RAM very much has a speed limit! It's unlikely you hit it at these very low sampling rates, though.
if you're using this low sampling rate, then it's not clear why you'd want to multithread. You need to very clearly write down why you think parallelizing solves your problem – and what exactly that problem is! Best regards, Marcus On 20.10.21 22:12, sp h wrote: > Hi Marcus Thanks, But my problem is not the memory limit speed......my data > is in Ram my > ram has no speed limit..? > I work with a sample rate of 100-500k, I expect my data is on RAM, not > storage memory... > > On Mon, Oct 18, 2021 at 6:57 PM Marcus Müller <[email protected] > <mailto:[email protected]>> > wrote: > > Dear stackprogrammer, > > you can write any software you like in a GNU Radio block's work > function. I'm sure you've done that before! > > However, multithreading a single GNU Radio block that only does a > multiplication or two is almost certainly worse than not doing it. GNU > Radio itself is a heavily multithreaded environment. So, if you take a > block that is not a bottleneck, and make it occupy multiple CPU cores, > you're reducing your overall system performance. > > Of course, if you already looked at your system workload, identified > that the block you want to parallelize is your bottleneck, AND already > know the type of algorithm you do inside can be done efficiently (!) in > parallel, well, that's something else. However, in the last email you > wrote about being throughput-limited, it seemed quite likely that your > storage speed is limiting you. > > Best regards, > Marcus > > On 16/10/2021 13.18, sp h wrote: > > I want to develop a Gnuradio block, in this block, I have a for, in this > > for I compute some multiplay and multiply conjugate ... > > > > my question is how I can implement a for (in Gnuradio block) so it is > > executed in parallel mode... > > > > thanks in advance >
smime.p7s
Description: S/MIME Cryptographic Signature
