scratch pad
diff --git a/sparse_strips/vello_dev_macros/src/test.rs b/sparse_strips/vello_dev_macros/src/test.rs index 9c57280..4fca01c 100644 --- a/sparse_strips/vello_dev_macros/src/test.rs +++ b/sparse_strips/vello_dev_macros/src/test.rs
@@ -161,8 +161,7 @@ // These tests currently don't work with `vello_hybrid`. skip_hybrid |= { - input_fn_name_str.contains("compose") - || input_fn_name_str.contains("gradient") + input_fn_name_str.contains("gradient") || input_fn_name_str.contains("layer_multiple_properties") || input_fn_name_str.contains("mask") || input_fn_name_str.contains("mix")
diff --git a/sparse_strips/vello_hybrid/src/lib.rs b/sparse_strips/vello_hybrid/src/lib.rs index 306acc3..eb78ac6 100644 --- a/sparse_strips/vello_hybrid/src/lib.rs +++ b/sparse_strips/vello_hybrid/src/lib.rs
@@ -29,7 +29,8 @@ //! //! See the individual module documentation for more details on usage and implementation. -#![no_std] +// DO NOT COMMIT - for println debugging. +// #![no_std] extern crate alloc;
diff --git a/sparse_strips/vello_hybrid/src/render/common.rs b/sparse_strips/vello_hybrid/src/render/common.rs index bb89822..fb38efb 100644 --- a/sparse_strips/vello_hybrid/src/render/common.rs +++ b/sparse_strips/vello_hybrid/src/render/common.rs
@@ -60,37 +60,40 @@ const COLOR_SOURCE_BLEND: u32 = 2; const PAINT_TYPE_SOLID: u32 = 0; const PAINT_TYPE_IMAGE: u32 = 1; - impl fmt::Debug for GpuStrip { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - let color_source = (self.paint >> 30) & 0x3; // Changed to 2 bits for 3 source types + let color_source = (self.paint >> 30) & 0x3; // Changed from 31 to 30 (2 bits) + let paint_type = (self.paint >> 28) & 0x3; // Changed from 29 to 28 let mut debug_struct = f.debug_struct("GpuStrip"); + // Add tile information based on y coordinate + let tile_index = self.y / 4 as u16; + let tile_local_y = self.y % 4 as u16; + debug_struct .field("x", &self.x) - .field("y", &self.y) + .field("y", &format!("{} (tile={}, local_y={})", self.y, tile_index, tile_local_y)) .field("width", &self.width) .field("dense_width", &self.dense_width) .field("col_idx", &self.col_idx); let paint_info = match color_source { - COLOR_SOURCE_PAYLOAD => { - let paint_type = (self.paint >> 28) & 0x3; // Adjusted bit position - if paint_type == PAINT_TYPE_SOLID { - format!("Solid(color_source=payload)") - } else if paint_type == PAINT_TYPE_IMAGE { - let paint_tex_id = self.paint & 0x0FFFFFFF; // Adjusted mask + 0 => { // COLOR_SOURCE_PAYLOAD + if paint_type == 0 { // PAINT_TYPE_SOLID + "Solid(color_source=payload)".to_string() + } else if paint_type == 1 { // PAINT_TYPE_IMAGE + let paint_tex_id = self.paint & 0x0FFFFFFF; format!("Image(color_source=payload, texture_id={})", paint_tex_id) } else { format!("Unknown(color_source=payload, type={})", paint_type) } } - COLOR_SOURCE_SLOT => { + 1 => { // COLOR_SOURCE_SLOT let opacity = self.paint & 0xFF; format!("Slot(color_source=slot, opacity={})", opacity) } - COLOR_SOURCE_BLEND => { + 2 => { // COLOR_SOURCE_BLEND let dest_slot = (self.paint >> 16) & 0x3FFF; let mix = (self.paint >> 8) & 0xFF; let compose = self.paint & 0xFF; @@ -101,17 +104,16 @@ debug_struct.field("paint", &paint_info); - // Decode payload based on paint configuration + // Enhanced payload info let payload_info = match color_source { - COLOR_SOURCE_PAYLOAD => { - let paint_type = (self.paint >> 28) & 0x3; - if paint_type == PAINT_TYPE_SOLID { + 0 => { // COLOR_SOURCE_PAYLOAD + if paint_type == 0 { // PAINT_TYPE_SOLID let r = (self.payload >> 0) & 0xFF; let g = (self.payload >> 8) & 0xFF; let b = (self.payload >> 16) & 0xFF; let a = (self.payload >> 24) & 0xFF; format!("Color(r={}, g={}, b={}, a={})", r, g, b, a) - } else if paint_type == PAINT_TYPE_IMAGE { + } else if paint_type == 1 { // PAINT_TYPE_IMAGE let x = self.payload & 0xFFFF; let y = self.payload >> 16; format!("ImageCoords(x={}, y={})", x, y) @@ -119,11 +121,10 @@ format!("Unknown(raw=0x{:08x})", self.payload) } } - COLOR_SOURCE_SLOT => { - format!("SlotIndex({})", self.payload) - } - COLOR_SOURCE_BLEND => { - format!("SourceSlot({})", self.payload) + 1 | 2 => { // COLOR_SOURCE_SLOT or COLOR_SOURCE_BLEND + let slot = self.payload; + let slot_y = slot as u16 * 4 as u16; + format!("SourceSlot({}) [maps to y={}]", slot, slot_y) } _ => format!("Unknown(raw=0x{:08x})", self.payload) };
diff --git a/sparse_strips/vello_hybrid/src/schedule.rs b/sparse_strips/vello_hybrid/src/schedule.rs index 67ed833..29f33e0 100644 --- a/sparse_strips/vello_hybrid/src/schedule.rs +++ b/sparse_strips/vello_hybrid/src/schedule.rs
@@ -216,15 +216,21 @@ /// The total number of slots in each slot texture. total_slots: usize, /// The slots that are free to use in each slot texture. - free: [Vec<usize>; 2], + free: [Vec<SlotOccupation>; 2], /// Slots that require clearing before subsequent draws for each slot texture. - clear: [Vec<u32>; 2], + clear: [Vec<SlotOccupation>; 2], /// Rounds are enqueued on push clip commands and dequeued on flush. rounds_queue: VecDeque<Round>, /// State for a single wide tile. tile_state: TileState, } +#[derive(Debug, Clone, Copy, PartialEq, PartialOrd)] +struct SlotOccupation { + slot_idx: usize, + texture: u8, +} + /// A "round" is a coarse scheduling quantum. /// /// It represents draws in up to three render targets; two for intermediate @@ -234,7 +240,7 @@ /// Draw calls scheduled into the two slot textures (0, 1) and the final target (2). draws: [Draw; 3], /// Slots that will be freed after drawing into the two slot textures [0, 1]. - free: [Vec<usize>; 2], + free: [Vec<SlotOccupation>; 2], } /// State for a single wide tile. @@ -245,8 +251,8 @@ #[derive(Clone, Copy, Debug)] struct TileEl { - slot_ix: usize, - slot_prime_ix: Option<usize>, + slot_ix: SlotOccupation, + slot_prime_ix: Option<SlotOccupation>, round: usize, opacity: f32, } @@ -256,8 +262,20 @@ impl Scheduler { pub(crate) fn new(total_slots: usize) -> Self { - let free0: Vec<_> = (0..total_slots).collect(); - let free1 = free0.clone(); + let free0: Vec<_> = (0..total_slots) + .map(|idx| SlotOccupation { + slot_idx: idx, + texture: 0, + }) + .collect(); + let free1 = free0 + .clone() + .into_iter() + .map(|SlotOccupation { slot_idx, .. }| SlotOccupation { + slot_idx, + texture: 1, + }) + .collect(); let free = [free0, free1]; let clear = [Vec::new(), Vec::new()]; Self { @@ -270,6 +288,7 @@ } } + /// This looks ahead to see if another PushBuf is incoming. If so - we will be "blending". fn should_allocate_prime(cmds: &[Cmd], current_idx: usize) -> bool { for cmd in &cmds[current_idx + 1..] { match cmd { @@ -295,7 +314,7 @@ // Left to right, top to bottom iteration over wide tiles. for wide_tile_row in 0..wide_tiles_per_col { for wide_tile_col in 0..wide_tiles_per_row { - let wide_tile = dbg!(scene.wide.get(wide_tile_col, wide_tile_row)); + let wide_tile = scene.wide.get(wide_tile_col, wide_tile_row); let wide_tile_x = wide_tile_col * WideTile::WIDTH; let wide_tile_y = wide_tile_row * Tile::HEIGHT; self.do_tile( @@ -321,8 +340,20 @@ #[cfg(debug_assertions)] { for i in 0..self.total_slots { - debug_assert!(self.free[0].contains(&i), "free[0] is missing slot {i}"); - debug_assert!(self.free[1].contains(&i), "free[1] is missing slot {i}"); + debug_assert!( + self.free[0].contains(&SlotOccupation { + slot_idx: i, + texture: 0 + }), + "free[0] is missing slot {i}" + ); + debug_assert!( + self.free[1].contains(&SlotOccupation { + slot_idx: i, + texture: 1 + }), + "free[1] is missing slot {i}" + ); } } debug_assert!