WGSL transcription of the paint interpreter. Takes the atlas and
palette as explicit storage-buffer pointer parameters (matching
the WGSL draw helper's calling convention):
fn hb_gpu_paint (renderCoord: vec2f, glyphLoc: u32,
foreground: vec4f,
hb_gpu_atlas: ptr<storage, array<vec4<i32>>, read>,
hb_gpu_palette: ptr<storage, array<vec4f>, read>) -> vec4f
hb_gpu_paint_shader_source() now returns real content for
HB_GPU_SHADER_LANG_WGSL fragment; HLSL still stubs to "".
util/gpu/web/hb-gpu-web-webgpu.cc:
- Concats the paint source alongside the draw source into the
WGSL shader module.
- Declares @group(0) @binding(2) for hb_gpu_palette.
- Allocates a 256 * vec4f storage buffer (zero-filled for now;
real palette upload is TODO), binds it at entry index 2.
- Fragment main switches from hb_gpu_draw to hb_gpu_paint.
util/gpu/web/gl-stub.h: adds glUniform4fv stub (demo-font.cc now
calls it for the GL palette path) and makes
glGetUniformLocation return -1 so the WebGPU build's
demo_font_upload_palette short-circuits.
Wasm sizes:
WebGL index.wasm : 443628 181597 gz
WebGPU webgpu.wasm : 461110 188045 gz
Non-color text should render through WebGPU. COLRv0 color glyphs
come up transparent until the cross-backend palette upload lands.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The draw-renderer fragment shaders defined utilities
(HB_GPU_UNITS_PER_EM, the hb_gpu_atlas uniform, hb_gpu_fetch) that
any future GPU renderer will also need. The draw-renderer vertex
shader likewise contained only the generic hb_gpu_dilate helper.
If a paint renderer were added today and a caller linked both
fragment sources together, hb_gpu_fetch would be defined twice.
Move these shared helpers into a new family of sources and expose
them via a new public entry point:
hb_gpu_shader_source (stage, lang);
The existing hb_gpu_draw_shader_source() now returns only the
draw-specific fragment code; it returns an empty string for the
vertex stage because all current vertex logic is shared. Callers
concatenate the shared source first, then the draw-specific
source, then their own main():
[#version, hb_gpu_shader_source (...), hb_gpu_draw_shader_source
(...), main]
Additions:
- src/hb-gpu-fragment.{glsl,msl,wgsl,hlsl} — shared fragment
- src/hb-gpu-vertex.{glsl,msl,wgsl,hlsl} — shared vertex,
renamed from hb-gpu-draw-vertex.*
- src/hb-gpu-shaders.cc — hb_gpu_shader_source() dispatcher
Removals:
- src/hb-gpu-draw-vertex.{glsl,msl,wgsl,hlsl} — moved to common
Shared sources are Behdad-only copyright; the Eric Lengyel / Slug
attribution stays on the draw-specific fragment sources where the
Slug algorithm actually lives.
Updates: test/api/test-gpu.cc, the Metal/D3D11/WebGPU/GL demo
shaders, the hb-gpu SECTION documentation, meson.build,
CMakeLists.txt, gen-gpu-shader-artifacts.py decl_map,
harfbuzz-world.cc, and docs/harfbuzz-sections.txt.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The previous commit renamed hb_gpu_darken to hb_gpu_draw_darken
for symmetry with hb_gpu_draw (formerly hb_gpu_render). But the
operation isn't draw-subsystem specific: it adjusts a coverage
value based on foreground brightness and ppem, which is just the
standard stem-darkening recipe. Name it after what it does.
Paint doesn't need its own variant — it can call this same
function on any alpha/coverage channel it produces.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Rename the public fragment-shader entry points provided by the
hb_gpu_draw_t renderer:
hb_gpu_render(...) -> hb_gpu_draw(...)
hb_gpu_darken(...) -> hb_gpu_draw_darken(...)
Also rename the internal helper:
_hb_gpu_render_single(...) -> _hb_gpu_draw_single(...)
A COLRv1 paint renderer is planned and will provide its own
fragment entry point (likely hb_gpu_paint returning a vec4); the
existing draw/coverage entry point gets a subsystem-qualified name
now so draw and paint can coexist.
