I think he means that the slowdown would occur just in general while looking at/using a menu. I had this issue as well with the bag in Ash:Phoenix; since the game was rewriting all the text every frame, the gray looked really crappy. I ended up writing a separate routine for the inventory menu that would only update the screen whenever a button was pressed, but it cost me a lot of space
Ah, that's not really much of an issue with E:SoR, since it manages its version of a textshadow buffer rather efficiently. It runs as a 4x4 tilemapper so the text is rather ... aligned... but what I get out of it is the ability to not write to areas to the screen that are not defined in this "textmapper". This is a real asset when updating cursor positions, since the only thing on the textshadow are the characters that form the cursor.
Still, when looking at the inventory, the textmapper *does* have to do a lot of work since it is updating more of the screen than it is usually used to. The real slowdown I'm worried about is what is happening with the printf macro when it is rebuilding the "textshadow" buffer during one of its frames.
To a normal player, this degradation of grayscale quality *might* be tolerable, since it only happens when something onscreen is changing. I'm a little worried about that one frame. Ought I be? Or am I just setting my expectations too high? I mean, I want to make this incarnation "cleaner" and from the scheme I'm coming up with, it has the potential of just being visually messier.
The only reason I'm adopting the scheme is because it makes managing data across pages almost infinitely easier. I'm sure it's a worthy goal to pursue if it means getting the game out earlier, but still. To sacrifice something for a quicker release is also rather upsetting in my eyes. On a slightly better note, I managed to make some core routines faster, so maybe I'm just imagining things.
If anyone wants to try to improve on this textmapper, give it a go. All registers are free for use, including shadow registers. I'm currently using the stack pointer to extract two 4x4 pairs at a time. Space is not (much of) an issue, though don't try unrolling loops. http://iambianet.pastebin.com/XCWgLzUK
Maybe "tomorrow," I'll lay down new groundwork for the next menu system. I've spent too much time reading fanfiction...