(self.rounds_queue.is_empty(), "rounds_queue is not empty"); @@ -353,7 +384,14 @@ LoadOp::Clear } else { // Some slots need to be preserved, so only clear the dirty slots. - renderer.clear_slots(i, self.clear[i].as_slice()); + renderer.clear_slots( + i, + self.clear[i] + .iter() + .map(|slot| slot.slot_idx as u32) + .collect::<Vec<u32>>() + .as_slice(), + ); self.clear[i].clear(); LoadOp::Load } @@ -396,7 +434,7 @@ // Sentinel `TileEl` to indicate the end of the stack where we draw all // commands to the final target. state.stack.push(TileEl { - slot_ix: usize::MAX, + slot_ix: SlotOccupation { slot_idx: usize::MAX, texture: 0 }, slot_prime_ix: None, round: self.round, opacity: 1., @@ -417,7 +455,7 @@ }); } } - for (cmd_idx, cmd) in tile.cmds.iter().enumerate() { + for (cmd_idx, cmd) in dbg!(tile).cmds.iter().enumerate() { // Note: this starts at 1 (for the final target) let clip_depth = state.stack.len(); match cmd { @@ -442,14 +480,14 @@ if effective_depth == 1 { scene_strip_y } else { - prime_slot as u16 * Tile::HEIGHT + prime_slot.slot_idx as u16 * Tile::HEIGHT } } else { // Write to main slot if clip_depth == 1 { scene_strip_y } else { - el.slot_ix as u16 * Tile::HEIGHT + el.slot_ix.slot_idx as u16 * Tile::HEIGHT } }; @@ -495,14 +533,13 @@ if effective_depth == 1 { scene_strip_y } else { - prime_slot as u16 * Tile::HEIGHT + prime_slot.slot_idx as u16 * Tile::HEIGHT } } else { - // Write to main slot - if clip_depth == 1 { + if effective_depth == 1 { scene_strip_y } else { - el.slot_ix as u16 * Tile::HEIGHT + el.slot_ix.slot_idx as u16 * Tile::HEIGHT } }; @@ -524,26 +561,26 @@ } Cmd::PushBuf => { if Self::should_allocate_prime(&tile.cmds, cmd_idx) { - let dest_ix = 1 - clip_depth % 2; // Destination texture + let dest_ix = (1 - clip_depth) % 2; // Destination texture (prime slot) let temp_ix = clip_depth % 2; // Temporary texture - while self.free[dest_ix].is_empty() || self.free[temp_ix].is_empty() { - if self.rounds_queue.is_empty() { - return Err(RenderError::SlotsExhausted); - } - self.flush(renderer); - } + // while self.free[dest_ix].is_empty() || self.free[temp_ix].is_empty() { + // if self.rounds_queue.is_empty() { + // return Err(RenderError::SlotsExhausted); + // } + // self.flush(renderer); + // } // Allocate both let dest_slot = self.free[dest_ix].pop().unwrap(); let temp_slot = self.free[temp_ix].pop().unwrap(); - self.clear[dest_ix].push(dest_slot as u32); - self.clear[temp_ix].push(temp_slot as u32); + self.clear[dest_ix].push(dest_slot); + self.clear[temp_ix].push(temp_slot); state.stack.push(TileEl { slot_ix: dest_slot, // Destination (prime) - blend result. - slot_prime_ix: Some(temp_slot), // Temporary accumulation + slot_prime_ix: Some(temp_slot), // Temporary accumulation of fills before blending... round: self.round, opacity: 1., }); @@ -556,7 +593,7 @@ self.flush(renderer); } let slot_ix = self.free[ix].pop().unwrap(); - self.clear[ix].push(slot_ix as u32); + self.clear[ix].push(slot_ix); state.stack.push(TileEl { slot_ix, @@ -582,12 +619,13 @@ "round must be in queue" ); - let main_tex_ix = 1 - clip_depth % 2; - self.rounds_queue[round - self.round].free[main_tex_ix] + + + self.rounds_queue[round - self.round].free[popped_buffer.slot_ix.texture as usize] .push(popped_buffer.slot_ix); + if let Some(slot_prime_ix) = popped_buffer.slot_prime_ix { - let prime_tex_ix = 1 - main_tex_ix; // Opposite texture - self.rounds_queue[round - self.round].free[prime_tex_ix] + self.rounds_queue[round - self.round].free[slot_prime_ix.texture as usize] .push(slot_prime_ix); } } @@ -604,7 +642,7 @@ } else { ( clip_fill.x as u16, - clip_target.slot_ix as u16 * Tile::HEIGHT, + clip_target.slot_ix.slot_idx as u16 * Tile::HEIGHT, ) }; // Opacity packed into the first 8 bits – pack full opacity (0xFF). @@ -615,7 +653,7 @@ width: clip_fill.width as u16, dense_width: 0, col_idx: 0, - payload: clip_source.slot_ix as u32, + payload: clip_source.slot_ix.slot_idx as u32, paint, }); } @@ -632,7 +670,7 @@ } else { ( clip_alpha_fill.x as u16, - clip_target.slot_ix as u16 * Tile::HEIGHT, + clip_target.slot_ix.slot_idx