Re: [Qemu-devel] [PATCH] tests: switch tests to accel=kvm:tcg

[Cornelia Huck]

On Tue, 15 Aug 2017 21:49:16 +0200
Thomas Huth <address@hidden> wrote:

> On 15.08.2017 17:47, Richard Henderson wrote:
> > On 08/14/2017 08:33 AM, Cornelia Huck wrote:  
> >> On Mon, 14 Aug 2017 17:34:15 +0300
> >> "Michael S. Tsirkin" <address@hidden> wrote:
> >>  
> >>> Speed up tests on host systems with kvm support.
> >>> In particular, this fixes tests with --disable-tcg.
> >>>
> >>> Cc: Paolo Bonzini <address@hidden>
> >>> Cc: Thomas Huth <address@hidden>
> >>> Cc: Laurent Vivier <address@hidden>
> >>> Suggested-by: Cornelia Huck <address@hidden>
> >>> Signed-off-by: Michael S. Tsirkin <address@hidden>
> >>> ---
> >>>
> >>> Tested on x86 only.
> >>>
> >>>  tests/boot-serial-test.c | 2 +-
> >>>  tests/pnv-xscom-test.c   | 4 ++--
> >>>  tests/prom-env-test.c    | 2 +-
> >>>  3 files changed, 4 insertions(+), 4 deletions(-)
> >>>
> >>> diff --git a/tests/boot-serial-test.c b/tests/boot-serial-test.c
> >>> index 11f48b0..c3b2e4e 100644
> >>> --- a/tests/boot-serial-test.c
> >>> +++ b/tests/boot-serial-test.c
> >>> @@ -78,7 +78,7 @@ static void test_machine(const void *data)
> >>>      fd = mkstemp(tmpname);
> >>>      g_assert(fd != -1);
> >>>  
> >>> -    args = g_strdup_printf("-M %s,accel=tcg -chardev 
> >>> file,id=serial0,path=%s"
> >>> +    args = g_strdup_printf("-M %s,accel=kvm:tcg -chardev 
> >>> file,id=serial0,path=%s"
> >>>                             " -no-shutdown -serial chardev:serial0 %s",
> >>>                             test->machine, tmpname, test->extra);  
> >>
> >> This has already been changed upstream.  
> > 
> > Ouch.  This is the only real smoke test we have for the tcg backend for the
> > host.  While it is still going to test tcg for whatever machines do not run
> > natively on the host, I can't help think we've lost testing.  
> 
> Ah, well, I didn't think of that when I suggested to use "kvm:tcg"
> here... I rather thought that kvm would be preferable to speed up the
> test. But since the test is already quite fast anyway, it's likely also
> OK to use "tcg:kvm" here. (But we should keep "kvm:tcg" in the pxe-test
> since that is quite slow with TCG, at least on ppc64).

That all means to me that we need more documentation, at least some
comments.

Will there be a -rc4? I can send a patch that reverses the accelerators
and adds a comment today.