Kenneth Graunke [Wed, 6 Apr 2016 02:33:04 +0000 (19:33 -0700)]
i965: Delete now dead persample_2x FS program key flag.
This was only used by the old gl_SampleID calculations. The new code
doesn't need to handle 2x specially.
v2: Delete it from the Vulkan driver, too.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Wed, 6 Apr 2016 02:29:36 +0000 (19:29 -0700)]
i965: Simplify gl_SampleID setup on Gen8+.
On Gen7+, the thread payload provides the sample ID - we can read it
in two instructions, without any elaborate calculations. We don't even
need a state dependency - this will properly produce zero in the
non-MSAA case. Unfortunately, we need the state flag anyway, so we
may as well continue to use it to produce a single MOV 0 instead of
SHR/AND.
For some reason, the sample ID field is always zero on Gen7/7.5, so
we can't use this yet. However, it works fine on Gen8+. So, land the
code and use it where it's working, and leave a TODO for later.
v2: Fix register types in the comment (caught by Matt Turner!).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Tue, 19 Apr 2016 01:11:01 +0000 (18:11 -0700)]
i965: Flip key->compute_sample_id check.
This just moves the simple case first.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Bas Nieuwenhuizen [Wed, 20 Apr 2016 13:31:22 +0000 (15:31 +0200)]
st/mesa: Use correct size for compute CAPs.
Some CAPs are stored as 64-bit value while Mesa stores
the related constant as 32-bit value.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Kenneth Graunke [Wed, 13 Apr 2016 23:58:10 +0000 (16:58 -0700)]
i965: Properly handle integer types in opt_vector_float().
Previously, opt_vector_float() always interpreted MOV sources as
floating point, and always created a MOV with a F-type destination.
This meant that we could mess up sequences of integer loads, such as:
mov vgrf6.0.x:D, 0D
mov vgrf6.0.y:D, 1D
mov vgrf6.0.z:D, 2D
mov vgrf6.0.w:D, 3D
Here, integer 0/1/2/3 become approximately 0.0f, so we generated:
mov vgrf6.0:F, [0F, 0F, 0F, 0F]
which is clearly wrong. We can properly handle this by converting
integer values to float (rather than bitcasting), and emitting a type
converting MOV:
mov vgrf6.0:D, [0F, 1F, 2F, 3F]
To do this, see first see if the integer values (converted to float)
are representable. If so, we use a D-type MOV. If not, we then try
the floating point values and an F-type MOV. We make zero not impose
type restrictions. This is important because 0D would imply a D-type
MOV, but is often used in sequences such as MOV 0D, MOV 0x3f800000D,
where we want to use an F-type MOV.
Fixes about 54 dEQP-GLES2 failures with the vec4 VS backend. This
recently became visible due to changes in opt_vector_float() which
made it optimize more cases, but it was a pre-existing bug.
Apparently it also manages to turn more integer loads into VFs,
producing the following shader-db statistics on Haswell:
total instructions in shared programs:
7084195 ->
7082191 (-0.03%)
instructions in affected programs: 246027 -> 244023 (-0.81%)
helped: 1937
total cycles in shared programs:
65669642 ->
65651968 (-0.03%)
cycles in affected programs: 531064 -> 513390 (-3.33%)
helped: 1177
v2: Handle the type of zero better.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Wed, 13 Apr 2016 23:39:54 +0000 (16:39 -0700)]
i965: Make opt_vector_float() only handle non-type-conversion MOVs.
We don't handle this properly - we'd have to perform the type conversion
before trying to convert the value to a VF.
While we could do that, it doesn't seem particularly useful - most
vector loads should be consistently typed (all float or all integer).
As a special case, we do allow type-converting MOVs of integer 0, as
it's represented the same regardless of the type. I believe this case
does actually come up.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Wed, 13 Apr 2016 23:04:04 +0000 (16:04 -0700)]
i965: Fold vectorize_mov() back into the one caller.
After the previous patch, this helper is only called in one place.
