Agner`s CPU blog

It would be nice to be able to provide hints: "I need a temporary chunk of memory that I'll give back soon", or, "I'll be needing a contiguous chunk of X size for a while."

I assume we're talking about a replacement for X86/X64, for servers and desktops. I can symphathize with the goal of optimizing, and wanting to keep everything simple, and that's a noble cause. But, before making it optimal, it has to work first (and by that, I mean it has to be practical, and be able to handle today's apps.)

Consider a web browser. Navigate to a text page (or a forum), and that browser may need a couple of Mb of memory. But, then navigate to an image-laden page, and BAM! You now need dozens of Mb to render all those images into. Some browsers delay the rendering of JPGs, for example, until you actually scroll down to where the JPG is visible. And if you keep scrolling, they toss that rendering from memory.

My point is that, some programs have very complicated memory allocation needs, and that does not suggest that they are written poorly. Those web browser designers are also optimizing, but they are concerned with memory usage, load time, render speed, and much more. Forcing them to allocate memory a different way limits their abilities. And, without paging, they are forced to be thrifty with memory. Which causes fragmentation. Full circle.

If this chip is to sit in people's homes across the world, you've got to expect that many thousands of different programs will be running on them. And each program will be built by people with varying levels of skill. People who will not respect the beauty behind the memory organization scheme. And, you have to cater to them, because if their program brings the system to its knees, your chip and the OS will be blamed.

In other words, some crappy app should be able to disrespect the entire memory space...and your chip and OS should handle it with some grace, I believe. And, even trying to be respectful of the machine's resources seems difficult, as I am torn between releasing no-longer-needed memory (which is supposed to be the right thing to do), and trying to prevent fragmentation (which is, essentially, the opposite, unless I've missed something).

Back to the subject of dreaming about my ideal CPU: These are not well thought out, or even practical, most likely. But they'd be nice Here goes:

1. 16Gb memory on the same chip as the CPU. I don't know what the limitations are, but, if all your PC's memory could be super-fast, on-chip, wow! Memory wait states slow down the CPU a lot, and, ,if it was all on the same chip, you could eliminate all the complex nightmare caching hardware.

2. If not #1, then some really good cache hint/directive instructions.

3. Instead of relying on branch prediction, why not take both branches, and provide the ability to swap pipelines to use the confirmed branch? This dual pipeline could be used for extra execution when branching was not occurring.

4. A hardware block move/fill/swap, page-based. Runs like a background job. An instruction could be used to test for completion by comparing a given address against each pending job's address range.

5. Instead of saving registers on task switch, use an array of registers, indexed by taskid.

I'm sure I could think up some more. As I stated before, I realize that most of these are far-fetched, and do not really fit into Mr. Fog's design. But, maybe one has merit, so I present them for thought. I find the whole project fascinating, and I hope one day that you can get chip to the fabrication stage! Best of luck!