Commit Graph

851 Commits

Author SHA1 Message Date
Hans-Kristian Arntzen
bf3793dd35 MSL: Improve handling of split tessellation access chains. 2021-05-21 16:32:03 +02:00
Hans-Kristian Arntzen
99ae0d32e9 MSL: Handle array with component when we cannot rely on user() attrib.
In these cases, we emit one variable per location, and so we must
flatten stuff.
2021-05-21 13:46:33 +02:00
Hans-Kristian Arntzen
a64ddcdd49 MSL: Handle array of IO variable with Component decoration. 2021-05-21 12:31:33 +02:00
Hans-Kristian Arntzen
e47a30e807 Honor NoContraction qualifier.
We'll need to force a temporary and mark it as precise.
MSL is a little weird here, but we can piggyback on top of the invariant
float math option here to force fma() operations everywhere.
2021-05-07 12:59:47 +02:00
Lukas Taparauskas
72a2ec4c1b
MSL: Fix '--msl-multi-patch-workgroup' out of bounds reads when dispatching more threads than control points (#1662)
* Fix '--msl-multi-patch-workgroup' cases where thread count exceeds data bounds

*Fix gl_PrimitiveID off by one error when computing last valid index
*Point gl_out to the last patch's data when threads exceed input data bounds
*Point patchOut to the last patch's data when threads exceed input data bounds

* Update MSL test expectations.

* Undo change to MSL multi-patch hull output bound checks

* Update MSL multi-patch test expectations.
2021-04-29 20:01:26 +02:00
Bill Hollings
098cdd64f6 MSL: Padding for Metal argument buffers should not double-count SampledImages. 2021-04-26 13:19:23 -04:00
Hans-Kristian Arntzen
82a77e534e MSL: Use proper array for quad tess levels.
We need to handle loads from array as well, so the float4 hack doesn't
work.
2021-04-23 14:12:00 +02:00
Hans-Kristian Arntzen
96ba044f01 HLSL: Fix automatic location assignment in block IO. 2021-04-20 13:04:26 +02:00
Hans-Kristian Arntzen
ae9ca7d73c MSL: Fix copy of arrays to/from stage IO variables.
Need to take into account effective storage classes and whether or not
we target stage IO blocks since native arrays are conditionally enabled.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
986196030d MSL: Don't use native arrays for tess level inputs. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
4a379a00f3 MSL: Don't emit native array for masked clip/cull distance. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
682a227f4b MSL: Make builtin argument type declaration context sensitive.
Sometimes we'll need array template, sometimes not 🤷.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
c1edd35d57 MSL: Use spvUnsafeArray for builtin arrays after all.
It will get too messy to deal with constant initializers any other way,
so just deal with complexity in argument_decl instead ...
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
7b9a591aa7 MSL: Hoist out to_tesc_invocation_id() in more places.
When emitting fixup code, we might not have gl_InvocationID yet.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
75ed73818c MSL: Handle loading Clip/CullDistance in TESE.
Need to allow the flattened space to go through in some edge cases where
we cannot reasonably unflatten.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
a159334895 MSL: Correctly analyze if builtin block is active.
Need to consider all members, bi_type is invalid for Blocks, need to
look at member decorations.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
5826298697 MSL: Handle CullDistance better. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
23da445bd4 MSL: Emit multiple threadgroup slices for multi-patch.
Multiple patches can run in the same workgroup when using multi-patch
mode, so we need to allocate enough storage to avoid false sharing.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
b442500204 MSL: Unroll initializations of CullDistance/ClipDistance control points. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
c9946296dd MSL: Fix initialization of masked threadgroup variables. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
faf80b08fc MSL: Don't report fallback location allocations as being "used".
It may shadow unused *real* inputs and confuse applications.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
5e9c2d060e MSL: Cleanup fallback IO block emission.
Need to emit in add_variable_to_iface(). Unifies the code paths a fair
bit.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
e32c474911 MSL: Handle masking of TESC IO block members. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
dc54f75eec MSL: Fixup gl_PerVertex names if we're emitting masked builtins. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
40f628f49c MSL: Add test for complex control point outputs. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
46c48ee6b5 MSL: Rewrite how IO blocks are emitted in multi-patch mode.
Firstly, never flatten inputs or outputs in multi-patch mode.
The main scenario where we do need to care is Block IO.
In this case, we should only flatten the top-level member, and after
that we use access chains as normal.

Using structs in Input storage class is now possible as well. We don't
need to consider per-location fixups at all here. In Vulkan, IO structs
must match exactly. Only plain vectors can have smaller vector sizes as
a special case.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
425e968720 MSL: Handle flattening of patch block outputs as well.
Always propagate InterfaceMember decoration.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
8e2dbe0d38 MSL: Do not declare patch variables on stack. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
ff3f5bcba5 MSL: Handle masking of builtin control points. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
6ecdd64a91 MSL: Emit a masked builtin IO block if necessary. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
22c9b63e78 MSL: Fix argument_decl check for builtin. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
c635c35c12 MSL: Temporarily allow empty output struct. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
a59e25db18 MSL: Small refactors. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
74b2acab9b MSL: Always emit block variable for block types. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
ae7bb41ef4 MSL: Test that we can mask location writes in TESC. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
3255d6cef0 MSL: Explicitly only consider masked variables to be thread-group-like. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
ba93b6518d MSL: Fix masking of vertex block outputs. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
a393de31e6 MSL: Refactor out variable/block member masking. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
e7b37392bf MSL: Emit correct address space for masked arguments. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
65b5ff7ece MSL: Don't emit weird reference type for spvUnsafeArray types. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
e7824c8b6e MSL: Handle masked outputs in extract_global_variables.
Need to conditionally add gl_in or gl_out.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
50a6bc058a MSL: Force builtin arrays for builtin array types.
Handles argument_decl() correctly.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
f2b5fb3f45 MSL: Emit threadgroup storage class for masked control point outputs.
Shader can still rely on writes to threadgroup memory to be visible.
2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
ea91579a7f MSL: Do not redirect tess access chains on masked outputs. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
f682e89188 MSL: Correctly emit array type for masked outputs. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
2a2d57df13 MSL: Sketch out API to aid LTO-style optimization. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
9a144bb2b9 Clean up member sorting. 2021-04-19 12:10:49 +02:00
Hans-Kristian Arntzen
0997e81118 MSL: Sort builtin IO block members by builtin type.
Ensures consistent block matching.
2021-04-19 12:10:49 +02:00
Bill Hollings
b3bfe22eaa MSL: Fixes to support padding Metal argument buffer entries based on argument index.
For buffers, support all MSLResourceBinding::basetype pointers, not just void*.
Rename MSLResourceBinding::base_type to basetype for consistent use in other structs.
2021-04-18 17:34:55 -04:00
Bill Hollings
daba0dfba6 MSL: Fixes to support padding Metal argument buffer entries based on argument index.
For completeness, add [[id(N)]] qualifier to padding struct members.
Run clang-format.
2021-04-17 15:20:53 -04:00