So, just fold it back in - there are a lot of parameters here and
not much code.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Kenneth Graunke [Wed, 13 Apr 2016 22:56:07 +0000 (15:56 -0700)]
i965: Rework opt_vector_float() control flow.
This reworks opt_vector_float() so that there's only one place that
flushes out any accumulated state and emits a VF.
v2: Don't break the sequence for non-representable numbers - just skip
recording their values. Only break it for non-MOVs or register
changes.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Jason Ekstrand [Tue, 19 Apr 2016 00:03:00 +0000 (17:03 -0700)]
anv: s/anv_batch_emit_blk/anv_batch_emit/
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 23:53:11 +0000 (16:53 -0700)]
anv: Remove the old emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 23:51:12 +0000 (16:51 -0700)]
anv/gen7_pipeline: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 23:33:46 +0000 (16:33 -0700)]
anv/gen7_cmd_buffer: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:32:29 +0000 (15:32 -0700)]
anv/device: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:24:59 +0000 (15:24 -0700)]
anv/state: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:53:14 +0000 (15:53 -0700)]
anv/gen8_pipeline: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:29:42 +0000 (15:29 -0700)]
anv/genX_pipeline: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 23:08:49 +0000 (16:08 -0700)]
anv/gen8_cmd_buffer: Use the new emit macro
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:20:06 +0000 (15:20 -0700)]
anv/cmd_buffer: Use the new emit macro for quaries
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:14:47 +0000 (15:14 -0700)]
anv/cmd_buffer: Use the new emit macro for DRAWING_RECTANGLE
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 22:11:43 +0000 (15:11 -0700)]
anv/cmd_buffer: Use the new emit macro for compute shader dispatch
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 21:55:10 +0000 (14:55 -0700)]
anv/cmd_buffer: Use the new emit macro for 3DSTATE_CONSTANT
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 21:48:33 +0000 (14:48 -0700)]
anv/cmd_buffer: Use the new emit macro for DEPTH/STENCIL_BUFFER
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 21:41:06 +0000 (14:41 -0700)]
anv/cmd_buffer: Use the new emit macro for PIPE_CONTROL and STATE_BASE_ADDRESS
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 21:27:29 +0000 (14:27 -0700)]
anv/cmd_buffer: Use the new emit macro for 3DPRIMITIVE commands
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Jason Ekstrand [Mon, 18 Apr 2016 21:25:03 +0000 (14:25 -0700)]
anv: Add a new block-based batch emit macro
This new macro uses a for loop to create an actual code block in which to
place the macro setup code. One advantage of this is that you syntatically
use braces instead of parentheses. Another is that the code in the block
doesn't even get executed if anv_batch_emit_dwords fails.
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Samuel Pitoiset [Wed, 20 Apr 2016 17:06:24 +0000 (19:06 +0200)]
gk110/ir: make use of IMUL32I for all immediates
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Samuel Pitoiset [Wed, 20 Apr 2016 17:47:37 +0000 (19:47 +0200)]
gk110/ir: do not overwrite def value with zero for EXCH ops
This is only valid for other atomic operations (including CAS). This
fixes an invalid opcode error from dmesg. While we are it, make sure
to initialize global addr to 0 for other atomic operations.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Marcin Ślusarz [Sat, 16 Apr 2016 20:48:09 +0000 (22:48 +0200)]
anv: fix build without Wayland platform
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Laurent Carlier [Sat, 16 Apr 2016 19:50:39 +0000 (21:50 +0200)]
anv: fix building on i686 with -mcpu=generic
mcpu=generic doesn't enable sse2, and anvil definitly needs it
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Jason Ekstrand [Wed, 20 Apr 2016 17:32:59 +0000 (10:32 -0700)]
spirv: Trivially handle the NonWriteable decoration
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Connor Abbott [Wed, 13 Apr 2016 20:25:34 +0000 (16:25 -0400)]
nir: rename nir_foreach_block*() to nir_foreach_block*_call()
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Samuel Pitoiset [Sun, 10 Apr 2016 20:08:34 +0000 (22:08 +0200)]
nvc0: avoid tex read fault from compute shaders on GK110
After some investigation, it seems like that disabling the UNK02C4
command avoid a read fault with texelFetch() from a compute shader.
I have no clue on what this method actually does, but this avoid the
GPU to hang with basic-texelFetch.shader_test without introducing any
compute-related regressions.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Jason Ekstrand [Tue, 19 Apr 2016 01:57:07 +0000 (18:57 -0700)]
i965/vec4: Always split uniforms in array_access_to_pull_constants
Normally, we split uniforms at the end but in Vulkan, we bail because we
don't want pull constants. However, we still need them split because
pack_uniforms relies on it.
I really don't like this patch not because it doesn't work (it does) but
because now that we're using MOV_INDIRECT, uniform numbers and sizes don't
really matter anymore. In the FS backend, uniform splitting and packing is
handled all at once (actual re-assignment of locations happens later) and
we really should do it that way in vec4 eventually as well.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94998
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95001
Jason Ekstrand [Tue, 19 Apr 2016 01:52:36 +0000 (18:52 -0700)]
i965/vec4: Use the correct offset for the swizzle shift in push constants
This was actually caught by Ken in review the first time around but somehow
didn't get fixed before the patches were pushed. :-(
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94998
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95001
Jason Ekstrand [Tue, 19 Apr 2016 01:51:50 +0000 (18:51 -0700)]
i965/vec4: Use nir_intrinsic_base in the load_uniform implementation
We shouldn't be reading the const_index directly
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94998
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95001
Jason Ekstrand [Tue, 19 Apr 2016 00:55:21 +0000 (17:55 -0700)]
anv/apply_dynamic_offsets: Provide a range on the load_uniform
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94998
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95001
Jason Ekstrand [Tue, 19 Apr 2016 00:17:31 +0000 (17:17 -0700)]
anv/lower_push_constants: Stop treating scalar specially
All of the code that did something special based on vec4 vs. scalar is
bogus. In the backend, everything is now in units of bytes and the vec4
backend can handle full std140 packing so we don't need to do anything
special anymore.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94998
Tim Rowley [Mon, 18 Apr 2016 23:10:39 +0000 (18:10 -0500)]
swr: fix resource backed constant buffers
Code was using an incorrect address for the base pointer.
v2: use swr_resource_data() utility function.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94979
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
Tested-by: Markus Wick <markus@selfnet.de>
Hans de Goede [Mon, 14 Mar 2016 12:57:07 +0000 (13:57 +0100)]
nouveau: codegen: Add support for OpenCL global memory buffers
Add support for OpenCL global memory buffers, note this has only
been tested with regular load and stores and likely needs more work
for e.g. atomic ops.
Tested with piglet on a gf119 and a gk107:
./piglit run -o shader -t '.*arb_shader_storage_buffer_object.*' results/shader
[9/9] pass: 9 /
./piglit run -o shader -t '.*arb_compute_shader.*' results/shader
[20/20] skip: 4, pass: 16 |
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Hans de Goede [Thu, 17 Mar 2016 14:50:00 +0000 (15:50 +0100)]
nouveau: codegen: Use FILE_MEMORY_BUFFER for buffers
Some of the lowering steps we currently do for FILE_MEMORY_GLOBAL only
apply to buffers, making it impossible to use FILE_MEMORY_GLOBAL for
OpenCL global buffers.
This commits changes the buffer code to use FILE_MEMORY_BUFFER at the
ir_from_tgsi and lowering steps, freeing use of FILE_MEMORY_GLOBAL
for use with OpenCL global buffers.
Note that after lowering buffer accesses use the FILE_MEMORY_GLOBAL
register file.
Tested with piglet on a gf119 and a gk107:
./piglit run -o shader -t '.*arb_shader_storage_buffer_object.*' results/shader
[9/9] pass: 9 /
./piglit run -o shader -t '.*arb_compute_shader.*' results/shader
[20/20] skip: 4, pass: 16 |
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Jose Fonseca [Wed, 20 Apr 2016 11:41:24 +0000 (12:41 +0100)]
scons: Build dri_common_interop.c.
