The code below is very slow: w=rand(300,300);x=rand(300,300); r=vec(rand(300,1));b=rand(300,1); j=12;N=300; function orig() sAll=0.0;k=1; while k<=1000000 sAll=sAll+sum(w[:,j].*(r + b[j]*x[:,j])); k=k+1; end return sAll; end And I write a helper function to sp…

Without information about the variables you’re using, it’s impossible to know. Can you produce a reproducible example?

Please take a look at Please read: make it easier to help you otherwise it will be very hard to provide meaningful help.

Thanks a lot. This is the test code: w=rand(300,300);x=rand(300,300); r=vec(rand(300,1));b=rand(300,1); j=12;N=300; function orig() sAll=0.0;k=1; while k<=1000000 sAll=sAll+sum(w[:,j].*(r + b[j]*x[:,j])); k=k+1; end return sAll; end function w_r_b_x_n(w,r,b,x,…

Please add three backtics like this ``` before and after your code. You needn’t write all your code in loops to achieve performance. You can use the sum intrinsic which is very fast and more accurate, just take care of unnecessary allocations. Write the sum like this: sum( @. @views w[:,j] * (r + …

You can also do sum(w[i, j] * (r + b[j] * x[i, j]) for i in 1:N) to avoid all allocations. (BTW, instead of while and the manual handling of i, better do for i in 1:N.)

Indeed currently it won’t, but depending on the situation it may or may not matter.

Seems like making all your globals const would fix performance more easily.

Can the compiler automatically optimize this code?

General Usage

Seif_Shebl December 22, 2018, 8:34pm 7

If I get it correctly, I think it will not give the exact same result as sum(array), see this for example.

Topic		Replies	Views
Slow code to compute b=A*x Performance question	7	212	July 3, 2024
Compiler Can't Optimize Away Unnecessary Memory Allocs with Convenience Variables Performance	14	739	May 3, 2023
Compiler optimizations with broadcast or map Performance compilation	5	2536	June 27, 2020
Small functions - best practices New to Julia question , compilation , function	5	869	February 8, 2021
Reduce memory allocated in array view and in place sum Performance question	12	813	November 10, 2023

Can the compiler automatically optimize this code?

Related topics