PPC [simd]: Implement vector population count

Change-Id: I88af87b611415753d1063d0b203f3c846fdecd57
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2778082Reviewed-by: Junliang Yan <junyan@redhat.com>
Commit-Queue: Milad Fa <mfarazma@redhat.com>
Cr-Commit-Position: refs/heads/master@{#73579}

src/codegen/ppc/constants-ppc.h

@@ -2424,7 +2424,9 @@ using Instr = uint32_t;
   /* Vector Unpack Low Signed Byte */      \
   V(vupklsb, VUPKLSB, 0x1000028E)          \
   /* Vector Unpack High Signed Byte */     \
-  V(vupkhsb, VUPKHSB, 0x1000020E)
+  V(vupkhsb, VUPKHSB, 0x1000020E)          \
+  /* Vector Population Count Byte */       \
+  V(vpopcntb, VPOPCNTB, 0x10000703)
 
 #define PPC_VX_OPCODE_UNUSED_LIST(V)                                      \
   /* Decimal Add Modulo */                                                \
@@ -2503,8 +2505,6 @@ using Instr = uint32_t;
   V(vpmsumh, VPMSUMH, 0x10000448)                                         \
   /* Vector Polynomial Multiply-Sum Word */                               \
   V(vpmsumw, VPMSUMW, 0x10000488)                                         \
-  /* Vector Population Count Byte */                                      \
-  V(vpopcntb, VPOPCNTB, 0x10000703)                                       \
   /* Vector Population Count Doubleword */                                \
   V(vpopcntd, VPOPCNTD, 0x100007C3)                                       \
   /* Vector Population Count Halfword */                                  \

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

@@ -3823,6 +3823,10 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       __ vinsertd(dst, kScratchSimd128Reg, Operand(lane_number));
       break;
     }
+    case kPPC_I8x16Popcnt: {
+      __ vpopcntb(i.OutputSimd128Register(), i.InputSimd128Register(0));
+      break;
+    }
     case kPPC_StoreCompressTagged: {
       ASSEMBLE_STORE_INTEGER(StoreTaggedField, StoreTaggedFieldX);
       break;

src/compiler/backend/ppc/instruction-codes-ppc.h

@@ -386,6 +386,7 @@ namespace compiler {
   V(PPC_I8x16Shuffle)                \
   V(PPC_I8x16Swizzle)                \
   V(PPC_I8x16BitMask)                \
+  V(PPC_I8x16Popcnt)                 \
   V(PPC_I64x2AllTrue)                \
   V(PPC_I32x4AllTrue)                \
   V(PPC_I16x8AllTrue)                \

src/compiler/backend/ppc/instruction-scheduler-ppc.cc

@@ -309,6 +309,7 @@ int InstructionScheduler::GetTargetInstructionFlags(
     case kPPC_I8x16Shuffle:
     case kPPC_I8x16Swizzle:
     case kPPC_I8x16BitMask:
+    case kPPC_I8x16Popcnt:
     case kPPC_I64x2AllTrue:
     case kPPC_I32x4AllTrue:
     case kPPC_I16x8AllTrue:

src/compiler/backend/ppc/instruction-selector-ppc.cc

@@ -2293,6 +2293,7 @@ void InstructionSelector::VisitInt64AbsWithOverflow(Node* node) {
   V(I16x8Abs)                  \
   V(I8x16Neg)                  \
   V(I8x16Abs)                  \
+  V(I8x16Popcnt)               \
   V(I16x8SConvertI8x16Low)     \
   V(I16x8SConvertI8x16High)    \
   V(I16x8UConvertI8x16Low)     \
@@ -2489,7 +2490,6 @@ void InstructionSelector::VisitS128Const(Node* node) {
   }
 }
 
-void InstructionSelector::VisitI8x16Popcnt(Node* node) { UNIMPLEMENTED(); }
 void InstructionSelector::VisitI64x2GtS(Node* node) { UNIMPLEMENTED(); }
 void InstructionSelector::VisitI64x2GeS(Node* node) { UNIMPLEMENTED(); }
 void InstructionSelector::VisitI64x2Abs(Node* node) { UNIMPLEMENTED(); }

src/execution/ppc/simulator-ppc.cc

@@ -4842,6 +4842,17 @@ void Simulator::ExecuteGeneric(Instruction* instr) {
       break;
     }
 #undef VECTOR_ROUNDING_AVERAGE
+    case VPOPCNTB: {
+      int t = instr->RTValue();
+      int b = instr->RBValue();
+      FOR_EACH_LANE(i, uint8_t) {
+        set_simd_register_by_lane<uint8_t>(
+            t, i,
+            base::bits::CountPopulation(
+                get_simd_register_by_lane<uint8_t>(b, i)));
+      }
+      break;
+    }
 #undef FOR_EACH_LANE
 #undef DECODE_VX_INSTRUCTION
 #undef GET_ADDRESS