!KeEjZhMhJhdPGFlLDY:matrix.org

CF General

13 Members
Welcome to ClusterFights! - Not to be confused with microcluster water.1 Servers

Load older messages


Timestamp Message
20 Feb 2019
23:59:26@di0_thedestroyer:matrix.orgdi0_thedestroyeri mean, i got the gigabit switch so that would be fast already
23:59:34@danielle.wis:matrix.orgdanielle.wiswhats got you slow then
23:59:38@di0_thedestroyer:matrix.orgdi0_thedestroyerbut i gotcha
23:59:53@danielle.wis:matrix.orgdanielle.wishook me up man i wanna help people now that i did all my cluster research and clusterclinic was a bust
23:59:55@di0_thedestroyer:matrix.orgdi0_thedestroyeri have 18 cores
23:59:59@danielle.wis:matrix.orgdanielle.wislink me some code
21 Feb 2019
00:00:17@danielle.wis:matrix.orgdanielle.wisit could be a limited number of cores issue
00:00:26@danielle.wis:matrix.orgdanielle.wislemme see that memcpy
00:00:30@danielle.wis:matrix.orgdanielle.wisi bet you got memcpy's going on
00:01:18@di0_thedestroyer:matrix.orgdi0_thedestroyeroh i got all kinds of memcpys mang you don't even know
00:01:22@di0_thedestroyer:matrix.orgdi0_thedestroyerhttps://bitbucket.org/di0/mrclustoffeleesv3/branch/feat-v3-rewrite
00:01:31@di0_thedestroyer:matrix.orgdi0_thedestroyermaster branch is old, stay on this branch
00:02:13@di0_thedestroyer:matrix.orgdi0_thedestroyercode is messy, full of comments and currently structured more like a script, so sorry about the style grossness
00:05:22@danielle.wis:matrix.orgdanielle.wisits perfect
00:11:27@danielle.wis:matrix.orgdanielle.wisman you wrote so much code
00:11:30@danielle.wis:matrix.orgdanielle.wisim so proud of you
00:11:30@danielle.wis:matrix.orgdanielle.wislol
00:22:39@danielle.wis:matrix.orgdanielle.wisok now that ive actually written a C solution
00:22:43@danielle.wis:matrix.orgdanielle.wisi understand your code lol
00:23:46@danielle.wis:matrix.orgdanielle.wis while(strlen(inMemoryCache) < textFileSize)

^^^ dont use strlen here, you already have that value in the subtotalSocketReadSize variable
00:24:05@danielle.wis:matrix.orgdanielle.wisthat's O(n) time checking memory while you read from the socket
00:27:37@danielle.wis:matrix.orgdanielle.wisstrcat(inMemoryCache, charReceiveBuffer); ^^^ this is an expensive call, dont make it, load the data directly into your buffer with the offset, try to populate the rest of your buffer: int socketReadSize = read(socket_desc, (inMemoryCache + subtotalSocketReadSize), (textFileSize-subtotalSocketReadSize));
00:29:29@danielle.wis:matrix.orgdanielle.wis if you rely on subtotalSocketReadSize and textFileSize to determine where the bounds of your strings are, you can skip the strlen() check, and the strlen() check was the only reason you needed to bzero() out the buffer, so you dont have to bzero() or memset() anymore
00:33:15@danielle.wis:matrix.orgdanielle.wis memcpy(&tempSearchWindow[0], &currentSearchWindow[1], (searchStringSize-1) * 8);

^^^ dont memcpy, it wastes time. find another md5 function to call that takes a pointer and a length as parameters, like Update(char* input, int inputLen), you will have to call Init() on it before Update()
00:35:43@danielle.wis:matrix.orgdanielle.wisdo something like: Md5 md5; // loop md5.Init(); md5.Update( inMemoryCache + offset, 19 ); md5.Final(); // read md5.digestChars
00:36:42@danielle.wis:matrix.orgdanielle.wisthat should get you some good improvements
00:38:09@danielle.wis:matrix.orgdanielle.wisthen, obviously, start 18 threads for serving files
00:41:46@danielle.wis:matrix.orgdanielle.wisalso i love the logic for fightAudio() lol
00:42:10@danielle.wis:matrix.orgdanielle.wisthis could be wicked fast, my dude!
00:47:44@danielle.wis:matrix.orgdanielle.wiswe can also compare the hashes character-by-character (or byte-by-byte) and save 31 cycles in most cases

There are no newer messages yet.


Back to Room List