Large Bitcoin Collider (Collision Finders Pool) - page 34.

SlarkBoy

member

Activity: 114

Merit: 11

Quote from: rico666 on February 21, 2017, 02:02:47 AM

Congrats! And ... what? ... better keyrate than me? I have to o.p.t.i.m.i.z.e. more Cool

I'm getting more and more feedback about GPU clients finally working with some heart-lung machine work.
I'll publish a new client soon which wraps up all the fixes, so GPU experience will be smooth.

Rico

Why isn't using all resources? Only 25%

Code:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 375.26                 Driver Version: 375.26                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1080    Off  | 0000:01:00.0      On |                  N/A |
| 29%   54C    P2    49W / 200W |   1217MiB /  8113MiB |     25%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0      1162    G   /usr/lib/xorg/Xorg                             181MiB |
|    0      2889    G   compiz                                         114MiB |
|    0     10005    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10008    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10011    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10012    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10013    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10014    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10015    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
|    0     10016    C   ./gen-hrdcore-avx2+gpu-linux64                 115MiB |
+-----------------------------------------------------------------------------+

Jude Austin

legendary

Activity: 1140

Merit: 1000

The Real Jude Austin

Quote from: rico666 on February 24, 2017, 04:29:46 PM

I managed to get the Bloom Filter check done on GPU also.

with some more optimizations total time for 16M keys went down about 0.2 secs on my machine.
Tested with -L 5 (generating 5 x 16M consecutive keys but just 1 startup cost), the time went down 0.5s per 16M keys so I'm at about 2396745 keys/s per core now.

Did some benchmarking and my M2000M Quadro - a midrange notebook GPU - does 60M of my BTC-optimized hash160 code + bloom-check per second. That's enough horsepower to check 30 Mkeys (uncompressed + compressed) per second.

If they are provided fast enough...

So how do I get from my current 7 Mkeys/s to 30 Mkeys/s?

I'd like to pursue some of the ideas arulbero threw in on ECC pubkey generation, but after having looked at his code and also at things like "supervanitygen" (which is now officially less than half the speed of LBC), It's pretty clear I will have to come up with something on my own.
And also that I will have to move ECC to the GPU.

Rico

This may be of some use: https://pdfs.semanticscholar.org/e1b1/1b509ac7c4f049ebab19caf57fcd0622b3c0.pdf

freemanjackal

sr. member

Activity: 378

Merit: 250

lets see if i understand the idea, are you trying to find another private key that should exists by a given public key(in this case public addresses), if i am not wrong this mathematically should take thouthands or millions of years, is that the idea behind the project? what about quantic processing that is actually working, i think that 's why this should be unfeasible at least for the moment, more processing capabilities needed.

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

I managed to get the Bloom Filter check done on GPU also.

with some more optimizations total time for 16M keys went down about 0.2 secs on my machine.
Tested with -L 5 (generating 5 x 16M consecutive keys but just 1 startup cost), the time went down 0.5s per 16M keys so I'm at about 2396745 keys/s per core now.

Did some benchmarking and my M2000M Quadro - a midrange notebook GPU - does 60M of my BTC-optimized hash160 code + bloom-check per second. That's enough horsepower to check 30 Mkeys (uncompressed + compressed) per second.

If they are provided fast enough...

So how do I get from my current 7 Mkeys/s to 30 Mkeys/s?

I'd like to pursue some of the ideas arulbero threw in on ECC pubkey generation, but after having looked at his code and also at things like "supervanitygen" (which is now officially less than half the speed of LBC), It's pretty clear I will have to come up with something on my own.
And also that I will have to move ECC to the GPU.

Rico

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: TooDumbForBitcoin on February 23, 2017, 09:07:41 AM

Hey, LBC!!! Did somebody just whup yo ass? (Philosophically, if not SHA-256ishly)

https://bitcointalk.org/index.php?topic=293382.20

SHA1 ... bah. That's like beating up a grown up MD5.