Marek Olšák [Mon, 29 Feb 2016 22:25:59 +0000 (23:25 +0100)]
st/dri: implement the GL interop DRI extension (v2.2)
v2: - set interop_version
- simplify the offset_after macro
v2.1: - use version numbers, remove offset_after
- set "out_driver_data_written"
v2.2: - set buf_offset & buf_size for GL_ARRAY_BUFFER too
- add whandle.offset to buf_offset
- disable the minmax cache for GL_TEXTURE_BUFFER
Marek Olšák [Thu, 3 Mar 2016 17:43:53 +0000 (18:43 +0100)]
glx: implement GLX part of interop interface (v2)
v2: - use const
Marek Olšák [Thu, 3 Mar 2016 14:59:48 +0000 (15:59 +0100)]
egl: implement EGL part of interop interface (v2)
v2: - use const
Marek Olšák [Mon, 29 Feb 2016 22:19:18 +0000 (23:19 +0100)]
dri_interface: add interface for GL interop with other APIs (v2)
v2: - use const
Marek Olšák [Fri, 26 Feb 2016 20:21:31 +0000 (21:21 +0100)]
include/GL: add mesa_glinterop.h for OpenGL-OpenCL interop (v4.2)
v2: - use "enum" to define stuff
v3: - more comments, define MESA_GLINTEROP_UNSUPPORTED
v4: - add mesa_glinterop_device_info::interop_version
- more comments
- remove #define MESA_GLINTEROP_VERSION
- use const for "in"
v4.1: - use version numbers for structures
- add "out_driver_data_written"
v4.2: - buf_offset & buf_size affect GL_ARRAY_BUFFER too, this is required
for sharing suballocations within a larger buffer
Nicolas Dufresne [Sun, 17 Apr 2016 00:49:44 +0000 (20:49 -0400)]
st/dri: Fix RGB565 EGLImage creation
When creating egl images we do a bytes to pixel conversion by deviding
by 4 regardless of the pixel format. This does not work for RGB565. In
this patch, we avoid useless conversion and use proper API when the
conversion cannot be avoided.
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Nicolas Dufresne [Sun, 17 Apr 2016 00:49:43 +0000 (20:49 -0400)]
st/dri: Factor out DRI2 to PIPE_FORMAT conversion
This code is already duplicated twice and will be useful again. This
will also help when adding formats.
Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Rob Clark [Tue, 19 Apr 2016 13:02:23 +0000 (09:02 -0400)]
freedreno/a4xx: lower srgb in shader for astc textures
This *seems* like a hw bug, and maybe only applies to certain a4xx
variants/revisions. But setting the SRGB bit in sampler view state
(texconst0) causes invalid alpha for ASTC textures. Work around this
by doing the srgb->linear conversion in the shader instead.
This fixes 392 dEQP tests: dEQP-GLES3.functional.texture.*astc*srgb*
(The remaining fails seem to be a bug w/ ASTC + linear filtering, also
possibly a420.0 specific.)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Rob Clark [Tue, 19 Apr 2016 12:28:22 +0000 (08:28 -0400)]
nir/lower-tex: add srgb->linear lowering
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Rob Clark [Tue, 19 Apr 2016 19:44:25 +0000 (15:44 -0400)]
nir/builder: const'ify swiz param
No need for it not to be const, and lets caller declare it const if
desired.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Rob Clark [Tue, 19 Apr 2016 11:46:50 +0000 (07:46 -0400)]
nir/lower-tex: make options a local var
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Rob Clark [Tue, 19 Apr 2016 19:52:18 +0000 (15:52 -0400)]
freedreno: cleanup fd_set_sampler_views
The separate FS/VS entrypoints are no longer used since
a3ed98f. So
just inline them.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Russell King [Wed, 13 Apr 2016 22:42:42 +0000 (18:42 -0400)]
tgsi/lowering: improved lowering for LRP
Provide an improved lowering for LRP, which can be implemented in two
MAD instructions with a bit of rearranging of the equation, rather
than the literal implementation of two multiplies, an add and a
subtract.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Russell King [Wed, 13 Apr 2016 22:42:41 +0000 (18:42 -0400)]
tgsi/lowering: improved lowering for XPD
Improve XPD lowering to consume less instructions by using the
MAD instruction to perform the multiply and subtraction together.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Russell King [Wed, 13 Apr 2016 22:42:40 +0000 (18:42 -0400)]
tgsi/lowering: add support for lowering TRUNC
Add support for lowering TRUNC using the following sequence:
FRC tmpA, |src|
SUB tmpA, |src|, tmpA
CMP dst, -tmpA, tmpA
Note that this is incompatible with FRC lowering.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Russell King [Wed, 13 Apr 2016 22:42:39 +0000 (18:42 -0400)]
tgsi/lowering: add support for lowering FLR and CEIL
Add support for lowering FLR and CEIL to FRC/SUB and FRC/ADD
instructions for GPUs that support FRC but not FLR or CEIL. Since
these uses FRC, it is invalid to ask for FLR or CEIL to be lowered
along with FRC, so add an assert to catch this invalid configuration.
