gs@vw25.chips.com (George Scolaro) (10/19/90)
[In the message entitled "memtest & bigger EPROM" on Oct 19, 8:47, Gary Jennejohn writes:] > I tried to write a simple memory sizing routine which wrote a pattern at > 1 MB boundaries and read it back to see if there was memory there. Much to > my astonishment I found memory at up to 48 MB, after which I stopped the test. > I only have 8 MB installed. Does the memory wrap or something? How are we > supposed to dynamically determine how much memory is installed? Any ideas? Yes, of course the memory wraps :-) I would have to have a very wide decoder, remember we have 32 bits of address. You'll notice that only A31-A27 go into the dec32 decoder pal. So, it will wrap all the way to 128 megabytes. The way to test for wrap is to write a pattern at each boundary and see if it affects memory in the base page, then you'll know that the memory that you can see is actually 'wrapped' memory. Though we support 1 mbit and 4 mbit SIMM memory, the smallest installed memory is 4 mbytes, thus you can check for memory on 4 mbyte boundaries. Initialise page 0, word 0, then go up 4 mbytes to page 1, jiggle the word there, see if it breaks page 0, word 0 and repeat... Note that if you only have 1/2 the memory installed, i.e. one bank, then the memory will not appear at the second bank (obvious right? and thus the normal memory good/bad will work, i.e. you wont read what you wrote). The sizing problem (due to wrap) only occurs when both banks are installed. Note also, that if you don't have both banks installed and the parity is enabled (i.e. you have initiated the parity clear function by reading PARCLU port) then you will get a parity error when reading for memory at the 2nd bank if it is not installed. Of course if both banks are installed you wont get a parity error, but memory will wrap. I hope the above makes sense! Looks a bit like I'm rambling :-) If you still can't see through the mud I've deposited up there, then email me again, and I'll get Dave to give you a software description.... best regards, -- George Scolaro (gs@vw25.chips.com) Chips & Technologies (408) 434-0600 3050 Zanker Road San Jose, CA 95134
garyj@neptun.pcs.com (Gary Jennejohn) (10/20/90)
Thanks for the info on how to use a bigger EPROM, George. It turns out that, with the memory tests integrated, there's still about 1k free. I'm going to give the new monitor to Jordan Hubbard so he can test it out too and then I'll provide the changes to the rest of the group. I tried to write a simple memory sizing routine which wrote a pattern at 1 MB boundaries and read it back to see if there was memory there. Much to my astonishment I found memory at up to 48 MB, after which I stopped the test. I only have 8 MB installed. Does the memory wrap or something? How are we supposed to dynamically determine how much memory is installed? Any ideas? Gary