6500 CPU-years & 100 GPU years Roll Eyes

Rico

TooDumbForBitcoin

legendary

Activity: 1638

Merit: 1001

Hey, LBC!!! Did somebody just whup yo ass? (Philosophically, if not SHA-256ishly)

https://bitcointalk.org/index.php?topic=293382.20

(or am I too dumb for bitcoin?)

Jude Austin

legendary

Activity: 1140

Merit: 1000

The Real Jude Austin

Quote from: rico666 on February 23, 2017, 03:09:41 AM

Quote from: Jude Austin on February 22, 2017, 09:12:22 AM

Code:

model name      : AMD A8-5500 APU with Radeon(tm) HD Graphics

Key rate:
Your maximum speed is 578988 keys/s per CPU core.
Averaging around 2.18Mkeys using a Sapphire 7970 reference card to assist 4 cores (-c 4).

As for OS, I used Ubuntu 14.04.4 because that is the last known version to work well with OpenCL.

So fresh Ubuntu install, then I sudo apt-get update && sudo apt-get install fglrx, RESTART PC, install LBC, profit.

Very, VERY interesting because of several reasons:

1) I'm still unable to get my AMD GPU working Undecided

Code:

root@rico666-desktop:/home/rico666/collider# time ./gen-hrdcore-sse42+gpu-linux64 -I 0000000000000000000000000000000000000000000000000000000000000001 -c 10000
Couldn't create a command queue: Out of host memory

Tried Arch, ditched that after half a day of error messages and install woes, Installed Ubuntu 16.04 with AMDGPU Pro on a USB stick -> same result. Aaargh. So you beat me there. I cannot run my own program, but you can.

2) I do have a R9 280X which is pretty near to your 7970, however when I look up the specs for the AMD A8-5500 APU, it says there's a Radeon 7560D built in there somewhere. So just to be clear. You have a system with that AMD APU + 7970? So you have 2 GPUs?
And you are sure LBC takes the 7970 to assist and not the 7560D? Because when I look at the diagnostics-OpenCL you sent me, there are - in fact - two Tahiti devices lurking there, and the 1st one is the 7560D.

So unless you are starting LBC with -gdev 2, you are using the 7560D to assist. Might want to cross check that.

3) Your AMD APU reminded me, I have something similar here sitting in my Kodi box:

AMD A6-5400K APU with Radeon(tm) HD Graphics - it claims to have a 7540D, I might try that.

Rico

1. Hmm, try setting environment variables for the GPU.

2. I tried using all the devices on my rig and got the same hash rate for each one.

I have 2 7970s and one 7950 on my rig, I went through devices 1-4 and achieved the same rate.

Thanks,
Jude

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: Jude Austin on February 22, 2017, 09:12:22 AM

Code:

model name      : AMD A8-5500 APU with Radeon(tm) HD Graphics

Key rate:
Your maximum speed is 578988 keys/s per CPU core.
Averaging around 2.18Mkeys using a Sapphire 7970 reference card to assist 4 cores (-c 4).

As for OS, I used Ubuntu 14.04.4 because that is the last known version to work well with OpenCL.

So fresh Ubuntu install, then I sudo apt-get update && sudo apt-get install fglrx, RESTART PC, install LBC, profit.

Very, VERY interesting because of several reasons:

1) I'm still unable to get my AMD GPU working Undecided

Code:

root@rico666-desktop:/home/rico666/collider# time ./gen-hrdcore-sse42+gpu-linux64 -I 0000000000000000000000000000000000000000000000000000000000000001 -c 10000
Couldn't create a command queue: Out of host memory

Tried Arch, ditched that after half a day of error messages and install woes, Installed Ubuntu 16.04 with AMDGPU Pro on a USB stick -> same result. Aaargh. So you beat me there. I cannot run my own program, but you can.

2) I do have a R9 280X which is pretty near to your 7970, however when I look up the specs for the AMD A8-5500 APU, it says there's a Radeon 7560D built in there somewhere. So just to be clear. You have a system with that AMD APU + 7970? So you have 2 GPUs?
And you are sure LBC takes the 7970 to assist and not the 7560D? Because when I look at the diagnostics-OpenCL you sent me, there are - in fact - two Tahiti devices lurking there, and the 1st one is the 7560D.

