Re: should we get rid of xf86-video-nv?
Rin Okuyama <[email protected]> Sat, 20 Nov 2021 16:19:47 +0900
| Newsgroups | gmane.os.netbsd.devel.x11 |
|---|---|
| Message-ID | <[email protected]> |
On 2021/11/20 16:13, matthew green wrote: > Rin Okuyama writes: >> Hi, >> >> On 2021/11/13 2:23, Michael wrote: >>> As long as we don't have drm on macppc, and we probably never will, >> >> Isn't this too pessimistic? ;) >> >> nouveau(4) is working on aarch64: >> >> https://dmesgd.nycbug.org/index.cgi?do=view&id=5834 >> >> I will buy a cheap card to check whether it can be used on aarch64eb. >> >> If it is no problem with big endian, IIUC, the missing part for powerpc >> should be __HAVE_PMAP_PV_TRACK support to oea pmap. I don't know how hard >> it is, but expect it is straightforward, c.f.: >> >> http://www.nerv.org/netbsd/?q=id:20201220T163826Z.809a29116767ce2d78a3ac3fae5edd5b908f26d6 > > while this would be a good option, what's the benefit of > enabling an very large kernel driver for hardware that is > pretty well supported by the existing driver? to give some > idea, the driver on amd64 is 1894223 bytes (size -t on > "nouveau*.o"), matched with 245851 bytes in nouveau_drv.so, > vs the 181321 byte nv_drv.o. (for reference, the macppc > GENERIC is currently in the 11833223 byte range.) > > i used this setup with a G4 powermac as my tv for several > years in the mid 2000s, and unless you're using a GL heavy > application, i'm not sure of the benefit, and see a bunch > of downsides. > > having it work for pcie and modern cards would be great, > but i don't think removing the "nv" driver would ever be > worth while. i can vaguely imagine having nouveau as a > default later.. Yeah, I don't mean to replace existing kernel/Xserver drivers. I wish to make nouveau(4) working on powerpc just for fun :) Thanks, rin