IRC channel logs

2023-02-09.log

back to list of logs

<stikonas>rickmasters: merged
<Mikaku>rickmasters: why are you using amd64 arch instead of i386? (<https://github.com/ironmeld/builder-hex0/commit/905f87dd643beae0df8584853c64200e70400d01>)
<Mikaku>I thought bootstrapping was only for i386
<Mikaku>also, why 4GB of memory?
<Mikaku>just curious :-)
<rickmasters>Oh it's just the qemu I found on my machine initially. qemu-system-i386 will work fine. It's all 32-bit code.
<Mikaku>OK
<rickmasters>With regards to 4GB of memory, that's probably a bit overkill. I'm using 3GB.
<Mikaku>3GB on an i386 without PAE?
<rickmasters>That's what's needed to hold the builder-hex0 file system, the Fiwix file system, and all the builder-hex0 working memory.
<Mikaku>I think 3GB is too much for a plain i386 arch
<rickmasters>With qemu I've got 3GB mapped without PAE. I'm not sure how realistic that is.
<Mikaku>are you sure you are using all this memory really?
<rickmasters>Yes
<Mikaku>OK
<Mikaku>rickmasters: builder-hex0 setup a GDT with a flat 4GB segment?
<rickmasters>Mikaku: yes
<Mikaku>I see, thanks!
<rickmasters>sure
<stikonas>rickmasters: I guess mes uses quite a bit of memory?
<rickmasters>stikonas: I believe it used 560MB compiling tcc
<stikonas>yeah, that's quite a lot...
<rickmasters>stikonas: And it spawns another process and with builder-hex0's single process design, it copies it when its suspended so its really double that.
<stikonas>well, we can't expect it to be optimal, it's a small bootstrap kernel
<rickmasters>the copy could be avoided with some page mapping code but I haven't had to do that.
<doras>fossy: do you have GCC10 already working reasonably well?
<doras>If so, do you think you could share your changes soon so we could try it out? It doesn't have to be final yet of course, but if it already works it can give us a headstart in attempting to build upon it.