diff src/cpu/x86/vm/x86_64.ad @ 7474:00af3a3a8df4

8005522: use fast-string instructions on x86 for zeroing Summary: use 'rep stosb' instead of 'rep stosq' when fast-string operations are available. Reviewed-by: twisti, roland
author kvn
date Thu, 03 Jan 2013 15:09:55 -0800
parents f6badecb7ea7
children b30b3c2a0cf2
line wrap: on
line diff
--- a/src/cpu/x86/vm/x86_64.ad	Sun Dec 23 17:08:22 2012 +0100
+++ b/src/cpu/x86/vm/x86_64.ad	Thu Jan 03 15:09:55 2013 -0800
@@ -10374,16 +10374,33 @@
 instruct rep_stos(rcx_RegL cnt, rdi_RegP base, rax_RegI zero, Universe dummy,
                   rFlagsReg cr)
 %{
+  predicate(!UseFastStosb);
   match(Set dummy (ClearArray cnt base));
   effect(USE_KILL cnt, USE_KILL base, KILL zero, KILL cr);
 
-  format %{ "xorl    rax, rax\t# ClearArray:\n\t"
-            "rep stosq\t# Store rax to *rdi++ while rcx--" %}
-  ins_encode(opc_reg_reg(0x33, RAX, RAX), // xorl %eax, %eax
-             Opcode(0xF3), Opcode(0x48), Opcode(0xAB)); // rep REX_W stos
+  format %{ "xorq    rax, rax\t# ClearArray:\n\t"
+            "rep     stosq\t# Store rax to *rdi++ while rcx--" %}
+  ins_encode %{ 
+    __ clear_mem($base$$Register, $cnt$$Register, $zero$$Register);
+  %}
   ins_pipe(pipe_slow);
 %}
 
+instruct rep_fast_stosb(rcx_RegL cnt, rdi_RegP base, rax_RegI zero, Universe dummy,
+                        rFlagsReg cr)
+%{
+  predicate(UseFastStosb);
+  match(Set dummy (ClearArray cnt base));
+  effect(USE_KILL cnt, USE_KILL base, KILL zero, KILL cr);
+  format %{ "xorq    rax, rax\t# ClearArray:\n\t"
+            "shlq    rcx,3\t# Convert doublewords to bytes\n\t"
+            "rep     stosb\t# Store rax to *rdi++ while rcx--" %}
+  ins_encode %{ 
+    __ clear_mem($base$$Register, $cnt$$Register, $zero$$Register);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
 instruct string_compare(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
                         rax_RegI result, regD tmp1, rFlagsReg cr)
 %{