On Sun, Jan 22, 2012 at 07:39:39PM +0100, Dominique Rodrigues wrote: > Le 22/01/2012 15:52, Dominique Rodrigues a écrit : > >Le 22/01/2012 15:46, Alon Levy a écrit : > >>On Sun, Jan 22, 2012 at 03:18:14PM +0100, Dominique Rodrigues wrote: > >>>Le 20/01/2012 20:04, Yaniv Kaul a écrit : > >>>>----- Original Message ----- > >>>>>>samples pcnt function DSO > >>>>>> > >>>>>>3796.00 60.4% hashlittle > >>>>>This sucks, unfortunately I don't have any quick fix for it. It's the > >>>>>hash computation done on each image. Needless to say it needs to be > >>>>>optimized - either made more efficient or reduce the number of calls. > >>>>> > >>>>>Soren, FYI. > >>>>You could move to murmur2, like Windows' QXL driver > >>>>(https://bugs.freedesktop.org/show_bug.cgi?id=37465) or > >>>>murmur3 or other faster ones > >>>>(https://bugs.freedesktop.org/show_bug.cgi?id=35775). > >>>>At least it's not hashed twice > >>>>now(https://bugs.freedesktop.org/show_bug.cgi?id=37977 - > >>>>worth updating bug status?!), but I'm not sure it should be > >>>>hashed at all, if it's not cached. > >>>>Y. > >>>That's true : when you draw in Gimp in a Windows Seven (64 bits) > >>>guest, you have absolutely no delay, at any speed. > >>> > >>>Soren has recently improved a lot the qxl driver for Linux, but if > >>>you accelerate the mouvement of your mouse, the drawing can't still > >>>follow your mouse in Gimp. > >>> > >>>I don't know why different functions have been choosen into the qxl > >>>drivers for Windows and Linux, but it could be interesting to > >>because it was done by different people at different times. > >> > > > >That's clear. I also thought that this was the main reason. > > > >However, since the qxl driver is the key to benefit from the best > >user experience in SPICE, my opinion is that it is important to > >optimize it as far as possible. Every effort for that goal is > >welcome. > > > >Dominique > > > >>>benefit from the best of these different versions. > >>> > >>> > >>>Dominique Rodrigues > > By the way, since last improvements from Soren on the qxl driver, > the new consuming function is "qxl_update_area".
Soren, would that be fixed by Xrender? > > Output of perf top (during rapid drawing in Gimp) : > > > PerfTop: 930 irqs/sec kernel:39.8% exact: 0.0% [1000Hz > cpu-clock], (all, 1 CPU) > ---------------------------------------------------------------------------------------------------------------------------------- > > samples pcnt function DSO > _______ _____ _____________________________ > __________________________________________ > > 1313.00 49.6% qxl_update_area > /usr/lib64/xorg/modules/drivers/qxl_drv.so > 294.00 11.1% retint_careful [kernel.kallsyms] > 122.00 4.6% qxl_ring_push > /usr/lib64/xorg/modules/drivers/qxl_drv.so > 92.00 3.5% finish_task_switch [kernel.kallsyms] > 75.00 2.8% _spin_unlock_irqrestore [kernel.kallsyms] > 33.00 1.2% g_type_check_instance_cast > /lib64/libgobject-2.0.so.0.2200.5 > 33.00 1.2% pixman_blt_sse2 > /usr/lib64/libpixman-1.so.0.18.4 > 31.00 1.2% _int_malloc > /lib64/libc-2.12.so > 30.00 1.1% __pthread_mutex_lock_internal > /lib64/libpthread-2.12.so > 30.00 1.1% g_type_check_instance_is_a > /lib64/libgobject-2.0.so.0.2200.5 > 28.00 1.1% g_hash_table_lookup > /lib64/libglib-2.0.so.0.2200.5 > 23.00 0.9% g_slice_alloc > /lib64/libglib-2.0.so.0.2200.5 > 21.00 0.8% __pthread_mutex_unlock > /lib64/libpthread-2.12.so > 21.00 0.8% unix_poll [kernel.kallsyms] > 19.00 0.7% native_flush_tlb [kernel.kallsyms] > 16.00 0.6% __do_page_fault [kernel.kallsyms] > 15.00 0.6% _int_free > /lib64/libc-2.12.so > 15.00 0.6% __memcpy_ssse3 > /lib64/libc-2.12.so > 14.00 0.5% __strstr_sse2 > /lib64/libc-2.12.so > 14.00 0.5% clear_page_c [kernel.kallsyms] > 11.00 0.4% do_select [kernel.kallsyms] > 9.00 0.3% qxl_surface_create > /usr/lib64/xorg/modules/drivers/qxl_drv.so > 8.00 0.3% g_atomic_int_exchange_and_add > /lib64/libglib-2.0.so.0.2200.5 > 8.00 0.3% g_param_spec_pool_lookup > /lib64/libgobject-2.0.so.0.2200.5 > 8.00 0.3% __GI___strcmp_ssse3 > /lib64/libc-2.12.so > 7.00 0.3% __GI_vfprintf > /lib64/libc-2.12.so > 7.00 0.3% g_slice_free1 > /lib64/libglib-2.0.so.0.2200.5 > 7.00 0.3% sock_poll [kernel.kallsyms] > 7.00 0.3% __GI___libc_malloc > /lib64/libc-2.12.so > > > Dominique > begin:vcard > fn:Dominique Rodrigues > n:Rodrigues;Dominique > org:nanoClouD > adr:;;8, rue Lemercier;Paris;;75017;France > email;internet:dominique.rodrig...@nanocloud.com > title:Directeur Technique > tel;work:+33 (0) 1 77 69 64 38 > tel;cell:+33 (0) 6 28 52 37 70 > url:www.nanocloud.com > version:2.1 > end:vcard > > _______________________________________________ > Spice-devel mailing list > Spice-devel@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/spice-devel _______________________________________________ Spice-devel mailing list Spice-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/spice-devel