as u16 * Tile::HEIGHT, ) }; // Opacity packed into the first 8 bits – pack full opacity (0xFF). @@ -645,7 +683,7 @@ col_idx: (clip_alpha_fill.alpha_idx / usize::from(Tile::HEIGHT)) .try_into() .expect("Sparse strips are bound to u32 range"), - payload: clip_source.slot_ix as u32, + payload: clip_source.slot_ix.slot_idx as u32, paint, }); } @@ -654,6 +692,7 @@ } Cmd::Blend(mode) => { let blend_source = state.stack.last().unwrap(); + let blend_target = &state.stack[state.stack.len() - 2]; // Determine where the actual content is for source @@ -675,16 +714,18 @@ let (x, y) = if clip_depth <= 2 { (wide_tile_x, wide_tile_y) } else { - (0, blend_target.slot_ix as u16 * Tile::HEIGHT) // Write to destination + (0, blend_target.slot_ix.slot_idx as u16 * Tile::HEIGHT) // Write to destination }; + // debug_assert_ne!(dest_temp_slot, source_content_slot, "cmd_idx: {cmd_idx}"); + // Encode blend operation let paint = (COLOR_SOURCE_BLEND << 30) - | ((dest_temp_slot as u32 & 0x3FFF) << 16) // Where dest content is + | ((dest_temp_slot.slot_idx as u32 & 0x3FFF) << 16) | ((mode.mix as u32) << 8) | (mode.compose as u32); - let payload = source_content_slot as u32; // Where source content is + let payload = source_content_slot.slot_idx as u32; // Where source content is draw.0.push(GpuStrip { x, @@ -706,7 +747,7 @@ let (x, y) = if clip_depth <= 2 { (wide_tile_x, wide_tile_y) } else { - (0, blend_target.slot_ix as u16 * Tile::HEIGHT) + (0, blend_target.slot_ix.slot_idx as u16 * Tile::HEIGHT) }; let opacity_u8 = (blend_source.opacity * 255.0) as u32; @@ -718,7 +759,7 @@ width: WideTile::WIDTH, dense_width: 0, col_idx: 0, - payload: source_content_slot as u32, // Use actual content location + payload: source_content_slot.slot_idx as u32, // Use actual content location paint, }); } @@ -727,6 +768,14 @@ } } + if state.stack.len() > 1 { + // Had nested operations + // Force a new round for the next tile + while !self.rounds_queue.is_empty() { + self.flush(renderer); + } + } + Ok(()) } @@ -756,7 +805,7 @@ ImageSource::OpaqueId(_) => { let paint_packed = (COLOR_SOURCE_PAYLOAD << 30) | (PAINT_TYPE_IMAGE << 28) - | (paint_tex_id & 0x1FFFFFFF); + | (paint_tex_id & 0x0FFFFFFF); let scene_strip_xy = ((scene_strip_y as u32) << 16) | (scene_strip_x as u32); (scene_strip_xy, paint_packed)
diff --git a/sparse_strips/vello_sparse_shaders/shaders/render_strips.wgsl b/sparse_strips/vello_sparse_shaders/shaders/render_strips.wgsl index 1a87d70..fbde354 100644 --- a/sparse_strips/vello_sparse_shaders/shaders/render_strips.wgsl +++ b/sparse_strips/vello_sparse_shaders/shaders/render_strips.wgsl
@@ -42,7 +42,22 @@ // Blend modes const MIX_NORMAL: u32 = 0u; + +// Composite modes - must match +const COMPOSE_CLEAR: u32 = 0u; +const COMPOSE_COPY: u32 = 1u; +const COMPOSE_DEST: u32 = 2u; const COMPOSE_SRC_OVER: u32 = 3u; +const COMPOSE_DEST_OVER: u32 = 4u; +const COMPOSE_SRC_IN: u32 = 5u; +const COMPOSE_DEST_IN: u32 = 6u; +const COMPOSE_SRC_OUT: u32 = 7u; +const COMPOSE_DEST_OUT: u32 = 8u; +const COMPOSE_SRC_ATOP: u32 = 9u; +const COMPOSE_DEST_ATOP: u32 = 10u; +const COMPOSE_XOR: u32 = 11u; +const COMPOSE_PLUS: u32 = 12u; +const COMPOSE_PLUS_LIGHTER: u32 = 13u; struct Config { // Width of the rendering target @@ -150,7 +165,7 @@ // NDC ranges from -1 to 1, with (0,0) at the center of the viewport let ndc_x = pix_x * 2.0 / f32(config.width) - 1.0; let ndc_y = 1.0 - pix_y * 2.0 / f32(config.height); - let paint_type = (instance.paint >> 29u) & 0x3u; + let paint_type = (instance.paint >> 28u) & 0x3u; if paint_type == PAINT_TYPE_IMAGE { let paint_tex_id = instance.paint & 0x1FFFFFFF; @@ -226,7 +241,7 @@ if paint_type == PAINT_TYPE_SOLID { final_color = alpha * unpack4x8unorm(in.payload); } else if paint_type == PAINT_TYPE_IMAGE { - let paint_tex_id = in.paint & 0x1FFFFFFF; + let paint_tex_id = in.paint & 0x0FFFFFFF; let