8318217: RISC-V: C2 VectorizedHashCode

Reviewed-by: mli, fyang
openjdk · Dec 12, 2023 · 6359b4e · 6359b4e · openjdk-notifier · Dec 12, 2023
1 parent ce4b257
commit 6359b4e
Show file tree

Hide file tree

Showing 4 changed files with 139 additions and 0 deletions.
diff --git a/src/hotspot/cpu/riscv/c2_MacroAssembler_riscv.cpp b/src/hotspot/cpu/riscv/c2_MacroAssembler_riscv.cpp
@@ -1459,6 +1459,112 @@ void C2_MacroAssembler::string_equals(Register a1, Register a2,
   BLOCK_COMMENT("} string_equals");
 }
 
+// jdk.internal.util.ArraysSupport.vectorizedHashCode
+void C2_MacroAssembler::arrays_hashcode(Register ary, Register cnt, Register result,
+                                        Register tmp1, Register tmp2, Register tmp3,
+                                        Register tmp4, Register tmp5, Register tmp6,
+                                        BasicType eltype)
+{
+  assert_different_registers(ary, cnt, result, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, t0, t1);
+
+  const int elsize = arrays_hashcode_elsize(eltype);
+  const int chunks_end_shift = exact_log2(elsize);
+
+  switch (eltype) {
+  case T_BOOLEAN: BLOCK_COMMENT("arrays_hashcode(unsigned byte) {"); break;
+  case T_CHAR:    BLOCK_COMMENT("arrays_hashcode(char) {");          break;
+  case T_BYTE:    BLOCK_COMMENT("arrays_hashcode(byte) {");          break;
+  case T_SHORT:   BLOCK_COMMENT("arrays_hashcode(short) {");         break;
+  case T_INT:     BLOCK_COMMENT("arrays_hashcode(int) {");           break;
+  default:
+    ShouldNotReachHere();
+  }
+
+  const int stride = 4;
+  const Register pow31_4 = tmp1;
+  const Register pow31_3 = tmp2;
+  const Register pow31_2 = tmp3;
+  const Register chunks  = tmp4;
+  const Register chunks_end = chunks;
+
+  Label DONE, TAIL, TAIL_LOOP, WIDE_LOOP;
+
+  // result has a value initially
+
+  beqz(cnt, DONE);
+
+  andi(chunks, cnt, ~(stride-1));
+  beqz(chunks, TAIL);
+
+  mv(pow31_4, 923521);           // [31^^4]
+  mv(pow31_3,  29791);           // [31^^3]
+  mv(pow31_2,    961);           // [31^^2]
+
+  slli(chunks_end, chunks, chunks_end_shift);
+  add(chunks_end, ary, chunks_end);
+  andi(cnt, cnt, stride-1);      // don't forget about tail!
+
+  bind(WIDE_LOOP);
+  mulw(result, result, pow31_4); // 31^^4 * h
+  arrays_hashcode_elload(t0,   Address(ary, 0 * elsize), eltype);
+  arrays_hashcode_elload(t1,   Address(ary, 1 * elsize), eltype);
+  arrays_hashcode_elload(tmp5, Address(ary, 2 * elsize), eltype);
+  arrays_hashcode_elload(tmp6, Address(ary, 3 * elsize), eltype);
+  mulw(t0, t0, pow31_3);         // 31^^3 * ary[i+0]
+  addw(result, result, t0);
+  mulw(t1, t1, pow31_2);         // 31^^2 * ary[i+1]
+  addw(result, result, t1);
+  slli(t0, tmp5, 5);             // optimize 31^^1 * ary[i+2]
+  subw(tmp5, t0, tmp5);          // with ary[i+2]<<5 - ary[i+2]
+  addw(result, result, tmp5);
+  addw(result, result, tmp6);    // 31^^4 * h + 31^^3 * ary[i+0] + 31^^2 * ary[i+1]
+                                 //           + 31^^1 * ary[i+2] + 31^^0 * ary[i+3]
+  addi(ary, ary, elsize * stride);
+  bne(ary, chunks_end, WIDE_LOOP);
+  beqz(cnt, DONE);
+
+  bind(TAIL);
+  slli(chunks_end, cnt, chunks_end_shift);
+  add(chunks_end, ary, chunks_end);
+
+  bind(TAIL_LOOP);
+  arrays_hashcode_elload(t0, Address(ary), eltype);
+  slli(t1, result, 5);           // optimize 31 * result
+  subw(result, t1, result);      // with result<<5 - result
+  addw(result, result, t0);
+  addi(ary, ary, elsize);
+  bne(ary, chunks_end, TAIL_LOOP);
+
+  bind(DONE);
+  BLOCK_COMMENT("} // arrays_hashcode");
+}
+
+int C2_MacroAssembler::arrays_hashcode_elsize(BasicType eltype) {
+  switch (eltype) {
+  case T_BOOLEAN: return sizeof(jboolean);
+  case T_BYTE:    return sizeof(jbyte);
+  case T_SHORT:   return sizeof(jshort);
+  case T_CHAR:    return sizeof(jchar);
+  case T_INT:     return sizeof(jint);
+  default:
+    ShouldNotReachHere();
+    return -1;
+  }
+}
+
+void C2_MacroAssembler::arrays_hashcode_elload(Register dst, Address src, BasicType eltype) {
+  switch (eltype) {
+  // T_BOOLEAN used as surrogate for unsigned byte
+  case T_BOOLEAN: lbu(dst, src);   break;
+  case T_BYTE:     lb(dst, src);   break;
+  case T_SHORT:    lh(dst, src);   break;
+  case T_CHAR:    lhu(dst, src);   break;
+  case T_INT:      lw(dst, src);   break;
+  default:
+    ShouldNotReachHere();
+  }
+}
+
 typedef void (Assembler::*conditional_branch_insn)(Register op1, Register op2, Label& label, bool is_far);
 typedef void (MacroAssembler::*float_conditional_branch_insn)(FloatRegister op1, FloatRegister op2, Label& label,
                                                               bool is_far, bool is_unordered);

