Broadcasting with bcast tutorial - Supercomputing and Parallel Programming in Python and MPI 8

Key Takeaways

hello everybody and welcome to another MP4 Pi programming tutorial video for parallel programming high performance Computing and supercomputing in this video what we're going to be discussing is broadcast so the idea of broadcast is so you just have to enter it one time and the object variable whatever is broadcast to all of the nodes so here's a quick example of how that works as usual from NPI 4 Pi we're going to import NPI com equals NPI doc oops np. comore World we're going to be using rank as well so rank equals com. rank now pretty simple if rank equals zero so if this is a master node we're going to say data equals and we're going to use the alphabet here we're going to say a and that corresponds to the first letter B corresponds to second second letter and C corresponds to the third letter so that's our data it's just a dictionary of ABC and what number they re or they rank in the alphabet now what we're going to say so if if we're the master node we specify hey this is the data else data equals none for now so if you wanted to have some of the nodes maybe change their mind on what data is you could but for now we're going to say data equals none on all of the other nodes and then what we're going to say is data equals uh com. bcast so bcast for broadcast and it's going to be data root equals z so where are we getting this from it's going to be zero and now what we want to say is print Rank and then just hit out the rank and then we're just going to after rank we'll spit out that data so uh let's go ahead and save that and I'm going to move it over to the pies now and head over here up arrow and8 and number of processor is five in the last video it only spit out two things just because we only specified things for two things um but just know like at least for us we have five so if you don't have five there your results might look different uh than mine but anyways hit enter sat8 uh let's see what we get sure enough we get the output rank 0 4 1 32 and they all have the exact same uh data here thank you to broadcast so you don't have to type it in all like over and over and over and over again it just adds one more layer of Dynamics plus broadcast is really useful for things like um probably the most real world example that most people have heard of would be something like Bitcoin where the you know current Stat or status of the entire network is broadcast if uh if a node um solves a block let's say it's going to broadcast it to all the other nodes and all of that so um you can do a lot of neat things um with broadcast uh but obviously this was just a pretty basic example of what you can do with broadcast so anyways hopefully you guys enjoyed if you have any questions or comments feel free to leave them below as always thanks for watching