encoded_image = unpack_encoded_image(paint_tex_id); let image_offset = encoded_image.image_offset; let image_size = encoded_image.image_size; @@ -287,16 +302,70 @@ let dest_y = (u32(in.position.y) & 3u) + dest_slot * config.strip_height; let dest_color = textureLoad(clip_input_texture, vec2(clip_x, dest_y), 0); - // Can if or switch over the compose modes.... - // if compose_mode == COMPOSE_SRC_OVER { - // // SrcOver: result = src + dest * (1 - src.a) - // } + switch compose_mode { + case COMPOSE_CLEAR: { + // Clear: result = 0 + final_color = vec4<f32>(0.0, 0.0, 0.0, 0.0); + } + case COMPOSE_COPY: { + // Copy: result = src + final_color = src_color; + } + case COMPOSE_DEST: { + // Dest: result = dest + final_color = dest_color; + } + case COMPOSE_SRC_OVER: { + // SrcOver: result = src + dest * (1 - src.a) + final_color = src_color + dest_color * (1.0 - src_color.a); + } + case COMPOSE_DEST_OVER: { + // DestOver: result = dest + src * (1 - dest.a) + final_color = dest_color + src_color * (1.0 - dest_color.a); + } + case COMPOSE_SRC_IN: { + // SrcIn: result = src * dest.a + final_color = src_color * dest_color.a; + } + case COMPOSE_DEST_IN: { + // DestIn: result = dest * src.a + final_color = dest_color * src_color.a; + } + case COMPOSE_SRC_OUT: { + // SrcOut: result = src * (1 - dest.a) + final_color = src_color * (1.0 - dest_color.a); + } + case COMPOSE_DEST_OUT: { + // DestOut: result = dest * (1 - src.a) + final_color = dest_color * (1.0 - src_color.a); + } + case COMPOSE_SRC_ATOP: { + // SrcAtop: result = src * dest.a + dest * (1 - src.a) + final_color = src_color * dest_color.a + dest_color * (1.0 - src_color.a); + } + case COMPOSE_DEST_ATOP: { + // DestAtop: result = dest * src.a + src * (1 - dest.a) + final_color = dest_color * src_color.a + src_color * (1.0 - dest_color.a); + } + case COMPOSE_XOR: { + // Xor: result = src * (1 - dest.a) + dest * (1 - src.a) + final_color = src_color * (1.0 - dest_color.a) + dest_color * (1.0 - src_color.a); + } + case COMPOSE_PLUS: { + // Plus: result = min(src + dest, 1) + final_color = clamp(src_color + dest_color, vec4<f32>(0.0), vec4<f32>(1.0)); + } + case COMPOSE_PLUS_LIGHTER: { + // PlusLighter: result = src + dest (unclamped) + final_color = src_color + dest_color; + } + default: { + // Fallback to SrcOver + final_color = src_color + dest_color * (1.0 - src_color.a); + } + } - // Hard coded SrcOver... - final_color = src_color + dest_color * (1.0 - src_color.a); final_color = alpha * final_color; - - } return final_color;
diff --git a/sparse_strips/vello_sparse_tests/log b/sparse_strips/vello_sparse_tests/log new file mode 100644 index 0000000..24a665f --- /dev/null +++ b/sparse_strips/vello_sparse_tests/log
@@ -0,0 +1,654 @@ + +running 6 tests +test compose::compose_wide_tile_nested_cpu_u8_scalar ... FAILED +test compose::compose_wide_tile_nested_cpu_u8_neon ... FAILED +test compose::compose_wide_tile_nested_cpu_multithreaded ... FAILED +test compose::compose_wide_tile_nested_cpu_f32_scalar ... FAILED +test compose::compose_wide_tile_nested_cpu_f32_neon ... FAILED +test compose::compose_wide_tile_nested_hybrid ... FAILED + +failures: + +---- compose::compose_wide_tile_nested_cpu_u8_scalar stdout ---- + +thread 'compose::compose_wide_tile_nested_cpu_u8_scalar' panicked at sparse_strips/vello_sparse_tests/tests/util.rs:314:9: +test didnt match reference image + +---- compose::compose_wide_tile_nested_cpu_u8_neon stdout ---- + +thread 'compose::compose_wide_tile_nested_cpu_u8_neon' panicked at sparse_strips/vello_sparse_tests/tests/util.rs:314:9: +test didnt match reference image +note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace + +---- compose::compose_wide_tile_nested_cpu_multithreaded stdout ---- + +thread 'compose::compose_wide_tile_nested_cpu_multithreaded' panicked at sparse_strips/vello_sparse_tests/tests/util.rs:314:9: +test didnt match reference image + +---- compose::compose_wide_tile_nested_cpu_f32_scalar stdout ---- + +thread 'compose::compose_wide_tile_nested_cpu_f32_scalar' panicked at sparse_strips/vello_sparse_tests/tests/util.rs:314:9: +test didnt match reference image + +---- compose::compose_wide_tile_nested_cpu_f32_neon stdout ---- + +thread 'compose::compose_wide_tile_nested_cpu_f32_neon' panicked at sparse_strips/vello_sparse_tests/tests/util.rs:314:9: +test didnt match reference image + +---- compose::compose_wide_tile_nested_hybrid stdout ---- +do_scene - tile: {width: 1, height: 2 } +self.tile_state len: TileState { stack: [] } +[sparse_strips/vello_hybrid/src/schedule.rs:458:31] tile = WideTile { + x: 0, + y: 0, + bg: PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 0, + b: 0, + a: 0, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 0.0, + 0.0, + 0.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + cmds: [ + AlphaFill( + CmdAlphaFill { + x: 0, + width: 4, + alpha_idx: 0, + thread_idx: 0, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 255, + g: 255, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 1.0, + 1.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Fill( + CmdFill { + x: 4, + width: 252, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 255, + g: 255, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 1.0, + 1.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + PushBuf, + AlphaFill( + CmdAlphaFill { + x: 0, + width: 4, + alpha_idx: 32, + thread_idx: 0, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 0, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 0.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Fill( + CmdFill { + x: 4, + width: 252, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 0, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 0.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + PushBuf, + AlphaFill( + CmdAlphaFill { + x: 48, + width: 4, + alpha_idx: 64, + thread_idx: 0, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 255, + b: 0, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 1.0, + 0.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Fill( + CmdFill { + x: 52, + width: 204, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 255, + b: 0, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 1.0, + 0.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Blend( + BlendMode { + mix: Normal, + compose: SrcOver, + }, + ), + PopBuf, + Blend( + BlendMode { + mix: Normal, + compose: SrcOver, + }, + ), + PopBuf, + ], + n_zero_clip: 0, + n_clip: 0, + n_bufs: 0, +} +[sparse_strips/vello_hybrid/src/schedule.rs:458:31] tile = WideTile { + x: 0, + y: 4, + bg: PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 0, + b: 0, + a: 0, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 0.0, + 0.0, + 0.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + cmds: [ + AlphaFill( + CmdAlphaFill { + x: 0, + width: 4, + alpha_idx: 16, + thread_idx: 0, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 255, + g: 255, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 1.0, + 1.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Fill( + CmdFill { + x: 4, + width: 252, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 255, + g: 255, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 1.0, + 1.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + PushBuf, + AlphaFill( + CmdAlphaFill { + x: 0, + width: 4, + alpha_idx: 48, + thread_idx: 0, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 0, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 0.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Fill( + CmdFill { + x: 4, + width: 252, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 0, + b: 255, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 0.0, + 1.