Dropping old processor support

[raspopov]

Intel Core Quad 6600 2.4GHz (Windows Score 5.6)

32-bit Assembler
MD4 hash: 10 MB by 18 ms (error 6%), 535 MB/s
MD5 hash: 10 MB by 26 ms (error 4%), 371 MB/s
SHA1 hash: 10 MB by 45 ms (error 4%), 219 MB/s

32-bit C++
MD4 hash: 10 MB by 21 ms (error 8%), 460 MB/s
MD5 hash: 10 MB by 31 ms (error 5%), 317 MB/s
SHA1 hash: 10 MB by 46 ms (error 4%), 213 MB/s

64-bit C++
MD4 hash: 10 MB by 20 ms (error 6%), 477 MB/s
MD5 hash: 10 MB by 29 ms (error 6%), 335 MB/s
SHA1 hash: 10 MB by 56 ms (error 3%), 175 MB/s

[ailurophobe]

These are without any SSE optimizations? I can't figure out the 64bit sha1 result. Compiler problem, maybe? You did say something like that IIRC.

[raspopov]

There are no places for SSE-operations (i.e. long text strings, matrix operations) in HashLib.

[ailurophobe]

Makes sense. Thanks.

[bosanek]

So, what is the final verdict? Shall the SSE-based operation be abandoned in the following versions of Shareaza (that's what I am hoping for)?

[cyko_01]

[bosanek]

That is an acceptable compromise, to keep everybody satisfied. Thanks to the programmers for their acknowledgment of real world users' needs.