Fixed gcc assembler constaints.

Added less efecient replacement functions for valgrind, to prevent warnings about uninitialized tail.

diff --git a/Zend/zend_string.c b/Zend/zend_string.c
index 8e0e8eb74b5df3d2f89829bf854442e276dfbd3a..e600b04a52101440058e13f4bb36c97d5e47efac 100644 (file)
--- a/Zend/zend_string.c
+++ b/Zend/zend_string.c
@@ -21,6 +21,10 @@
 #include "zend.h"
 #include "zend_globals.h"
 
+#ifdef HAVE_VALGRIND
+# include "valgrind/callgrind.h"
+#endif
+
 ZEND_API zend_string *(*zend_new_interned_string)(zend_string *str);
 ZEND_API zend_string *(*zend_string_init_interned)(const char *str, size_t size, int permanent);
 
@@ -30,7 +34,7 @@ static zend_string *zend_string_init_interned_permanent(const char *str, size_t
 static zend_string *zend_string_init_interned_request(const char *str, size_t size, int permanent);
 
 /* Any strings interned in the startup phase. Common to all the threads,
-   won't be free'd until process exit. If we want an ability to 
+   won't be free'd until process exit. If we want an ability to
    add permanent strings even after startup, it would be still
    possible on costs of locking in the thread safe builds. */
 static HashTable interned_strings_permanent;
@@ -312,75 +316,141 @@ ZEND_API void zend_interned_strings_switch_storage(zend_bool request)
 #if defined(__GNUC__) && defined(__i386__)
 ZEND_API zend_bool ZEND_FASTCALL zend_string_equal_val(zend_string *s1, zend_string *s2)
 {
+       char *ptr = ZSTR_VAL(s1);
+       size_t delta = (char*)s2 - (char*)s1;
+       size_t len = ZSTR_LEN(s1);
        zend_ulong ret;
 
        __asm__ (
-               "subl %1, %2\n\t"
                "ll0%=:\n\t"
-               "movl (%1,%2), %0\n\t"
-               "xorl (%1), %0\n\t"
+               "movl (%2,%3), %0\n\t"
+               "xorl (%2), %0\n\t"
                "jne ll1%=\n\t"
-               "addl $0x4, %1\n\t"
-               "subl $0x4, %3\n\t"
+               "addl $0x4, %2\n\t"
+               "subl $0x4, %1\n\t"
                "ja ll0%=\n\t"
                "movl $0x1, %0\n\t"
                "jmp ll3%=\n\t"
                "ll1%=:\n\t"
-               "cmpl $0x4,%3\n\t"
+               "cmpl $0x4,%1\n\t"
                "jb ll2%=\n\t"
                "xorl %0, %0\n\t"
                "jmp ll3%=\n\t"
                "ll2%=:\n\t"
-               "negl %3\n\t"
-               "lea 0x1c(,%3,8), %3\n\t"
-               "shll %b3, %0\n\t"
+               "negl %1\n\t"
+               "lea 0x1c(,%1,8), %1\n\t"
+               "shll %b1, %0\n\t"
                "sete %b0\n\t"
-               "andl $0x1, %0\n\t"
+               "movzbl %b0, %0\n\t"
                "ll3%=:\n"
-               : "=&a"(ret)
-               : "d"(ZSTR_VAL(s2)),
-                 "r"(ZSTR_VAL(s1)),
-                 "c"(ZSTR_LEN(s1))
+               : "=&a"(ret),
+                 "+c"(len),
+                 "+r"(ptr)
+               : "r"(delta)
+               : "cc");
+       return ret;
+}
+
+#ifdef HAVE_VALGRIND
+ZEND_API zend_bool ZEND_FASTCALL I_WRAP_SONAME_FNNAME_ZU(NONE,zend_string_equal_val)(zend_string *s1, zend_string *s2)
+{
+       size_t len = ZSTR_LEN(s1);
+       char *ptr1 = ZSTR_VAL(s1);
+       char *ptr2 = ZSTR_VAL(s2);
+       zend_ulong ret;
+
+       __asm__ (
+               "test %1, %1\n\t"
+               "jnz ll1%=\n\t"
+               "movl $0x1, %0\n\t"
+               "jmp ll2%=\n\t"
+               "ll1%=:\n\t"
+               "cld\n\t"
+               "rep\n\t"
+               "cmpsb\n\t"
+               "sete %b0\n\t"
+               "movzbl %b0, %0\n\t"
+               "ll2%=:\n"
+               : "=a"(ret),
+                 "+c"(len),
+                 "+D"(ptr1),
+                 "+S"(ptr2)
+               :
                : "cc");
        return ret;
 }
+#endif
+
 #elif defined(__GNUC__) && defined(__x86_64__)
 ZEND_API zend_bool ZEND_FASTCALL zend_string_equal_val(zend_string *s1, zend_string *s2)
 {
+       char *ptr = ZSTR_VAL(s1);
+       size_t delta = (char*)s2 - (char*)s1;
+       size_t len = ZSTR_LEN(s1);
        zend_ulong ret;
 