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + PushBuf, + AlphaFill( + CmdAlphaFill { + x: 48, + width: 4, + alpha_idx: 80, + thread_idx: 0, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 255, + b: 0, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 1.0, + 0.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Fill( + CmdFill { + x: 52, + width: 204, + paint: Solid( + PremulColor { + premul_u8: PremulRgba8 { + r: 0, + g: 255, + b: 0, + a: 255, + }, + premul_f32: PremulColor { + components: [ + 0.0, + 1.0, + 0.0, + 1.0, + ], + cs: PhantomData<color::colorspace::Srgb>, + }, + }, + ), + blend_mode: None, + }, + ), + Blend( + BlendMode { + mix: Normal, + compose: SrcOver, + }, + ), + PopBuf, + Blend( + BlendMode { + mix: Normal, + compose: SrcOver, + }, + ), + PopBuf, + ], + n_zero_clip: 0, + n_clip: 0, + n_bufs: 0, +} +Drawing a round: +[sparse_strips/vello_hybrid/src/schedule.rs:370:9] &round = Round { + draws: [ + Draw( + [ + GpuStrip { + x: 0, + y: "8188 (tile=2047, local_y=0)", + width: 4, + dense_width: 4, + col_idx: 8, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=0, b=255, a=255)", + }, + GpuStrip { + x: 4, + y: "8188 (tile=2047, local_y=0)", + width: 252, + dense_width: 0, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=0, b=255, a=255)", + }, + GpuStrip { + x: 48, + y: "8184 (tile=2046, local_y=0)", + width: 4, + dense_width: 4, + col_idx: 16, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=255, b=0, a=255)", + }, + GpuStrip { + x: 52, + y: "8184 (tile=2046, local_y=0)", + width: 204, + dense_width: 0, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=255, b=0, a=255)", + }, + GpuStrip { + x: 0, + y: "8184 (tile=2046, local_y=0)", + width: 4, + dense_width: 4, + col_idx: 12, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=0, b=255, a=255)", + }, + GpuStrip { + x: 4, + y: "8184 (tile=2046, local_y=0)", + width: 252, + dense_width: 0, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=0, b=255, a=255)", + }, + GpuStrip { + x: 48, + y: "8176 (tile=2044, local_y=0)", + width: 4, + dense_width: 4, + col_idx: 20, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=255, b=0, a=255)", + }, + GpuStrip { + x: 52, + y: "8176 (tile=2044, local_y=0)", + width: 204, + dense_width: 0, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=0, g=255, b=0, a=255)", + }, + ], + ), + Draw( + [ + GpuStrip { + x: 0, + y: "8188 (tile=2047, local_y=0)", + width: 256, + dense_width: 0, + col_idx: 0, + paint: "Blend(dest_slot=2047, mix=0, compose=3)", + payload: "SourceSlot(2046) [maps to y=8184]", + }, + GpuStrip { + x: 0, + y: "8180 (tile=2045, local_y=0)", + width: 256, + dense_width: 0, + col_idx: 0, + paint: "Blend(dest_slot=2046, mix=0, compose=3)", + payload: "SourceSlot(2044) [maps to y=8176]", + }, + ], + ), + Draw( + [ + GpuStrip { + x: 0, + y: "0 (tile=0, local_y=0)", + width: 4, + dense_width: 4, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=255, g=255, b=255, a=255)", + }, + GpuStrip { + x: 4, + y: "0 (tile=0, local_y=0)", + width: 252, + dense_width: 0, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=255, g=255, b=255, a=255)", + }, + GpuStrip { + x: 0, + y: "0 (tile=0, local_y=0)", + width: 256, + dense_width: 0, + col_idx: 0, + paint: "Slot(color_source=slot, opacity=255)", + payload: "SourceSlot(2047) [maps to y=8188]", + }, + GpuStrip { + x: 0, + y: "4 (tile=1, local_y=0)", + width: 4, + dense_width: 4, + col_idx: 4, + paint: "Solid(color_source=payload)", + payload: "Color(r=255, g=255, b=255, a=255)", + }, + GpuStrip { + x: 4, + y: "4 (tile=1, local_y=0)", + width: 252, + dense_width: 0, + col_idx: 0, + paint: "Solid(color_source=payload)", + payload: "Color(r=255, g=255, b=255, a=255)", + }, + GpuStrip { + x: 0, + y: "4 (tile=1, local_y=0)", + width: 256, + dense_width: 0, + col_idx: 0, + paint: "Slot(color_source=slot, opacity=255)", + payload: "SourceSlot(2046) [maps