So unless you are starting LBC with -gdev 2, you are using the 7560D to assist. Might want to cross check that.

3) Your AMD APU reminded me, I have something similar here sitting in my Kodi box:

AMD A6-5400K APU with Radeon(tm) HD Graphics - it claims to have a 7540D, I might try that.

Rico

Jude Austin

legendary

Activity: 1140

Merit: 1000

The Real Jude Austin

Quote from: rico666 on February 22, 2017, 01:35:46 AM

Quote from: Jude Austin on February 21, 2017, 04:49:56 PM

I just used Ubuntu 14.04.4 and installed fglrx.

Then the normal LBC install.

As for the GPU client: Can you write more about hardware you use and keyrate you get out of it?

I'm adding some overview of speed for complete system configurations
https://lbc.cryptoguru.org/man/admin#generator-speed

Rico

Yep, I will do that after work.

It's after work:

Code:

processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 21
model           : 16
model name      : AMD A8-5500 APU with Radeon(tm) HD Graphics
stepping        : 1
microcode       : 0x6001116
cpu MHz         : 1400.000
cache size      : 2048 KB
physical id     : 0
siblings        : 4
core id         : 0
cpu cores       : 2
apicid          : 16
initial apicid  : 0
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov                                                                                                                                                              pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp                                                                                                                                                              lm constant_tsc rep_good nopl nonstop_tsc extd_apicid aperfmperf pni pclmulqdq                                                                                                                                                              monitor ssse3 fma cx16 sse4_1 sse4_2 popcnt aes xsave avx f16c lahf_lm cmp_legac                                                                                                                                                             y svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs xop skinit                                                                                                                                                              wdt lwp fma4 tce nodeid_msr tbm topoext perfctr_core perfctr_nb arat cpb hw_pst                                                                                                                                                             ate npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists p                                                                                                                                                             ausefilter pfthreshold vmmcall bmi1
bugs            : fxsave_leak sysret_ss_attrs
bogomips        : 6388.03
TLB size        : 1536 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb eff_freq_ro

processor       : 1
vendor_id       : AuthenticAMD
cpu family      : 21
model           : 16
model name      : AMD A8-5500 APU with Radeon(tm) HD Graphics
stepping        : 1
microcode       : 0x6001116
cpu MHz         : 1400.000
cache size      : 2048 KB
physical id     : 0
siblings        : 4
core id         : 1
cpu cores       : 2
apicid          : 17
initial apicid  : 1
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov                                                                                                                                                              pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp                                                                                                                                                              lm constant_tsc rep_good nopl nonstop_tsc extd_apicid aperfmperf pni pclmulqdq                                                                                                                                                              monitor ssse3 fma cx16 sse4_1 sse4_2 popcnt aes xsave avx f16c lahf_lm cmp_legac                                                                                                                                                             y svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs xop skinit                                                                                                                                                              wdt lwp fma4 tce nodeid_msr tbm topoext perfctr_core perfctr_nb arat cpb hw_pst                                                                                                                                                             ate npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists p                                                                                                                                                             ausefilter pfthreshold vmmcall bmi1
bugs            : fxsave_leak sysret_ss_attrs
bogomips        : 6388.03
TLB size        : 1536 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb eff_freq_ro

processor       : 2
vendor_id       : AuthenticAMD
cpu family      : 21
model           : 16
model name      : AMD A8-5500 APU with Radeon(tm) HD Graphics
stepping        : 1
cpu MHz         : 1400.000
cache size      : 2048 KB
physical id     : 0
siblings        : 4
core id         : 2
cpu cores       : 2
apicid          : 18
initial apicid  : 2
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov                                                                                                                                                              pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp                                                                                                                                                              lm constant_tsc rep_good nopl nonstop_tsc extd_apicid aperfmperf pni pclmulqdq                                                                                                                                                              monitor ssse3 fma cx16 sse4_1 sse4_2 popcnt aes xsave avx f16c lahf_lm cmp_legac                                                                                                                                                             y svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs xop skinit                                                                                                                                                              wdt lwp fma4 tce nodeid_msr tbm topoext perfctr_core perfctr_nb arat cpb hw_pst                                                                                                                                                             ate npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists p                                                                                                                                                             ausefilter pfthreshold vmmcall bmi1
bugs            : fxsave_leak sysret_ss_attrs
bogomips        : 6388.03
TLB size        : 1536 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb eff_freq_ro

processor       : 3
vendor_id       : AuthenticAMD
cpu family      : 21
model           : 16
model name      : AMD A8-5500 APU with Radeon(tm) HD Graphics
stepping        : 1
cpu MHz         : 1400.000
cache size      : 2048 KB
physical id     : 0
siblings        : 4
core id         : 3
cpu cores       : 2
apicid          : 19
initial apicid  : 3
fpu             : yes
fpu_exception   : yes
cpuid level     : 13
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov                                                                                                                                                              pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp                                                                                                                                                              lm constant_tsc rep_good nopl nonstop_tsc extd_apicid aperfmperf pni pclmulqdq                                                                                                                                                              monitor ssse3 fma cx16 sse4_1 sse4_2 popcnt aes xsave avx f16c lahf_lm cmp_legac                                                                                                                                                             y svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs xop skinit                                                                                                                                                              wdt lwp fma4 tce nodeid_msr tbm topoext perfctr_core perfctr_nb arat cpb hw_pst                                                                                                                                                             ate npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists p                                                                                                                                                             ausefilter pfthreshold vmmcall bmi1
bugs            : fxsave_leak sysret_ss_attrs
bogomips        : 6388.03
TLB size        : 1536 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb eff_freq_ro

Key rate:

Your maximum speed is 578988 keys/s per CPU core.

Averaging around 2.18Mkeys using a Sapphire 7970 reference card to assist 4 cores (-c 4).

As for OS, I used Ubuntu 14.04.4 because that is the last known version to work well with OpenCL.

So fresh Ubuntu install, then I sudo apt-get update && sudo apt-get install fglrx, RESTART PC, install LBC, profit.

Thanks,
Jude

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

There is a new version, which fixes the rogue "Death Kiss" events.

Please update as follows:

End your LBC client
Do a LBC -u (you should have a 1.031 or newer)
Start your LBC as usual

If you haven't updated yet (I see clients as old as 0.993), you are also missing out quite some keyrate.
There are reports of 1.6 Mkeys -> 2.1 Mkeys for CPU clients on VMs

Rico

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: Jude Austin on February 21, 2017, 04:49:56 PM

I just used Ubuntu 14.04.4 and installed fglrx.

Then the normal LBC install.

As for the GPU client: Can you write more about hardware you use and keyrate you get out of it?

I'm adding some overview of speed for complete system configurations
https://lbc.cryptoguru.org/man/admin#generator-speed

Rico

Jude Austin

legendary

Activity: 1140

Merit: 1000

The Real Jude Austin

Quote from: rico666 on February 21, 2017, 04:26:59 AM

I have feedback, that at least one user got his AMD GPU to run with LBC and the sse42+gpu generator.
Which makes him the 1st, because I was not lucky so far. May he come forth and bath in glory (and answer questions/give support Tongue

)

The new generators have a new parameter -L you may see in the process table. In case you're wondering what that is:

Quote

-L is loops

earlier versions of the generator started up, searched 16M keys, terminated, next startup, searched 16M keys...
so per 'o' one run.

-L will now tell the generator to run loops, i.e. x 16M
this is especially important for GPU generators as they have a high startup cost, but also the CPU generators profit from this.

So if you e.g. see -L 25, this means, the generator will startup and run from a certain offset a search of 25 x 16M keys, which means only 1 startup cost for 400M keys instead of 25 startups.
There is a drawback to this - unfortunately. If you want to end LBC by pressing "e", it will take longer. Namely at the end of the next ask-for-work-block.

Code:

Ask for work... got blocks [426612921-426615288] (2483 Mkeys)
oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooe <--- here we pressed "e"
ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo (7.33 Mkeys/s)
Ask for work... got blocks [426626393-426628760] (2483 Mkeys)
<-- here LBC found out
END requested. (Ending this loop) Waiting for children to finish...
oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo  <--- here it ends

So max. time from "e" to end is twice the time given in -t

Also, you may see a -d 1 parameter in the process table with the GPU generators.

Quote

-d is simply the GPU device to use. If there are several GPUs on the system, this says which one to use. Default is 1

When does one use that? Again, as a LBC user, most of the time you do not need to take care of this, but in case you have a really big iron (say 32 physical CPUs and 4 GPUs), the LBC parameter is called "gdev" aka "GPU device".

Howto - you open 4 windows/terminals:

terminal1: ./LBC --gpu -c 8 -gdev 1
terminal2: ./LBC --gpu -c 8 -gdev 2
terminal3: ./LBC --gpu -c 8 -gdev 3
terminal4: ./LBC --gpu -c 8 -gdev 4

=> you should have 4 LBCs running, each taming 8 generators (8 CPUs + 1 GPU)
used if e.g. 8 CPUs are right to saturate 1 GPU

Rico

Hi,

I just used Ubuntu 14.04.4 and installed fglrx.

Then the normal LBC install.

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

New client 1.029 - upon restart your LBC should auto-update.

Also, most generators (except generic and sse41) have been updated to accept the -L option.
These too, should auto update.

Everyone who is GPU auth, should now be plug and play via --gpu.

The names of the generators are now in a more canonical form.

gen-hrdcore-skylake-linux64 (CPU)
gen-hrdcore-skylake+gpu-linux64 (CPU+GPU)
gen-hrdcore-avx2-linux64 (CPU)
gen-hrdcore-avx2+gpu-linux64 (CPU+GPU)
...etc. - you get the idea

So basically if you had a gen-hrdcore-avx2-linux64 generator up to now, with --gpu LBC would fetch the gen-hrdcore-avx2+gpu-linux64 counterpart and use that.

Even if you are CPU only, an update makes sense, as the -L will lower the generator startup overhead and should give you some more keys/s

Rico

shifty252

full member

Activity: 177

Merit: 101

Quote from: rico666 on February 21, 2017, 04:45:11 AM

Quote from: shifty252 on February 19, 2017, 12:24:17 PM

but it also threw about 4 error lines about an unknown command "-L" but it went on asking for work so I let it be.

And that is something you shouldn't have done. When I asked for GPU beta testers, there were several prerequisites.
Among them there were

If you have GPU authorization
If you are bold

You had no GPUauth, but you were extremely bold, which means you can handle the consequences.
At the moment I handle the consequences too, because I put my client on rectifying your 278+ Gkeys of invalid mess.
If you really are interested to help out in the project and not making things worse, PM me.

BTW - this (PM me) applies to everyone who feels he can't make it into top30, cannot fork out 0.1BTC and would like to explore the 3rd option

Quote from: rico666 on February 19, 2017, 10:31:46 AM

How does that gpuauth=1 happen?
...

You get a gpuauth set to 1 by decree (i.e. for special services)

Rico

You're right, I deserve what came after the mess , BUT:

Quote from: rico666 on February 19, 2017, 08:05:51 AM

Ok guys - this is it:

Everyone who has a working OpenCL install, is authorized for GPU, has an AVX2 capable CPU (I will provide sse42 GPU clients later) and feels bold enough, please download the beta from.

This was way misleading and it read like anyone with gpu and opencl installed may try the beta.

This is were my error is.

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: sana98212 on February 19, 2017, 12:39:57 PM

Quote

No I haven't. I have sworn to myself to pull this project off without any thinking at all.

Well let's see how the LBC evolves, thankyou.

