Merge branch 'openjdk:master' into backport-sendaoYan-c0c76703-master

Author: sendaoYan

make/autoconf/boot-jdk.m4

@@ -444,6 +444,9 @@ AC_DEFUN_ONCE([BOOTJDK_SETUP_BOOT_JDK_ARGUMENTS],
   # Force en-US environment
   UTIL_ADD_JVM_ARG_IF_OK([-Duser.language=en -Duser.country=US],boot_jdk_jvmargs,[$JAVA])
 
+  UTIL_ADD_JVM_ARG_IF_OK([-Xlog:all=off:stdout],boot_jdk_jvmargs,[$JAVA])
+  UTIL_ADD_JVM_ARG_IF_OK([-Xlog:all=warning:stderr],boot_jdk_jvmargs,[$JAVA])
+
   if test "x$BOOTJDK_USE_LOCAL_CDS" = xtrue; then
     # Use our own CDS archive
     UTIL_ADD_JVM_ARG_IF_OK([$boot_jdk_cds_args -Xshare:auto],boot_jdk_jvmargs,[$JAVA])

src/hotspot/cpu/aarch64/aarch64.ad

@@ -3401,7 +3401,7 @@ encode %{
     }
   %}
 
-  /// mov envcodings
+  // mov encodings
 
   enc_class aarch64_enc_movw_imm(iRegI dst, immI src) %{
     C2_MacroAssembler _masm(&cbuf);

src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp

@@ -5913,7 +5913,7 @@ void MacroAssembler::fill_words(Register base, Register cnt, Register value)
 // - sun/nio/cs/ISO_8859_1$Encoder.implEncodeISOArray
 //     return the number of characters copied.
 // - java/lang/StringUTF16.compress
-//     return zero (0) if copy fails, otherwise 'len'.
+//     return index of non-latin1 character if copy fails, otherwise 'len'.
 //
 // This version always returns the number of characters copied, and does not
 // clobber the 'len' register. A successful copy will complete with the post-
@@ -6130,15 +6130,15 @@ address MacroAssembler::byte_array_inflate(Register src, Register dst, Register
 }
 
 // Compress char[] array to byte[].
+// Intrinsic for java.lang.StringUTF16.compress(char[] src, int srcOff, byte[] dst, int dstOff, int len)
+// Return the array length if every element in array can be encoded,
+// otherwise, the index of first non-latin1 (> 0xff) character.
 void MacroAssembler::char_array_compress(Register src, Register dst, Register len,
                                          Register res,
                                          FloatRegister tmp0, FloatRegister tmp1,
                                          FloatRegister tmp2, FloatRegister tmp3,
                                          FloatRegister tmp4, FloatRegister tmp5) {
   encode_iso_array(src, dst, len, res, false, tmp0, tmp1, tmp2, tmp3, tmp4, tmp5);
-  // Adjust result: res == len ? len : 0
-  cmp(len, res);
-  csel(res, res, zr, EQ);
 }
 
 // java.math.round(double a)

src/hotspot/cpu/ppc/ppc.ad

@@ -12799,16 +12799,8 @@ instruct string_compress(rarg1RegP src, rarg2RegP dst, iRegIsrc len, iRegIdst re
   ins_cost(300);
   format %{ "String Compress $src,$dst,$len -> $result \t// KILL $tmp1, $tmp2, $tmp3, $tmp4, $tmp5" %}
   ins_encode %{
-    Label Lskip, Ldone;
-    __ li($result$$Register, 0);
-    __ string_compress_16($src$$Register, $dst$$Register, $len$$Register, $tmp1$$Register,
-                          $tmp2$$Register, $tmp3$$Register, $tmp4$$Register, $tmp5$$Register, Ldone);
-    __ rldicl_($tmp1$$Register, $len$$Register, 0, 64-3); // Remaining characters.
-    __ beq(CCR0, Lskip);
-    __ string_compress($src$$Register, $dst$$Register, $tmp1$$Register, $tmp2$$Register, Ldone);
-    __ bind(Lskip);
-    __ mr($result$$Register, $len$$Register);
-    __ bind(Ldone);
+    __ encode_iso_array($src$$Register, $dst$$Register, $len$$Register, $tmp1$$Register, $tmp2$$Register,
+                        $tmp3$$Register, $tmp4$$Register, $tmp5$$Register, $result$$Register, false);
   %}
   ins_pipe(pipe_class_default);
 %}

src/hotspot/cpu/riscv/c2_MacroAssembler_riscv.cpp

@@ -1768,22 +1768,20 @@ void C2_MacroAssembler::byte_array_inflate_v(Register src, Register dst, Registe
 }
 
