| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
Reviewed-by: Emil Velikov <[email protected]>
Reviewed-by: Kristian Høgsberg <[email protected]>
|
|
|
|
|
|
|
|
| |
Right now libglsl.la depends on libnir.la so putting it in libnir.la
adds a dependency on libglsl.la that goes the wrong direction.
Reviewed-by: Emil Velikov <[email protected]>
Reviewed-by: Kristian Høgsberg <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The stated bug describes a scenario in which a post sync write operation for
depth or timestamp can be ignored. There are two workarounds suggested, the
first and easier is to simply do a cs stall when we do these type of writes.
The second option is to do a PIPE_CONTROL flush after the post sync but before
the data is required.
Generally, I believe the data written out is consumed by the application on the
CPU side and so doing the easier of the two is ideal. Furthermore, these queries
aren't tremendously common in the perf sensitive apps I have looked at. However,
there could be cases where a shader stage might directly consume the data, and
as a result option 2 may be desirable.
This patch goes with the easier solution for now.
gen9lp bug_de_id=2137196
By itself, this does *not* fix any of the GT4 hangs we're currently
experiencing.
Cc: Mika Kuoppala <[email protected]>
Signed-off-by: Ben Widawsky <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
|
|
|
|
|
|
| |
Cc: [email protected]
Signed-off-by: Ben Widawsky <[email protected]>
Reviewed-by: Kristian Høgsberg <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The R_028B50_VGT_TESS_DISTRIBUTION value is copied from
amdgpu-pro. Smaller values in the ACCUM fields seem to
decrease the performance advantage from this patch, higher
values don't seem to matter.
v2: Add distribution mode field enums.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Using more than 1 wave per threadgroup does increase performance
generally. Not using too many patches per threadgroup also
increases performance. Both catalyst and amdgpu-pro seem to
use 40 patches as their maximum, but I haven't really seen
any performance increase from limiting the number of patches
to 40 instead of 64.
Note that the trick where we overlap the input and output LDS
does not work anymore as the insertion of the tess factors
changes the patch stride.
v2: - Add comment about LDS assumptions.
- Add constant for buffer size.
- Fix code style.
v3: - Correct limits for not splitting patches between waves.
- Set max num_patches to 40 as in the proprietary driver.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
| |
The factors may be stored to LDs by another invocation than
the invocation for vertex 0.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
This allows running the TES on different CU's than the
TCS which results in performance improvements.
v2: Only write the control word from one invocation.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
| |
They are unused.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We always try to use 4-component loads, as LLVM does not combine loads
and they bypass the L1 cache.
We can't use a similar strategy for stores and this is especially
notable with the tess factors, as they are often set with separate
MOV's per component in the TGSI.
We keep storing to LDS and the LDS space, so we can load the outputs
later, either due to the shader, of for wrting the tess factors.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We need to copy the VS outputs to memory. I decided to do this
using a shader key, as the value depends on other shaders.
I also switch the fixed function TCS over to monolithic, as
otherwisze many of the user SGPR's need to be passed to the
epilog, which increases register pressure, or complexity to
avoid that. The main body of the fixed function TCS is not
that interesting to precompile anyway, since we do it on
demand and it is very small.
v2: Use u_bit_scan64.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Instead of creating a memory area per patch and per vertex, we put
the same attribute of every vertex & patch together. Most loads
and stores access the same attribute across all lanes, only for
different patches and vertices.
For the TCS this results in tightly packed data for 4-component
stores.
For the TES this is not the case as within a patch the loads
often also access the same vertex. However if there are < 4
vertices/patch, this still results in a reduction of the number
of cache lines. In the LDS situation we only do better than worst
case if the data per patch < 64 bytes, which due to the
tessellation factors is pretty much never.
We do not use hardware swizzling for this. It would slightly reduce
the number of executed VALU instructions, but I had issues with
increased wait times that I haven't been able to solve yet.
Furthermore, the tbuffer_store intrinsic does not support both
VGPR offset and an index, so we have a problem storing
indirectly indexed outputs. This can be solved by temporarily
storing arrays in LDS and then copying them, but I don't think
that is worth the effort. The difference in VALU cycles
hardware swizzling gives is about 0.2% of total busy cycles.
That is without handling the array case.
I chose for attributes instead of components as they are often
accessed together, and the software swizzling takes VALU cycles
for calculating offsets.
v2: - Rename functions to get_tcs_tes_buffer_address.
- multiply by 16 as late as possible.
- Use tgsi_full_src_register_from_dst.
- Remove some bad comments.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This happens to be in the right position, but that changes
when TCS/TES get new parameters.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
v2: - Use llvm.admgcn.buffer.load instrinsics for new LLVM.
