Re: Test of 0.9.8 under DJGPP (was Re: [openssl.org #932] Patch for 0.9.8 (DJGPP))

>>OpenSSL 0.9.8-dev XX xxx XXXX
>>built on: Mon Sep  6 09:04:18 PST 2004
>>platform: DJGPP
>>options:  bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) blowfish(idx) 
>>compiler: gcc -DOPENSSL_NO_IDEA -DOPENSSL_NO_RC5 -DOPENSSL_NO_MDC2 
>>-I/dev/env/WATT_ROOT/inc -DTERMIOS -DL_ENDIAN -fno-strict-aliasing 
>>-fomit-frame-pointer -O2 -Wall -DOPENSSL_BN_ASM_PART_WORDS -DSHA1_ASM -DMD5_ASM 
>>-DRMD160_ASM
>>OPENSSLDIR: "/dev/env/DJDIR/ssl"
> 
> To be able to use SSE2 code on a Pentium 4(?) , you need to add
> -DOPENSSL_IA32_SSE2 -DOPENSSL_CPUID_OBJ and have an up-to-date
> version of "as". (v2.13?).

./Configure should take care of it. Naturally not of as version, but the 
required preprocessor flags. Doug has probably configured with 386 flag.

> Then you'll notice a quadrupling in speed (for SHA at least).

For SHA-512 only, not for all SHAs.

> BTW. I don't see any way to turn off run-time use of SSE2. For comparison
> it could been nice to see the speed gain. Andy, how about turning SSE2 off
> with "$OPENSSL_ia32cap=none" etc.

The plan is to have OPENSSL_ia32cap=0 to provide for this and I tried to 
make it work and it was verified to work on several platforms. You 
mentioned earlier that you use own Makefile you've prepared long time 
ago, but can you try to build openssl "from scratch" starting with 
'./Configure DJGPP'? Actually if something goes wrong it SHA-512 
shouldn't go faster... Collect output from make and send it over (to 
save public bandwidth it's more than appropriate to send log file to me 
personally). A.

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org