Strange, I swear some flash APPS runs fine even when you're almost out of user RAM, on a 83+. Example, you got 200 bytes of free RAM and still can run MirageOS. I know there are SafeRAM areas but I doubt only about 2 KB of code is copied to the calc at a time. It might be good that someone like Calc84maniac, BrandonW, FloppusMaximus, KermMartian or ThePenguin77 confirms this.

Also I thought only rewriting the flash chip sectors weared it out, not reading?

Quote from: critor on October 11, 2010, 08:36:36 pm

It seems there's been some confusion bewteen RAM and ROM in some posts above.

TI's applications are stored in ROM (Flash-ROM).
Casio's add-ins are equivalent, and are stored in ROM too.

So the geometry add-in has nothing to do with the RAM of the Prizm.

Yes, Casio addins are stored in ROM (Flash). But to be executed, it has to be loaded into the RAM. My experience with programming on the fx-9860G tells me this. I believe this is true actually for TI and Casio calcs. Flash is good for storage, but not good for running programs off of and is very slow (compared to RAM). Read and write cycles will shorten down the life of the flash chip and kill it. TI's I think do this too: copies program from Flash and places them into a space of RAM reserved by the OS to be executed. I'm pretty sure the TI-89 does that for sure. On Z80 based TI-8X's, portions of a program are copied into RAM when necessary by code banking (or as some of you call code swapping) due to the 16-bit addressing limit on the Z80.

I think a lot of confusion comes from the way TI uses it's RAM (to store user data), whereas Casio uses it's RAM in a more modern computer-ish way. (The 61KB of BASIC program memory is actually a tiny chunk of Flash. On the fx-9860G, a hack was done to expand the available BASIC programming storage from 64KB to I think 300KB since it's just Flash memory, not RAM. The RAM is completely separate from the 61KB)

Interesting...
So it's like the TI-Nspire.


Thank you very much for sharing your knowledge.

Quote from: critor on October 11, 2010, 09:17:30 pm

Thank you very much for sharing your knowledge.

Yes, thank you.  I've been learning a lot with every word you post.

Ditto. I think there needs to be a unified Casio-TI forum. The calcs are similar enough for it.

EDIT: Also, there are types of Flash that can execute almost as quickly as RAM. I doubt TI spent the money to implement them, but perhaps one of the people mentioned previously could comment.

I'm just guessing and I could be totally wrong, because I'm not a TI guy, but the portion of RAM used by the OS is temporarily replaced by your flash app. When you close or exit out of your flash app, a system call to some portion of RAM (your SafeRAM probably?) that still contains a tiny piece of the OS that will load back the OS from the ROM to the RAM. That would explain your case of why APPS still run fine when you're almost out of RAM.

Also I thought only rewriting the flash chip sectors weared it out, not reading?

I believe that is true for most flash chips. But it would be slower to run off a program that was on Flash, than on RAM since Flash typically have longer reading cycles than RAM.

I am 100% sure that the TI-83+ to TI-84+SE run Apps directly from Archive.

Hmmm, that actually wouldn't seem very smart in a first glance, but it might make a little sense actually. It's a cheap solution. Z80 is kind of slow (compared to today's ARM and SuperH) so the CPU wouldn't be outrunning the Flash and the programs aren't extremely large (like 100KB or bigger) so not much reading would have to be done.

EDIT: Wait a sec, I don't see how this is actually possible. A Z80 has a 16-bit parallel address bus with an 8-bit parallel bus to load in 1 byte for each cycle. With 16-bit's for addressing, the Z80 would only be able to access the lowest 64KB of Flash max. Well, unless the Flash is split into 64KB memory banks and you access each one separately. Or the Z80 has been modified to have it's address bus extended beyond? And since it's running off Flash, is the data bus on the Z80 swapped or disconnected by a multiplexer and connected to the Flash's data output? But then the RAM would be disconnected and there would be no memory for stacks and variables. Or is it 32KB for program instructions and 32KB for program data (variables and stacks)? Or a 48KB/16KB memory map? I need someone here to explain what's happening. 

Interesting...
So it's like the TI-Nspire.

Yes, I'm pretty sure the TI-nSpire would follow the same scheme as the Prizm or fx-9860G since it uses a modern ARM processor. The TI-nSpire, needs to "boot up" by copying the OS to the RAM. If power is lost to the RAM (bad batteries, or something), it's need to be loaded back in again. That's why the TI-nSpire goes into some sort of sleep mode (instead of having power cut off to everything) so the OS stays within the RAM.

EDIT: Yep, I've been reading your guy's wiki: http://hackspire.unsads.com/wiki/index.php/Memory_layout Quote: "Since 32MB of RAM is available, which is quite a lot, the whole OS code is decrypted from the OS image and copied to RAM at boot time, when the message "Loading Operating System..." is displayed during ~8 seconds. The RAM is also used as temporary storage transparently for the user as described above." The fx-9860G and most likely the Prizm do the same thing, except Casio's don't have that 8 second boot time.

Most of today's CPU's are built using Harvard architecture where the program data and instruction memory are separate. Typically, the program data are stored on a hard drive (or some device for storage such as Flash) and the instruction memory is the RAM. All PCs are built this way and almost all computers are like this.

For the safeRAM, there are 3 sections with 768 bytes of RAM used for the screen buffer and stuff like that, another 128 bytes section for storing content to the home screen (16x8 characters) and another section I forgot the size off (it was around 500 bytes if I remember)

Btw I forgot to mention, Kucalc, but this wiki has a lot of good calc documentation for the TI-83+: http://wikiti.brandonw.net . In case no one like BrandonW himself checks this topic soon, it might be good to go check there. I think this page might explain how Flash APPs works: http://wikiti.brandonw.net/index.php?title=83Plus:OS:Variable_Storage_in_the_User_Archive

These models don't actually use a physical Z80, but rather a SoC that has additional hardware that does things beyond what a normal Z80 would be doing. Example: Memory-mapping "pages" of FlashROM in 16KB chunks to sit evenly with the Z80's 64KB physical address space.

Ah, ok a System-on-Chip. Yeah, because there would be no way a regular Z80 would be able to do that unless you did some fancy stuff with the address and data buses.

Quote from: kucalc on October 11, 2010, 09:51:53 pm

Hmmm, that actually wouldn't seem very smart in a first glance, but it might make a little sense actually. It's a cheap solution. Z80 is kind of slow (compared to today's ARM and SuperH) so the CPU wouldn't be outrunning the Flash and the programs aren't extremely large (like 100KB or bigger) so not much reading would have to be done.

To address another concern, the calculator *is* slow enough (even at 15MHz) to run programs directly from FlashROM without needing to insert "wait cycles". Or rather, the Flash is fast enough to keep up with the Z80.

Heh, yeah, that's what I was saying.

Thanks for clearing this up Iambian. I guess I was just embarrassing myself trying to guess how TI apps get executed, thought it would be same way as Casio. I guess we now know for sure that Casio and TI (at least the TI-8x ones) load their apps differently.

One thing I didn't knew is that the 83+ didn't use a real z80 :O