Updates the .glsl/.msl/.wgsl/.hlsl sources, regenerated .hh
artifacts, the documentation block in hb-gpu.cc, and the util/gpu
demo shaders. No behavior change.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace the two separate getters
hb_gpu_shader_fragment_source (lang)
hb_gpu_shader_vertex_source (lang)
with a single
hb_gpu_draw_shader_source (stage, lang)
where stage is the new hb_gpu_shader_stage_t enum (VERTEX or
FRAGMENT). Parameter order is stage-first because "the vertex
shader in GLSL" reads more naturally than the reverse, and stage
is the primary axis (it determines which shader object the source
is compiled into).
The `_draw_` in the new name makes room for the incoming
hb_gpu_paint_shader_source() once the COLRv1 encoder lands —
that family will have its own (likely different) shader sources
per stage and language.
Updated the four in-tree callers: test/api/test-gpu.cc, and the
GPU demos under util/gpu/. ABI break on a 14.0-only API; the
gpu library is still young enough that this is acceptable.
Marked new API with XSince: REPLACEME.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Check navigator.gpu synchronously in main() before calling
the WebGPU C API. Previously the error was reported in an
async callback that raced with onRuntimeInitialized, which
hid the loading div before the error message could appear.
Also re-show the loading div and set window._webgpuError in
the adapter callback as a fallback.
Assisted-by: Claude <noreply@anthropic.com>
The GL renderer cached the vertex buffer upload by comparing
the data pointer and count. After clearing and refilling the
buffer with a new font's glyphs, std::vector reused the same
allocation, so the pointer and count matched — the upload was
skipped and the GPU drew with stale vertex data containing old
atlas offsets.
Fix by adding a generation counter to demo_buffer_t that
increments on clear(). The renderer compares generation and
count to decide whether to re-upload.
Assisted-by: Claude <noreply@anthropic.com>
Library function for stem darkening: pow(2, brightness-0.5)
scaled by ppem. Added to GLSL, MSL, WGSL, HLSL with consistent
brief docs. Demo shaders now call the library function.
Section docs in hb-gpu.cc updated with usage example.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace boolean light_on_dark with continuous brightness-driven
exponent. pow(2, b-0.5) gives sqrt(2)≈1.414 for white text
(thin) and 1/sqrt(2)≈0.707 for black text (fatten), perfectly
symmetric around 1.0 at 50% gray.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The WebGPU touch path was still using the older gesture state machine.
It handled 3->2 reseeding, but it did not track whether a pinch was
active and it had no explicit 2->1 single-finger pan handoff.
Carry over the proven transition rules from the WebGL path without
changing desktop input handling: track pinch-active and synthetic
single-pan state, preserve the 3->2 reseed, and continue pan after a
pinch when one finger remains. touchcancel now also clears those touch
states consistently.
Testing:
- source /home/behdad/emsdk/emsdk_env.sh >/dev/null &&
EM_CACHE=/tmp/emscripten-cache ./util/gpu/web/build.sh
- source /home/behdad/emsdk/emsdk_env.sh >/dev/null &&
EM_CACHE=/tmp/emscripten-cache ./util/gpu/web/build-webgpu.sh
Assisted-by: OpenAI Codex
Track the last gesture centroid and explicitly handle touchcancel
for both web gesture paths. This clears active gesture state
without turning browser-driven cancellation into a tap, fling, or
stuck drag.
Testing:
- sh -n util/gpu/web/build.sh
- sh -n util/gpu/web/build-webgpu.sh
- EM_CACHE=/tmp/emscripten-cache ./util/gpu/web/build.sh
- EM_CACHE=/tmp/emscripten-cache ./util/gpu/web/build-webgpu.sh
(blocked offline fetching emdawnwebgpu)
Assisted-by: OpenAI Codex
Properly handle finger count transitions with cooldown and
state reinitialization to prevent jumps in zoom/rotation.
Co-Authored-By: Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Prevents jumps during 3→2→1 finger transitions by dropping
one frame and reinitializing pinch state for the new finger
count.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Reinitialize pinch state (distance, angle, centroid) when
exiting 3-finger mode with 2 fingers remaining. Prevents
large zoom/rotate delta from stale pinch values.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add demo_view_request_redraw() and export as _web_request_redraw.
Shell calls it on visibilitychange, focus, pageshow, and
webglcontextrestored events to force a redraw after phone
lock/unlock or tab switch.
