +3038,10 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kS390_I8x16GeS: {
-      __ vceq(kScratchDoubleReg, i.InputSimd128Register(0),
-              i.InputSimd128Register(1), Condition(0), Condition(0));
-      __ vch(i.OutputSimd128Register(), i.InputSimd128Register(0),
-             i.InputSimd128Register(1), Condition(0), Condition(0));
-      __ vo(i.OutputSimd128Register(), i.OutputSimd128Register(),
-            kScratchDoubleReg, Condition(0), Condition(0), Condition(0));
+      __ vch(kScratchDoubleReg, i.InputSimd128Register(1),
+             i.InputSimd128Register(0), Condition(0), Condition(0));
+      __ vno(i.OutputSimd128Register(), kScratchDoubleReg, kScratchDoubleReg,
+             Condition(0), Condition(0), Condition(0));
       break;
     }
     case kS390_I8x16GtU: {