|
October 31, 2013, 10:20:36 AM |
|
our ASIC miner hardware is suffering a very big BUG,here it is
Our mcu environment is cortex-m3, os is rt thread, tcp / ip protocol stack is lwip. The reason of the problem: lwip connect/close cause memory leaks, which makes MCU crashed about every 60 seconds. So the mining performance can not match the hardware. The efficiency is lower than we expected. our ideas: 1 fix the lwip's memory leak when connect/close socket, solve the reset problem .
2 reduces the times of connect / close socket , you need to change the stratum proxy side, get more works at one time.
mcu side: need to initiate a request to the stratum proxy side, and the work number could be configured. stratum proxy side: needs to acknowledge the request from mcu side and find out how many works need to be transfer to mcu and then transfer them. For example, There could be a "getwork2" which has many "getwork" implemented in it. So, once mcu client has a request (e.g. 10 works), then stratum could transfer 10 works to mcu by only one connetion transaction, rather than 10 times connection.
BTW:since we need to same time so much, we prefer the idea 2) very much,thank you for your time and if you think you can help us to code&achieve it,plz msg me.
|