Re: [Xen-devel] [GSoc] Adding Floating Point support to Mini-OS

[Juergen Gross <jgross@xxxxxxxx>]

On 29/03/17 20:53, Felix Schmoll wrote:
> Hi,
> 
> while looking at this some more I came to the following
> questions/assumptions, so I'd be grateful if you could shortly address them:
> 
> -While implementing our own kernel last semester me and my team-mate
> came to believe that pusha/popa were faster that pushing/popping the
> individual registers, since it is just a single command. The Mini-OS
> kernel however does the latter. Is that a conscience performance-trade
> for something or did we just underly a misconception, in that it
> compiles to the same thing in the end?

pusha/popa are 32-bit (and real mode) only. They waste space on the
stack for the stack pointer. As the Linux kernel isn't using them
either I'd suggest to not use pusha/popa.

> -Lazy floating point register saving is similar to Copy-on-write, is
> that correct?

Samuel has already answered this question.

> -There is nothing preventing me from using some floating-point library
> for the user-space test program, right?

I wouldn't bother too much with libraries. Just use the FP types (float,
double) of gcc and do some basic math.

> I'd also appreciate if you could have a quick glance my updated proposal
> (on the GSoC portal) and give me some more feedback on it.

I'd replace the FP library part with basic FP operations based on gcc
support.

While I don't expect major problems I suggest adding a note about
testing in pv- and pvh-mode (both should work in 32- and 64-bit mode).


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel