hi..im new to all this 8086 coding lark...havent looked at assembler since
the amiga.. - anyway to the point. I was wondering whether in general it is faster
to access stack data than global/heap data?...

what about something like a matrix? - would i be better off reserving
stack space for matrices for potentially faster access or would i be better
off using temp matrix objects on the heap + simply uses the address in
functions/procedures?

sorry if the questions are stupid..

I ony have a very basic understanding of 8086 assembler and how windows
works internally..so any feedback would be greatly appreciated

cheers!
matt.

There tends to be a bit better chance of the stack being in the CPU cache because, well, the stack is used all the time. On the other hand, the stack is somewhat limited in size, so you shouldn't allocate too large objects there (I tend to go heap once I need more than a few kilobytes).

Stack variables have the advantage of being easier to make thread-safe than use of global variables. But of course you can HeapAlloc and save pointers in stack variables.

thanks - i dont know anything about the cache, last time i coded in assembler
there was no instruction or data cache..and no fpu/mmu :p
..will probably need to hunt out some docs on modern processor architecture..
its all very interesting, if a little confusing..i think i much prefered 68k assembler

http://www.agner.org/optimize/ - here are many useful docs

Also worth reading this presentation by Felix von Leitner, see the bit about alloca vs malloc/new.

http://www.fefe.de/dietlibc/diet.pdf


Me too
But then again x86's have advantages too, they come with GHz+ clock speeds at dirt cheap prices