Co-Authored-By: Codex <noreply@openai.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Long-press uses _web_reset directly with 10px jitter threshold.
Double-tap-drag undoes first tap's animation toggle via
_web_toggle_animation, then preventDefault blocks GLFW from
seeing the second tap.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Single demo_view_pinch() handles all three components atomically:
- Pan from centroid movement
- Zoom around centroid (proven middle-drag math)
- Screen-space Z rotation (separate from 3D quaternion)
Pinch rotation uses a screen_angle field applied as a post-MVP
2D rotation, keeping it independent from the 3D trackball.
Zoom centering works correctly; rotation centering is approximate.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds demo_view_zoom_around() which scales while compensating
translation so the midpoint between fingers stays fixed.
Same maps-like behavior as the pinch rotation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds demo_view_rotate_z_around() which compensates translation
so the point between the two fingers stays fixed during rotation,
like maps apps.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two-finger pinch now does both zoom (distance change) and Z-axis
rotation (angle change) simultaneously. Adds demo_view_rotate_z()
for direct Z-axis rotation from touch input.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Now that the WGSL shader has _hb_gpu_curve_counts, wire up
the debug uniform and heatmap visualization. Same R/G/B output
as the GL demo.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Check canvas size each frame and reconfigure the WebGPU surface
when it changes. Fixes black screen when toggling dev tools or
resizing the browser window.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Same per-pixel coverage adjustment as the GL demo. Replaces
unused debug uniform with stem_darkening in the uniform buffer.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Map DOM button 1 (middle) to GLFW middle button so middle-click
reset and middle-drag zoom work.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The WGSL shader compilation fails with _hb_gpu_curve_counts.
Debug mode is available in desktop and WebGL demos.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Refactor library fragment shaders (GLSL, MSL, WGSL) to extract
_hb_gpu_decode_glyph and _hb_gpu_curve_counts helpers, used by
both hb_gpu_render and the debug visualization.
Debug mode (d key) shows per-pixel curve counts as a heatmap:
R = horizontal band curves, G = vertical band curves, B = text
coverage. Scale 0-8 curves maps to 0-1 intensity.
Also auto-generates demo shader headers via meson, and adds
set_debug to the renderer interface.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Three-finger drag on mobile now does trackball rotation, matching
right-drag on desktop. Uses centroid of three touch points.
Release with momentum triggers spin animation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Don't call demo_view_motion_func in on_mouseup — it zeros dx/dy
(same position as last mousemove), killing the velocity needed
for drag-release spin detection. Also remove debug printfs.
Track active_buttons to filter spurious document-level mouseup
events from UI controls while still catching drag releases
outside the canvas.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Refactor demo_view to work without GLFW (guarded by HB_GPU_NO_GLFW):
- Add demo_view_create_headless() for non-GLFW use
- Store fb/window dimensions in the view struct
- Use emscripten_get_now() for timing when on Emscripten
- Guard GLFW-specific calls (fullscreen, window close, cursor query)
Rewrite WebGPU entry point to use demo_view and demo_renderer_t
interface instead of duplicated view/input code. This gives the
WebGPU demo identical trackball rotation, drag-release animation,
keyboard shortcuts, and gamma cycling as the GL demos.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Show helpful error with browser support link when WebGPU is
unavailable (JS-level check before WASM loads)
- Suggest Linux Chrome flags for Vulkan/WebGPU
- Prevent onRuntimeInitialized from hiding the error message
- Set page title to WebGL/WebGPU variant
- Make adapter failure in C++ also show HTML error
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Display browser support link and Linux Chrome flags when
adapter or device creation fails.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Default to light mode matching GL demo
- Gamma cycles between 2.2 and none (no sRGB framebuffer)
- Gamma depends on dark_mode (1/2.2 for dark, 2.2 for light)
- Add console logging for key presses
- Blur backend dropdown after selection to prevent space capture
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a WebGPU variant of the web demo using Emscripten's
emdawnwebgpu port. Uses the WGSL shaders from the library API,
a storage buffer for the atlas (i16→i32 expansion on upload),
and Emscripten HTML5 APIs for input (no GLFW).
Also moves demo_shader_add_glyph_vertices to demo-shader.h as
static inline so demo-buffer.cc can be compiled without GL.
Build: bash util/gpu/web/build-webgpu.sh → out/webgpu.html
The existing WebGL build continues to work unchanged.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>