search

tromp / cuckoo

a memory-bound graph-theoretic proof-of-work system
Other
822 stars 173 forks source link

mean30sx8 error #51

Closed thepiwo closed 6 years ago

thepiwo commented 6 years ago

When running the build instructions on arch linux I get the following error:

./mean30sx8 -n 63
Looking for 42-cycle on cuckoo30("",63) with 50% edges
Using 3168MB bucket memory at 7f6bc0763010,
1x21MB thread memory at 7f6bbf272010,
8-way siphash, and 128 buckets.
nonce 63
genUnodes round  0 size 536870912 rdtsc: 5862158714
mean30sx8: mean_miner.hpp:198: u32 zbucket<BUCKETSIZE>::setsize(const u8*) [with unsigned int BUCKETSIZE = 202752; u32 = unsigned int; u8 = unsigned char]: Assertion `size <= BUCKETSIZE' failed.

The system is a i7 8700k with 32GB of DDR4 Linux philipp-pc 4.14.57-1-MANJARO #1 SMP PREEMPT Sun Jul 22 19:25:51 UTC 2018 x86_64 GNU/Linux

avx and avx2 are supported as stated in the flags from /proc/cpuinfo

is there any more debugging information I can provide?

tromp commented 6 years ago

dear Philipp,

When running the build instructions on arch linux I get the following error:

./mean30sx8 -n 63 Looking for 42-cycle on cuckoo30("",63) with 50% edges Using 3168MB bucket memory at 7f6bc0763010, 1x21MB thread memory at 7f6bbf272010, 8-way siphash, and 128 buckets. nonce 63 genUnodes round 0 size 536870912 rdtsc: 5862158714 mean30sx8: mean_miner.hpp:198: u32 zbucket::setsize(const u8*) [with unsigned int BUCKETSIZE = 202752; u32 = unsigned int; u8 = unsigned char]: Assertion `size <= BUCKETSIZE' failed.

The system is a i7 8700k with 32GB of DDR4 Linux philipp-pc 4.14.57-1-MANJARO #1 SMP PREEMPT Sun Jul 22 19:25:51 UTC 2018 x86_64 GNU/Linux

avx and avx2 are supported as stated in the flags from /proc/cpuinfo

is there any more debugging information I can provide?

Hmm, I cannot reproduce that on my system. Can you check whether this also happens with mean30x8 and mean30x1? Is there a way for me to debug this on your machine?

regards, -John

thepiwo commented 6 years ago

Thank you, I did email you about the possibility to debug on my machine.

tromp commented 6 years ago

determined to be compiler issue (generation of incorrect SSE2 code)