Chad Versace [Fri, 22 Jan 2016 17:46:48 +0000 (09:46 -0800)]
isl/tests: Give tests less cryptic names
Chad Versace [Fri, 22 Jan 2016 16:32:00 +0000 (08:32 -0800)]
isl: Fix isl_surf_get_image_offset_sa for gen4_3d layout
Bug found by unit test
test_bdw_3d_r8g8b8a8_unorm_256x256x256_levels09_tiley0.
Chad Versace [Fri, 22 Jan 2016 16:31:29 +0000 (08:31 -0800)]
isl/tests: Add test for bdw 3d surface
test_bdw_3d_r8g8b8a8_unorm_256x256x256_levels09_tiley0
Currently fails.
Chad Versace [Fri, 22 Jan 2016 15:17:34 +0000 (07:17 -0800)]
isl/tests: Remove copy-paste assertion
Chad Versace [Fri, 22 Jan 2016 15:17:06 +0000 (07:17 -0800)]
isl/tests: Fix build
isl_device_init() acquired a new param for bit6 swizzling.
Francisco Jerez [Thu, 21 Jan 2016 22:14:07 +0000 (14:14 -0800)]
anv/batch_chain: Fix patching up of block pool relocations on Gen8+.
Relocations are 64 bits on Gen8+. Most CTS tests that send
non-trivial work to the GPU would fail when run from a single deqp-vk
invocation because they were effectively relying on reloc presumed
offsets to be wrong so the kernel would come and apply relocations
correctly.
Jason Ekstrand [Fri, 22 Jan 2016 00:03:06 +0000 (16:03 -0800)]
nir/spirv: Ignore cull distance
Jason Ekstrand [Fri, 22 Jan 2016 00:00:47 +0000 (16:00 -0800)]
nir/lower_system_values: Use the correct invication id for CS
Jason Ekstrand [Thu, 21 Jan 2016 20:41:45 +0000 (12:41 -0800)]
nir/spirv: Properly assign locations to split structures
Jason Ekstrand [Thu, 21 Jan 2016 18:58:53 +0000 (10:58 -0800)]
nir/spirv: Improve handling of variable loads and copies
Before we were asuming that a deref would either be something in a block or
something that we could pass off to NIR directly. However, it is possible
that someone would choose to load/store/copy a split structure all in one
go. We need to be able to handle that.
Jason Ekstrand [Thu, 21 Jan 2016 18:58:03 +0000 (10:58 -0800)]
nir/spirv: Make vectors a proper array time with an array_element
This makes dealing with single-component derefs easier
Jason Ekstrand [Thu, 21 Jan 2016 18:20:50 +0000 (10:20 -0800)]
nir/spirv: Rework access chains a bit to allow for literals
This makes them much easier to construct because you can also just specify
a literal number and it doesn't have to be a valid SPIR-V id.
Jason Ekstrand [Thu, 21 Jan 2016 16:53:04 +0000 (08:53 -0800)]
vtn/variables: Compact local loads/stores into one function
This is similar to what we did for block loads/stores.
Jason Ekstrand [Thu, 21 Jan 2016 10:09:38 +0000 (02:09 -0800)]
nir/spirv: Add an actual variable struct to spirv_to_nir
This allows us, among other things, to do structure splitting on-the-fly to
more correctly handle input/output structs.
Jason Ekstrand [Thu, 21 Jan 2016 01:40:49 +0000 (17:40 -0800)]
nir/spirv: Split variable handling out into its own file
It's 1300 lines all by itself and it will only grow.
Jason Ekstrand [Thu, 21 Jan 2016 01:11:55 +0000 (17:11 -0800)]
nir/spirv: Rework access chains
Previously, we were creating nir_deref's immediately. Now, instead, we
have an intermediate vtn_access_chain structure. While a little more
awkward initially, this will allow us to more easily do structure splitting
on-the-fly.
Kenneth Graunke [Thu, 21 Jan 2016 22:56:34 +0000 (14:56 -0800)]
nir/spirv: Implement ModfStruct opcode.
Kenneth Graunke [Thu, 21 Jan 2016 22:53:44 +0000 (14:53 -0800)]
nir/spirv: Delete stray fmod remnants.
Jason left these stray code fragments in
22804de110b97dce1415318fd02c1003e16ef14a.
Kristian Høgsberg Kristensen [Thu, 21 Jan 2016 22:14:01 +0000 (14:14 -0800)]
vk: Do render cache flush for GEN8+
This is needed for SKL as well.
Kristian Høgsberg Kristensen [Thu, 21 Jan 2016 21:52:41 +0000 (13:52 -0800)]
vk: Emit surface state base address before renderpass
If we're continuing a render pass, make sure we don't emit the depth and
stencil buffer addresses before we set the state base addresses.
Fixes crucible func.cmd-buffer.small-secondaries
Kristian Høgsberg Kristensen [Thu, 21 Jan 2016 18:43:32 +0000 (10:43 -0800)]
vk: Fix indirect push constants
This currently sets the base and size of all push constants to the
entire push constant block. The idea is that we'll use the base and size
to eventually optimize the amount we actually push, but for now we don't
do that.
Kristian Høgsberg Kristensen [Thu, 21 Jan 2016 19:07:09 +0000 (11:07 -0800)]
Merge remote-tracking branch 'jekstrand/wip/i965-uniforms' into vulkan
Jordan Justen [Sun, 10 Jan 2016 19:12:17 +0000 (11:12 -0800)]
nir/spirv: Handle compute shared atomics
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Sat, 9 Jan 2016 03:45:45 +0000 (19:45 -0800)]
nir/spirv: Support workgroup (shared) variable translation
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Sat, 9 Jan 2016 09:28:40 +0000 (01:28 -0800)]
anv/gen8: Set SLM size in interface descriptor
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Thu, 21 Jan 2016 01:26:41 +0000 (17:26 -0800)]
anv/gen8+9: Invalidate color calc state when switching to the GPGPU pipeline
Port
044acb9256046bebec890cac7e42043754459fc2 to anv.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Sat, 9 Jan 2016 09:18:24 +0000 (01:18 -0800)]
anv/gen8: Enable SLM in L3 cache control register
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Mon, 18 Jan 2016 18:00:49 +0000 (10:00 -0800)]
anv/pipeline: Set size of shared variables in prog_data
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Mon, 18 Jan 2016 17:45:46 +0000 (09:45 -0800)]
i965/nir: Lower nir compute shader shared variables
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Mon, 18 Jan 2016 17:59:19 +0000 (09:59 -0800)]
nir: Lower shared var atomics during nir_lower_io
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Mon, 18 Jan 2016 17:44:31 +0000 (09:44 -0800)]
nir: Add support for lowering load/stores of shared variables
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Mon, 18 Jan 2016 17:53:44 +0000 (09:53 -0800)]
nir: Add atomic operations on variables
This allows us to first generate atomic operations for shared
variables using these opcodes, and then later we can lower those to
the shared atomics intrinsics with nir_lower_io.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Sat, 9 Jan 2016 01:16:29 +0000 (17:16 -0800)]
nir: Add compute shader shared variable storage class
Previously we were receiving shared variable accesses via a lowered
intrinsic function from glsl. This change allows us to send in
variables instead. For example, when converting from SPIR-V.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Sun, 17 Jan 2016 07:11:16 +0000 (23:11 -0800)]
nir/print: Add space after shader_storage var mode
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Jordan Justen [Sat, 16 Jan 2016 02:38:19 +0000 (18:38 -0800)]
i965/fs/nir: Move shared variable load/store to nir_emit_cs_intrinsic
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Chad Versace [Wed, 20 Jan 2016 23:41:25 +0000 (15:41 -0800)]
anv/image: Stop including gen8_pack.h in common file
Chad Versace [Fri, 15 Jan 2016 18:00:13 +0000 (10:00 -0800)]
isl: Add a README
Most of the file-level comment in isl.h is moved to the README.
