rc4_skey.c: remove dead/redundant code (it's never compiled) and
authorAndy Polyakov <appro@openssl.org>
Mon, 6 Jun 2011 20:04:33 +0000 (20:04 +0000)
committerAndy Polyakov <appro@openssl.org>
Mon, 6 Jun 2011 20:04:33 +0000 (20:04 +0000)
misleading/obsolete comment [from HEAD].

crypto/rc4/rc4_skey.c

index d8108cf19cc4c6d7389eb24d2a0bd2f47828c68f..fda27636e7b3477e75cf543658cd2a4105c3cf1d 100644 (file)
@@ -104,40 +104,6 @@ void private_RC4_set_key(RC4_KEY *key, int len, const unsigned char *data)
                d[(n)]=d[id2]; \
                d[id2]=tmp; }
 
-#if defined(OPENSSL_CPUID_OBJ) && !defined(OPENSSL_NO_ASM)
-# if   defined(__i386)   || defined(__i386__)   || defined(_M_IX86) || \
-       defined(__INTEL__) || \
-       defined(__x86_64) || defined(__x86_64__) || defined(_M_AMD64)
-       if (sizeof(RC4_INT) > 1) {
-               /*
-                * Unlike all other x86 [and x86_64] implementations,
-                * Intel P4 core [including EM64T] was found to perform
-                * poorly with wider RC4_INT. Performance improvement
-                * for IA-32 hand-coded assembler turned out to be 2.8x
-                * if re-coded for RC4_CHAR! It's however inappropriate
-                * to just switch to RC4_CHAR for x86[_64], as non-P4
-                * implementations suffer from significant performance
-                * losses then, e.g. PIII exhibits >2x deterioration,
-                * and so does Opteron. In order to assure optimal
-                * all-round performance, let us [try to] detect P4 at
-                * run-time by checking upon HTT bit in CPU capability
-                * vector and set up compressed key schedule, which is
-                * recognized by correspondingly updated assembler
-                * module...
-                *                              <appro@fy.chalmers.se>
-                */
-               if (OPENSSL_ia32cap_P & (1<<28)) {
-                       unsigned char *cp=(unsigned char *)d;
-
-                       for (i=0;i<256;i++) cp[i]=i;
-                       for (i=0;i<256;i++) SK_LOOP(cp,i);
-                       /* mark schedule as compressed! */
-                       d[256/sizeof(RC4_INT)]=-1;
-                       return;
-               }
-       }
-# endif
-#endif
        for (i=0; i < 256; i++) d[i]=i;
        for (i=0; i < 256; i+=4)
                {