Switch Binding Arrays on Metal to Argument Buffers #6751
Conversation
cwfitzgerald
force-pushed
the
cw/argument-buffer-metal
branch
from
1f67c9f to
05236bc
Compare
cwfitzgerald
changed the title
cwfitzgerald
force-pushed
the
cw/argument-buffer-metal
branch
3 times, most recently
from
7be0e51 to
8986532
Compare
cwfitzgerald
force-pushed
the
cw/argument-buffer-metal
branch
3 times, most recently
from
cb8d1d7 to
793f047
Compare
| CARGO_TERM_COLOR: always | ||
| WGPU_DX12_COMPILER: dxc | ||
| RUST_LOG: info | ||
| RUST_LOG: debug |
There was a problem hiding this comment.
I think this is a better setting anyway
| { | ||
| return; | ||
| } |
There was a problem hiding this comment.
Missed a feature check on this benchmark
| use crate::auxil::map_naga_stage; | ||
| use crate::TlasInstance; | ||
|
|
||
| use metal::foreign_types::ForeignType; |
| (entry, layout) | ||
| }); | ||
| for (entry, layout) in layout_and_entry_iter { | ||
| // Bindless path |
There was a problem hiding this comment.
The bindless path is entirely new, the bindful path is entirely the same.
cwfitzgerald
force-pushed
the
cw/argument-buffer-metal
branch
from
6857fc9 to
4ce012c
Compare
| /// | ||
| /// As such, we wrap all argument buffers in a struct that has a single generic `<T>` field. | ||
| /// This allows `NagaArgumentBufferWrapper<metal::texture<..>>*` to work. The astute among | ||
| /// you have noticed that this should be exactly the same to the compiler, and you're correct. |
| && self.supports_arrays_of_textures_write | ||
| { | ||
| features.insert(F::STORAGE_RESOURCE_BINDING_ARRAY); | ||
| } |
There was a problem hiding this comment.
I don't fully understand the old comment but removing this means that write only storage textures are no longer supported. Did this work before?
There was a problem hiding this comment.
Lines 573 to 578 in f6f9233
Seems to be listed here.
There was a problem hiding this comment.
I don't fully understand the old comment but removing this means that write only storage textures are no longer supported. Did this work before?
No they hit a nice fat assert in naga and generated bad msl once through
2 participants
Connections
Closes #3334
Closes #3640
Depends on #6732
Commits cannot be merged separately, but can be reviewed separately