intrigeri wrote (09 May 2014 16:27:14 GMT) :
> I think it raises the bar way to high in terms of hardware needed to
> run the test suite. I also tried:
> * Westmere: my CPU isn't good enough
Works for me on another (sid) host system.
> * Nehalem: fails to start X
Confirmed on another (sid) host system.
> * kvm64: seems to work,
Confirmed on another (sid) host system.
> but likely suboptimal
Not *that* suboptimal, apparently.
The only features Westmere has on top of kvm64 are: aes, lahf_lm,
popcnt, ssse3, sse4.2, sse4.1. Then, SandyBridge also adds: pclmuldq,
x2apic, tsc-deadline, xsave, avx, rdtscp.
So, I propose we:
1. Go back to cirrus graphics, until we have time to sort out the
issues I'm seeing with QXL (ticket++, please)
2. Use kvm64 as a basis (with match='minimum'), and add Westmere- and
SandyBridge-specific features in "optional" mode, that is
"supported by virtual CPU if and only if it is supported by host
CPU". Hopefully it will work on everyone's tester systems, and
still use the most out of recent CPUs. I'll try this on my tester
system, and report back.
Cheers!
--
intrigeri
| GnuPG key @ https://gaffer.ptitcanardnoir.org/intrigeri/intrigeri.asc
| OTR fingerprint @ https://gaffer.ptitcanardnoir.org/intrigeri/otr.asc