Of course my answer was sarcasto-ironic (which I often do to stimulate thinking in my discussion counterpart, but it almost never works).
So full verbatim, my answer translates to this:

There are currently P2PKH and P2SH addresses. We search only P2PKH, but we may extend our search to P2SH.

This "If you find collisions BTC will die/will be worth nothing" is FUD. Bitcoin is an evolving system. And yes, LBC may put evolutionary pressure on the P2PKH part right now.
If you look at https://github.com/bitcoin/bips/blob/master/README.mediawiki, you will see that there are other address formats already proposed.

E.g. the now deferred BIP142 defining a P2WPKH format.

Others may follow (I'd propose a true-512bit address format in a BIP myself, but am lacking the time atm. Also I prefer to interact with core devs as few as possible).

In other words: There are so many levels between LBC finding collisions and BTC dying, one can safely assume these two events are not connected. Also I do not believe in the 2nd one.

Rico

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: shifty252 on February 19, 2017, 12:24:17 PM

but it also threw about 4 error lines about an unknown command "-L" but it went on asking for work so I let it be.

And that is something you shouldn't have done. When I asked for GPU beta testers, there were several prerequisites.
Among them there were

If you have GPU authorization
If you are bold

You had no GPUauth, but you were extremely bold, which means you can handle the consequences.
At the moment I handle the consequences too, because I put my client on rectifying your 278+ Gkeys of invalid mess.
If you really are interested to help out in the project and not making things worse, PM me.

BTW - this (PM me) applies to everyone who feels he can't make it into top30, cannot fork out 0.1BTC and would like to explore the 3rd option

Quote from: rico666 on February 19, 2017, 10:31:46 AM

How does that gpuauth=1 happen?
...

You get a gpuauth set to 1 by decree (i.e. for special services)

Rico

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

I have feedback, that at least one user got his AMD GPU to run with LBC and the sse42+gpu generator.
Which makes him the 1st, because I was not lucky so far. May he come forth and bath in glory (and answer questions/give support Tongue

)

The new generators have a new parameter -L you may see in the process table. In case you're wondering what that is:

Quote

-L is loops

earlier versions of the generator started up, searched 16M keys, terminated, next startup, searched 16M keys...
so per 'o' one run.

-L will now tell the generator to run loops, i.e. x 16M
this is especially important for GPU generators as they have a high startup cost, but also the CPU generators profit from this.

So if you e.g. see -L 25, this means, the generator will startup and run from a certain offset a search of 25 x 16M keys, which means only 1 startup cost for 400M keys instead of 25 startups.
There is a drawback to this - unfortunately. If you want to end LBC by pressing "e", it will take longer. Namely at the end of the next ask-for-work-block.

Code:

Ask for work... got blocks [426612921-426615288] (2483 Mkeys)
oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooe <--- here we pressed "e"
ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo (7.33 Mkeys/s)
Ask for work... got blocks [426626393-426628760] (2483 Mkeys)
<-- here LBC found out
END requested. (Ending this loop) Waiting for children to finish...
oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo  <--- here it ends

So max. time from "e" to end is twice the time given in -t

Also, you may see a -d 1 parameter in the process table with the GPU generators.

Quote

-d is simply the GPU device to use. If there are several GPUs on the system, this says which one to use. Default is 1

When does one use that? Again, as a LBC user, most of the time you do not need to take care of this, but in case you have a really big iron (say 32 physical CPUs and 4 GPUs), the LBC parameter is called "gdev" aka "GPU device".

Howto - you open 4 windows/terminals:

terminal1: ./LBC --gpu -c 8 -gdev 1
terminal2: ./LBC --gpu -c 8 -gdev 2
terminal3: ./LBC --gpu -c 8 -gdev 3
terminal4: ./LBC --gpu -c 8 -gdev 4

=> you should have 4 LBCs running, each taming 8 generators (8 CPUs + 1 GPU)
used if e.g. 8 CPUs are right to saturate 1 GPU

Rico

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: SlarkBoy on February 20, 2017, 01:03:15 PM

it works

Code:

root@Meh:~/LBCGPU# ./LBC --id XXXX --secret XXXX --gpu -t 1 -l 1 --no_update
Your maximum speed is 2232588 keys/s per CPU core.