- Code style fixes.
v3: - Code style fix.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Nicolai Hähnle <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
The buffer is quite large, but should only be allocated if the
application uses tessellation. Most non-games don't.
v2: - Use the correct register for SI.
- Add define for block size.
Signed-off-by: Bas Nieuwenhuizen <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
| |
CID 1271532 (#1 of 1): Out-of-bounds read (OVERRUN)34. overrun-local:
Overrunning array of 2 16-byte elements at element index 2 (byte offset
32) by dereferencing pointer &inst.Dst[i].
Signed-off-by: Rob Clark <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Not sure why coverity calls this an out-of-bounds read vs out-of-bounds
write.
CID 1358920 (#1 of 1): Out-of-bounds read (OVERRUN)9. overrun-local:
Overrunning array r of 3 16-byte elements at element index 3 (byte
offset 48) using index chan (which evaluates to 3).
Signed-off-by: Rob Clark <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
XY_FAST_COPY_BLT command doesn't have a field for raster operation. So, fall
back to using XY_SRC_COPY_BLT to handle those cases.
Fixes piglit test gl-1.1-xor-copypixels when fast copy blit is enabled
for all tiling formats.
Signed-off-by: Anuj Phogat <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Experimentation with different values of src/dst horizontal/vertical
alignment showed that these fileds are not used on gen9 hardware.
A recent update in graphics specs has removed these fields from
XY_FAST_COPY_BLT command.
Cc: Ben Widawsky <[email protected]>
Cc: Chad Versace <[email protected]>
Signed-off-by: Anuj Phogat <[email protected]>
Reviewed-by: Ben Widawsky <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
To read out MP perf counters we use a compute shader and need to upload
input data like a 64-bits addr used to store the values and a sequence
ID for synchronization. Currently, this input data is uploaded as user
uniforms which means that it's sticked to c0[], but if a compute shader
from a real application is used, monitoring those performance counters
will just overwrite some data and miserably crash.
Instead, sticking the 64-bits addr and the sequence into the driver
constant buffer seems like much better and will allow to monitor
counters with GL 4.3 apps.
Tested on GF119 and GK110, but should not hurt anything on GK104.
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Kristian Høgsberg Kristensen <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Kristian Høgsberg Kristensen <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
eglCreatePbufferSurface should generate an EGL_BAD_MATCH error if:
1: The EGL_TEXTURE_FORMAT attribute is EGL_NO_TEXTURE and EGL_TEXTURE_TARGET
is something other than EGL_NO_TEXTURE
2: EGL_TEXTURE_FORMAT is something other than EGL_NO_TEXTURE and
EGL_TEXTURE_TARGET is EGL_NO_TEXTURE.
This fixes the dEQP-EGL.functional.negative_api.create_pbuffer_surface test.
Signed-off-by: Plamena Manolova <[email protected]>
Reviewed-by: Ben Widawsky <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Example:
Gallium 0.4 on AMD TONGA (DRM 3.2.0 / 4.5.0, LLVM 3.9.0)
My kernel version is pretty long already (4.5.0-amd-01025-g32791c1)
and adding "kernel" into the string would make too it long for glxinfo
to display.
Reviewed-by: Michel Dänzer <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Ported from the initial amdgpu winsys from the private AMD branch.
The thread creates the buffer list, submits IBs, and cleans up
the submission context, which can also destroy buffers.
3-5% reduction in CPU overhead is expected for apps submitting a lot
of IBs per frame. This is most visible with DMA IBs.
v2: use a semaphore instead of a busy loop in amdgpu_ws_queue_cs
add another amdgpu_cs_sync_flush call into amdgpu_bo_map
Reviewed-by: Nicolai Hähnle <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fixes the following piglit tests (for softpipe):
/spec/glsl-1.30/execution/built-in-functions/...
fs-roundeven-float
fs-roundeven-vec2
fs-roundeven-vec3
fs-roundeven-vec4
vs-roundeven-float
vs-roundeven-vec2
vs-roundeven-vec3
vs-roundeven-vec4
/spec/glsl-1.50/execution/built-in-functions/...
gs-roundeven-float
gs-roundeven-vec2
gs-roundeven-vec3
gs-roundeven-vec4
Signed-off-by: Lars Hamre <[email protected]>
Reviewed-by: Matt Turner <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
| |
The VMware one is bouncing.
Signed-off-by: Emil Velikov <[email protected]>
|
|
|
|
|
|
|
| |
Not piping this all the way through yet, but no better place to note
this down. This will can be used with NV_viewport_array2.
Signed-off-by: Ilia Mirkin <[email protected]>
|
|
|
|
|
|
|
|
|
| |
For fermi, this likely will require use of linked tsc mode. However on
bindless architectures, we can have as many as we want. As it stands,
the AUX_TEX_INFO has 32 teture handles reserved.
