IRC channel logs
2020-10-21.log
back to list of logs
<OriansJ>M1 completed 8hours of fuzzing with no segfaults and one false positive for hanging; I think that is stable enough for end users <OriansJ>nothing like 10 times faster and just as stable as ever <OriansJ>now I just need to repeast that miracle with hex2 (certainly can't do it with blood-elf) <OriansJ>unless anyone looking at this: https://paste.debian.net/1168048/ has a solution to reduce the sub 1/10 second runtime needed to process M2.M1 even further because at this speed profiling tools stop working so well. <nimaje>well, make some bigger example to profile and hope making the bigger example faster helps with the normal usecase or lable it fast enought <OriansJ>nimaje: well the biggest example is the largest program MesCC can compile and as a single pass process that ignores everything until it sees a :label and then generates the output and then just dumps it all quickly; the only thing to make it faster is to steal the grep solution to fast pattern matching and I am not sure M2-Planet could build that <OriansJ>So right now Disk I/O R/W is the speed limiter; so something to make file_print faster would likely be the optimization possible to implement but I don't see it yet <OriansJ>unless the solution is add read/write buffering to M2-Planet (then implement fflush properly) <fossy>not without much more complex kernel support requirements