Age | Commit message (Collapse) | Author |
|
Swizzles are now defined everywhere as a field with 12 bits that contains
4 channels worth of meaningful information. Any channel that is unused is
set to RC_SWIZZLE_UNUSED. This change is necessary because rgb instructions
and alpha instructions were initializing channels that would never be used
(channel 3 for rgb and channels 1-3 for alpha) with 0 (aka RC_SWIZZLE_X).
This made it impossible to use generic helper functions for swizzles,
because sometimes a channel value of 0 meant unused and other times it
meant RC_SWIZZLE_X.
All hacks that tried to guess how many channels were relevant have
also been removed.
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=31193
NOTE: This is a candidate for the 7.9 branch.
|
|
Previously, presubtract operations where only being used by instructions
with less than three source source registers.
|
|
|
|
|
|
|
|
Fixes this GCC warning with linux-x86 build.
radeon_dataflow.c: In function 'get_readers_normal_read_callback':
radeon_dataflow.c:472: warning: ISO C90 forbids mixed declarations and code
|
|
This makes rc_get_readers_normal() more conservative than it needs to be,
but it fixes some incorrect behavior in the optimization passes.
|
|
rc_get_readers_normal() supplies a list of readers for a given
instruction. This function is now being used by the copy propagate
optimization and will eventually be used by most other optimization
passes as well.
|
|
rc_for_all_reads_src() and rc_pair_for_all_reads_arg() pass references to
instruction sources to the callback so they can be modified directly.
|
|
The r300 compiler can now emit instructions that select from the presubtract
source. A peephole optimization has been added to convert instructions like:
ADD Temp[0].x, none.1, -Temp[1].x into the INV (1 - src0) presubtract
operation.
|
|
Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
|
|
Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
|
|
Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
|
|
instructions
Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
|
|
Note that control flow instruction support isn't actually fully functional yet.
Signed-off-by: Nicolai Hähnle <nhaehnle@gmail.com>
|
|
This replaces the old NQSSADCE code with the same functionality, but quite
different design. Instead of doing a single integerated pass, we now build
explicit data structures representing the dataflow.
This will enable analysis of flow control instruction, and could potentially
open an avenue for several dataflow based optimizations, such as peephole
optimization, fusing MUL+ADD to MAD, and so on.
|