| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
Integer textures generate invalid operation in glGenerateMipmap.
So, the code related to integer textures is now redundant.
Note: This is a candidate for stable branches.
Signed-off-by: Anuj Phogat <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
| |
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=55030
|
|
|
|
|
|
|
|
|
|
| |
It's usually forced to 1 by the surface format, but sometimes we actually have
alpha present because it's the only format available.
Fixes piglit texwrap bordercolor tests for OpenGL 1.1, GL_EXT_texture_sRGB and
GL_ARB_texture_float.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
If the index buffer is full of values like "0 1 2 3", but basevertex is 4, we
need to upload at least vertex data for elements 4 5 6 7. Whether we also
upload 0 1 2 3 is a question of whether there are VBOs present or not -- see
the code setting start_vertex_bias in brw_draw_upload.c.
Fixes piglit draw-elements*base-vertex user_varrays
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
| |
Otherwise, if we had a set of prims passed in with a num_instances varying
between them, we wouldn't upload enough (or too much!) from user vertex
arrays.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
| |
The brw_draw_upload.c start_vertex_bias code has support for doing the rebase
without rewriting the index buffer by applying a basevertex. It looks like
vbo_rebase_prims() is not equipped to handle basevertex.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
| |
We haven't been only tracking raw GRF-GRF moves since the constant propagation
merge, and also the extension for source modifiers and uniforms.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Given that we handle similarly-regioned GRFs registers for our copy
propagation from our UNIFORM file, there's no reason not to allow it.
The shader-db impact is negligible -- +90 instructions total, 2 shaders helped
and 7 hurt (slightly increased register pressure increased spilling), but this
is to prevent regression in other shaders when fixing copy_propagation to
reduce register pressure in the shaders that are hurt here.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If we put the register coalescing in between the two, then we end up with code
sequences involving dead writes that the dead code elimination doesn't know
how to remove. In place of making dead code elimination smart (which we
should do, too), make it less important for the moment.
shader-db results:
total instructions in shared programs: 722240 -> 721275 (-0.13%)
instructions in affected programs: 50573 -> 49608 (-1.91%)
(no shaders regressed).
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
During code generation, we create tons of temporary variables, many of
which get immediately killed and are never used. Later optimization and
analysis passes, such as compute_live_intervals, loop over all the
virtual GRFs. By compacting them, we can save a lot of overhead.
Reduces compilation time in L4D2's largest fragment shader from 10.2
seconds to 5.2 seconds (50%). Drops compute_live_variables() from
10-12% of another game's startup time to 8%.
Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
| |
Fixes compiler warnings about unused variables.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Variable indexing of non-uniform arrays only exists in GLSL. Likewise,
OPCODE_CAL/OPCODE_RET only existed to try and support GLSL's function
calls. We don't use Mesa IR for GLSL, and these features are explicitly
disallowed by ARB_vertex_program/ARB_fragment_program and never
generated by ffvertex_prog.c.
Since they'll never happen, there's no need to check for them, which
saves us from walking through all the Mesa IR instructions.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
| |
brw_vec4_emit.cpp implements this directly; only the old backend used
the brw_eu_emit.c code.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
| |
The old VS backend doesn't exist, but I believe these still need to be
delivered to the clipper thread.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
| |
It was only used for the old backend.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
| |
Only the old backend used it.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
| |
It's no longer used for anything.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Rather than having two separate backends, just create a small layer that
translates the subset of Mesa IR used for ARB_vertex_program and fixed
function programs to the Vec4 IR. This allows us to use the same
optimization passes, code generator, register allocator as for GLSL.
v2: Incorporate Eric's review comments.
- Fix use of uninitialized src_swiz[] values in the SWIZZLE_ZERO/ONE
case: just initialize it to 0 (.x) since the value doesn't matter
(those channels get writemasked out anyway).
- Properly reswizzle source register's swizzles, rather than overwriting
the swizzle.
- Port the old brw_vs_emit code for computing .x of the EXP2 opcode.
- Update comments, removing mention of NV_vertex_program, etc.
- Delete remaining #warning lines and debug comments.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
| |
v2: Properly use "conditionalmod" pre-Gen6, rather than the incorrectly
copy-and-pasted "BRW_CONDITIONAL_G".
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
| |
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
| |
This will become necessary once we start supporting ARB programs and
fixed function in this backend.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
| |
Don't forget to set depth_mt even if !hiz_mt.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
| |
This was missing and got labeled "Something else".
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When copy_array_to_vbo_array encountered an array with src_stride == 0
and dst_stride != 0, we would replicate out the single element to the
whole size (max - min + 1). This is unnecessary: we can simply upload
one copy and set the buffer's stride to 0.
Decreases vertex upload overhead in an upcoming Steam for Linux title.
Prior to this patch, copy_array_to_vbo_array appeared very high in the
profile (Eric quoted 20%). After the patch, it disappeared completely.
Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This essentially reverts the following:
commit c625aa19cb53ed27f91bfd16fea6ea727e9a5bbd
Author: Chris Wilson <[email protected]>
Date: Fri Feb 18 10:37:43 2011 +0000
intel: extend current vertex buffers
While working on optimizing an upcoming Steam title, I broke this code.
Eric expressed his doubts about this optimization, and noted that the
original commit offered no performance data.
I ran before and after benchmarks on Xonotic and Citybench, and found
that this code made no difference. So, remove it to reduce complexity
and make future work simpler.
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
| |
For Intel, expose it only if gen >= 4.
For Gallium, expose it only if PIPE_CAP_SM3 is advertised.
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
| |
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>
|
|
|
|
|
|
|
|
| |
Fixes build error on Cygwin and Solaris. _R, _G, and _B are used in
ctype.h on those platforms.
Signed-off-by: Vinson Lee <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
|
|
|
|
|
|
|
|
|
| |
Patch adds additional singlesample config with 565 color buffer,
24 bit depth and 8 bit stencil buffer. This makes Quadrant benchmark
work on Android. Tested with Sandybridge and Ivybridge machines.
Signed-off-by: Tapani Pälli <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
| |
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
| |
This is instead of the pair of GLenums for format and type that were
previously used. This is necessary for the Intel drivers to expose sRGB
framebuffer formats.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
|
|
|
|
|
|
|
|
| |
With fewer formats to support, it's kind of useless.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
| |
None of the remaining DRI drivers in Mesa use this.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
| |
There is no gl_format in Mesa that corresponds to this arrangement, so I
have a very hard time believing that this works.
Signed-off-by: Ian Romanick <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Reviewed-by: Brian Paul <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
| |
This is a leftover from when we had to split those two functions due to
the separate BO validation step.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
| |
"Active" is an already-used term for the query being between
glBeginQuery() and glEndQuery(), while this is tracking whether the
start of the packet pair for emitting state has been inserted into the
current batchbuffer.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
| |
Helpful for debugging.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Consider the following code, which reinterprets a register as a
different type:
mov(8) g6<1>F g1.4<0,4,1>.xF
and(8) g5<1>.xUD g6<4,4,1>.xUD 0x7fffffffUD
Copy propagation would notice that we can replace the use of g6 with
g1.4 and eliminate the MOV. Unfortunately, it failed to preserve the UD
type, incorrectly generating:
and(8) g5<1>.xUD g6<4,4,1>.xF 0x7fffffffUD
Found while debugging Ian's uncommitted ARB_vertex_program LOG opcode
test with my new Mesa IR -> Vec4 IR translator.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Consider the following code sequence:
mul(8) g4<1>F g1<0,4,1>.wzwwF g3<4,4,1>.wzwwF
mov.sat(8) m1<1>.xyF g4<4,4,1>F
mul(8) g4<1>F g1<0,4,1>.xxyxF g3<4,4,1>.xxyxF
mov.sat(8) m1<1>.zwF g4<4,4,1>F
The compute-to-MRF pass will discover the first mov.sat and attempt to
replace it by rewriting earlier instructions. Everything works out,
so it replaces scan_inst's destination file, reg, and reg_offset,
resulting in:
mul(8) m1<1>F g1<0,4,1>.wzwwF g3<4,4,1>.wzwwF
mul(8) g4<1>F g1<0,4,1>.xxyxF g3<4,4,1>.xxyxF
mov.sat(8) m1<1>.zwF g4<4,4,1>F
Unfortunately, it loses the .xy writemask on the mov.sat's MRF
destination. While this doesn't pose an immediate problem, it then
proceeds to transform the second mov.sat, resulting in:
mul(8) m1<1>F g1<0,4,1>.wzwwF g3<4,4,1>.wzwwF
mul(8) m1<1>F g1<0,4,1>.xxyxF g3<4,4,1>.xxyxF
Instead of writing both halves of the vector (like the original code),
it overwrites the full vector both times, clobbering the desired .xy
values.
