8271567: AArch64: AES Galois CounterMode (GCM) interleaved implementation using vector instructions

Reviewed-by: ngasson, adinn, xliu

Author: Andrew Haley

src/hotspot/cpu/aarch64/assembler_aarch64.hpp

@@ -647,22 +647,22 @@ typedef enum {
 
 class Assembler : public AbstractAssembler {
 
+public:
+
 #ifndef PRODUCT
   static const uintptr_t asm_bp;
 
-  void emit_long(jint x) {
+  void emit_int32(jint x) {
     if ((uintptr_t)pc() == asm_bp)
       NOP();
     AbstractAssembler::emit_int32(x);
   }
 #else
-  void emit_long(jint x) {
+  void emit_int32(jint x) {
     AbstractAssembler::emit_int32(x);
   }
 #endif
 
-public:
-
   enum { instruction_size = 4 };
 
   //---<  calculate length of instruction  >---

src/hotspot/cpu/aarch64/macroAssembler_aarch64.hpp

@@ -1296,11 +1296,37 @@ class MacroAssembler: public Assembler {
   void kernel_crc32c_using_crc32c(Register crc, Register buf,
         Register len, Register tmp0, Register tmp1, Register tmp2,
         Register tmp3);
+
+  void ghash_modmul (FloatRegister result,
+                     FloatRegister result_lo, FloatRegister result_hi, FloatRegister b,
+                     FloatRegister a, FloatRegister vzr, FloatRegister a1_xor_a0, FloatRegister p,
+                     FloatRegister t1, FloatRegister t2, FloatRegister t3);
+  void ghash_load_wide(int index, Register data, FloatRegister result, FloatRegister state);
 public:
   void multiply_to_len(Register x, Register xlen, Register y, Register ylen, Register z,
                        Register zlen, Register tmp1, Register tmp2, Register tmp3,
                        Register tmp4, Register tmp5, Register tmp6, Register tmp7);
   void mul_add(Register out, Register in, Register offs, Register len, Register k);
+  void ghash_multiply(FloatRegister result_lo, FloatRegister result_hi,
+                      FloatRegister a, FloatRegister b, FloatRegister a1_xor_a0,
+                      FloatRegister tmp1, FloatRegister tmp2, FloatRegister tmp3);
+  void ghash_multiply_wide(int index,
+                           FloatRegister result_lo, FloatRegister result_hi,
+                           FloatRegister a, FloatRegister b, FloatRegister a1_xor_a0,
+                           FloatRegister tmp1, FloatRegister tmp2, FloatRegister tmp3);
+  void ghash_reduce(FloatRegister result, FloatRegister lo, FloatRegister hi,
+                    FloatRegister p, FloatRegister z, FloatRegister t1);
+  void ghash_reduce_wide(int index, FloatRegister result, FloatRegister lo, FloatRegister hi,
+                    FloatRegister p, FloatRegister z, FloatRegister t1);
+  void ghash_processBlocks_wide(address p, Register state, Register subkeyH,
+                                Register data, Register blocks, int unrolls);
+
+
+  void aesenc_loadkeys(Register key, Register keylen);
+  void aesecb_encrypt(Register from, Register to, Register keylen,
+                      FloatRegister data = v0, int unrolls = 1);
+  void aesecb_decrypt(Register from, Register to, Register key, Register keylen);
+  void aes_round(FloatRegister input, FloatRegister subkey);
 
   // Place an ISB after code may have been modified due to a safepoint.
   void safepoint_isb();