Fix two r.adf regressions: copper-chunky border COLOR00 + hi-res window-edge flush#154
Merged
Merged
Conversation
The horizontal display-window flip-flop's closed intervals are repainted with the border pixel after compositing. That repaint walked only the control (BPLCON/DDF) segments to bound each fill run and sampled COLOR00 once at the run's start. When the copper rewrites COLOR00 across the border to draw a colour bar -- the copper-chunky technique behind demo logo banners -- the write lands mid-run, so the whole border run was painted with the frame-start COLOR00 and the colour change was dropped, clipping the bar's edge back to the previous control boundary. Denise applies COLOR00 changes per pixel in the border exactly as it does inside the display window, so end each repaint run at the next control OR palette (colour) segment. The border now follows COLOR00 write-for-write. Regression example: the "binary" demo (r.adf) red logo bar rendered short on the left and grew stray colour fragments as the copper's border COLOR00 writes were sampled at the wrong x.
Standard hi-res ($81 DIWSTRT / $3C DDFSTRT) put its first fetched sample two framebuffer pixels inside the display window instead of flush at the window edge, so a 40-word row overran the window's closing edge and its rightmost fetched pixel was masked as border. On real hardware (and vAmiga) the picture fills the window flush to both edges. The 2H-196 comparator move set DIW_HSTART_FB0 to 0x62 and bumped both fetch references +1 "in lockstep". Lo-res stayed flush, but hi-res has one framebuffer pixel per sample where lo-res has two, so the same +1 bump shifted hi-res twice as far and left it two pixels into the window. Give hi-res the same 0x81 reference as lo-res so both place sample 0 at the window edge; late-DDF hi-res (its border shows the pre-window samples) follows the corrected reference too. Regression example: the AmigaDOS window's right vertical border on KS1.3 (the "binary" demo, r.adf) reappears, matching vAmiga. vAmigaTS Denise, Agnus/DDF and BPLCON1 buckets improve with no regressions; the standard hi-res BPLCON0/modes references reach 0.000%.
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.
Fixes two rendering regressions in the "binary" demo (r.adf, KS1.3), both
in the Denise display path. Verified pixel-for-pixel against vAmiga.
1. Copper-chunky border follows per-pixel COLOR00 (
0c85c8e)The horizontal display-window flip-flop's closed intervals are repainted
with the border pixel after compositing. That repaint bounded each fill run
on the control (BPLCON/DDF) segments only and sampled COLOR00 once at the
run's start. When the copper rewrites COLOR00 across the border to draw a
colour bar (the copper-chunky technique behind the demo's red logo banner),
the write lands mid-run, so the whole border run was painted with the
frame-start COLOR00 and the colour change was dropped -- clipping the bar's
left edge and leaving stray colour fragments.
Denise applies COLOR00 per pixel in the border exactly as inside the
display window, so the repaint now ends each run at the next control or
palette (colour) segment. Regression examples: the red banner rendered
short on the left and grew red noise lines as the copper's border COLOR00
writes were sampled at the wrong x.
2. Hi-res bitplane content flush at the 2H-196 window edge (
1e476cb)Standard hi-res (
$81DIWSTRT /$3CDDFSTRT) placed its first fetchedsample two framebuffer pixels inside the display window instead of flush
at the edge, so a 40-word row overran the window's closing edge and its
rightmost fetched pixel was masked as border. The 2H-196 comparator move
bumped both fetch references +1 "in lockstep"; lo-res stayed flush, but
hi-res has one framebuffer pixel per sample where lo-res has two, so the
same +1 shifted hi-res twice as far. Give hi-res the same
0x81referenceas lo-res so both place sample 0 at the window edge.
Regression example: the AmigaDOS window's right vertical border reappears,
matching vAmiga (fb 62/700).
Validation
78 improved, 0 regressed -- Denise 39 (standard hi-res
BPLCON0/modesreach 0.000%), Agnus/DDF 32 (late-DDF hi-res), BPLCON1 7,DIW unchanged.
old "2 hi-res px into the window" placement updated to the flush values,
plus a new regression test for the closed-DIW border following COLOR00.