diff --git a/src/hotspot/cpu/riscv/c2_MacroAssembler_riscv.hpp b/src/hotspot/cpu/riscv/c2_MacroAssembler_riscv.hpp
@@ -82,6 +82,15 @@
                      Register result, Register cnt1,
                      int elem_size);
 
+  void arrays_hashcode(Register ary, Register cnt, Register result,
+                       Register tmp1, Register tmp2,
+                       Register tmp3, Register tmp4,
+                       Register tmp5, Register tmp6,
+                       BasicType eltype);
+  // helper function for arrays_hashcode
+  int arrays_hashcode_elsize(BasicType eltype);
+  void arrays_hashcode_elload(Register dst, Address src, BasicType eltype);
+
   void string_equals(Register r1, Register r2,
                      Register result, Register cnt1,
                      int elem_size);

diff --git a/src/hotspot/cpu/riscv/riscv.ad b/src/hotspot/cpu/riscv/riscv.ad
@@ -10371,6 +10371,26 @@ instruct array_equalsC(iRegP_R11 ary1, iRegP_R12 ary2, iRegI_R10 result,
   ins_pipe(pipe_class_memory);
 %}
 
+// fast ArraysSupport.vectorizedHashCode
+instruct arrays_hashcode(iRegP_R11 ary, iRegI_R12 cnt, iRegI_R10 result, immI basic_type,
+                         iRegLNoSp tmp1, iRegLNoSp tmp2,
+                         iRegLNoSp tmp3, iRegLNoSp tmp4,
+                         iRegLNoSp tmp5, iRegLNoSp tmp6, rFlagsReg cr)
+%{
+  match(Set result (VectorizedHashCode (Binary ary cnt) (Binary result basic_type)));
+  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, TEMP tmp5, TEMP tmp6,
+         USE_KILL ary, USE_KILL cnt, USE basic_type, KILL cr);
+
+  format %{ "Array HashCode array[] $ary,$cnt,$result,$basic_type -> $result   // KILL all" %}
+  ins_encode %{
+    __ arrays_hashcode($ary$$Register, $cnt$$Register, $result$$Register,
+                       $tmp1$$Register, $tmp2$$Register, $tmp3$$Register,
+                       $tmp4$$Register, $tmp5$$Register, $tmp6$$Register,
+                       (BasicType)$basic_type$$constant);
+  %}
+  ins_pipe(pipe_class_memory);
+%}
+
 // ============================================================================
 // Safepoint Instructions
 

diff --git a/src/hotspot/cpu/riscv/vm_version_riscv.cpp b/src/hotspot/cpu/riscv/vm_version_riscv.cpp
@@ -315,6 +315,10 @@ void VM_Version::c2_initialize() {
     }
   }
 
+  if (FLAG_IS_DEFAULT(UseVectorizedHashCodeIntrinsic)) {
+    FLAG_SET_DEFAULT(UseVectorizedHashCodeIntrinsic, true);
+  }
+
   if (!UseZicbop) {
     if (!FLAG_IS_DEFAULT(AllocatePrefetchStyle)) {
       warning("Zicbop is not available on this CPU");