News for hackers. Stuff that matters.
Re:Memory Bandwidth Napkin Math by Taniwha on Monday February 10, 2020 @04:14PM (#22293442)
Unlike just reading a bunch of random data the read instructions can't be pipelined, the instruction that uses the read pointer can't be dispatched to the load-store unit until after it's address has arrived in the CPU (two reads where you know the address can just be queued, and even finished out of order if the second one hits in a closer cache than the first one)
Re:Memory Bandwidth Napkin Math by fulafel on Tuesday February 11, 2020 @12:07AM (#22296228)
I think you are describing the pessimal pointer chasing case, that should be the smaller figure.
Re:Memory Bandwidth Napkin Math by Taniwha on Tuesday February 11, 2020 @04:32AM (#22297197)
More I'm trying to explain why pointer chasing is going to be slower than random integer accesses
Re:Memory Bandwidth Napkin Math by fulafel on Tuesday February 11, 2020 @12:05PM (#22300057)
Yes, it should be lower. Yet there is a lower figure estimated for a mixed worlkoad.

"A more realistic application might consume 5 GB/s at 144 Hz which is just 69 MB per frame."

But, now I see what it's about: tje 7 GB/s figure is for 6 threads, for 1 thread he gets 2 GB/s.