It’s hard to predict cache behavior without an actual workload, so i would recommend using cachegrind with a real program (not a benchmark) to evaluate the cost of doing things one way or the other.
On Tue, Apr 28, 2020 at 05:24 Guido Stepken <gstep...@gmail.com> wrote: > Certainly. But how is it *implemented* internally? Mostly you suffer > massive performance loss when prepending, because complete linked list gets > moved to a new place in memory. If internal representaion ist a double > cell, one value, pointer to next, then you quickly suffer CPU cache misses. > Wild jumps across memory with upto 18 CPU waitstates for random access. > Means: Your proud 4 GHz machine gets slower than a 250 MHz embedded ESP32 > ARM CPU. E.g Python dqueue doesn't show any performance loss here. > > Have fun! > > Am Dienstag, 28. April 2020 schrieb Wilhelm Fitzpatrick <raf...@well.com>: > > On 4/27/20 2:42 PM, Guido Stepken wrote: > > > >> In most Lisp languages, you only can "append" to a list, never > "prepend".: > > > > "Prepend", aka "add to the beginning" seems the natural (and > non-destructive) operation of Lisp, e.g. > > > > (cons 9 (1 2 3)) -> (9 1 2 3) > > > > ..perhaps that is what you meant? > > > > -wilhelm > > > > > > -- > > UNSUBSCRIBE: mailto:picolisp@software-labde <picolisp@software-lab.de> > ?subject=Unsubscribe > > -- John Duncan