When encountering a MOV, the compute-to-MRF code scans for instructions
which generate channels of the MOV source. It ensures that all
necessary channels are available (possibly written by several
instructions). In this case, *more* channels are available than
necessary, so we want to take the subset that's actually used.
Taking the bitwise and of both writemasks should accomplish that.
This was discovered by analyzing an ARB_vertex_program test
(glean/vertProg1/MUL test (with swizzle and masking)) with my new
Mesa IR -> Vec4 IR translator code. However, it should be possible
with GLSL programs as well.
NOTE: This is a candidate for stable release branches.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
While copying the values into the batch space, we advance the param
pointer. The debug code then tries to iterate over all the uploaded
values, starting at param...which is now the end of the uploaded data,
rather than the start.
This patch saves a pointer to the start of push constant space before
it gets altered and switches the debug code to use that.
Tested by uncommenting the code and examining the output of
glsl-vs-clamp-1.shader_test. Previously all values appeared to be zero.
Signed-off-by: Kenneth Graunke <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
|
|
|
|
|
|
| |
v2: Fix mangled sentence in the comment, and make the loop exit early.
Reviewed-by: Ian Romanick <[email protected]> (v1)
|
|
|
|
|
|
|
| |
Given the usecase we have of trying to measure timestamps across individual
draw calls, flushing will totally mess up what people are trying to measure.
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The theory I had when I wrote the code was that you wanted to minimize latency
on your queries because the app was going to ask soon. Only, it turns out
that everybody batches up their queries and asks for the results later (often
after the next SwapBuffers!), so this was a pessimization.
Until now, I had no workload where it mattered enough to benchmark. Recently
I started playing some Minecraft, which uses tons of queries to decide whether
to render chunks of the terrain. For that app, avoiding the flush in the
query-generation loop improves performance 22.7% +/- 4.7% (n=3) on an apitrace
capture of it (confirmed in game by watching the fps meter found by pressing
F3, 15/16 -> 20/21 fps).
Reviewed-by: Kenneth Graunke <[email protected]>
|
|
|
|
|
| |
I'm amazed that my usual warnings check didn't catch this, and that this
passed piglit.
|
|
|
|
|
|
|
|
| |
otherwise some compilers will throw error
"error: format not a string literal and no format arguments"
Signed-off-by: Tapani Pälli <[email protected]>
Reviewed-by: Chad Versace <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fixes piglit tests "unpack-teximage2d --pbo=* --format=GL_BGRA" on
Sandybridge+.
The fastpath was checking an incomplete set of pixel unpack state. This
patch adds checks for all the fields of gl_pixelstore_attrib that affect
2D texture uploads. Also, it begins permitting the case where
GL_UNPACK_ROW_LENGTH is 0.
Ideally, we would just ask a unicorn to JIT this fastpath for us in
a way that safely handles the unpacking state. Until then, it's safer if
only a small set of situations activate the fastpath.
v2: Use _mesa_is_bufferobj(), per Anholt.
Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Chad Versace <[email protected]>
|