sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 12, 2014, 07:52:52 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) ![](https://ip.bitcointalk.org/?u=http%3A%2F%2Fi62.tinypic.com%2F4g5kn.png&t=665&c=DbPj5SaLj-GGVw)
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 12, 2014, 08:03:36 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 12, 2014, 09:25:30 PM |
|
I have merged all the files from tvpouvet (1.5.1git version) New beta from github: Should support old cards as well. Compute 3.0 or newer. http://www.filedropper.com/release19x11 is down a bit, but hope that it is more stable source https://github.com/sp-hash/ccminer/
|
|
|
|
Schleicher
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 13, 2014, 12:00:38 AM |
|
I think you should check your files on github. There's some stuff missing.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 13, 2014, 10:18:06 AM |
|
I fixed it now. I am currently working with implementing uint2 into some of the hashingfunctions. (idea from DJM34 lyra implementation)
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 13, 2014, 11:36:07 AM |
|
The uint2 keccak made x11 20-25 KHASH faster. Checked in now.
|
|
|
|
Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 14, 2014, 01:32:31 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary djm your version is the optimized one with 970-980?
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 14, 2014, 01:53:59 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary djm your version is the optimized one with 970-980? actually, I think the main difference is in the compute 3.0 kernel which weren't there in that prelim. version. It also add an adjustable throughput, which might be able to increase a little the hashrate over that version (in the limitation of memory bandwidt... ) or decrease it if it is plugged into a monitor (my gtx980 were causing some lag when they were plugged into a monitor) The current implementation is already optimized for 9xx serie (that's the interest of the calculation using uint2 type actually) however because of the memory bandwidth limitation in the 9xx serie, the 780ti is doing better...
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 14, 2014, 03:57:41 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary djm your version is the optimized one with 970-980? actually, I think the main difference is in the compute 3.0 kernel which weren't there in that prelim. version. It also add an adjustable throughput, which might be able to increase a little the hashrate over that version (in the limitation of memory bandwidt... ) or decrease it if it is plugged into a monitor (my gtx980 were causing some lag when they were plugged into a monitor) The current implementation is already optimized for 9xx serie (that's the interest of the calculation using uint2 type actually) however because of the memory bandwidth limitation in the 9xx serie, the 780ti is doing better... what number your version is doing? with sp version i'm currently at 7800 hash with one 970(1400 core)
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 14, 2014, 04:11:10 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary djm your version is the optimized one with 970-980? actually, I think the main difference is in the compute 3.0 kernel which weren't there in that prelim. version. It also add an adjustable throughput, which might be able to increase a little the hashrate over that version (in the limitation of memory bandwidt... ) or decrease it if it is plugged into a monitor (my gtx980 were causing some lag when they were plugged into a monitor) The current implementation is already optimized for 9xx serie (that's the interest of the calculation using uint2 type actually) however because of the memory bandwidth limitation in the 9xx serie, the 780ti is doing better... what number your version is doing? with sp version i'm currently at 7800 hash with one 970(1400 core) on x11 ? I don't know, but sp_ version should be better on x11, I haven't looked so far on x11 (the point of this version was Lyra2RE)
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
Amph
Legendary
Offline
Activity: 3248
Merit: 1070
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 14, 2014, 10:00:36 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary djm your version is the optimized one with 970-980? actually, I think the main difference is in the compute 3.0 kernel which weren't there in that prelim. version. It also add an adjustable throughput, which might be able to increase a little the hashrate over that version (in the limitation of memory bandwidt... ) or decrease it if it is plugged into a monitor (my gtx980 were causing some lag when they were plugged into a monitor) The current implementation is already optimized for 9xx serie (that's the interest of the calculation using uint2 type actually) however because of the memory bandwidth limitation in the 9xx serie, the 780ti is doing better... what number your version is doing? with sp version i'm currently at 7800 hash with one 970(1400 core) on x11 ? I don't know, but sp_ version should be better on x11, I haven't looked so far on x11 (the point of this version was Lyra2RE) oh that's a new algo, a fast google search tells me that is vertcoin releated, it's something anti-asic?
