Important to note that GEM performs one order of magnitude better than VGEM or
DRM.
This commit also sets by default the printing of profiling numbers.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
This commit wraps all VGEM related code with 'USE_VGEM' macro, which gives the
chance to enable a different running path using DRM fd without VGEM. This is
particularly useful to test and compare the characteristics of VGEM and DRM.
One can switch VGEM or DRM in compilation time by defining 'USE_VGEM' (which is
by default on).
Self annotation: if we opt out VGEM, just like the vgem_fb_test is now doing
here, we achieve the same feature promised of map graphics buffers imported
from a different process that in turn had allocated them. I.e. there's no GEM
(or driver) specific optimizations such as tiling, memory coherency,
whatsoever. Am I missing something here? Reference: http://crbug.com/426172
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
This CL introduces -p option to vgem_fb_test. For now it only prints the
time to mmap imported vgem buffers, which is particularly useful to
benchmark and compare with other driver implementations.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
