| Age | Commit message (Collapse) | Author |
|---|
|
NOTE: This is a candidate for the 7.9 and 7.10 branches.
|
|
(1, -_, ...) was converted to (-1, ...) because of the negation
in the second component.
Masking out the unused bits fixes this.
Piglit:
- glsl-fs-texture2d-branching
NOTE: This is a candidate for the 7.9 and 7.10 branches.
|
|
This finishes the implementation of the fragment color clamp control
for ARB_color_buffer_float. I don't wanna keep this stuff in a branch...
|
|
This bug can only be triggered if the source texture is either signed or float.
|
|
and rename them.
|
|
|
|
|
|
|
|
Now the expression V==0 generates one instruction instead of two.
|
|
This fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=31159
NOTE: This is a candidate for the 7.9 and 7.10 branches.
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=34030
NOTE: This is a candidate for the 7.10 branch.
|
|
The scheduler and the register allocator are not good enough yet to deal
with the effects of the register rename pass. This was causing a 50%
performance drop in Lightsmark. The pass can be re-enabled once the
scheduler and the register allocator are more mature. r300 and r400
still need this pass, because it prevents a lot of shaders from using
too many texture indirections.
NOTE: This is a candidate for the 7.10 branch.
|
|
|
|
|
|
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.
|
|
The same number of shaders is now printed regardless of optimizations being
enabled or not, so that we can compare shader stats side by side easily.
|
|
r400 fragment shaders now support up to 64 temporary registers,
512 ALU instructions, and 512 TEX instructions.
|
|
The hw can't do it and the code was useless anyway (it's lowered
in the GLSL compiler).
|
|
|
|
|
|
This workaround fixes rendering of kwin thumbnails.
NOTE: This is a candidate for the 7.9 and 7.10 branches.
|
|
rc_inst_can_use_presub() wasn't checking for too many RGB sources in
Alpha instructions or too many Alpha sources in RGB instructions.
Note: This is a candidate for the 7.10 branch.
|
|
If a source operand has a non-native swizzle (e.g. the KIL instruction
cannot have a swizzle other than .xyzw), the lowering pass uses one or more
MOV instructions to move the operand to an intermediate temporary with
native swizzles.
This commit fixes that the presubtract information was lost during
the lowering.
NOTE: This is a candidate for both the 7.9 and 7.10 branches.
|
|
This fixes broken rendering of trees in ETQW. The trees still disappear
for an unknown reason when they are close.
Broken since:
2ff9d4474bdf5f05852ad4963d0b597d20743678
r300/compiler: make lowering passes possibly use up to two less temps
NOTE: This is a candidate for the 7.10 branch.
|
|
|
|
CMP may now use two less temps, other non-native instructions may end up
using one less temp, except for SIN/COS/SCS, which I am leaving unchanged
for now.
This may reduce register pressure inside loops, because the register
allocator doesn't do a very good job there.
|
|
This bug can only be triggered if you put deadcode before native rewrite.
|
|
|
|
|
|
|
|
The same check is already in a later pass (translate_vertex_program).
|
|
It was a no-op because all swizzles are native there.
|
|
Needed for st/vega.
|
|
The compiler seriously needs a cleanup as far as the arrangement of functions
is concerned. It's hard to know whether some function was implemented or not
because there are so many places to search in and it can be anywhere and
named anyhow.
|
|
|
|
|
|
It looks like the function was originally written for ARB_fragment_program.
NOTE: This is a candidate for the 7.9 branch.
|
|
Fixes this GCC warning with linux-x86 build.
radeon_pair_regalloc.c: In function ‘compute_live_intervals’:
radeon_pair_regalloc.c:222: warning: ISO C90 forbids mixed declarations and code
|
|
Fixes this GCC warning with linux-x86 build.
radeon_pair_regalloc.c: In function ‘compute_live_intervals’:
radeon_pair_regalloc.c:221: warning: ISO C90 forbids mixed declarations and code
|
|
https://bugs.freedesktop.org/show_bug.cgi?id=31193
NOTE: This is a candidate for the 7.9 branch.
|
|
|
|
In addition, the rename_reg pass has been rewritten to use
rc_get_readers().
|
|
Previously, presubtract operations where only being used by instructions
with less than three source source registers.
|
|
|
|
|
|
|
|
|
|
When the result of the alpha instruction is being replicated to the RGB
destination register, we do not need to use alpha's destination register.
This fixes an invalid "Too many hardware temporaries used" error in
the case where a transcendent operation writes to a temporary register
greater than max_temp_regs.
NOTE: This is a candidate for the 7.9 branch.
|
|
This fixes an invalid "Too many hardware temporaries used" error in the
case where a source reads from a temporary register with an index greater
than max_temp_regs and then the source is marked as unused before the
register allocation pass.
NOTE: This is a candidate for the 7.9 branch.
|
|
Reads of registers that where not written to within the same block were
not being tracked. So in a situations like this:
0: IF
1: ADD t0, t1, t2
2: MOV t2, t1
Instruction 2 didn't know that instruction 1 read from t2, so
in some cases instruction 2 was being scheduled before instruction 1.
NOTE: This is a candidate for the 7.9 branch.
|
