mfaktc bug?

[Prime95]

This was submitted by a reliable user:

no factor for M384000053 from 2^76 to 2^77 [mfaktc 0.21 barrett76_mul32_gs]

Note that mfaktc used a barrett76 kernel to TF to 2^77. Bug?

[srow7]

I got barrett87_mul32_gs
using mfaktc 0.21
CUDA driver 9.10
eta 1hr

[kriesel]

Originally Posted by Prime95

This was submitted by a reliable user:

no factor for M384000053 from 2^76 to 2^77 [mfaktc 0.21 barrett76_mul32_gs]

Note that mfaktc used a barrett76 kernel to TF to 2^77. Bug?

Failed to replicate here, with GTX 1080 Ti, mfaktc v0.21 CUDA 8. I tried to tempt it with
Factor=384000053,75,77.

Apr 25 19:47 | 4612  99.9% |  1.562   0m02s |   1148.18    90677    n.a.%
Apr 25 19:47 | 4615 100.0% |  1.571   0m00s |   1141.60    90677    n.a.%
no factor for M384000053 from 2^75 to 2^76 [mfaktc 0.21 barrett76_mul32_gs]
tf(): total time spent: 25m  4.678s

Starting trial factoring M384000053 from 2^76 to 2^77 (39.85 GHz-days)
 k_min =  98382621478680
 k_max =  196765242961865
Using GPU kernel "barrett87_mul32_gs"
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
Apr 25 19:47 |    0   0.1% |  3.196  51m05s |   1122.31    90677    n.a.%
  Apr 25 19:47 |    3   0.2% |  3.398  54m15s |   1055.60    90677    n.a.%

Is it possible it's a copy/paste error?

UID: Kriesel/condorette-gtx1070, no factor for M153500033 from 2^77 to 2^78 [mfaktc 0.21 barrett87_mul32_gs]
UID: Kriesel/condorette-gtx1070, no factor for M149447533 from 2^73 to 2^74 [mfaktc 0.21 barrett76_mul32_gs]

becomes

UID: Kriesel/condorette-gtx1070, no factor for M153500033 from 2^77 to 2^78 [mfaktc 0.21 barrett76_mul32_gs]

via an accidental deletion of the bold text.

[Madpoo]

Originally Posted by kriesel

...Is it possible it's a copy/paste error?...

I don't know. It seems unlikely, and the user kept trying to submit the results over and over (because the server was correctly rejecting the impossible result).

Tried submitting it 4 times I think, and the last attempt had a mix of other TF results. The others were TF to 72 - 75 bits with the same "mfaktc 0.21 barrett76_mul32_gs" client string, but that one result to 77 bits from the 76-bit client ... that's definitely strange.

Like George said, it's a user that we wouldn't expect to try anything fishy, thus wondering if it's some strange bug in the client that mysteriously allowed it.

[dcheuk]

got assignment: exp=384000053 bit_min=76 bit_max=77 (39.85 GHz-days)
Starting trial factoring M384000053 from 2^76 to 2^77 (39.85 GHz-days)
 k_min =  98382621478680
 k_max =  196765242961865
Using GPU kernel "barrett87_mul32_gs"
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
Apr 25 22:55 |    0   0.1% |  1.504  24m02s |   2384.92    82485    n.a.%
Apr 25 22:55 |    3   0.2% |  1.548  24m43s |   2317.13    82485    n.a.%
Apr 25 22:55 |    7   0.3% |  1.455  23m12s |   2465.23    82485    n.a.%

Like Kriesel said, probably a copy and paste error.

Just read Madpoo's comment. Maybe the user submitted it with lots of other ones but found the server rejected that specific one? Then tried to submit it like 4 more times?
Maybe ask the user directly to see if the result.txt is still intact?

[LaurV]

Can't replicate it here either.

Was an IP address check done? As it is not difficult to send results in the name of other user (we did it in the past, for "credit trade"), maybe someone is trying to discredit somebody else?
Or contact the user and ask?