Kristian Høgsberg Kristensen [Wed, 20 Jan 2016 22:36:52 +0000 (14:36 -0800)]
vk: Make maxSamplerAllocationCount more reasonable
We can't allocate 4 billion samplers. Let's go with 64k.
Kristian Høgsberg Kristensen [Wed, 20 Jan 2016 21:46:54 +0000 (13:46 -0800)]
vk/tests: Add stub for anv_gem_get_bit6_swizzle()
Kristian Høgsberg Kristensen [Wed, 20 Jan 2016 21:45:25 +0000 (13:45 -0800)]
vk/tests: Add isl include path
Kenneth Graunke [Wed, 20 Jan 2016 20:09:29 +0000 (12:09 -0800)]
nir/spirv/glsl450: Use fabs not iabs in ldexp.
This was just wrong.
Kristian Høgsberg Kristensen [Wed, 20 Jan 2016 20:02:48 +0000 (12:02 -0800)]
isl: Add ish.h to libsil_la_SOURCES
Jason Ekstrand [Wed, 20 Jan 2016 19:36:41 +0000 (11:36 -0800)]
nir/spirv/glsl450: Implement FrexpStruct
Jason Ekstrand [Wed, 20 Jan 2016 19:36:26 +0000 (11:36 -0800)]
spirv/nir/glsl450: Use vtn_create_ssa_value to create SSA values
Jason Ekstrand [Wed, 20 Jan 2016 19:16:44 +0000 (11:16 -0800)]
anv/device: Default to scalar GS on BDW+
Jason Ekstrand [Wed, 20 Jan 2016 19:11:30 +0000 (11:11 -0800)]
nir/spirv: Pull texture dimensionality out of the image when available
Jason Ekstrand [Wed, 20 Jan 2016 15:56:48 +0000 (07:56 -0800)]
anv/meta: fix UpdateBuffer in the case where we do multiple updates
Jason Ekstrand [Wed, 20 Jan 2016 15:33:41 +0000 (07:33 -0800)]
anv/meta: Fix a finishme
Jason Ekstrand [Wed, 20 Jan 2016 02:58:31 +0000 (18:58 -0800)]
nir/spirv: Move OpPhi handling to vtn_cfg.c
Phi handling is somewhat intrinsically tied to the CFG. Moving it here
makes it a bit easier to handle that. In particular, we can now do SSA
repair after we've done the phi node second-pass. This fixes 6 CTS tests.
Jason Ekstrand [Wed, 20 Jan 2016 02:44:44 +0000 (18:44 -0800)]
nir/spirv: Handle OpLine and OpNoLine in foreach_instruction
This way we don't have to explicitly handle them everywhere.
Kenneth Graunke [Wed, 20 Jan 2016 01:40:58 +0000 (17:40 -0800)]
nir: Lower ldexp to arithmetic.
This is a port of Matt's GLSL IR lowering pass to NIR. It's required
because we translate SPIR-V directly to NIR, bypassing GLSL IR.
I haven't introduced a lower_ldexp flag, as I believe all current NIR
consumers would set the flag. i965 wants this, vc4 doesn't implement
this feature, and st_glsl_to_tgsi currently lowers ldexp
unconditionally anyway.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Kenneth Graunke [Wed, 20 Jan 2016 01:34:24 +0000 (17:34 -0800)]
nir: Let nir_opt_algebraic rules contain unsigned constants > INT_MAX.
struct.pack('i', val) interprets `val` as a signed integer, and dies
if `val` > INT_MAX. For larger constants, we need to use 'I' which
interprets it as an unsigned value.
This patch makes us use 'I' for all values >= 0, and 'i' for negative
values. This should work in all cases.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Jason Ekstrand [Wed, 20 Jan 2016 00:39:48 +0000 (16:39 -0800)]
anv/meta: Implement UpdateBuffer
Jason Ekstrand [Wed, 20 Jan 2016 00:17:13 +0000 (16:17 -0800)]
anv/meta: Implement CmdFillBuffer
Jason Ekstrand [Tue, 19 Jan 2016 23:52:03 +0000 (15:52 -0800)]
anv/meta_clear: Call emit_clear directly in ClearImage
Using the load op means that we end up with recursive meta. We shouldn't
be doing that.
