Like we have done for the VS and for constant-index uniform loads, we use
the sampler engine to get caching in front of the L3 to avoid tickling the
IVB L3 bug. This is also a bit of a functional change, as we're now
loading a vec4 instead of a single dword, though we're not taking
advantage of the other 3 components of the vec4 (yet).
With the driver hacked to always take the varying-index path for all
uniforms, improves performance of my old GLSL demo by 315% +/- 2% (n=4).
This a major fix for some blur shaders in compositors from the
varying-index uniforms support I introduced in 9.1.
v2: Move old offset computation into the pre-gen7 path.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=61554
NOTE: This is a candidate for the 9.1 branch.
Right now we don't have anything with regs_written() > 1 and !inst->mlen,
but that's about to change.
NOTE: This is a candidate for the 9.1 branch.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We want to load vec4s, since loading a vec4 instead of a dword is
basically no increased latency. But for variable indexed access, the
previous requirement of aligned vec4s for a sampler LD was hard to
implement.
Note that this change only affects those messages that use the surface
format, like sampler LDs, but not to the untyped data cache loads we've
used in other cases.
No significant performance difference on my GLSL demo with uniforms forced
to take the varying pull constants path (n=4).
NOTE: This is a candidate for the 9.1 branch.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This puts the rounding-up logic into the function itself instead of all
the callers having to manage it. Also drop an "unused" comment in gen4,
as the stride *is* used for texbos (and will be for uniforms soon).
NOTE: This is a candidate for the 9.1 branch.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
I'm going to want to change the math for gen7 using sampler LD
instructions in a way that gets CSE to occur like we'd hope.
NOTE: This is a candidate for the 9.1 branch.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We weren't inserting it into the list, so it did nothing. This line was
replaced by the MOV/MUL block above.
NOTE: This is a candidate for the 9.1 branch.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This happens quite a bit with varying-index uniform loads. We could also
do better by avoiding the MACH entirely, but there's no reason not to at
least take this step.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
If the active uniform is an array, then the length of the uniform name should
include the three extra characters for the "[0]" suffix, which is required by
the GL 4.2 spec to be appended to the uniform name in glGetActiveUniform().
This avoids the situation where the output buffer does not have enough space
to hold the "[0]" suffix, resulting in an incomplete array specification like
"foobar[0".
NOTE: This is a candidate for the 9.1 branch.
Change-Id: I41e87ba347a7169eec8c575596cc3416adbe0728
Signed-off-by: Haixia Shi <hshi@chromium.org>
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This is a more aggressive version of the old brw_optimize() path. Reduces
cycles spent in the vertex shader on minecraft by 18.6% +/- 10.0% (n=15).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We dump shader source in ir_to_mesa.cpp, and we dump linked programs here,
but we had no reference from the linked programs to their source. This
was preventing improvement of shader-db to use linked shader programs
instead of individual shader files (which is bogus, because it means we
optimize out VS outputs, and don't interpolate FS inputs!)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
If we're drawing to a surface that's 2048 x 2048 pixels or larger there's
danger of fixed-point overflow in the triangle rasterization code. That
leads to various rendering glitches.
Rather than implement some intricate changes to the rasterization code,
simply subdivide triangles into smaller subtriangles to avoid the issue.
Only do this when the drawing surface is larger than 2048 by 2048.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Use the __builtin_ffs, __builtin_ffsll functions whenever we have GCC,
not just for specific platforms. Fixes Solaris build.
Note: This is a candidate for the stable branches.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=62868
Signed-off-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
This can be enabled everywhere that ARB_texture_multisample is
supported -- ARB_texture_storage is supported on everything.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Brian Paul <brianp@vmware.com>
ARB_texture_storage_multisample allows texture parameters to be
queried for TEXTURE_2D_MULTISAMPLE and TEXTURE_2D_MULTISAMPLE_ARRAY
targets.
Some parameters may also be set, with the following exceptions:
- TEXTURE_BASE_LEVEL may not be set to a nonzero value; generates
INVALID_OPERATION
- any state which appears in the `per-sampler` state table may not
be set; generates INVALID_OPERATION
V2: Don't introduce bogus handling of TEXTURE_MAX_LEVEL
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Brian Paul <brianp@vmware.com>
Now that there are 4 variants, just pass the function name into
teximagemultisample rather than reconstructing it.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Brian Paul <brianp@vmware.com>
Adds XML for the extension, dispatch_sanity enabling, and the two new
entrypoints. These are both implemented by calling the shared
teximagemultisample() with immutable=GL_TRUE.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Brian Paul <brianp@vmware.com>
The new entrypoints will come later, but this adds the actual logic for
supporting immutable multisample textures:
- The immutability flag is set as desired.
- Attempting to modify an immutable multisample texture produces
INVALID_OPERATION.
Note: The extension spec does not mention adding this behavior to
TexImage*Multisample, but it seems like the reasonable thing to do.
