Optimize resource upload in D3D11#48282
Merged
reflectronic merged 2 commits intozed-industries:mainfrom Feb 3, 2026
Merged
Conversation
adb-sh
pushed a commit
to adb-sh/zed
that referenced
this pull request
Feb 5, 2026
Currently, each time we draw a primitive batch, we fully overwrite the instance buffer with the contents of the new batch. Since we use a write-only mapping to do this, the GPU driver may handle synchronization hazards by transparently creating new allocations if the previous allocation is still in use. We draw many primitive batches in one frame, which stress-tests this mechanism somewhat. If internal driver limits are hit, the resource update will start to block until the GPU catches up and releases in-use allocations. This would result in a significant reduction in framerate. To avoid this, we upload the data for all primitive batches at once at the beginning of the frame. Each primitive batch draw then binds the relevant sub-array of the instance buffer. This way, there are no mid-frame resource updates. Release Notes: - N/A
rtfeldman
pushed a commit
that referenced
this pull request
Feb 5, 2026
Currently, each time we draw a primitive batch, we fully overwrite the instance buffer with the contents of the new batch. Since we use a write-only mapping to do this, the GPU driver may handle synchronization hazards by transparently creating new allocations if the previous allocation is still in use. We draw many primitive batches in one frame, which stress-tests this mechanism somewhat. If internal driver limits are hit, the resource update will start to block until the GPU catches up and releases in-use allocations. This would result in a significant reduction in framerate. To avoid this, we upload the data for all primitive batches at once at the beginning of the frame. Each primitive batch draw then binds the relevant sub-array of the instance buffer. This way, there are no mid-frame resource updates. Release Notes: - N/A
naaiyy
added a commit
to Glass-HQ/Glass
that referenced
this pull request
Feb 16, 2026
Key changes: - Reduce monomorphizations in GPUI app.rs (zed-industries#48014) - Entities no longer implement Element directly, go through AnyElement (zed-industries#48217) - D3D11 resource upload optimization (zed-industries#48282) - Migrate features.edit_prediction_provider to edit_predictions.provider (zed-industries#48224) - Make mercury and sweep non-experimental (zed-industries#48227) - CompanionView consolidation in block_map (zed-industries#48223) - Show memory used by language servers (zed-industries#48226) - Settings links open sub pages (zed-industries#48212) - Tool security rules that can't be overridden (zed-industries#48209) - Add sweep_ai privacy mode setting (zed-industries#48220) - Configurable REPL output size limits (zed-industries#47114) - Fix .editorconfig files in subdirectories (zed-industries#48203) - Security updates: bytes v1.11.1, jsonwebtoken v10 - Git UI: hide "View on GitHub" for stashes (zed-industries#48271) - Indent guide fix in tree view with collapsed folders (zed-industries#48194) - Edit prediction fixes and improvements Conflict resolution: - collab/completion.rs, collab/rpc.rs: deleted (collab removed) - vim/search.rs: deleted (vim removed) - livekit_api/Cargo.toml: deleted (livekit removed) - GPUI files: deleted from Glass (handled in Obsydian-HQ/gpui) - migrations: combined both ours (m_2026_02_06) and upstream (m_2026_02_02, m_2026_02_03) - project.rs: kept collab functions removed, restored handle_create_file_for_peer for remote dev - lsp_store.rs: removed collab-only set_language_server_statuses_from_proto - proto.rs: kept CreateFileForPeer, removed CreateChannel/CreateChannelResponse - remote_servers.rs: merged import lists (kept native button imports + added Action) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or 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
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.
Currently, each time we draw a primitive batch, we fully overwrite the instance buffer with the contents of the new batch. Since we use a write-only mapping to do this, the GPU driver may handle synchronization hazards by transparently creating new allocations if the previous allocation is still in use. We draw many primitive batches in one frame, which stress-tests this mechanism somewhat. If internal driver limits are hit, the resource update will start to block until the GPU catches up and releases in-use allocations. This would result in a significant reduction in framerate.
To avoid this, we upload the data for all primitive batches at once at the beginning of the frame. Each primitive batch draw then binds the relevant sub-array of the instance buffer. This way, there are no mid-frame resource updates.
Release Notes: