| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In core profile, we support up to 16 viewports. However, in the
majority of cases, only 1 of them is actually used - we only need
the others if the last shader stage prior to the rasterizer writes
gl_ViewportIndex.
Processing all 16 viewports adds additional CPU overhead, which hurts
CPU-intensive workloads such as Glamor. This meant that switching to
core profile actually penalized Glamor to an extent, which is
unfortunate.
This patch tracks the number of relevant viewports, switching between
1 and ctx->Const.MaxViewports if gl_ViewportIndex is written. A new
BRW_NEW_VIEWPORT_COUNT flag tracks this. This could mean re-emitting
viewport state when switching, but hopefully this is offset by doing
1/16th of the work in the common case. The new flag is also lighter
weight than BRW_NEW_VUE_MAP_GEOM_OUT, which we were using in one case.
According to Eric Anholt, x11perf -copypixwin10 performance improves by
11.5094% +/- 3.10841% (n=10) on his Skylake.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
Acked-by: Anuj Phogat <[email protected]>
|
|
|
|
|
|
|
|
| |
This just translates to the correct cull distance slot.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
| |
We need these for spir-v/nir shaders.
Reviewed-by: Edward O'Callaghan <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, we were saving off the last nir_block in a vtn_block before
moving on so that we could find the nir_block again when it came time to
handle phi sources. Unfortunately, NIR's control flow modification code is
inconsistent when it comes to how it splits blocks so the block pointer we
saved off may point to a block somewhere else in the shader by the time we
get around to handling phi sources. In order to get around this, we insert
a nop instruction and use that as the logical end of our block. Since the
control flow manipulation code respects instructions, the nop will keeps
its place like any other instruction and we can easily find the end of our
block when we need it.
This fixes a bug triggered by a couple of vkQuake shaders.
Signed-off-by: Jason Ekstrand <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97233
Cc: "12.0" <[email protected]>
Tested-by: Dave Airlie <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
This intrinsic has no destination, no sources, no variables, and can be
eliminated. In other words, it does nothing and will always get deleted by
dead code elimination. However, it does provide a quick-and-easy way to
temporarily tag a particular location in a NIR shader.
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
Cc: "12.0" <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
We already do those checks in filter_tiling. There's no good reason to
repeat them in choose_msaa_layout. If anything they should have been
asserts and not "return false" checks. Also, this check was causing us to
outright reject multisampled HiZ surfaces which wasn't intended.
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The HiZ and CCS tiling formats are always used for HiZ and CCS surfaces
respectively. There's no reason why we should go through filter_tiling and
it's much easier to always get HiZ and CCS right if we just handle them
directly.
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Topi Pohjolainen <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
HiZ buffers can be multisampled and, on Broadwell and earlier, simply using
interleaved multisampling with a compression block size of 8x4 samples
yields the correct HiZ surface size calculations. Unfortunately,
choose_msaa_layout was rejecting multisampled HiZ buffers because of format
checks. Now that we have a simple helper for determining if a format
supports multisampling, that's an easy enough issue to fix.
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
Compressed 1-D textures are not well-defined thing in either GL or Vulkan.
However, auxiliary surfaces are treated as compressed textures in ISL and
we can do HiZ and CCS with 1-D so we need to be able to create them. In
order to prevent actually using them (the docs say no), we assert in the
state setup code.
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The assertion that a format is uncompressed in the multisample layouts
isn't quite right. What we really want to assert is that the format
supports multisampling which is a bit more complicated query. We also want
to assert that it has a block size of 1x1 since we do nothing with the
block size in the phys_level0_sa assignment.
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Topi Pohjolainen <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
Reviewed-by: Nanley Chery <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Nayan Deshmukh <[email protected]>
Reviewed-by: Leo Liu <[email protected]>
|
|
|
|
|
|
|
|
| |
Previous fundamental change in stats gathering added a temporary
SwrWaitForIdle to begin_query and end_query. Code has been reworked to
remove stall.
Reviewed-by: George Kyriazis <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
Create worker pool now computes number of worker threads based on
things like topologies, etc. and creates the pool but doesn't actually
launch the threads. Instead there is a separate start thread pool
function. This allows thread resources to be constructed first before
threads start.
Signed-off-by: Tim Rowley <[email protected]>
|
|
|
|
|
|
| |
Canonicalize to prevent unnecessary JIT compiles.
Signed-off-by: Tim Rowley <[email protected]>
|
|
|
|
|
|
|
| |
- Immediately sleep threads until thread data is initialized
- Fix some compile bugs with AR enabled
Signed-off-by: Tim Rowley <[email protected]>
|
|
|
|
|
|
|
| |
- Move most jitter functionality into SwrJit namespace
- Avoid global "using namespace llvm" in headers
Signed-off-by: Tim Rowley <[email protected]>
|
|
|
|
|
|
| |
Generalize compute dispatch mechanism to support other types of dispatches.
Signed-off-by: Tim Rowley <[email protected]>
|
|
|
|
|
|
|
| |
- Fix conflict between windows MemoryFence and llvm::sys::MemoryFence
- Declare gettid()
Signed-off-by: Tim Rowley <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
gcc-4 and earlier don't allow compound literals where a constant
is required in -std=c99/gnu99 mode, so we can't use ISL_SWIZZLE()
when populating the anv_formats[] array. There are a few ways around
it: First one would be -std=c89/gnu89, but the rest of the code
depends on c99 so it's not really an option. The second option
would be to upgrade to gcc-5+ where the compiler behaviour was relaxed
a bit [1]. And the third option is just to avoid using compound
literals. I chose the last option since it keeps gcc-4 and earlier
working.