        __asm__ (
-               "subq %1, %2\n\t"
                "ll0%=:\n\t"
-               "movq (%1,%2), %0\n\t"
-               "xorq (%1), %0\n\t"
+               "movq (%2,%3), %0\n\t"
+               "xorq (%2), %0\n\t"
                "jne ll1%=\n\t"
-               "addq $0x8, %1\n\t"
-               "subq $0x8, %3\n\t"
+               "addq $0x8, %2\n\t"
+               "subq $0x8, %1\n\t"
                "ja ll0%=\n\t"
                "movq $0x1, %0\n\t"
                "jmp ll3%=\n\t"
                "ll1%=:\n\t"
-               "cmpq $0x8,%3\n\t"
+               "cmpq $0x8,%1\n\t"
                "jb ll2%=\n\t"
                "xorq %0, %0\n\t"
                "jmp ll3%=\n\t"
                "ll2%=:\n\t"
-               "negq %3\n\t"
-               "lea 0x3c(,%3,8), %3\n\t"
-               "shlq %b3, %0\n\t"
+               "negq %1\n\t"
+               "lea 0x3c(,%1,8), %1\n\t"
+               "shlq %b1, %0\n\t"
                "sete %b0\n\t"
-               "andq $0x1, %0\n\t"
+               "movzbl %b0, %0\n\t"
                "ll3%=:\n"
-               : "=&a"(ret)
-               : "r"(ZSTR_VAL(s2)),
-                 "r"(ZSTR_VAL(s1)),
-                 "c"(ZSTR_LEN(s1))
+               : "=&a"(ret),
+                 "+c"(len),
+                 "+r"(ptr)
+               : "r"(delta)
+               : "cc");
+       return ret;
+}
+
+#ifdef HAVE_VALGRIND
+ZEND_API zend_bool ZEND_FASTCALL I_WRAP_SONAME_FNNAME_ZU(NONE,zend_string_equal_val)(zend_string *s1, zend_string *s2)
+{
+       size_t len = ZSTR_LEN(s1);
+       char *ptr1 = ZSTR_VAL(s1);
+       char *ptr2 = ZSTR_VAL(s2);
+       zend_ulong ret;
+
+       __asm__ (
+               "test %1, %1\n\t"
+               "jnz ll1%=\n\t"
+               "movq $0x1, %0\n\t"
+               "jmp ll2%=\n\t"
+               "ll1%=:\n\t"
+               "cld\n\t"
+               "rep\n\t"
+               "cmpsb\n\t"
+               "sete %b0\n\t"
+               "movzbl %b0, %0\n\t"
+               "ll2%=:\n"
+               : "=a"(ret),
+                 "+c"(len),
+                 "+D"(ptr1),
+                 "+S"(ptr2)
+               :
                : "cc");
        return ret;
 }
 #endif
 
+#endif
+
 /*
  * Local variables:
  * tab-width: 4