V2: - Cover missing error cases (unsized formats; texture object zero)
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
[V1] Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Brian Paul <brianp@vmware.com>
This is about to be used in teximagemultisample() when immutable=true.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Brian Paul <brianp@vmware.com>
These haven't been used since we deleted NV_vertex_program support.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
We are intentionally not allocating a slot for gl_ClipVertex. But by
leaving the bit set in the slots_valid, the fragment shader's computation
of where varyings are in urb entry coming out of the SF would be off by
one. Fixes rendering in Freespace 2 SCP, and improves rendering in TF2.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=62830
Tested-by: Joaquín Ignacio Aramendía <samsagax@gmail.com>
NOTE: This is a candidate for the 9.1 branch.
Reviewed-and-tested-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Paul Berry <stereotype441@gmail.com>
Alessandro Pignotti noted when I added this code in commit
0e723b135b that it's in the else block for
"if (busy)", so this debug print couldn't happen.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
When reading a column from a row-major matrix, we would slot the single
value read into the vector using an ir_dereference_array of the vector
with a constant index. This will (eventually) get optimized to a
masked-write, so just generate the masked write in the first place.
v2: Remove unused variable 'chan'. Suggested by Ken.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: Eric Anholt <eric@anholt.net>
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: Eric Anholt <eric@anholt.net>
Cc: Paul Berry <stereotype441@gmail.com>
Search and replace:
][0] -> ].x
][1] -> ].y
][2] -> ].z
][3] -> ].w
Fixes piglit tests inverse-mat[234].{vert,frag}. These tests call the
inverse function with constant parameters and expect proper constant
folding to happen. My suspicion is that this patch papers over some bug
in constant propagation involving array accesses.
Either way, all of these accesses eventually get lowered to swizzles.
This cuts out the middle man (saving a trivial amount of CPU).
NOTE: This is a candidate for the 9.1 branch.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Cc: Eric Anholt <eric@anholt.net>
Cc: Paul Berry <stereotype441@gmail.com>
Since the case was missing bec4->get_scalar_type() would return bvec4,
but vec4->get_scalar_type() would return float.
NOTE: This is a candidate for stable branches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
"discard" instructions generate HALT instructions which jump to a final
HALT near the end of the shader. Previously, fs_generator created this
final jump target when it saw the first FS_OPCODE_FB_WRITE, causing it
to jump right before the FB write epilogue. This is normally good.
However, INTEL_DEBUG=shader_time also has an epilogue section which
records the final timestamp. The frontend emits IR for this just before
FS_OPCODE_FB_WRITE. Unfortunately, this led to the following ordering:
1. Shader Time Epilogue
2. Final HALT (where discards jump)
3. Framebuffer Write Epilogue
This meant that discarded pixels completely skipped the shader time
epilogue, causing no ending timestamp to be written. This obviously
led to inaccurate results.
This patch adds a new FS_OPCODE_PLACEHOLDER_HALT in the IR stream just
before any epilogue sections. This is where the final HALT should be
generated, and makes it easy to ensure the correct ordering:
1. Final HALT
2. Shader Time Epilogue
3. Framebuffer Write Epilogue
For shaders that don't discard, this opcode compiles away to nothing.
The scheduler adds barrier dependencies to make sure that it doesn't
get moved above any FS_OPCODE_DISCARD_JUMP instructions.
One 8-wide shader in GLBenchmark 2.7 dropped from 2291.67 Gcycles to
a mere 5.13 Gcycles.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
I'd previously added the minimum names to understand my dumps, but this
makes dumps in general much easier to read.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
v2 [mattst88]:
- Rebase.
- #define GL_ARB_texture_query_lod to 1.
- Remove comma after ir_lod in ir.h for MSVC.
- Handled ir_lod in ir_hv_accept.cpp, ir_rvalue_visitor.cpp,
opt_tree_grafting.cpp.
- Rename textureQueryLOD to textureQueryLod, see
https://www.khronos.org/bugzilla/show_bug.cgi?id=821
- Fix ir_reader of (lod ...).
v3 [mattst88]:
- Rename textureQueryLod to textureQueryLOD, pending resolution of
Khronos 821.
- Add ir_lod case to ir_to_mesa.cpp.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
x before
+ after
+------------------------------------------------------------------------------+
| x x + |
| xx ++ x + |
| xx ++ + xx ++ |
|x xxx x+++++ + xxx x*x+*+++ + x +|
| |_____|____________A______A____M____M_|_______| |
+------------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 23 8083.78 8287.83 8205.55 8162.7461 68.307951
+ 23 8107.56 8358.74 8224.33 8186.1765 71.506301
No difference proven at 95.0% confidence
Reviewed-by: Eric Anholt <eric@anholt.net>
Set its latency to what happens to be the default floating-point
instruction latency. One day we may want to handle latency based on
register bank information.
Reviewed-by: Eric Anholt <eric@anholt.net>
Eric Anholt
Vincent Lejeune
Haixia Shi
Matt Turner
Mike Lothian
Brian Paul
Adam Jackson
Dave Airlie
Dave Airlie
Chris Forbes
Kenneth Graunke
Ian Romanick