src: Add vk_format_utils to UtilityHeaders #89
Conversation
charles-lunarg
requested review from
ziga-lunarg,
juan-lunarg,
spencer-lunarg and
artem-lunarg
charles-lunarg
force-pushed
the
format_utils
branch
from
82e4221
to
1256e07
Compare
There was a problem hiding this comment.
I guess before we merge here, a good exercise would be go in the Validation Layers
- Remove the current format utils (so it doesn't generate)
- Replace the header with this file
- Make sure CI will pass and this didn't forget anything subtle a test there would have caught
(the goal not being to merge that PR, but to prove this new interface works)
I started to do that on Friday and did not have much trouble with the conversion but had a lot of trouble running tests (w and w/o mock icd). It felt like I setup the repo wrong but I was on a different machine than I normally develop on, so I’ll try again come the morn on that machine. |
| inline bool FormatIsUSCALED(VkFormat format); | ||
|
|
||
| // Returns whether a format type is any type from "Interpretation of Numeric Format" table (OpTypeInt) | ||
| inline bool FormatIsSampledInt(VkFormat format); |
There was a problem hiding this comment.
Is "Interpretation of Numeric Format" table about the formats that can be sampled? Not sure I understand semantics of this function and in which scenarios it can be used.
There was a problem hiding this comment.
I wasn't super clear on that myself - but I'll look at the format table in question and double check that.
There was a problem hiding this comment.
so for the formats, awhile ago I tried to unify the wording
https://registry.khronos.org/vulkan/specs/1.3-extensions/html/vkspec.html#formats-numericformat
in short:
- Numeric Format = [
SFLOAT,SINT,SNORM,SRGB,SSCALED,UFLOAT,UINT,UNORM,USCALED] - Numeric Type = [float, signed int, unsigned int]
- SPIR-V Type = [ [
OpTypeFloat,OpTypeInt], width, sign ]
.... basically we should re-write this comment to whatever @charles-lunarg thinks "makes sense" as I am now to tainted with spec lawyer knowledge
There was a problem hiding this comment.
Let me know if the comments make more sense now.
|
Did some minimal integration testing with VulkanTools. Looks good on that front 👍🏾 |
charles-lunarg
force-pushed
the
format_utils
branch
from
1256e07
to
815fdc8
Compare
charles-lunarg
requested review from
spencer-lunarg,
juan-lunarg and
artem-lunarg
charles-lunarg
force-pushed
the
format_utils
branch
2 times, most recently
from
bd36e90
to
659a524
Compare
juan-lunarg
force-pushed
the
format_utils
branch
from
659a524
to
97fccae
Compare
charles-lunarg
force-pushed
the
format_utils
branch
from
97fccae
to
6030460
Compare
charles-lunarg
force-pushed
the
format_utils
branch
4 times, most recently
from
a749af3
to
2376416
Compare
|
Apologies for the force push spam - the C compatibility tests & add_subdirectory/find_package tests are easy to forget about since they aren't compiled normally. The build.gn file should be correct now as well @juan-lunarg |
The library originates from Vulkan-ValidationLayers, but is being moved into this repo to make it easier for others to use it. The library has also been modified to be header only and C compatible, which allows more developers to be able to use it. This does require some changes, but only affects the vkuFormatElementSize and vkuFormatTexelSize functions which used default parameters. Two new functions, vkuFormatElementSizeWithAspect and vkuFormatTexelSizeWithAspect have been added to handle the non-default image aspect case (the default was COLOR_BIT). Renaming was done using the following convention: * public header files begin with `vk_` * enums begin with VKU_FORMAT_ * functions begin with vku
charles-lunarg
force-pushed
the
format_utils
branch
from
2376416
to
b267b0f
Compare
The library originates from Vulkan-ValidationLayers, but is being moved into this repo to make it easier for others to use it.
The library has also been modified to be header only and C compatible, which allows more developers to be able to use it. This does require some changes, but only affects the FormatElementSize and FormatTexelSize functions which used default parameters. Two new functions, FormatElementSizeWithAspect and FormatTexelSizeWithAspect have been added to handle the non-default image aspect case (the default was COLOR_BIT).