Finally put my code up on github.
https://github.com/uncle-bob/quarkcoin-cpuminer
I'm looking for suggestions or ideas. Groestl-Hash seem to be the deal breaker right now. It's taking up as much time as all other 5 hashes combined. Any speed up for Groestl would be a win. Currently the V1 release confuses the compiler slightly less and is a smidgen faster. However I think this code is currently the best option for speeding up beyond that. It's still around twice as fast as the default miner.
Added bonus, almost everything for quark now fits a single function. While it won't fit on a GPU until the code size is reduced and memory is shrunk, it's just yet. Consider this my last update for cpuminer.
I wrote this cause I'm greedy, and so I leave my tipjar.
BTC: 1DaSUSqcr13G4QJYamsecW6yHh95QTT39R
QRK: QQRNpg7aGHcvgEFvXEk1DoCtLydboVsAzf
https://github.com/uncle-bob/quarkcoin-cpuminer
I'm looking for suggestions or ideas. Groestl-Hash seem to be the deal breaker right now. It's taking up as much time as all other 5 hashes combined. Any speed up for Groestl would be a win. Currently the V1 release confuses the compiler slightly less and is a smidgen faster. However I think this code is currently the best option for speeding up beyond that. It's still around twice as fast as the default miner.
Added bonus, almost everything for quark now fits a single function. While it won't fit on a GPU until the code size is reduced and memory is shrunk, it's just yet. Consider this my last update for cpuminer.
I wrote this cause I'm greedy, and so I leave my tipjar.
BTC: 1DaSUSqcr13G4QJYamsecW6yHh95QTT39R
QRK: QQRNpg7aGHcvgEFvXEk1DoCtLydboVsAzf
Is this faster than the other quarkcoin cpuminer?
By a long shot, yes. I get an 85% speedup from using it.
Something I've only just noticed: the Quarkcoin splash screen has a copyright year of 20,013.
Huh. I guess Quarkcoin really *is* the future.
As for AES-NI - don't forget that code won't run on processors that don't support it.
Matthew:out
well, almost any CPU thats worth it has AES-NI support, a simple check in the configure script can handle it while in the code its handled with #ifdef Huh. I guess Quarkcoin really *is* the future.
As for AES-NI - don't forget that code won't run on processors that don't support it.
Matthew:out
So you've never heard of the Core 2 family then?
At least two people in this thread use Core 2 based Xeon chips. For £35 each, they are a bargain.
Matthew:out