perf(graphics): comprehensive framebuffer rendering optimizations #116
+421
−86
Conversation
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
Three performance improvements for framebuffer rendering: 1. Character-level dirty marking: Reduced mark_dirty() calls from ~80 per character (one per pixel) to 1 per character. Added write_pixel_no_mark() for batch operations with single mark_dirty_rect() call at the end. 2. Multiple dirty rectangles: Replace single bounding box with 4 separate dirty rects. Uses intelligent merging: - Merge if rects overlap or are within 32 pixels - When full, merge closest pair to make room - Result: sparse updates (cursor + text) flush independently 3. Scroll optimization: Remove redundant flush_if_dirty() before scroll - pending dirty regions scroll with content, only the cleared bottom line needs flushing. Expected improvements: - Character rendering: 80x fewer dirty region operations - Sparse updates: cursor blink + text no longer flush giant bbox - Scroll: ~1.92MB less buffer copying per scroll operation Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Additional optimizations for framebuffer rendering: 1. Add mark_rect_dirty() to DirtyRegionTracker - marks entire rectangle in one operation instead of per-row calls. For a 16-pixel character, this is 16 merge operations → 1. 2. Add mark_region_dirty_rect() to DoubleBufferedFrameBuffer as wrapper for the new efficient method. 3. Add fast path in flush() for full-width dirty rectangles - if x_start == 0 && x_end == stride, copy entire vertical block in one memcpy instead of per-row copies. 4. Update mark_dirty_rect() in logger.rs to use the new batch method. Expected improvements: - Character dirty marking: 16x fewer operations - Scroll/clear flush: Nx fewer memcpy calls (N = number of rows) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Critical performance fixes for graphics rendering: 1. Create write_bytes_to_shell_internal() that doesn't flush - Render thread uses this non-flushing version - Render thread handles flush once after all batches - Before: render_batch -> write_bytes_to_shell [FLUSH] -> flush_framebuffer [FLUSH AGAIN] - After: render_batch -> write_bytes_to_shell_internal [NO FLUSH] -> flush_framebuffer [FLUSH ONCE] 2. Batch cursor operations in write_bytes_to_shell() - Hide cursor ONCE at start of batch - Write all bytes - Show cursor ONCE at end of batch - Before: 80 chars = 240 cursor ops (hide/draw/show per char) - After: 80 chars = 2 cursor ops (hide once, show once) 3. Update run-interactive.sh to kill existing containers - Prevents port conflict errors - Cleaner startup experience Expected improvement: 10-50x for batch text output. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
CRITICAL FIX: The Canvas::set_pixel implementation was calling mark_region_dirty for EVERY SINGLE PIXEL, completely defeating all our batching optimizations. For a character with ~100 drawn pixels: - Before: 100 dirty region operations (with merge logic each time) - After: 1 dirty region operation for the entire glyph Changes: 1. Remove mark_region_dirty from set_pixel in logger.rs - Pixel data still written, just not marked dirty per-pixel 2. Add mark_dirty_region at end of fill_rect in primitives.rs - Marks entire filled rectangle once 3. Add mark_dirty_region at end of draw_glyph in primitives.rs - Marks entire glyph bounding box once This was the root cause of the slow rendering - O(n) dirty marking per character where n = pixels in glyph. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Change flush_full() to flush() in switch_terminal - only flush dirty regions instead of copying entire 8MB framebuffer - Reduce LOG_BUFFER_SIZE from 200 to 50 lines for faster tab switching - Skip offscreen log lines during replay - only render visible rows, eliminating expensive scroll operations entirely Previously switching to Logs tab would render all 200 lines causing ~160 scroll operations. Now we calculate visible rows and skip lines that would scroll off, reducing to zero scroll operations. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
ryanbreen
changed the title
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.
Summary
Key Changes
Dirty Region Tracking (double_buffer.rs)
mark_rect_dirty()for batch dirty markingPer-Pixel Fix (logger.rs, primitives.rs)
mark_region_dirtyfrom Canvasset_pixelfill_rectanddraw_glyphCascading Flush Elimination (terminal_manager.rs, render_queue.rs)
write_bytes_to_shell_internal()non-flushing API for render threadTerminal Switching (terminal_manager.rs)
flush_full()toflush()- only dirty regions, not entire 8MB bufferPerformance Impact
Test plan
--features interactive./docker/qemu/run-interactive.shand verify:🤖 Generated with Claude Code