defined(__INTEL__) || \
defined(__x86_64) || defined(__x86_64__) || defined(_M_AMD64) || defined(_M_X64)
-unsigned long OPENSSL_ia32cap_P=0;
-unsigned long *OPENSSL_ia32cap_loc(void) { return &OPENSSL_ia32cap_P; }
+unsigned int OPENSSL_ia32cap_P[2];
+unsigned long *OPENSSL_ia32cap_loc(void)
+{ if (sizeof(long)==4)
+ /*
+ * If 32-bit application pulls address of OPENSSL_ia32cap_P[0]
+ * clear second element to maintain the illusion that vector
+ * is 32-bit.
+ */
+ OPENSSL_ia32cap_P[1]=0;
+ return (unsigned long *)OPENSSL_ia32cap_P;
+}
#if defined(OPENSSL_CPUID_OBJ) && !defined(OPENSSL_NO_ASM) && !defined(I386_ONLY)
#define OPENSSL_CPUID_SETUP
+#if defined(_WIN32)
+typedef unsigned __int64 IA32CAP;
+#else
+typedef unsigned long long IA32CAP;
+#endif
void OPENSSL_cpuid_setup(void)
{ static int trigger=0;
- unsigned long OPENSSL_ia32_cpuid(void);
+ IA32CAP OPENSSL_ia32_cpuid(void);
+ IA32CAP vec;
char *env;
if (trigger) return;
trigger=1;
if ((env=getenv("OPENSSL_ia32cap")))
- OPENSSL_ia32cap_P = strtoul(env,NULL,0)|(1<<10);
+#if defined(_WIN32)
+ { if (!sscanf(env,"%I64i",&vec)) vec = strtoul(env,NULL,0); }
+#else
+ vec = strtoull(env,NULL,0);
+#endif
else
- OPENSSL_ia32cap_P = OPENSSL_ia32_cpuid()|(1<<10);
+ vec = OPENSSL_ia32_cpuid();
/*
* |(1<<10) sets a reserved bit to signal that variable
* was initialized already... This is to avoid interference
* with cpuid snippets in ELF .init segment.
*/
+ OPENSSL_ia32cap_P[0] = (unsigned int)vec|(1<<10);
+ OPENSSL_ia32cap_P[1] = (unsigned int)(vec>>32);
}
#endif
#define HEX_SIZE(type) (sizeof(type)*2)
void OPENSSL_cpuid_setup(void);
-extern unsigned long OPENSSL_ia32cap_P;
+extern unsigned int OPENSSL_ia32cap_P[];
void OPENSSL_showfatal(const char *,...);
void *OPENSSL_stderr(void);
extern int OPENSSL_NONPIC_relocated;
sub ::file_end
{ if (grep {/\b${nmdecor}OPENSSL_ia32cap_P\b/i} @out) {
- my $tmp=".comm\t${nmdecor}OPENSSL_ia32cap_P,4";
+ my $tmp=".comm\t${nmdecor}OPENSSL_ia32cap_P,8";
if ($::elf) { push (@out,"$tmp,4\n"); }
else { push (@out,"$tmp\n"); }
}
CPUs this is actually faster! */
# endif
# define GO_FOR_MMX(ctx,inp,num) do { \
- extern unsigned long OPENSSL_ia32cap_P; \
+ extern unsigned int OPENSSL_ia32cap_P[]; \
void whirlpool_block_mmx(void *,const void *,size_t); \
- if (!(OPENSSL_ia32cap_P & (1<<23))) break; \
+ if (!(OPENSSL_ia32cap_P[0] & (1<<23))) break; \
whirlpool_block_mmx(ctx->H.c,inp,num); return; \
} while (0)
# endif