We also need to deal with FLR instructions emitted by the lowering
code. Fix these up with the FRC+SUB equivalent when FLR lowering is
enabled.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Bas Nieuwenhuizen [Sat, 19 Mar 2016 14:16:50 +0000 (15:16 +0100)]
radeonsi: enable TGSI support cap for compute shaders
v2: Use chip_class instead of family.
v3: Check kernel version for SI.
v4: Preemptively allow amdgpu winsys for SI.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Tue, 19 Apr 2016 12:08:13 +0000 (14:08 +0200)]
radeonsi: Consider input SGPR count for compute shader SGPR count.
si_shader_create corrects the SGPR count with si_fix_num_sgprs. We then
recompute the rsrc1 register to use the new SGPR count.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Tue, 19 Apr 2016 11:52:32 +0000 (13:52 +0200)]
radeonsi: Add CE synchronization for compute dispatches.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 11:39:54 +0000 (13:39 +0200)]
mesa/st: enable compute shaders if images are also supported
v2: Also depend on atomic counters.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 09:37:06 +0000 (11:37 +0200)]
radeonsi: clean up compute flush
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Sun, 27 Mar 2016 09:14:34 +0000 (11:14 +0200)]
radeonsi: do not do two full flushes on every compute dispatch
v2: Add more CS_PARTIAL_FLUSH events.
Essentially every place with waits on finishing for pixel shaders
also has a write after read hazard with compute shaders.
Invalidating L2 waits implicitly on pixel and compute shaders,
so, we don't need a CS_PARTIAL_FLUSH for switching FBO.
v3: Add CS_PARTIAL_FLUSH events even if we already have INV_GLOBAL_L2.
According to Marek the INV_GLOBAL_L2 events don't wait for compute
shaders to finish, so wait for them explicitly.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Sat, 19 Mar 2016 12:56:29 +0000 (13:56 +0100)]
radeonsi: split setting graphics and compute descriptors
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Sat, 19 Mar 2016 17:41:20 +0000 (18:41 +0100)]
radeonsi: split texture decompression for compute shaders
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Tue, 5 Apr 2016 15:38:38 +0000 (17:38 +0200)]
radeonsi: update predicate condition for compute dispatches
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Sat, 19 Mar 2016 14:15:20 +0000 (15:15 +0100)]
radeonsi: implement TGSI compute dispatch
v2: - Use radeon_set_sh_reg_seq.
- Set predicate bit for conditional rendering.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 11:19:42 +0000 (13:19 +0200)]
radeonsi: only emit compute shader state when switching shaders
v2: - Do check if anything changed earlier
- Use emitted_program instead of emitted_bo to prevent
shaders with shader->bo = NULL confusing the check
- Use radeon_set_sh_reg*
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 11:04:18 +0000 (13:04 +0200)]
radeonsi: rework compute scratch buffer
Instead of having a scratch buffer per program, have one per
context.