|
|
|
|
K1773R
Legendary
Offline
Activity: 1792
Merit: 1008
/dev/null
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 06:05:15 PM |
|
HEAD@windows never submits any shares.
|
[GPG Public Key]BTC/DVC/TRC/FRC: 1 K1773RbXRZVRQSSXe9N6N2MUFERvrdu6y ANC/XPM A K1773RTmRKtvbKBCrUu95UQg5iegrqyeA NMC: N K1773Rzv8b4ugmCgX789PbjewA9fL9Dy1 LTC: L Ki773RBuPepQH8E6Zb1ponoCvgbU7hHmd EMC: E K1773RxUes1HX1YAGMZ1xVYBBRUCqfDoF BQC: b K1773R1APJz4yTgRkmdKQhjhiMyQpJgfN
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 06:35:19 PM |
|
HEAD@windows never submits any shares.
you need to use --diff 128 (if it is related to lyra2re)
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 06:36:13 PM |
|
Nice, you finally updated your fork ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) the good thing with my complicated relationship with github is that I update only when necessary djm your version is the optimized one with 970-980? actually, I think the main difference is in the compute 3.0 kernel which weren't there in that prelim. version. It also add an adjustable throughput, which might be able to increase a little the hashrate over that version (in the limitation of memory bandwidt... ) or decrease it if it is plugged into a monitor (my gtx980 were causing some lag when they were plugged into a monitor) The current implementation is already optimized for 9xx serie (that's the interest of the calculation using uint2 type actually) however because of the memory bandwidth limitation in the 9xx serie, the 780ti is doing better... what number your version is doing? with sp version i'm currently at 7800 hash with one 970(1400 core) on x11 ? I don't know, but sp_ version should be better on x11, I haven't looked so far on x11 (the point of this version was Lyra2RE) oh that's a new algo, a fast google search tells me that is vertcoin releated, it's something anti-asic? it is anti-asic for now...
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
K1773R
Legendary
Offline
Activity: 1792
Merit: 1008
/dev/null
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 08:34:30 PM |
|
HEAD@windows never submits any shares.
you need to use --diff 128 (if it is related to lyra2re) tested with x11 and x13. i see a big hashrate increase, tough never any submitted shares.
|
[GPG Public Key]BTC/DVC/TRC/FRC: 1 K1773RbXRZVRQSSXe9N6N2MUFERvrdu6y ANC/XPM A K1773RTmRKtvbKBCrUu95UQg5iegrqyeA NMC: N K1773Rzv8b4ugmCgX789PbjewA9fL9Dy1 LTC: L Ki773RBuPepQH8E6Zb1ponoCvgbU7hHmd EMC: E K1773RxUes1HX1YAGMZ1xVYBBRUCqfDoF BQC: b K1773R1APJz4yTgRkmdKQhjhiMyQpJgfN
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 09:45:44 PM |
|
I just submitted an improved BMW rewrite.
32% faster. Using UINT2 instead of Uint64_t. Added new operators for uint2 operations. SHL2 and SHR2, and minus. (in cuda_helper.h)
Around +30-40 KHASH on x11, x13 and x15
Now BMW is spilling 0 bytes to slow memory...
I also tried to use uint2 in blake, but it was slower.
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 09:50:40 PM |
|
HEAD@windows never submits any shares.
you need to use --diff 128 (if it is related to lyra2re) tested with x11 and x13. i see a big hashrate increase, tough never any submitted shares. Works here. Is this on linux? wich card?
|
|
|
|
Epsylon3
Legendary
Offline
Activity: 1484
Merit: 1082
ccminer/cpuminer developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 10:05:44 PM |
|
I had this while testing keccak uint2 changes with a sm30 only binary on the 750 ti... No errors but no found nonces neither.. So i commited the optimisation differently
|
|
|
|
|
sp_ (OP)
Legendary
Offline
Activity: 2926
Merit: 1087
Team Black developer
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
December 15, 2014, 10:08:17 PM |
|
My version is probobly failing on sm30 cards.
|
|
|
|
|