to y=8184]", + }, + ], + ), + ], + free: [ + [ + SlotOccupation { + slot_idx: 2046, + texture: 0, + }, + SlotOccupation { + slot_idx: 2047, + texture: 0, + }, + SlotOccupation { + slot_idx: 2044, + texture: 0, + }, + SlotOccupation { + slot_idx: 2045, + texture: 0, + }, + ], + [ + SlotOccupation { + slot_idx: 2047, + texture: 1, + }, + SlotOccupation { + slot_idx: 2046, + texture: 1, + }, + ], + ], +} + +thread 'compose::compose_wide_tile_nested_hybrid' panicked at sparse_strips/vello_sparse_tests/tests/util.rs:314:9: +test didnt match reference image + + +failures: + compose::compose_wide_tile_nested_cpu_f32_neon + compose::compose_wide_tile_nested_cpu_f32_scalar + compose::compose_wide_tile_nested_cpu_multithreaded + compose::compose_wide_tile_nested_cpu_u8_neon + compose::compose_wide_tile_nested_cpu_u8_scalar + compose::compose_wide_tile_nested_hybrid + +test result: FAILED. 0 passed; 6 failed; 0 ignored; 0 measured; 1584 filtered out; finished in 0.14s +
diff --git a/sparse_strips/vello_sparse_tests/snapshots/compose_wide_tile_nested.png b/sparse_strips/vello_sparse_tests/snapshots/compose_wide_tile_nested.png index 00f5e7c..35b9641 100644 --- a/sparse_strips/vello_sparse_tests/snapshots/compose_wide_tile_nested.png +++ b/sparse_strips/vello_sparse_tests/snapshots/compose_wide_tile_nested.png
@@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ab11bef3cf105e5117d115622f9463fbcd7752643e0e9e2b4f8add92c79c0846 -size 101 +oid sha256:a020f1465bb609cb18276a01e2ebcdaedb3bf674d048a527859302cb5deabead +size 99
diff --git a/sparse_strips/vello_sparse_tests/tests/compose.rs b/sparse_strips/vello_sparse_tests/tests/compose.rs index 0c46eed..0e15490 100644 --- a/sparse_strips/vello_sparse_tests/tests/compose.rs +++ b/sparse_strips/vello_sparse_tests/tests/compose.rs
@@ -26,41 +26,41 @@ #[vello_test(height = 8)] fn compose_wide_tile_nested(ctx: &mut impl Renderer) { const WIDTH: f64 = 100.0; - const HEIGHT: f64 = 4.0; - const OFFSET: f64 = 20.0; - + const HEIGHT: f64 = 8.0; + const OFFSET: f64 = 50.0; + // Pure colors: max values only - let red = Color::from_rgb8(255, 0, 0); - let green = Color::from_rgb8(0, 255, 0); let blue = Color::from_rgb8(0, 0, 255); - let cyan = Color::from_rgb8(0, 255, 255); // Green + Blue maxed - + let green = Color::from_rgb8(0, 255, 0); + // let blue = Color::from_rgb8(0, 0, 255); + // let cyan = Color::from_rgb8(0, 255, 255); // Green + Blue maxed + ctx.push_blend_layer(BlendMode::new(Mix::Normal, Compose::SrcOver)); - - // Draw base red rectangle - ctx.set_paint(red); + + // Draw base blue rectangle + ctx.set_paint(blue); ctx.fill_rect(&Rect::new(0.0, 0.0, WIDTH, HEIGHT)); - + // Start nesting - green with Xor ctx.push_blend_layer(BlendMode::new(Mix::Normal, Compose::SrcOver)); ctx.set_paint(green); ctx.fill_rect(&Rect::new(OFFSET, 0.0, OFFSET + WIDTH, HEIGHT)); - - // Nest blue with Plus - ctx.push_blend_layer(BlendMode::new(Mix::Normal, Compose::SrcOver)); - ctx.set_paint(blue); - ctx.fill_rect(&Rect::new(OFFSET * 2.0, 0.0, OFFSET * 2.0 + WIDTH, HEIGHT)); - - // Nest cyan with Multiply - ctx.push_blend_layer(BlendMode::new(Mix::Normal, Compose::SrcOver)); - ctx.set_paint(cyan); - ctx.fill_rect(&Rect::new(OFFSET * 3.0, 0.0, OFFSET * 3.0 + WIDTH, HEIGHT)); - ctx.pop_layer(); // cyan - - ctx.pop_layer(); // blue - + + // Nest blue with Plus + // ctx.push_blend_layer(BlendMode::new(Mix::Normal, Compose::SrcOver)); + // ctx.set_paint(blue); + // ctx.fill_rect(&Rect::new(OFFSET * 2.0, 0.0, OFFSET * 2.0 + WIDTH, HEIGHT)); + + // // Nest cyan with Multiply + // ctx.push_blend_layer(BlendMode::new(Mix::Normal, Compose::SrcOver)); + // ctx.set_paint(cyan); + // ctx.fill_rect(&Rect::new(OFFSET * 3.0, 0.0, OFFSET * 3.0 + WIDTH, HEIGHT)); + // ctx.pop_layer(); // cyan + + // ctx.pop_layer(); // blue + ctx.pop_layer(); // green - + ctx.pop_layer(); // base }
