Workgroup size a compute entry point declares.
More...
#include "donner/gpu/Descriptors.h"
|
|
uint32_t | x = 1 |
| | Invocations along X. Must be nonzero.
|
|
uint32_t | y = 1 |
| | Invocations along Y. Must be nonzero.
|
|
uint32_t | z = 1 |
| | Invocations along Z. Must be nonzero.
|
Workgroup size a compute entry point declares.
The runtime carries it in the pipeline descriptor rather than reading it back from the shader because Metal takes the threadgroup size at dispatch time, not from the compiled pipeline state; every backend then dispatches with the same declared shape. Device::createComputePipeline checks the value against the size the shader module reports for the same entry point, so the two cannot disagree.
◆ operator==()
| bool donner::gpu::WorkgroupSize::operator== |
( |
const WorkgroupSize & | other | ) |
const |
|
default |
Equality operator.
- Parameters
-
| other | Size to compare against. |
◆ operator<<
| std::ostream & operator<< |
( |
std::ostream & | os, |
|
|
const WorkgroupSize & | value ) |
|
friend |
Ostream output operator, e.g. 8x8x1.
- Parameters
-
| os | Output stream. |
| value | Size to output. |
The documentation for this struct was generated from the following file: