Mahdi (Dec 29 2017 at 02:30): Mahdi (Dec 29 2017 at 02:30):Yeah! I finished my first post a couple days ago and am hoping to get some more done in the next couple of days. Here's the link if you want to check it out: https://googlecodein2017.wordpress.com/
@Mitesh @Lucas Prieels
Sean (Dec 29 2017 at 15:47):That's an outstanding article Mahdi. :)
Sean (Dec 29 2017 at 15:47):Interesting that you get those results without -msse2 -DHAVE_SSE2
Lucas Prieels (Dec 30 2017 at 00:20):That's really well written: I'm impressed!
Lucas Prieels (Dec 30 2017 at 00:26):@Sean @Mahdi I've also solved this task and I ran my program with 1000000 iterations too so that we can compare our results. I've found approximately the same time for bn_randmt and SFMT (which is interesting because I used msse2 but it doesn't seem to be faster). On the other hand, for the STL rand I found a number totally different: 4114 instead of 11461 (even faster than SFMT)! That's because I didn't use the same implementation as you.
Lucas Prieels (Dec 30 2017 at 00:28):Actually, I've checked and my implementation of bn_randmt is much faster (7287 instead of 19560) even though I used the exact same implementation as you, probably because I didn't compile it with CMake
Sean (Dec 30 2017 at 06:26):You'd have to compile with more than just -msse2 .. SFMT takes a flag to use SSE2. Without it, I found it to be on par with STL's, but with it, cuts the time way down.
Mahdi (Dec 31 2017 at 03:42):Could you send me your .cpp file? I'd like to check it out and see what I'm doing wrong
Lucas Prieels (Dec 31 2017 at 09:10):Sure! But I don't think you're doing anything wrong, maybe it's just the compilation process which change something
Last updated: Oct 09 2024 at 00:44 UTC