8329538: Accelerate P256 on x86_64 using Montgomery intrinsic

Reviewed-by: ihse, ascarpino, sviswanathan

Author: vpaprotsk

make/jdk/src/classes/build/tools/intpoly/FieldGen.java

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018, 2022, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2018, 2024, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -778,7 +778,7 @@ private String generate(FieldParams params) throws IOException {
         result.appendLine("}");
 
         result.appendLine("@Override");
-        result.appendLine("protected void mult(long[] a, long[] b, long[] r) {");
+        result.appendLine("protected int mult(long[] a, long[] b, long[] r) {");
         result.incrIndent();
         for (int i = 0; i < 2 * params.getNumLimbs() - 1; i++) {
             result.appendIndent();
@@ -804,6 +804,9 @@ private String generate(FieldParams params) throws IOException {
             }
         }
         result.append(");\n");
+        result.appendIndent();
+        result.append("return 0;");
+        result.appendLine();
         result.decrIndent();
         result.appendLine("}");
 
@@ -833,7 +836,7 @@ private String generate(FieldParams params) throws IOException {
         //      }
         //  }
         result.appendLine("@Override");
-        result.appendLine("protected void square(long[] a, long[] r) {");
+        result.appendLine("protected int square(long[] a, long[] r) {");
         result.incrIndent();
         for (int i = 0; i < 2 * params.getNumLimbs() - 1; i++) {
             result.appendIndent();
@@ -874,6 +877,9 @@ private String generate(FieldParams params) throws IOException {
             }
         }
         result.append(");\n");
+        result.appendIndent();
+        result.append("return 0;");
+        result.appendLine();
         result.decrIndent();
         result.appendLine("}");
 

make/test/BuildMicrobenchmark.gmk

@@ -109,6 +109,8 @@ $(eval $(call SetupJavaCompilation, BUILD_JDK_MICROBENCHMARK, \
         --add-exports java.base/jdk.internal.vm=ALL-UNNAMED \
         --add-exports java.base/sun.invoke.util=ALL-UNNAMED \
         --add-exports java.base/sun.security.util=ALL-UNNAMED \
+        --add-exports java.base/sun.security.util.math=ALL-UNNAMED \
+        --add-exports java.base/sun.security.util.math.intpoly=ALL-UNNAMED \
         --enable-preview \
         -XDsuppressNotes \
         -processor org.openjdk.jmh.generators.BenchmarkProcessor, \

src/hotspot/cpu/x86/macroAssembler_x86.hpp

@@ -1549,6 +1549,8 @@ class MacroAssembler: public Assembler {
       Assembler::evpsrlvd(dst, mask, nds, src, merge, vector_len);
     }
   }
+
+  using Assembler::evpsrlq;
   void evpsrlq(XMMRegister dst, KRegister mask, XMMRegister nds, XMMRegister src, bool merge, int vector_len, bool is_varshift) {
     if (!is_varshift) {
       Assembler::evpsrlq(dst, mask, nds, src, merge, vector_len);
@@ -1570,6 +1572,7 @@ class MacroAssembler: public Assembler {
       Assembler::evpsravd(dst, mask, nds, src, merge, vector_len);
     }
   }
+  using Assembler::evpsraq;
   void evpsraq(XMMRegister dst, KRegister mask, XMMRegister nds, XMMRegister src, bool merge, int vector_len, bool is_varshift) {
     if (!is_varshift) {
       Assembler::evpsraq(dst, mask, nds, src, merge, vector_len);

src/hotspot/cpu/x86/stubGenerator_x86_64.cpp

@@ -4255,6 +4255,11 @@ void StubGenerator::generate_compiler_stubs() {
     StubRoutines::_poly1305_processBlocks = generate_poly1305_processBlocks();
   }
 
+  if (UseIntPolyIntrinsics) {
+    StubRoutines::_intpoly_montgomeryMult_P256 = generate_intpoly_montgomeryMult_P256();
+    StubRoutines::_intpoly_assign = generate_intpoly_assign();
+  }
+
   if (UseMD5Intrinsics) {
     StubRoutines::_md5_implCompress = generate_md5_implCompress(false, "md5_implCompress");
     StubRoutines::_md5_implCompressMB = generate_md5_implCompress(true, "md5_implCompressMB");

src/hotspot/cpu/x86/stubGenerator_x86_64.hpp

@@ -483,6 +483,9 @@ class StubGenerator: public StubCodeGenerator {
                                const XMMRegister P2L, const XMMRegister P2H,
                                const XMMRegister YTMP1, const Register rscratch);
 
+  address generate_intpoly_montgomeryMult_P256();
+  address generate_intpoly_assign();
+
   // BASE64 stubs
 
   address base64_shuffle_addr();