Re: PCB's internals


fpga_toys@xxxxxxxxx writes:
set. If it takes 0.1 seconds (as you state below) to redraw the
screen while running cleanly out of l1/L2/L3 cache, then you can
accept mouse events at 100 per second

10 per second.

and keep up with the users motion without creating a backlog that
grows queue length. The problem starts when the working set is
exceeded,

Even simpler. If you're getting 10 mouse events per second, and it
takes 0.099 seconds to process them, you're OK. If it takes 0.101
seconds to process them, you're toast. The trip point isn't filling
cache, it's just that the redraw finally takes longer than the
available time, even if it's still running cleanly out of cache.

By combining motion events you slow down the rate of context
switching to the X server, and do more work per context switch.

More importantly, I redraw the board less often. Redraws are
expensive no matter how much cache you have.

2) use segmented tables (arrays, vectors) instead of single object

PCB does this.
.

Relevant Pages

  • Re: PCBs internals
    ... If it takes 0.1 seconds to redraw the screen ... while running cleanly out of l1/L2/L3 cache, ... that is the first step in countering working set problems ... ... Linked lists and trees frequently have a very poor memory usage ...
    (sci.electronics.cad)
  • Re: Difference between first call of site and refresh?
    ... Try Tools-Advanced-Reload from Cache if that also works then I'm sure ... it is a redraw bug.. ...
    (alt.html)
  • Re[2]: How does disk caching work?
    ... >> While you are correct that when cache is emtry kenrel will dip into the inactive queue. ... Pages on the cache queue still have the association. ... > You may want to study the kernel sources some more, ...
    (freebsd-performance)
  • Re: Best database for implementing a cache
    ... However i need to limit the size of the cache so that it may not grow ... metric for evicting existing stored images. ... simply rebuilding meta data at startup (list all files on disk ... with a file and the priority queue will implement file aviction ...
    (comp.lang.java.programmer)
  • Re: [RFC] [PATCH 0/3] ioat: DMA engine support
    ... > the queue of copies long (to amortize the cost of ... Don't forget that there are benefits of not polluting the cache with the ... userland data on the return-to-userspace code path. ... send the line "unsubscribe linux-kernel" in ...
    (Linux-Kernel)