Hi everybody,
I noticed in DirectFB (I'm working with 1.4.2, but can be found in all
versions) the following lines in src/core/surface_buffer.c (e.g. in function
dfb_surface_buffer_lock) :
if (accessor == CSAID_CPU) {
/* If hardware has written or is writing... */
if (allocation->accessed[CSAID_GPU] & CSAF_WRITE) {
/* ...wait for the operation to finish. */
dfb_gfxcard_sync(); /* TODO: wait for serial instead */
/* Software read access after hardware write requires flush
of the (bus) read cache. */
dfb_gfxcard_flush_read_cache();
(*/* TODO: wait for serial instead */* can be found in
src/core/surface_buffer.c and src/core/layer_region.c)
I don't know that much DirectFB internal implementation, but I'm reviewing
an internal patch and I would like to know what you think about it:
Replace the line :
dfb_gfxcard_sync(); /* TODO: wait for serial instead */
using the following function :
dfb_gfxcard_wait_serial( const CoreGraphicsSerial *serial )
That gives something like:
E.g. in src/core/surface_buffer.c, dfb_surface_buffer_lock :
/* Software write access... */
if (access & CSAF_WRITE) {
/* ...if hardware has (to) read... */
if (allocation->accessed[CSAID_GPU] & CSAF_READ) {
/* ...wait for the operation to finish. */
dfb_gfxcard_wait_serial(&allocation->gfxSerial);
/* ...clear hardware read access. */
if (!buffer->locked)
allocation->accessed[CSAID_GPU] &= ~CSAF_READ;
}
Where gfxSerial is a CoreGraphicsSerial that has been added to the struct
__DFB_CoreSurfaceAllocation.
So what do you think about that? Can you confirm that this modification is
doing what the TODO was about?
Thanks in advance,
Fabien, Vallée
_______________________________________________
directfb-users mailing list
[email protected]
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users