 // Compress char[] array to byte[].
-// result: the array length if every element in array can be encoded; 0, otherwise.
+// Intrinsic for java.lang.StringUTF16.compress(char[] src, int srcOff, byte[] dst, int dstOff, int len)
+// result: the array length if every element in array can be encoded,
+// otherwise, the index of first non-latin1 (> 0xff) character.
 void C2_MacroAssembler::char_array_compress_v(Register src, Register dst, Register len,
                                               Register result, Register tmp) {
-  Label done;
   encode_iso_array_v(src, dst, len, result, tmp, false);
-  beqz(len, done);
-  mv(result, zr);
-  bind(done);
 }
 
 // Intrinsic for
 //
 // - sun/nio/cs/ISO_8859_1$Encoder.implEncodeISOArray
 //     return the number of characters copied.
 // - java/lang/StringUTF16.compress
-//     return zero (0) if copy fails, otherwise 'len'.
+//     return index of non-latin1 character if copy fails, otherwise 'len'.
 //
 // This version always returns the number of characters copied. A successful
 // copy will complete with the post-condition: 'res' == 'len', while an

src/hotspot/cpu/s390/s390.ad

@@ -10177,7 +10177,7 @@ instruct string_compress(iRegP src, iRegP dst, iRegI result, iRegI len, iRegI tm
   format %{ "String Compress $src->$dst($len) -> $result" %}
   ins_encode %{
     __ string_compress($result$$Register, $src$$Register, $dst$$Register, $len$$Register,
-                       $tmp$$Register, false, false);
+                       $tmp$$Register, true, false);
   %}
   ins_pipe(pipe_class_dummy);
 %}

src/hotspot/cpu/x86/macroAssembler_x86.cpp

@@ -8837,23 +8837,27 @@ void MacroAssembler::crc32c_ipl_alg2_alt2(Register in_out, Register in1, Registe
 #undef BLOCK_COMMENT
 
 // Compress char[] array to byte[].
-//   ..\jdk\src\java.base\share\classes\java\lang\StringUTF16.java
+// Intrinsic for java.lang.StringUTF16.compress(char[] src, int srcOff, byte[] dst, int dstOff, int len)
+// Return the array length if every element in array can be encoded,
+// otherwise, the index of first non-latin1 (> 0xff) character.
 //   @IntrinsicCandidate
-//   private static int compress(char[] src, int srcOff, byte[] dst, int dstOff, int len) {
+//   public static int compress(char[] src, int srcOff, byte[] dst, int dstOff, int len) {
 //     for (int i = 0; i < len; i++) {
-//       int c = src[srcOff++];
-//       if (c >>> 8 != 0) {
-//         return 0;
+//       char c = src[srcOff];
+//       if (c > 0xff) {
+//           return i;  // return index of non-latin1 char
 //       }
-//       dst[dstOff++] = (byte)c;
+//       dst[dstOff] = (byte)c;
+//       srcOff++;
+//       dstOff++;
 //     }
 //     return len;
 //   }
 void MacroAssembler::char_array_compress(Register src, Register dst, Register len,
   XMMRegister tmp1Reg, XMMRegister tmp2Reg,
   XMMRegister tmp3Reg, XMMRegister tmp4Reg,
   Register tmp5, Register result, KRegister mask1, KRegister mask2) {
-  Label copy_chars_loop, return_length, return_zero, done;
+  Label copy_chars_loop, done, reset_sp, copy_tail;
 
   // rsi: src
   // rdi: dst
@@ -8868,28 +8872,28 @@ void MacroAssembler::char_array_compress(Register src, Register dst, Register le
   assert(len != result, "");
 
   // save length for return
-  push(len);
+  movl(result, len);
 
   if ((AVX3Threshold == 0) && (UseAVX > 2) && // AVX512
     VM_Version::supports_avx512vlbw() &&
     VM_Version::supports_bmi2()) {
 
-    Label copy_32_loop, copy_loop_tail, below_threshold;
+    Label copy_32_loop, copy_loop_tail, below_threshold, reset_for_copy_tail;
 
     // alignment
     Label post_alignment;
 
-    // if length of the string is less than 16, handle it in an old fashioned way
+    // if length of the string is less than 32, handle it the old fashioned way
     testl(len, -32);
     jcc(Assembler::zero, below_threshold);
 
     // First check whether a character is compressible ( <= 0xFF).
     // Create mask to test for Unicode chars inside zmm vector
-    movl(result, 0x00FF);
-    evpbroadcastw(tmp2Reg, result, Assembler::AVX_512bit);
+    movl(tmp5, 0x00FF);
+    evpbroadcastw(tmp2Reg, tmp5, Assembler::AVX_512bit);
 
     testl(len, -64);
-    jcc(Assembler::zero, post_alignment);
+    jccb(Assembler::zero, post_alignment);
 
     movl(tmp5, dst);
     andl(tmp5, (32 - 1));
@@ -8898,18 +8902,19 @@ void MacroAssembler::char_array_compress(Register src, Register dst, Register le
 
