On Mon, May 10, 2010 at 6:28 PM, Melwyn Lobo <melwyn.l...@gmail.com> wrote: > On Mon, May 10, 2010 at 12:22 PM, Ben Lau <xben...@gmail.com> wrote: >> Hi Melwyn, >> >> On Fri, May 7, 2010 at 6:18 PM, Melwyn Lobo <melwyn.l...@gmail.com> wrote: >>> Hi , >>> I must say I was really impressed with performances of DirectFB when >>> compiling Qtopia 4.6.2 and using DirectFB as Gfx driver.... I tested this >>> release on our platform containing an SH4 processor and 2d blitter using >>> underlying hardware acclerated graphics functions in Directfb. The >>> performace gain for a sample QT application, "forever" (which blits random >>> rectangles on screen ASAP) was 10 times (302 rects/per sec) vis-vis simple >>> stock software QWS sans DirectFb (30 rects per sec). And even if I switch >>> off all accleration from DirectFB (DFBARGS=no-hardware), I managed to get 3 >>> times better performnace (90 rects per sec). >> >> If I understand correctly , you should be using Sti7109 processor. Did >> you build > Yes, an mb442 board with 7109 processor. > >> 4.6.2 with the patches for 4.6.0(available in the SRPM of 4.6.0) > I did not apply any patches, used the basic release 4.6.2
hmm.. No any error for missing mksepcs of linux-sh4-g++ ? >> Moreover, do you >> mind to share your build option? > > Configured with: > ./configure -qt-gfx-directfb -dwarf2 -continue -confirm-license > -verbose -verbose -prefix /home/guest/qt_4.6.2_libs_DFB_sprof/ > -no-qt3support -exceptions -profile -no-rpath -stl -force-pkg-config > -system-sqlite -system-zlib -system-libpng -system-libmng > -system-libjpeg -no-mmx -no-3dnow -no-sse -no-sse2 -no-cups > -little-endian -host-little-endian -depths 8,16,24,32 -opensource > -embedded sh4 -arch sh4 Thx! But I am not able to get an acceptable performance (it is lagging even for a single but large rectangle moving around) Are you start your application with : ./app -qws -display directfb >> >> I am using 4.6.0 on 7105 , the performance over directfb is quite >> disappointed.. >> >> Thanks. >> >>> I would like to understand from DirectFB experts via these queries. >>> My queries: >>> 1. How would you explain this difference in performance, i.e. stock QWS >>> vis-a-vis QT over DirectFB (unaccelerated) >>> 2. For a UMA system with system memory acting as framebuffer, what are >>> benefits when choosing DSCAPS_VIDEO while creating a surface. Is this >>> surface present on the framebuffer or elsewhere? >>> 3. While blitting an updated rectangle, is there an implicit back buffer in >>> DirectFB where this rectangle is first copied before the entire surface is >>> blitted to the framebuffer. >>> 4. I did not use the fusion kernel module from DirectFB for running this QT >>> app. In which use case is it useful and what is its purpose? >>> Thanks so much!! >>> I would really appreciate any explanation/advice. >>> >>> Regards, >>> M. >>> >>> _______________________________________________ >>> directfb-users mailing list >>> directfb-users@directfb.org >>> http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users >>> >>> >> > > Regards, > M. > _______________________________________________ directfb-users mailing list directfb-users@directfb.org http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users
