Hello, I love openssl's crypto library, but getting it to perform its best is befuddling me.
I have an Intel Core 650 running amd64 debian squeeze with the openssl 0.9.8n package installed. When I run my benchmark that depends heavily on calls to AES_Encrypt I get > i5> gcc -march=native -O2 -static ocb3bis.c -lcrypto -ldl > i5> a.out > Validation string: A1998382DABC73EB5302BC0809FF6206 > Should be: A1998382DABC73EB5302BC0809FF6206 > OCB > 9.95 seconds. > 15.55 cpb. When I compile and install a local copy of version 1.0.0a using ./config -march=native, I get a reasonable looking CFLAG > CFLAG= -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H > -march=native - > m64 -DL_ENDIAN -DTERMIO -O3 -Wall -DMD32_REG_T=int -DOPENSSL_IA32_SSE2 > -DOPENSSL > _BN_ASM_MONT -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DAES_ASM > -DWHIRLPOO > L_ASM but, the result speed is half that of the debian package. > i5> gcc -march=native -O2 -static ocb3bis.c -I/usr/local/ssl/include > -L/usr/local/ssl/lib64 -lcrypto -ldl > i5> a.out > Validation string: A1998382DABC73EB5302BC0809FF6206 > Should be: A1998382DABC73EB5302BC0809FF6206 > OCB > 20.85 seconds. > 32.58 cpb. Are there some options that I should be passing along to ./config to make it faster? I looked hard for configuration and/or compiler options that might help, but couldn't find much of use. Thanks, Ted Krovetz______________________________________________________________________ OpenSSL Project http://www.openssl.org Development Mailing List [email protected] Automated List Manager [email protected]