Congrats! And ... what? ... better keyrate than me? I have to o.p.t.i.m.i.z.e. more Cool

I'm getting more and more feedback about GPU clients finally working with some heart-lung machine work.
I'll publish a new client soon which wraps up all the fixes, so GPU experience will be smooth.

Rico

SlarkBoy

member

Activity: 114

Merit: 11

it works

Code:

root@Meh:~/LBCGPU# ./LBC --id XXXX --secret XXXX --gpu -t 1 -l 1 --no_update
GPU authorized: yes
Will use 4 CPUs.
Benchmark info not found - benchmarking... done.
Your maximum speed is 2232588 keys/s per CPU core.
Ask for work... got blocks [424129113-424129624] (536 Mkeys)
oooooooooooooooooooooooooooooooo (7.00 Mkeys/s)
root@Meh:~/LBCGPU# ./LBC --id XXXX --secret XXXX --gpu -t 1 -l 1 --no_update --cpus 8
GPU authorized: yes
Ask for work... got blocks [424136153-424137176] (1073 Mkeys)
oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo (7.63 Mkeys/s)

rico666

legendary

Activity: 1120

Merit: 1037

฿ → ∞

Quote from: Jude Austin on February 19, 2017, 10:32:44 PM

Patiently waiting, hehe.

JFYI: The "hrd-core" binary below is a sse42+gpu version with a AMD GPU (R9-280X - I wonder why it's named Hainan, I thought they are Tahiti XTL)

No idea why it's throwing that "Out of host memory" on me. The host is sitting there with 8GB unused memory...

Code:

[root@localhost HRD-GPU]# time ./hrd-core -I 0000000000000000000000000000000000000000000000000000000000000001 -c 10000 -d 1
OpenCL device chosen: Advanced Micro Devices, Inc. (Hainan)
256/16/1
256/1/1
Couldn't create a command queue: Out of host memory

As soon as I crack this, you will have your toy. Wink

edit: needless to say, it works on NVIDIA platforms without problems:

ubuntu@ip-172-31-44-82:~/collider$ ./LBC --gpu -x GPU authorized: yes Testing mode. Using page 0, turning off looping. Benchmark info not found - benchmarking... done. Your maximum speed is 1308421 keys/s per CPU core. Generator chosen: gen-hrdcore-sse42+gpu-linux64 o Test ok. Your test results were stored in FOUND.txt. Have a look and then you may want to remove the file. 2d17543d32448acc7a1c43c5f72cd5be459ab302:c:priv:0000000000000000000000000000000000000000000000000000000000001001 + 0x5e 02e62151191a931d51cdc513a86d4bf5694f4e51:u:priv:0000000000000000000000000000000000000000000000000000000000001001 + 0x66 9d74ffdb31068ca2a1feb8e34830635c0647d714:c:priv:00000000000000000000000000000000000000000000000000000000000fa001 + 0xf8c 3d6871076780446bd46fc564b0c443e1fd415beb:u:priv:00000000000000000000000000000000000000000000000000000000000fa001 + 0xf8d Ending test run.

1308421 keys/s per core - that's on a g2.xlarge (xeon v2 + K80)

Also in the news: ftp://ftp.cryptoguru.org/LBC/generators/

Quote from: SlarkBoy on February 19, 2017, 12:36:21 PM

Quote from: rico666 on February 19, 2017, 09:35:17 AM

And we are sure the CPU is AVX2 capable?

My CPU: Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz

You could try some hacking by using the 170220-3a3e82c4efe4a75e07c9c43a46206135.gen-hrdcore-avx2+gpu-linux64.bz2 generator, unpacking and renaming it to gen-hrdcore-gpu-linux64
and starting LBC with all options according to the instructions (as you did) and additionally --no_update (so it doesn't overwrite your generator).
It might work, because it has been made on Haswell.

Rico

Topic: Large Bitcoin Collider (Collision Finders Pool) - page 34. (Read 193404 times)