Commit Graph

596 Commits

Author SHA1 Message Date
Chip Davis
f3c0942d10 MSL: Use vectors for the tessellation level builtins in tese shaders.
The tessellation levels in Metal are stored as a densely-packed array of
half-precision floating point values. But, stage-in attributes in Metal
have to have offsets and strides aligned to a multiple of four, so we
can't add them individually. Luckily for us, the arrays have lengths
less than 4. So, let's use vectors for them!

Triangles get a single attribute with a `float4`, where the outer levels
are in `.xyz` and the inner levels are in `.w`. The arrays are unpacked
as though we had added the elements individually. Quads get two: a
`float4` with the outer levels and a `float2` with the inner levels.
Further, since vectors can be indexed as arrays, there's no need to
unpack them in this case.

This also saves on precious vertex attributes. Before, we were using up
to 6 of them. Now we need two at most.
2019-02-22 12:18:51 -06:00
Hans-Kristian Arntzen
a4ac27546a MSL: Fix textures which are sampled and compared against.
depth2d in MSL only returns float, not float4, even for normal sampling.
We need to conditionally remap-swizzle back to float4.
2019-02-22 12:27:40 +01:00
Chip Davis
7a7e210515 MSL: Force unnamed array builtin attributes to have a name.
That way, when we refer to them, they'll have the name that we're
expecting.
2019-02-20 22:16:51 -06:00
Chip Davis
285ca4c2b1 MSL: Don't bother fixing up triangle tess coords.
Instead, I'm going to have MoltenVK reverse the winding order in the
lower-left case. This seems to be what the test suite expects to happen
anyhow.
2019-02-20 14:30:44 -06:00
Chip Davis
8095434dc4 MSL: Drop stores to nonexistent tess levels.
In SPIR-V, there are always two inner levels and four outer levels, even
if the input patch isn't a quad patch. But in MSL, due to requirements
imposed by Metal, only one inner level and three outer levels exist when
the input patch is a triangle patch. We must explicitly ignore any write
to the nonexistent second inner and fourth outer levels in this case.
2019-02-20 09:11:24 -06:00
Chip Davis
c8ee9fbe76 MSL: Expand quad gl_TessCoord to a float3.
This is the actual SPIR-V type of the builtin. We forced to a `float2`
in the declaration because that's what Metal wants.
2019-02-20 09:11:24 -06:00
Hans-Kristian Arntzen
58f264c99d
Merge pull request #865 from KhronosGroup/fix-863
Always value-cast FP16 constants instead of using literals.
2019-02-20 14:58:44 +01:00
Hans-Kristian Arntzen
4ef51331b2 Always value-cast FP16 constants instead of using literals.
GL_NV_gpu_shader5 doesn't support "hf", so to avoid lots of complicated
workarounds, just value-cast the half literals.
2019-02-20 12:30:01 +01:00
Hans-Kristian Arntzen
056a0ba27e Fix case where a struct is loaded which contains a row-major matrix. 2019-02-20 12:19:00 +01:00
Hans-Kristian Arntzen
d2cc43e667 Fix edge case where opaque types can be declared on stack.
In the bizarre case where the ID of a loaded opaque type aliased with a
literal which was used as part of another texturing instruction, we
could end up with a case where domination analysis assumed the loaded
opaque type needed to be moved to a different scope.

Fix the issue by never doing dominance analysis for opaque temporaries,
and be more robust when analyzing texturing instructions.

Also make sure reflection output is deterministic.
This patch slightly alterered output for some unknown reason, but it came from an
unordered_map, so it's fine.
2019-02-19 17:28:31 +01:00
Chip Davis
41d9424233 MSL: Add an option to set the tessellation domain origin.
This is intended to be used to support `VK_KHR_maintenance2`'s
tessellation domain origin feature. If `tess_domain_origin_lower_left`
is `true`, the `v` coordinate will be inverted with respect to the
domain. Additionally, in `Triangles` mode, the `v` and `w` coordinates
will be swapped. This is because the winding order is interpreted
differently in lower-left mode.
2019-02-18 14:25:42 -06:00
Chip Davis
08863c1e28 Don't set any aliases or do any flattening for arrayed per-vertex I/O.
We already handle all that specially.
2019-02-15 17:24:16 -06:00
Chip Davis
6b7988046d Handle blocks of patch I/O.
In this case, each member of the block will be decorated with
`DecorationPatch`, rather than the block variable having the decoration.
2019-02-15 17:21:38 -06:00
Chip Davis
e75add42c9 MSL: Add support for tessellation evaluation shaders.
These are mapped to Metal's post-tessellation vertex functions. The
semantic difference is much less here, so this change should be simpler
than the previous one. There are still some hairy parts, though.

In MSL, the array of control point data is represented by a special
type, `patch_control_point<T>`, where `T` is a valid stage-input type.
This object must be embedded inside the patch-level stage input. For
this reason, I've added a new type to the type system to represent this.