Also removed the per kernel wave count calculations, but
that only helped if the total number of waves in the dispatch
was smaller than sctx->scratch_waves.
v2: Fix style issue.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 10:35:36 +0000 (12:35 +0200)]
radeonsi: do per cs setup for compute shaders once per cs
Also removes PKT3_CONTEXT_CONTROL as that is already being done
by si_begin_new_cs, when emitting init_config.
v2: - Use radeon_set_sh_reg_seq.
- Also set COMPUTE_STATIC_THREAD_MGMT_SE2 / SE3 for CIK+
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 10:48:05 +0000 (12:48 +0200)]
radeonsi: don't pass scratch buffer to user SGPRs
As far as I can see we use relocations for clover too.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Sat, 2 Apr 2016 10:04:15 +0000 (12:04 +0200)]
radeonsi: split input upload off from si_launch_grid
Also uses a dynamically allocated buffer using u_upload_alloc.
The old buffer per program approach required serializing all
dispatches of the same program.
v2: - Clarified commit message.
- Use radeon_set_sh_reg_seq.
- Also upload input buffer for clover kernels, even when
input_size is 0, as it contains grid parameters.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Sat, 19 Mar 2016 13:07:33 +0000 (14:07 +0100)]
radeonsi: implement TGSI compute shader creation
v2: Moved scratch_enabled initialization after compile.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Sat, 19 Mar 2016 12:54:55 +0000 (13:54 +0100)]
radeonsi: update shader count for compute shaders
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Mon, 28 Mar 2016 01:01:56 +0000 (03:01 +0200)]
radeonsi: set maximum work group size based on block size
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Tue, 29 Mar 2016 11:20:26 +0000 (13:20 +0200)]
radeonsi: implement shared atomics
v2: - Use single region
- Use get_memory_ptr
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Tue, 29 Mar 2016 11:17:40 +0000 (13:17 +0200)]
radeonsi: implement shared memory load/store
v2: - Use single region
- Combine address calculation
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Tue, 29 Mar 2016 15:51:49 +0000 (17:51 +0200)]
radeonsi: add shared memory
Declares the shared memory as a global variable so that
LLVM is aware of it and it does not conflict with passes
like AMDGPUPromoteAlloca.
v2: - Use ctx->i8.
- Dropped null-check for declare_memory_region.
- Changed memory region array to single region.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Bas Nieuwenhuizen [Thu, 17 Mar 2016 13:12:21 +0000 (14:12 +0100)]
radeonsi: lower compute shader arguments
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Bas Nieuwenhuizen [Thu, 10 Mar 2016 20:39:20 +0000 (21:39 +0100)]
radeonsi: Use CE for all descriptors.
v2: Load previous list for new CS instead of re-emitting
all descriptors.
v3: Do radeon_add_to_buffer_list in si_ce_upload.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Wed, 13 Apr 2016 21:30:55 +0000 (23:30 +0200)]
gallium/util: Add u_bit_scan_consecutive_range64.
For use by radeonsi.
v2: Make sure that it works for all 64 bits set.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Thu, 14 Apr 2016 23:00:41 +0000 (01:00 +0200)]
radeonsi: Replace list_dirty with a mask.
We can then upload only the dirty ones with the constant engine.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Thu, 10 Mar 2016 20:23:49 +0000 (21:23 +0100)]
radeonsi: Add CE uploader.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Thu, 10 Mar 2016 20:19:37 +0000 (21:19 +0100)]
radeonsi: Allocate chunks of CE ram.
v2: Use 32 byte alignment.
v3: Don't allocate CE space for vertex buffer descriptors.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Thu, 10 Mar 2016 20:01:39 +0000 (21:01 +0100)]
radeonsi: Add CE synchronization.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Thu, 10 Mar 2016 19:59:16 +0000 (20:59 +0100)]
radeonsi: Add CE packet definitions.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Wed, 13 Apr 2016 20:31:17 +0000 (22:31 +0200)]
radeonsi: Create CE IB.
Based on work by Marek Olšák.
v2: Add preamble IB.
Leaves the load packet in the space calculation as the
radeon winsys might not be able to support a premable.