[1] https://gcc.gnu.org/gcc-5/porting_to.html
Cc: Jason Ekstrand <[email protected]>
Cc: Topi Pohjolainen <[email protected]>
Fixes: 7ddb21708c80 ("intel/isl: Add an isl_swizzle structure and use it for isl_view swizzles")
Signed-off-by: Ville Syrjälä <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes a crash in egl-create-msaa-pbuffer-surface Piglit test
and same crash in many dEQP EGL tests.
I also found that some Qt example did a workaround because of this
crash: https://bugreports.qt.io/browse/QTBUG-47509
v2: Ian pointed out that v1 removed support for all multisample
configs, including window ones. This one removes pbuffer bit
when adding configs, now only pbuffer+msaa gets rejected and
window+msaa continues to work. Fixed also comment (Emil)
Signed-off-by: Tapani Pälli <[email protected]>
Cc: [email protected]
Reviewed-by: Marek Olšák <[email protected]>
Reviewed-by: Emil Velikov <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
| |
Using consistent naming allows us to create macros more easily.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
| |
Using consistent naming allows us to create macros more easily.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
| |
There's an assert right above this.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
| |
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
While the current CFG code is valid in the case where a switch break also
happens to be a loop continue, it's a bit suboptimal. Since hardware is
capable of handling the continue as a direct jump, it's better to use a
continue instruction when we can than to bother with all of the nasty
switch break lowering.
Signed-off-by: Jason Ekstrand <[email protected]>
Cc: "12.0" <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
It is possible that the break block of a switch is actually the continue of
the loop containing the switch. In this case, we need to identify the
break block as a continue and break out of current level of CFG handling.
If we don't, the continue portion of the loop will get handled twice, once
by following after the break and a second time by the loop handling code
handling it explicitly.
This fixes 6 of the new Vulkan CTS tests:
- dEQP-VK.spirv_assembly.instruction.graphics.opphi.out_of_order*
- dEQP-VK.spirv_assembly.instruction.graphics.selection_block_order.out_of_order*
Signed-off-by: Jason Ekstrand <[email protected]>
Cc: "12.0" <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Eric Engestrom <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Eric Engestrom <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Eric Engestrom <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
| |
CovID: 1373369
Signed-off-by: Eric Engestrom <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
| |
These structs will be written to disk as part of the shader cache
so use uint32_t just to be safe.
Reviewed-by: Jason Ekstrand <[email protected]>
|
|
|
|
|
|
|
| |
The warning was also the wrong location, it should have been
in the else.
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
| |
Almost all of the other drawing validation code is in api_validate.c
so put this function there as well.
Reviewed-by: Anuj Phogat <[email protected]>
|
|
|
|
|
|
|
|
| |
Detect all of the CPUs in the system. Expose metrics
for min, max and current frequency in Hz.
Signed-off-by: Steven Toth <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
|
|
| |
This reverts commit dbfeb0ec12d6550e68de1bcd164e422e79bccf2d.
With max_value being rounded to 100, it's often wrong.
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
| |
Reviewed-by: Emil Velikov <[email protected]>
|
|
|
|
|
|
|
| |
Regression introduced by "gallium/radeon: zero all query buffers".
Cc: Michel Dänzer <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
| |
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
|
|
| |
This is only needed for r600 which doesn't have ARB_query_buffer_object and
therefore wouldn't really need the fences, but let's be optimistic about
filling in this feature gap eventually.
Cc: Dieter Nützel <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>
|
|
|
|
|
|
|
| |
We don't plan to use sub-allocated buffers with UVD, but just in case one
slips through, this increases the chances of things working out anyway.
Reviewed-by: Christian König <[email protected]>
|
|
|
|
|
|
|
| |
We don't plan to use sub-allocated buffers with VCE, but just in case one
slips through, this increases the chances of things working out anyway.
Reviewed-by: Christian König <[email protected]>
|
|
|
|
|
|
|
| |
Cc: Christian König <[email protected]>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97976
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97969
Reviewed-by: Christian König <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Implement support for power based sensors, reporting units in
milli-watts and watts.
Also, minor cleanup - change the related if block to a switch.
Tested with two different power sensors, including the nouveau
'power1' sensors on a GTX950 card.
Signed-off-by: Steven Toth <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Commutativity is not allowed with SHLADD, but src2 can accept
loads. To allow the load propagation pass to do its job, add a
special case like for SUCLAMP because src1 is always an immediate.
This IMAD to SHLADD optimization helps a bunch of shaders from Tomb
Raider, Victor Vran, UE4 demos (+15% perf with Elemental) and Shadow
Warrior.
GF100/GK104:
total instructions in shared programs :2838045 -> 2834712 (-0.12%)
total gprs used in shared programs :396684 -> 396386 (-0.08%)
total local used in shared programs :34416 -> 34416 (0.00%)
local gpr inst bytes
helped 0 326 1105 1105
hurt 0 55 3 3
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
|
|
|
|
|
|
|
| |
Only and only if src1 is a power of 2 we can replace IMAD by SHLADD.
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Ilia Mirkin <[email protected]>
|