     // bail out when there is nothing to be done
     testl(tmp5, 0xFFFFFFFF);
-    jcc(Assembler::zero, post_alignment);
+    jccb(Assembler::zero, post_alignment);
 
     // ~(~0 << len), where len is the # of remaining elements to process
-    movl(result, 0xFFFFFFFF);
-    shlxl(result, result, tmp5);
-    notl(result);
-    kmovdl(mask2, result);
+    movl(len, 0xFFFFFFFF);
+    shlxl(len, len, tmp5);
+    notl(len);
+    kmovdl(mask2, len);
+    movl(len, result);
 
     evmovdquw(tmp1Reg, mask2, Address(src, 0), /*merge*/ false, Assembler::AVX_512bit);
     evpcmpw(mask1, mask2, tmp1Reg, tmp2Reg, Assembler::le, /*signed*/ false, Assembler::AVX_512bit);
     ktestd(mask1, mask2);
-    jcc(Assembler::carryClear, return_zero);
+    jcc(Assembler::carryClear, copy_tail);
 
     evpmovwb(Address(dst, 0), mask2, tmp1Reg, Assembler::AVX_512bit);
 
@@ -8924,7 +8929,7 @@ void MacroAssembler::char_array_compress(Register src, Register dst, Register le
     movl(tmp5, len);
     andl(tmp5, (32 - 1));    // tail count (in chars)
     andl(len, ~(32 - 1));    // vector count (in chars)
-    jcc(Assembler::zero, copy_loop_tail);
+    jccb(Assembler::zero, copy_loop_tail);
 
     lea(src, Address(src, len, Address::times_2));
     lea(dst, Address(dst, len, Address::times_1));
@@ -8934,55 +8939,60 @@ void MacroAssembler::char_array_compress(Register src, Register dst, Register le
     evmovdquw(tmp1Reg, Address(src, len, Address::times_2), Assembler::AVX_512bit);
     evpcmpuw(mask1, tmp1Reg, tmp2Reg, Assembler::le, Assembler::AVX_512bit);
     kortestdl(mask1, mask1);
-    jcc(Assembler::carryClear, return_zero);
+    jccb(Assembler::carryClear, reset_for_copy_tail);
 
     // All elements in current processed chunk are valid candidates for
     // compression. Write a truncated byte elements to the memory.
     evpmovwb(Address(dst, len, Address::times_1), tmp1Reg, Assembler::AVX_512bit);
     addptr(len, 32);
-    jcc(Assembler::notZero, copy_32_loop);
+    jccb(Assembler::notZero, copy_32_loop);
 
     bind(copy_loop_tail);
     // bail out when there is nothing to be done
     testl(tmp5, 0xFFFFFFFF);
-    jcc(Assembler::zero, return_length);
+    jcc(Assembler::zero, done);
 
     movl(len, tmp5);
 
     // ~(~0 << len), where len is the # of remaining elements to process
-    movl(result, 0xFFFFFFFF);
-    shlxl(result, result, len);
-    notl(result);
+    movl(tmp5, 0xFFFFFFFF);
+    shlxl(tmp5, tmp5, len);
+    notl(tmp5);
 
-    kmovdl(mask2, result);
+    kmovdl(mask2, tmp5);
 
     evmovdquw(tmp1Reg, mask2, Address(src, 0), /*merge*/ false, Assembler::AVX_512bit);
     evpcmpw(mask1, mask2, tmp1Reg, tmp2Reg, Assembler::le, /*signed*/ false, Assembler::AVX_512bit);
     ktestd(mask1, mask2);
-    jcc(Assembler::carryClear, return_zero);
+    jcc(Assembler::carryClear, copy_tail);
 
     evpmovwb(Address(dst, 0), mask2, tmp1Reg, Assembler::AVX_512bit);
-    jmp(return_length);
+    jmp(done);
+
+    bind(reset_for_copy_tail);
+    lea(src, Address(src, tmp5, Address::times_2));
+    lea(dst, Address(dst, tmp5, Address::times_1));
+    subptr(len, tmp5);
+    jmp(copy_chars_loop);
 
     bind(below_threshold);
   }
 
   if (UseSSE42Intrinsics) {
-    Label copy_32_loop, copy_16, copy_tail;
+    Label copy_32_loop, copy_16, copy_tail_sse, reset_for_copy_tail;
 
-    movl(result, len);
+    // vectored compression
+    testl(len, 0xfffffff8);
+    jcc(Assembler::zero, copy_tail);
 
     movl(tmp5, 0xff00ff00);   // create mask to test for Unicode chars in vectors
+    movdl(tmp1Reg, tmp5);
+    pshufd(tmp1Reg, tmp1Reg, 0);   // store Unicode mask in tmp1Reg
 