Signed-off-by: Ilia Mirkin <[email protected]>
Reviewed-by: Samuel Pitoiset <[email protected]>
|
|
|
|
|
|
| |
v2: fix 025 test. Add three more tests (Ian Romanick)
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
It executes compiler-glsl on all the available shaders, and it checks
that the outcome is the expected.
Bash code based on the already existing optimization-test
v2: rebasing: use --version option
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Add an option in order to ask to just print the InfoLog, without any
header or separator. Useful if we want to use the standalone compiler
to track only the warning/error messages.
v2: all printfs goes on its own line (Ian Romanick)
v3: rebasing: move just_log to standalone.h/cpp
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It silence by default warnings with function parameters, as the
parameters need to be processed in order to have the actual and the
formal parameter, and the function signature. Then it raises the
warning if needed at verify_parameter_modes where other in/out/inout modes
checks are done.
v2: fix comment style, multi-line condition style, simplify check,
remove extra blank (Ian Romanick)
v3: inout function parameters can raise the warning too (Ian
Romanick)
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
| |
Just to allow to call set_is_lhs on any ast_node without a casting. Useful
when processing a ast_node list that we know it contain ast_expression.
v2: comment out new_value to avoid unused parameter warning (Ian Romanick)
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The current code disallows unsized arrays except at the end of
an SSBO but it is a bit overzealous in doing so.
struct a {
int b[];
int f[4];
};
is valid as long as b is implicitly sized within the shader,
i.e. it is accessed only by integer indices.
I've submitted some piglit tests to test for this.
This also has no regressions on piglit on my Haswell.
This fixes:
GL45-CTS.shader_storage_buffer_object.basic-syntax
GL45-CTS.shader_storage_buffer_object.basic-syntaxSSO
This patch moves a chunk of the linker code down, so
that we don't link the uniform blocks until after we've
merged all the variables. The logic went something like:
Removing the checks for last ssbo member unsized from
the compiler and into the linker, meant doing the check
in the link_uniform_blocks code. However to do that the
array sizing had to happen first, so we knew that the
only unsized arrays were in the last block. But array
sizing required the variable to be merged, otherwise
you'd get two different array sizes in different
version of two variables, and one would get lost
when merged. So the solution was to move array sizing
up, after variable merging, but before uniform block
visiting.
Reviewed-by: Samuel Iglesias Gonsálvez <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
| |
This looks like a cut-paste from above.
Reviewed-by: Anuj Phogat <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes:
GL44-CTS.tessellation_shader.tessellation_control_to_tessellation_evaluation.data_pass_through
As the OUT_TC interface structures weren't matching because
one of them had explicit_xfb_buffer set when it shouldn't.
Reviewed-by: Timothy Arceri <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
| |
Indexed primitives were always using cut-aware primitive assembly,
whether primitive_restart was enabled or not. Correctly pass down
primitive_restart and select optimized PA when possible.
Reviewed-by: Tim Rowley <[email protected]>
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
For now, only enable it on platforms that actually support ETC2.
At this point, Broadwell is only failing 5 (out of 8358) dEQP tests:
dEQP-GLES31.functional.copy_image.non_compressed.viewclass_32_bits.
srgb8_alpha8_r11f_g11f_b10f.renderbuffer_to_texture3d
srgb8_alpha8_rgb10_a2ui.renderbuffer_to_cubemap
srgb8_alpha8_rgb10_a2ui.renderbuffer_to_renderbuffer
srgb8_alpha8_rgb10_a2.renderbuffer_to_texture2d
srgb8_alpha8_rgb9_e5.renderbuffer_to_texture3d
These fail with all methods (meta, blorp, blitter, memcpy).
All are blacklisted from the Android mustpass list, which makes me
wonder whether there's an issue with the tests. The formats in
question work with other targets, and the targets in question work
with other formats...
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
We're dropping Meta in favor of BLORP everywhere we can.
This also fixes bugs when copying cubemaps to 2D, which is currently
broken in the meta pass. BLORP just works.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94198
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
The BLT can't handle S8 because it's W-tiled (at least without
additional funny business, and I'm not sure we care). Disallow
it so it falls back to the CPU path, which works.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
| |
The Meta path handles this, but the CPU/BLT fallbacks did not.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently, it only contains the BLT/CPU fallbacks, so the name is a bit
too generic. But eventually this will use BLORP as well, at which point
the name will make more sense.
The next patch will introduce a second call.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This simplifies things a little - now we only have one (tex or rb?)
if-ladder for src, and a second for dst, rather than four.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Fixes Piglit's arb_copy_image-texview test with the Meta path disabled
(so we hit the blitter/CPU fallback paths).
v2: Add MinLayer even for cube maps (suggested by Ilia).
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Anuj Phogat <[email protected]>
Reviewed-by: Chris Forbes <[email protected]>
|