The added space calculation may look expensive, but
is converted to a constant with (at least) -O2 and -O3.
v3: - Fix code style.
- Remove needed space for vertex buffer descriptors.
- Fail when the preamble cannot be created.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bas Nieuwenhuizen [Thu, 14 Apr 2016 00:11:07 +0000 (02:11 +0200)]
winsys/amdgpu: Enlarge const IB size.
Necessary to prevent performance regressions due to extra flushing.
Probably should enlarge it even further when also updating
uniforms through the CE, but this seems large enough for now.
v2: Add preamble IB.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Marek Olšák [Sat, 8 Aug 2015 12:02:02 +0000 (14:02 +0200)]
winsys/amdgpu: Add support for const IB.
v2: Use the correct IB to update request (Bas Nieuwenhuizen)
v3: Add preamble IB. (Bas Nieuwenhuizen)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Marek Olšák [Sat, 8 Aug 2015 11:27:38 +0000 (13:27 +0200)]
winsys/amdgpu: split IB data into a new structure in preparation for CE
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Marek Olšák [Sat, 8 Aug 2015 12:12:10 +0000 (14:12 +0200)]
gallium/radeon: move ring_type into winsyses
Not used by drivers.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Jose Fonseca [Tue, 19 Apr 2016 11:08:01 +0000 (12:08 +0100)]
llvmpipe: Call LLVMShutdown before exiting.
So that LLVM frees its globals.
Trivial.
Jose Fonseca [Tue, 19 Apr 2016 11:07:16 +0000 (12:07 +0100)]
llvmpipe: Avoid LLVMGetGlobalContext in tests.
Trivial.
Jose Fonseca [Fri, 15 Apr 2016 10:02:06 +0000 (11:02 +0100)]
llvmpipe: Skip false exp2 failure in lp_test_arit due to buggy MSVCRT.
64bits MSVCRT's exp2f(-inf) returns -inf instead of 0. Tested with
MSVC 2013's CRT. (I haven't tried 2015 yet.)
Also this does not happen with MinGW.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Jose Fonseca [Fri, 15 Apr 2016 13:41:19 +0000 (14:41 +0100)]
llvmpipe: Test more vector lengths.
All power of two of up native vector length.
There is actually a bug in lp_build_round for v2, whereby it doesn't
round to nearest. Fixing is left to the future, but the test is now
able to expect it to fail.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Jose Fonseca [Fri, 15 Apr 2016 11:05:09 +0000 (12:05 +0100)]
gallivm: Avoid llvm::sys::getProcessTriple().
Just use LLVM_HOST_TRIPLE, which is available at least from LLVM 3.3
onwards, and is pretty much what llvm::sys::getProcessTriple() does anyway,
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Jose Fonseca [Fri, 15 Apr 2016 10:27:02 +0000 (11:27 +0100)]
gallivm: Remove lp_get_module_id.
Just keep a copy of the module_name in gallivm.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Jose Fonseca [Thu, 14 Apr 2016 15:47:14 +0000 (16:47 +0100)]
gallivm: Fix MCJIT with LLVM 3.3.
One needs to call setJITMemoryManager for LLVM 3.3, instead of
setMCJITMemoryManager.
This regressed in commits
065256df/
75ad4fe7 when trying to make the
code to build with LLVM 3.6.
Tested MCJIT with LLVM 3.3 to 3.6.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Jose Fonseca [Thu, 14 Apr 2016 12:42:37 +0000 (13:42 +0100)]
gallivm: Make MCJIT a runtime option.
On the LLVM versions that support it, so we can easily switch between
MCJIT/old-jit for testing.
The new option is GALLIVM_MCJIT.
Unfortunately setting GALLIVM_MCJIT=1 for LLVM 3.3 or 3.4 causes
segfault, both on Linux and Windows. I'm almost certain this used to
work, so there probably is a regression somewhere.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Jose Fonseca [Thu, 14 Apr 2016 12:41:33 +0000 (13:41 +0100)]
scons: Show the unit test full path.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>