| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Grigori Goronzy <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Guard band clipping speeds up rasterization for primitives that are
partially off-screen. This change in particular results in small
framerate improvements in a wide range of games.
Started by Grigori Goronzy <[email protected]>.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Grigori Goronzy <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
|
|
| |
and clamp it right before emitting. This is a prerequisite for computing
the guard band.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Grigori Goronzy <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
|
| |
Just for consistency. This should have no effect, because OpenGL textures
always go to VRAM.
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Edward O'Callaghan <[email protected]>
|
|
|
|
|
|
|
| |
This makes Tonga with vramlimit=128 2x faster in Heaven.
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Edward O'Callaghan <[email protected]>
|
|
|
|
|
|
|
| |
A better solution will be used.
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Edward O'Callaghan <[email protected]>
|
|
|
|
|
|
| |
this is the default behavior of the closed driver when running on VI
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
| |
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
|
|
| |
The closed driver does this, but it looks at base_level and last_level
and uses a conditional assignment, which LLVM can't generate on SGPRs.
That led me to invent this solution that abuses the image descriptor.
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
| |
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
| |
copied from the Vulkan driver
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
| |
We were overdoing it for non-power-of-two values.
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
Just for consistency. This is actually not a problem, because both addrlib
and radeon check and fix this.
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
This is a remnant of the times when the DDX was allocating depth-stencil
buffers for windows. Now, st/dri allocates them and doesn't share them.
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
Only CMASK and DCC clears can use this, because only textures can be so
large.
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
| |
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
| |
print as unsigned instead of signed
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
| |
v2: fail in radeon_winsys_bo_create if size > 32 bits
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
| |
being able to allocate more than 4 GB may be useful
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
not used anymore
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
All of them are paused only between IBs.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
r600_set_active_query_state does it better.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
Use ZPASS_INCREMENT_DISABLE everywhere.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
| |
The caller does the same checking.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
| |
Acked-by: Edward O'Callaghan <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Grigori Goronzy <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Roland Scheidegger <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Roland Scheidegger <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
| |
Add missing PIPE_SHADER_CAP_INTEGERS for frag shaders to
nv30_screen_get_shader_param().
Signed-off-by: Hans de Goede <[email protected]>
Reviewed-by: Samuel Pitoiset <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
It is incorrect to assume BGRA byte order for the GLES3 sRGB workaround.
v2: use _mesa_get_srgb_format_linear to handle all formats
Signed-off-by: Haixia Shi <[email protected]>
Reviewed-by: Stéphane Marchesin <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Rhys Kidd <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
This adds support for the features requires for ARB_shader_storage_buffer_object
and ARB_shader_atomic_counters, ARB_shader_atomic_counter_ops.
[airlied: some cleanups applied]
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Roland Scheidegger <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
| |
Like the image code, but for shader buffers this time.
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Roland Scheidegger <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This adds support for doing load/store/atomic operations on
buffer objects.
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Roland Scheidegger <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
For atomic operations we really need to avoid executing unnecessary shaders, so for some
tests that just draw a single point we only want one vertex to get processed not 4,
this fixes a number of the atomic counters tests.
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Roland Scheidegger <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
| |
Trivial.
Signed-off-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
| |
Dolphin uses them a lot. Range tracking would be better in the long term,
but this two lines works fine for now.
Signed-off-by: Markus Wick <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This makes the extra multiply visible to NIR's algebraic optimizations
(for constant reassociation) as well as constant folding. This means
that when the result of sin/cos are multiplied by an constant, we can
eliminate the extra multiply altogether, reducing the cost of the
workaround.
It also means we only have to implement it one place, rather than in
both backends.
This makes INTEL_PRECISE_TRIG=1 cost nothing on GPUTest/Volplosion,
which has a ton of sin() calls, but always multiplies them by an
immediate constant. The extra multiply gets folded away.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eduardo Lima Mitev <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
I want to be able to read other fields.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Alejandro Piñeiro <[email protected]>
Reviewed-by: Eduardo Lima Mitev <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Some passes may not refer to options->..., at which point the compiler
will warn about an unused variable. Just cast to void unconditionally
to shut it up.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eduardo Lima Mitev <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Many shaders contain expression trees of the form:
const_1 * (value * const_2)
Reorganizing these to
(const_1 * const_2) * value
will allow constant folding to combine the constants. Sometimes, these
constants are 2 and 0.5, so we can remove a multiply altogether. Other
times, it can create more immediate constants, which can actually hurt.
Finding a good balance here is tricky. While much more could be done,
this simple patch seems to have a lot of positive benefit while having
a low downside.
shader-db results on Broadwell:
total instructions in shared programs: 8963768 -> 8961369 (-0.03%)
instructions in affected programs: 438318 -> 435919 (-0.55%)
helped: 1502
HURT: 245
total cycles in shared programs: 71527354 -> 71421516 (-0.15%)
cycles in affected programs: 11541788 -> 11435950 (-0.92%)
helped: 3445
HURT: 1224
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eduardo Lima Mitev <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Boyuan Zhang <[email protected]>
Reviewed-by: Christian König <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
Instead of using an array indexed by SYSTEM_VALUE_x, just use a
switch statement. This fixes a regression caused by inserting new
SYSTEM_VALUE_ enums but not updating the mapping to TGSI semantics.
v2: fix a few switch statement mistakes for compute-related enums
Reviewed-by: Ilia Mirkin <[email protected]>
|
|
|
|
| |
Reviewed-by: Rob Clark <[email protected]>
|