On Mac, the number of input control points to the function must be
specified in the `patch()` attribute. This is optional on iOS.
SPIRV-Cross takes this from the `OutputVertices` execution mode; the
intent is that if it's not set in the shader itself, MoltenVK will set
it from the tessellation control shader. If you're translating these
offline, you'll have to update the control point count manually, since
this number must match the number that is passed to the
`drawPatches:...` family of methods.

Fixes #120.
2019-02-14 10:00:08 -06:00
Hans-Kristian Arntzen
aab4a5632b Merge branch 'tesc-shader' of git://github.com/cdavis5e/SPIRV-Cross 2019-02-14 09:17:32 +01:00
Chip Davis
13df78bebf Unflatten inputs when copying to outputs.
This should fix a whole host of issues related to structs in the `Input`
class in a tessellation control shader.

Also, use pointer arithmetic instead of dereferencing the `ops` array.
This is critical in case we wind up stepping beyond the bounds of the
array.
2019-02-13 12:37:24 -06:00
Hans-Kristian Arntzen
d7090b8322 GLSL: Fix block name shenanigans in edge cases.
When we force recompile, the old var.self name we used as a fallback
name might have been disturbed, so we should recover certain names back
to their original form in case we are forced to take a recompile to make
the naming algorithm more deterministic.
2019-02-13 16:39:59 +01:00
Chip Davis
0bb6bbda22 Never flatten outputs when capturing them.
There's no need to do so, since these are not stage-out structs being
returned, but regular structures being written to a buffer. This also
neatly avoids issues writing to composite (e.g. arrayed) per-patch
outputs from a tessellation control shader.
2019-02-11 17:18:54 -06:00
Chip Davis
1919eb1b46 Pass the original pointer type to ensure_correct_attribute_type().
This prevents us from overwriting the variable's type with a non-pointer
type.
2019-02-11 16:07:43 -06:00
Chip Davis
eb89c3a428 MSL: Add support for tessellation control shaders.
These are transpiled to kernel functions that write the output of the
shader to three buffers: one for per-vertex varyings, one for per-patch
varyings, and one for the tessellation levels. This structure is
mandated by the way Metal works, where the tessellation factors are
supplied to the draw method in their own buffer, while the per-patch and
per-vertex varyings are supplied as though they were vertex attributes;
since they have different step rates, they must be in separate buffers.

The kernel is expected to be run in a workgroup whose size is the
greater of the number of input or output control points. It uses Metal's
support for vertex-style stage input to a compute shader to get the
input values; therefore, at least one instance must run per input point.
Meanwhile, Vulkan mandates that it run at least once per output point.
Overrunning the output array is a concern, but any values written should
either be discarded or overwritten by subsequent patches. I'm probably
going to put some slop space in the buffer when I integrate this into
MoltenVK to be on the safe side.
2019-02-07 08:51:22 -06:00
Hans-Kristian Arntzen
d9ed3dcc7a
Merge pull request #848 from cdavis5e/capture-output-buffer
MSL: Add a setting to capture vertex shader output to a buffer.
2019-02-07 15:11:41 +01:00
Chip Davis
056c0e207d Take the vertex count from any indirect parameters passed.
This is necessary to deal with indirect draws, where the draw parameters
are given in a buffer instead of passed by the CPU. For normal draws,
the draw parameters are set with Metal's `setVertexBytes:` method.

This undoes the change to add the vertex count to the aux buffer,
rendering that entire discussion largely moot. Oh well. It was a
discussion that needed to happen anyway.
2019-02-06 15:17:14 -06:00
Chip Davis
546f1ccbb5 Test that out variables still work in leaf functions with capture on. 2019-02-06 10:49:25 -06:00
Hans-Kristian Arntzen
d5385190ff
Merge pull request #850 from KhronosGroup/fix-846
Support LUTs in single-function CFGs on Private storage class.
2019-02-06 11:34:30 +01:00
Hans-Kristian Arntzen
3e584f2c3f Support LUTs in single-function CFGs on Private storage class.
Fairly common pattern in unoptimized SPIR-V. Support this case as well.
2019-02-06 10:38:59 +01:00
Chip Davis
0757fae511 MSL: Stop passing the aux buffer around.
Since we pass the component swizzle around now, there's no need to pass
it to every function that takes a sampled image.
2019-02-05 20:04:32 -06:00
Chip Davis
c51e5b7911 MSL: Add a setting to capture vertex shader output to a buffer.
This will be necessary to support transform feedback, as well as
tessellation shaders.
2019-02-05 20:00:10 -06:00
Hans-Kristian Arntzen
fcbe999d99 MSL: Fix another test incompatibility. 2019-01-30 17:22:38 +01:00
Hans-Kristian Arntzen
b78ffa1cc7 Fixup MSL iOS test. 2019-01-30 16:26:41 +01:00
Hans-Kristian Arntzen
2ed171e525 GLSL/MSL: Implement 8-bit part of VK_KHR_shader_float16_int8.
Storage was in place already, so mostly just dealing with bitcasts and
constants.

