| Age | Commit message (Collapse) | Author |
|---|
|
Instead of leaving pixel values in src_sample registers, compute the pixel
values directl to the data port to save 8 moves. This cannot work when no
computation is done as there is both no way to wait for the sampler to
finish and because the sampler returns data in a different order from that
required by the data port (sigh).
|
|
|
|
|
|
Performance change is in the noise. Also from Carl Worth.
|
|
|
|
This reduces the CPU overhead of memcpying them in every time, for a speedup
in aa24text of around 30%. This is based on work by Carl Worth which is
in the intel-batchbuffer branch.
|
|
Hopefully some day we can be rid of libtool entirely.
|
|
The Intel driver appears to be coded to only work with displays
expecting 18 bit pixels. However I have an application using a LCD
display that expects pixel data in 24 bit format. The difference is
only 2 bits in a single GPU register. This patch implements that
change, controlled by a new driver option, "LVDS24Bit". The default
value is false, which is the previous behavior. When set to true,
then 24 bit panels should work (at least the one I'm testing here
does).
Fd.o bug #15201
Signed-off-by: Mike Isely <isely@pobox.com>
(cherry picked from commit e031cc02e65acfbafb48136dad414751e04425c5)
|
|
(cherry picked from commit 79b18980ac9f60b1978abe421352df965aed1681)
|
|
|
|
|
|
|
|
Based on a patch by Will Wiseman.
|
|
The Intel driver appears to be coded to only work with displays
expecting 18 bit pixels. However I have an application using a LCD
display that expects pixel data in 24 bit format. The difference is
only 2 bits in a single GPU register. This patch implements that
change, controlled by a new driver option, "LVDS24Bit". The default
value is false, which is the previous behavior. When set to true,
then 24 bit panels should work (at least the one I'm testing here
does).
Fd.o bug #15201
Signed-off-by: Mike Isely <isely@pobox.com>
|
|
|
|
Disable panel fitting on 855GM, and fix dither setting.
|
|
Disable panel fitting on 855GM, and fix dither setting.
(cherry picked from commit 2b720262e1235f1c9da860ba3e9181f0c377aa5e)
|
|
Using the updated factors even when BT709 conversion isn't available
(non-965) should still give us better color reproduction. Tested on a
945GM, examining the +/-5% of black bars of videotestsrc.
(cherry picked from commit af92f4f885c8eae2211d09080b2289aa5bab9ec1)
|
|
Using the updated factors even when BT709 conversion isn't available
(non-965) should still give us better color reproduction. Tested on a
945GM, examining the +/-5% of black bars of videotestsrc.
|
|
FD bug #15353. Launchpad bug ID is available for reference.
(cherry picked from commit 24248097988775d62f6c416f2988e74d31c91cd0)
|
|
FD bug #15353. Launchpad bug ID is available for reference.
|
|
(cherry picked from commit 0147c1c84872f7a109721a53d88a539932d9be81)
|
|
(cherry picked from commit 90886f9a602d58b754e9a8d0f1a9c40803d34fa2)
|
|
(cherry picked from commit 0836373dc6e2f8612f120074980561f7ac11f6f7)
|
|
Saving registers means we can run more in parallel.
(cherry picked from commit bfd803e085e938866efb45c67a79facef78ec399)
|
|
Clean up register allocation to never overlap
Always write 4 values for each texture vertex.
(cherry picked from commit a6492661ae07310128eb73c3ef037c42ce7ab184)
|
|
(cherry picked from commit f8081178eb6fda0e405967cbacad532561619262)
|
|
(cherry picked from commit 879f8717b09f79156b218ee9cc2107700190d586)
|
|
(cherry picked from commit 6db8faeb754897b21af045d00f50db9640b080bb)
|
|
(cherry picked from commit 6bb92213374f278387c539bbe05b773e87e11b90)
|
|
(cherry picked from commit 05710145b6fc4ed2c528128b2e6022591a53d050)
|
|
Use macros for register names, modularize functions into separate files.
(cherry picked from commit 08500507284f13ad7084eb231b43e117e9728129)
|
|
(cherry picked from commit 949d73271d7100c1f028fd60f185f4929461304e)
|
|
sf_mask is the same as sf except that it must compute both src and mask uvw
cooefficients, which are conveniently adjacent in the same registers, and so
need only an extended execution width
(cherry picked from commit 492ff1494f782240e6ca68919b2d0b9aa400fc53)
|
|
This involves correctly computing u/v locations based on x/y vectors and
line constants computed in new sf program.
Also, use fewer instructions to make this go a bit faster (2X for 500x500
composite).
(cherry picked from commit 6304b38423f99190a5e54f1a7dcaa75adfad4f2a)
|
|
(cherry picked from commit 771a56b1ed0df69345c723cb62a73b6842cd8227)
|
|
(cherry picked from commit 4f469189fed541549e5d470b2529275a29cc2f20)
|
|
|
|
|
|
|
|
The 2-bit input_mask was actually an input count -- in0 is always there, and
in1 is optional.
The output flags weren't being reported in the log, so I mistakenly took
controlled_output == RGB0 to mean that the device only reported an RGB0,
while it actually reported RGB0|SVID0|YPRPB0|misc|other. Move SVID0 up
in priority and remove the RGB-is-it-really-TV hack I had just come up with.
Finally, set the input/output mapping at mode set time. We're always
supposed to do this, but haven't had to so far as we've never handled
devices with more than one output.
|
|
Still doesn't light anything up.
|
|
Doesn't include properties for selecting TV formats or picture enhancements,
and totally untested.
|
|
|
|
|
|
|
|
Saving registers means we can run more in parallel.
|
|
Clean up register allocation to never overlap
Always write 4 values for each texture vertex.
|
|
|
|
|