-    // vectored compression
-    andl(len, 0xfffffff0);    // vector count (in chars)
-    andl(result, 0x0000000f);    // tail count (in chars)
-    testl(len, len);
-    jcc(Assembler::zero, copy_16);
+    andl(len, 0xfffffff0);
+    jccb(Assembler::zero, copy_16);
 
     // compress 16 chars per iter
-    movdl(tmp1Reg, tmp5);
-    pshufd(tmp1Reg, tmp1Reg, 0);   // store Unicode mask in tmp1Reg
     pxor(tmp4Reg, tmp4Reg);
 
     lea(src, Address(src, len, Address::times_2));
@@ -8995,59 +9005,60 @@ void MacroAssembler::char_array_compress(Register src, Register dst, Register le
     movdqu(tmp3Reg, Address(src, len, Address::times_2, 16)); // load next 8 characters
     por(tmp4Reg, tmp3Reg);
     ptest(tmp4Reg, tmp1Reg);       // check for Unicode chars in next vector
-    jcc(Assembler::notZero, return_zero);
+    jccb(Assembler::notZero, reset_for_copy_tail);
     packuswb(tmp2Reg, tmp3Reg);    // only ASCII chars; compress each to 1 byte
     movdqu(Address(dst, len, Address::times_1), tmp2Reg);
     addptr(len, 16);
-    jcc(Assembler::notZero, copy_32_loop);
+    jccb(Assembler::notZero, copy_32_loop);
 
     // compress next vector of 8 chars (if any)
     bind(copy_16);
-    movl(len, result);
-    andl(len, 0xfffffff8);    // vector count (in chars)
-    andl(result, 0x00000007);    // tail count (in chars)
-    testl(len, len);
-    jccb(Assembler::zero, copy_tail);
+    // len = 0
+    testl(result, 0x00000008);     // check if there's a block of 8 chars to compress
+    jccb(Assembler::zero, copy_tail_sse);
 
-    movdl(tmp1Reg, tmp5);
-    pshufd(tmp1Reg, tmp1Reg, 0);   // store Unicode mask in tmp1Reg
     pxor(tmp3Reg, tmp3Reg);
 
     movdqu(tmp2Reg, Address(src, 0));
     ptest(tmp2Reg, tmp1Reg);       // check for Unicode chars in vector
-    jccb(Assembler::notZero, return_zero);
+    jccb(Assembler::notZero, reset_for_copy_tail);
     packuswb(tmp2Reg, tmp3Reg);    // only LATIN1 chars; compress each to 1 byte
     movq(Address(dst, 0), tmp2Reg);
     addptr(src, 16);
     addptr(dst, 8);
+    jmpb(copy_tail_sse);
 
-    bind(copy_tail);
+    bind(reset_for_copy_tail);
+    movl(tmp5, result);
+    andl(tmp5, 0x0000000f);
+    lea(src, Address(src, tmp5, Address::times_2));
+    lea(dst, Address(dst, tmp5, Address::times_1));
+    subptr(len, tmp5);
+    jmpb(copy_chars_loop);
+
+    bind(copy_tail_sse);
     movl(len, result);
+    andl(len, 0x00000007);    // tail count (in chars)
   }
   // compress 1 char per iter
+  bind(copy_tail);
   testl(len, len);
-  jccb(Assembler::zero, return_length);
+  jccb(Assembler::zero, done);
   lea(src, Address(src, len, Address::times_2));
   lea(dst, Address(dst, len, Address::times_1));
   negptr(len);
 
   bind(copy_chars_loop);
-  load_unsigned_short(result, Address(src, len, Address::times_2));
-  testl(result, 0xff00);      // check if Unicode char
-  jccb(Assembler::notZero, return_zero);
-  movb(Address(dst, len, Address::times_1), result);  // ASCII char; compress to 1 byte
+  load_unsigned_short(tmp5, Address(src, len, Address::times_2));
+  testl(tmp5, 0xff00);      // check if Unicode char
+  jccb(Assembler::notZero, reset_sp);
+  movb(Address(dst, len, Address::times_1), tmp5);  // ASCII char; compress to 1 byte
   increment(len);
-  jcc(Assembler::notZero, copy_chars_loop);
+  jccb(Assembler::notZero, copy_chars_loop);
 
-  // if compression succeeded, return length
-  bind(return_length);
-  pop(result);
-  jmpb(done);
-
-  // if compression failed, return 0
-  bind(return_zero);
-  xorl(result, result);
-  addptr(rsp, wordSize);
+  // add len then return (len will be zero if compress succeeded, otherwise negative)
+  bind(reset_sp);
+  addl(result, len);
 
   bind(done);
 }