Simplies some of the bitcasting logic, and this exposed some bugs in the
implementation. Refactor to use correct width integers with explicit bitcast opcodes.
2019-01-30 15:45:24 +01:00
Hans-Kristian Arntzen
4e7777c443 Update to latest glslang/SPIRV-Tools.
Fix various bugs along the way.
2019-01-30 13:41:57 +01:00
Hans-Kristian Arntzen
3e09879131 Support initializers on StorageClassOutput. 2019-01-30 10:29:08 +01:00
Hans-Kristian Arntzen
912fde95f1 MSL: Use correct size for structs.
Need to align the size of structs to the natural alignment.
2019-01-28 15:20:30 +01:00
Hans-Kristian Arntzen
217eb5b5f9 MSL: Add a preliminary check for bad arrays of structs.
ArrayStride can be larger than the declared struct size.
We have no obvious solution for now, but warn about it in the MSL output
for the time being.
2019-01-28 15:20:30 +01:00
Hans-Kristian Arntzen
8c632da461 MSL: Use correct alignment rule for whole structs.
Structs are aligned as you would expect in MSL (maximum member
alignment), and it is not minimum 16 bytes like in std140.

Also rename the dummy "pad" members to a reserved naming scheme.
2019-01-28 15:20:30 +01:00
Hans-Kristian Arntzen
18a4accd2f HLSL/MSL: Fix texture projection with Dref.
We need to divide the Dref by q.
2019-01-28 10:25:13 +01:00
Hans-Kristian Arntzen
437fc87a89 MSL: Deal with resource name aliasing.
Apparently we didn't use those yet. MSL seems to be able to alias struct
types and variable types to a degree, so that's why it has escaped
testing until now.
2019-01-18 16:27:57 +01:00
Hans-Kristian Arntzen
1040cf6cc1
Merge pull request #831 from cdavis5e/force-recompile-hooks
MSL: Hoist fixup hooks in entry_point_args() out of the compile loop.
2019-01-17 19:42:05 +01:00
Chip Davis
f500d2f70c MSL: Hoist fixup hooks in entry_point_args() out of the compile loop.
Otherwise, in the event of a forced recompile, we could end up adding
them twice.
2019-01-17 10:18:38 -06:00
Hans-Kristian Arntzen
3aa08f764e MSL: Fix image load/store for short vectors.
Same fixes as for GLSL.
2019-01-17 14:54:29 +01:00
Hans-Kristian Arntzen
432aaed737 Need to know the original packed type when unpacking loads. 2019-01-17 11:39:46 +01:00
Hans-Kristian Arntzen
f4b74f2f4f Add basic test for std140 small vector arrays. 2019-01-17 11:29:09 +01:00
Hans-Kristian Arntzen
de7e5ccd8b Refactor out packed expressions to extended decorations.
Can't safely just cast to the original enum without lots of hacks.
2019-01-17 11:28:51 +01:00
Hans-Kristian Arntzen
f4026a5618 Refactor access_chain_internal to be more readable from callsite. 2019-01-17 10:30:13 +01:00
Hans-Kristian Arntzen
15b52bee48 Deal with packing/unpacking on store.
Still a bit buggy, since we cannot deduce between float2[] and
packed_float2. Need a deeper refactor to plumb this through ...
2019-01-17 10:06:23 +01:00
Hans-Kristian Arntzen
9e3a41ad00
Merge pull request #821 from cdavis5e/pass-sampled-images
MSL: Fix passing a sampled image to a function.
2019-01-15 09:05:54 +01:00
Chip Davis
664df22d12 MSL: Fix passing a sampled image to a function.
In the past, SPIRV-Cross threw an error in this case because it couldn't
work out which swizzle from the auxiliary buffer needs to be passed.
Now, we pass the swizzle around with the texture object, like a combined
image-sampler and its associated sampler.
2019-01-14 09:29:31 -06:00
Hans-Kristian Arntzen
b8033d7525 MSL: Add option to pad fragment outputs.
If not enough components are provided in the shader,
the shader MSL compiler throws an error rather than make components
undefined. This hurts portability, so we need to add explicit padding
here.
2019-01-14 15:11:52 +01:00
Hans-Kristian Arntzen
7ee04936ac MSL: Fix case where we pass arrays to functions by value.
MSL does not support value semantics for arrays (sigh), so we need to
force constant references and deal with copies if we have a different
address space than what we end up guessing.
2019-01-14 11:00:14 +01:00
Hans-Kristian Arntzen
4394249b86
Merge pull request #815 from KhronosGroup/fix-811
HLSL: Support dual-source blending.
2019-01-11 12:52:26 +01:00