Jason Ekstrand [Tue, 19 Jan 2016 23:27:20 +0000 (15:27 -0800)]
anv/meta_clear: Do save/restore in actual entry points
Jason Ekstrand [Tue, 19 Jan 2016 23:01:10 +0000 (15:01 -0800)]
anv: Add support for VK_WHOLE_SIZE several places
Kenneth Graunke [Wed, 20 Jan 2016 00:46:03 +0000 (16:46 -0800)]
nir/spirv/glsl450: Implement Frexp.
Kenneth Graunke [Tue, 19 Jan 2016 23:56:50 +0000 (15:56 -0800)]
nir/spirv/glsl450: Blindly implement Atan2.
This is untested and probably broken.
We already passed the atan2 CTS tests before implementing this opcode.
Presumably, glslang or something was giving us a plain Atan opcode
instead of Atan2. I don't know why.
Kenneth Graunke [Tue, 19 Jan 2016 22:30:02 +0000 (14:30 -0800)]
nir/spirv/glsl450: Implement Atan.
Kenneth Graunke [Tue, 19 Jan 2016 20:15:36 +0000 (12:15 -0800)]
nir/spirv/glsl450: Implement Asin and Acos.
Jason Ekstrand [Tue, 19 Jan 2016 20:02:53 +0000 (12:02 -0800)]
anv/pipeline: Fix point size
Jason Ekstrand [Tue, 19 Jan 2016 01:45:12 +0000 (17:45 -0800)]
getX/state: Set LOD pre-clamp to OpenGL mode
This gets us another couple hundred sampler tests
Jason Ekstrand [Tue, 5 Jan 2016 21:53:05 +0000 (13:53 -0800)]
isl/device: Add a flag for bit 6 swizzling
Jason Ekstrand [Mon, 18 Jan 2016 23:42:41 +0000 (15:42 -0800)]
anv/gem: Add a helper for getting bit6 swizzling information
Jason Ekstrand [Tue, 19 Jan 2016 00:01:00 +0000 (16:01 -0800)]
nir/spirv: Patch through image qualifiers
Jason Ekstrand [Tue, 5 Jan 2016 22:03:58 +0000 (14:03 -0800)]
nir/spirv: Implement ImageQuerySize for storage iamges
SPIR-V only has one ImageQuerySize opcode that has to work for both
textures and storage images. Therefore, we have to special-case that one a
bit and look at the type of the incoming image handle.
Jason Ekstrand [Mon, 18 Jan 2016 23:00:01 +0000 (15:00 -0800)]
nir/spirv: Insert movs around image intrinsics
Image intrinsics always take a vec4 coordinate and always return a vec4.
This simplifies the intrinsics a but but also means that they don't
actually match the incomming SPIR-V. In order to compensate for this, we
add swizzling movs for both source and destination to get the right number
of components.
Jason Ekstrand [Mon, 18 Jan 2016 22:07:46 +0000 (14:07 -0800)]
anv/meta: Improve meta clear cleanup a bit
Jason Ekstrand [Mon, 18 Jan 2016 22:04:13 +0000 (14:04 -0800)]
anv: Misc allocation scope fixes
Jason Ekstrand [Mon, 18 Jan 2016 22:03:20 +0000 (14:03 -0800)]
anv/meta: Add a meta allocator that uses SCOPE_DEVICE
The Vulkan spec requires all allocations that happen for device creation to
happen with SCOPE_DEVICE. Since meta calls into other things that allocate
memory, the easiest way to do this is with an allocator.
Jason Ekstrand [Mon, 18 Jan 2016 21:05:02 +0000 (13:05 -0800)]
anv/meta: Initialize a handle to null
Jason Ekstrand [Mon, 18 Jan 2016 20:16:31 +0000 (12:16 -0800)]
gen8: Fix border color
The border color packet is specified as a 64-byte aligned address relative
to dynamic state base address. The way the packing functions are currently
set up, we need to provide it with (offset >> 6) because it just shoves the
bits in where the PRM says they go and isn't really aware that it's an
address.
Jason Ekstrand [Mon, 18 Jan 2016 19:35:04 +0000 (11:35 -0800)]
genX/pack: Add a __gen_fixed helper and use it for TextureLODBias
The __gen_fixed helper properly clamps the value and also handles negative
values correctly. Eventually, we need to make the scripts generate this
and use it for more things.
Jason Ekstrand [Mon, 18 Jan 2016 18:36:53 +0000 (10:36 -0800)]
anv/pack: Make TextureLODBias a proper 4.8 float
XXX: We need to update the generators so this doesn't get stompped.
Jason Ekstrand [Mon, 18 Jan 2016 18:10:47 +0000 (10:10 -0800)]
nir/spirv: Handle if's where the merge is also a break or continue
Jason Ekstrand [Mon, 18 Jan 2016 17:33:56 +0000 (09:33 -0800)]
nir/spirv: Hanle continues that use SSA values from the loop body
Instead of emitting the continue before the loop body we emit it
afterwards. Then, once we've finished with the entire function, we run
nir_repair_ssa to add whatever phi nodes are needed.
Jason Ekstrand [Mon, 18 Jan 2016 17:24:07 +0000 (09:24 -0800)]
nir/lower_returns: Repair SSA after doing return lowering
Jason Ekstrand [Mon, 18 Jan 2016 17:19:01 +0000 (09:19 -0800)]
nir: Add a pass to repair SSA form
Jason Ekstrand [Mon, 18 Jan 2016 16:40:16 +0000 (08:40 -0800)]
nir/vars_to_ssa: Use the new nir_phi_builder helper
The efficiency should be approximately the same. We do a little more work
per phi node because we have to sort the predecessors. However, we no
longer have to walk the blocks a second time to pop things off the stack.
The bigger advantage, however, is that we can now re-use the phi placement
and per-block SSA value tracking in other passes.
Jason Ekstrand [Mon, 18 Jan 2016 16:38:57 +0000 (08:38 -0800)]
nir: Add a phi node placement helper
Right now, we have phi placement code in two places and there are other
places where it would be nice to be able to do this analysis. Instead of
repeating it all over the place, this commit adds a helper for placing all
of the needed phi nodes for a value.
Jason Ekstrand [Sun, 17 Jan 2016 05:26:10 +0000 (21:26 -0800)]
util/bitset: Allow iterating over const bitsets
Jason Ekstrand [Sat, 16 Jan 2016 00:15:50 +0000 (16:15 -0800)]
nir/lower_system_values: Lower vertexID to id+base if needed
Jason Ekstrand [Fri, 15 Jan 2016 23:59:58 +0000 (15:59 -0800)]
anv/batch_chain: Remove padding from the BO before emitting BUFFER_END
Jason Ekstrand [Fri, 15 Jan 2016 23:48:18 +0000 (15:48 -0800)]
anv/batch_chain: Don't call current_batch_bo() again
We call it once at the top of the function and then hold on to the pointer.
It shouldn't have changed, so there's no reason to query for it again.
Jason Ekstrand [Fri, 15 Jan 2016 19:10:58 +0000 (11:10 -0800)]
nir/spirv: Stop trusting the SPIR-V for the number of texture coordinates
Chad Versace [Fri, 15 Jan 2016 18:54:05 +0000 (10:54 -0800)]
anv: Populate SURFACE_STATE more safely
genX_image_view_init allocates up to 3 separate SURFACE_STATE structures,
and populates each from a single template. Stop mutating the template
between each final SURFACE_STATE.
Chad Versace [Fri, 15 Jan 2016 18:07:16 +0000 (10:07 -0800)]
anv/meta: Stop leaking renderpass and framebuffer
Chad Versace [Fri, 15 Jan 2016 15:43:31 +0000 (07:43 -0800)]
anv/meta: Reuse code for vkCmdClear{Color,DepthStencil}Image
The two function bodies were very similar. Move common code to
anv_cmd_clear_image().
Fixes all 'dEQP-VK.renderpass.formats.*' on Skylake.
Chad Versace [Fri, 15 Jan 2016 05:29:46 +0000 (21:29 -0800)]
anv/gen8: Fix SF_CLIP_VIEWPORT's Z elements
SF_CLIP_VIEWPORT does not clamp Z values. It only scales and shifts
them. Clamping to VkViewport::minDepth,maxDepth is instead handled by
CC_VIEWPORT.
Fixes dEQP-VK.renderpass.simple.depth on Broadwell.
Chad Versace [Fri, 15 Jan 2016 04:09:38 +0000 (20:09 -0800)]
anv/meta: Implement vkCmdClearDepthStencilImage
Chad Versace [Thu, 14 Jan 2016 02:37:27 +0000 (18:37 -0800)]
anv/meta: Implement vkCmdClearAttachments
Chad Versace [Thu, 14 Jan 2016 02:34:20 +0000 (18:34 -0800)]
anv/meta: Add VkClearRect param to emit_clear()
Prepares for vkCmdClearAttachments.
Chad Versace [Thu, 14 Jan 2016 23:18:20 +0000 (15:18 -0800)]
anv: Distinguish between subpass setup and subpass start
vkCmdBeginRenderPass, vkCmdNextSubpass, and vkBeginCommandBuffer with
VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT, all *setup* the
command buffer for recording commands for some subpass. But only the
first two, vkCmdBeginRenderPass and vkCmdNextSubpass, can *start*
a subpass.
Therefore, calling anv_cmd_buffer_begin_subpass() inside
vkCmdBeginCommandBuffer is misleading. Clarify its purpose by renaming
it to anv_cmd_buffer_set_subpass() and adding comments.
Chad Versace [Wed, 13 Jan 2016 22:47:51 +0000 (14:47 -0800)]
anv: Emit load clears at start of each subpass
This should improve cache residency for render targets.
Pre-patch, vkCmdBeginRenderPass emitted all the meta clears for
VK_ATTACHMENT_LOAD_OP_CLEAR before any subpass began. Post-patch,
vCmdBeginRenderPass and vkCmdNextSubpass emit only the clears needed for
that current subpass.
Chad Versace [Thu, 14 Jan 2016 00:29:45 +0000 (16:29 -0800)]
anv/meta: Create 8 pipelines for color clears
This prepares for moving the clear ops from the start of the render pass
into each subpass.
Pipeline N will be used to clear color attachment N of the current
subpass. Currently meta color clears still create a throwaway subpass
with exactly one attachment, so currently only pipeline 0 is used.
This is an ugly hack to workaround the compiler's current inability to
dynamically set the render target index in the render target write
message.
Chad Versace [Thu, 14 Jan 2016 02:24:18 +0000 (18:24 -0800)]
anv: Allow override of pipeline color attachment count
Add anv_graphics_pipeline_create_info::color_attachment_count. If
non-negative, then it overrides the color attachment count in the
pipeline's subpass. Useful for meta. (All the hacks for meta!)
Chad Versace [Thu, 14 Jan 2016 02:09:01 +0000 (18:09 -0800)]
anv/meta: Name the nir shaders
The names appear in debug output.
Chad Versace [Thu, 14 Jan 2016 00:03:09 +0000 (16:03 -0800)]
anv: Move MAX_* defs to top of anv_private.h
Because I need to use MAX_RTS in struct anv_meta_state.
Chad Versace [Thu, 14 Jan 2016 00:24:14 +0000 (16:24 -0800)]
anv: Define zero() macro
zero(x) memsets x to zero. Eliminates bugs due to errors in memset's
size param.