[mtl] cache depth/stencil states #2195
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
22 tasks
grovesNL
approved these changes
Jun 30, 2018
There was a problem hiding this comment.
This looks really good! 👍 I agree it's much easier to follow now.
My only concern would be that the hashmap caching depth stencil states could potentially grow fairly large over time, because we never really release it. But we could improve that in the future.
There seems to be some regression within the CTS, but it doesn't seem related to these commits. I'll look into it.
| error!("Using {:?} with no depth attachment!", ds_desc); | ||
| ds_desc.depth = pso::DepthTest::Off; | ||
| } | ||
| pso::DepthTest::On { .. } => {} |
|
Confirmed the CTS failure seems unrelated to this PR, created #2199 |
|
Thank you!
With pattern matching, we try to avoid more generic patterns if we know a more specific one.
Bors r=grovesNL
… On Jun 30, 2018, at 02:00, Josh Groves ***@***.***> wrote:
Confirmed the CTS failure seems unrelated to this PR, created #2199
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.
|
bors bot
added a commit
that referenced
this pull request
Jun 30, 2018
2195: [mtl] cache depth/stencil states r=grovesNL a=kvark
Fixes the dynamic allocation of depth/stencil states in Metal.
PR checklist:
- [x] `make` succeeds (on *nix)
- [x] `make reftests` succeeds
- [x] tested examples with the following backends: metal
The new logic goes as follows:
- there is a central storage for depth-stencil states in our service pipes
- a command buffer keeps a hold of just 2 things: the DS state that came from the last graphics pipeline and the set of stencil masks & reference values
- a command buffer then assembles a full DS state from those things in the following sync points:
- binding a render pipeline
- setting stencil masks
- starting a render pass
I find this logic easier to understand than the previous solution. It allows us to avoid re-creating Metal depth-stencil states. As a small bonus, it becomes possible to work around Dota's bug of trying to use a depth state without a proper attachment (included in the PR).
Co-authored-by: Dzmitry Malyshau <[email protected]>
bors bot
added a commit
to gfx-rs/portability
that referenced
this pull request
Jul 1, 2018
109: [WIP] Handle sanitation and gfx-hal update r=msiglreith,grovesNL a=kvark Depends on gfx-rs/gfx#2195 Depends on gfx-rs/metal-rs#59 Fixes #103 Co-authored-by: Dzmitry Malyshau <[email protected]>
bors bot
added a commit
to gfx-rs/portability
that referenced
this pull request
Jul 1, 2018
109: [WIP] Handle sanitation and gfx-hal update r=msiglreith,grovesNL a=kvark Depends on gfx-rs/gfx#2195 Depends on gfx-rs/metal-rs#59 Fixes #103 Co-authored-by: Dzmitry Malyshau <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes the dynamic allocation of depth/stencil states in Metal.
PR checklist:
makesucceeds (on *nix)make reftestssucceedsThe new logic goes as follows:
I find this logic easier to understand than the previous solution. It allows us to avoid re-creating Metal depth-stencil states. As a small bonus, it becomes possible to work around Dota's bug of trying to use a depth state without a proper attachment (included in the PR).