Open
Conversation
* Initial Metal backend * Adds image support * Metal Cursor blinking * Strikethrough support * Dirty‑row caching is in place for the Metal renderer. It now caches per‑row draw data and only rebuild rows in the update range, with cache invalidation when scale/view/font/buffer/kitty state changes. Also moved row rendering into a dedicated builder, and AppleTerminalView now hands the absolute dirty range to Metal so blinks don’t trigger full rebuilds. * Debug FPS counter * Implemented an atlas eviction/resize policy with a max size of 2048 per side for now * Split atlas, a grayscale one for text, a color one for emojis. We determine this by scanning the resulting rendered image * Two modes for caching the row data caches. New property in TerminalView.metalBufferingMode perRowPersistent: caches MTLBuffer so we no longer rebuild big vertex buffers every frame; we now create persistent buffers only when a row is rebuilt and then reuse them across frames. MetalTerminalRenderer now stores RowDrawBuffers in rowCache and builds ImageDrawBuffer/vertex buffers once per dirty row (makeRowBuffers + makeStaticBuffer). The draw path iterates cached row buffers per category to preserve layer ordering while avoiding per‑frame buffer allocation churn. The cursor rendering still uses the per‑frame BufferPool, which is appropriate because it can change every frame. The second mode of operation is closer to what Ghostty does, krow‑level CPU caching + per‑frame GPU buffer sync, instead of persistent per‑row GPU buffers. Concretely: - Rebuilds only dirty rows into RowDrawData (CPU arrays). - Concatenates all visible row arrays into single per‑pass arrays (background, gray glyphs, color glyphs, decorations, images). - Uploads those arrays into one GPU buffer per pass using the per‑frame BufferPool. - Draws in the same layered order as before, but from those frame‑level buffers. * Implemented a Ghostty-style shaper cache and moved Metal row building onto shaped runs (CPU row model), so we reuse glyph/position shaping across redraws instead of recreating CTLine each time. MetalTerminalRenderer now has ShaperCache (keyed by font + text) and buildShapedSegments, and both background + text/ decoration paths consume shaped runs instead of per-run CTLine creation. Row rebuilds still use the existing dirty-row cache, but shaping is now reused across selection/attribute-only changes. * Implemented the Ghostty-style cell buffer model so backgrounds, text glyphs, and decorations are now emitted as per-cell structs and expanded to quads in the GPU; cursor/images still use the existing per-vertex paths. This reduces CPU-side vertex churn and aligns SwiftTerm’s renderer with the Ghostty approach. * Add Metal support for iOS * Close a small feature gap with the CoreGraphics version * Do not run async the data reception * Metal: Do not render every time a change, use a poor man's display-link, like we do with CoreGraphics * Rendering: Add additional underline styles * Fix build * Fix * Set throttling during resize * Add CustomBlockGlyph support for the GPU renderer * Document the GPU features
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
1 participant
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.
No description provided.