1. I found a hash algorithm that may be faster than the currently used Hsieh hash (as of 0.78 - I don't have a newer version of the source code). It can be found at http://murmurhash.googlepages.com/. It is faster at least on the Core 2 Duo. MurmurHash 2.0 calculates a hash at about 2 GB/s, more than twice as fast as the next fastest hash algorithm listed (SuperFastHash).
2. Since the sqrt() function makes up around 10% of Spring's CPU consumption (see <a href=http://springrts.com/phpbb/viewtopic.ph ... 16953>this thread</a>), are any optimizations planned? I know SSE has a sqrt instruction, and I heard that SSE would be used in the next release of Spring.
Possible performance improvements
Moderator: Moderators
-
- Spring Developer
- Posts: 1254
- Joined: 24 Jun 2007, 08:34
Re: Possible performance improvements
U didn't read the soruce right:
Code: Select all
/** @brief a fast hash function
*
* This hash function is roughly 4x as fast as CRC32, but even that is too slow.
* We use a very simplistic add/xor feedback scheme when not debugging. */
Re: Possible performance improvements
2. Since the sqrt() function makes up around 10% of Spring's CPU consumption (see <a href=http://springrts.com/phpbb/viewtopic.ph ... 16953>this thread</a>), are any optimizations planned? I know SSE has a sqrt instruction, and I heard that SSE would be used in the next release of Spring.
^^
^^
Re: Possible performance improvements
It uses SSE right now. x87 has a square root opcode too; does the SSE version have a specification for what it outputs down to the bit, so it'll sync?