150 lines
7.4 KiB
GLSL
150 lines
7.4 KiB
GLSL
// Crafter.Graphics UI shader contract — shared by every standard UI compute
|
||
// shader and intended to be #included by user-authored shaders that want to
|
||
// dispatch alongside them. Layouts here are FROZEN: only additive changes
|
||
// (using the reserved `flags` bits or `_pad`).
|
||
|
||
#extension GL_EXT_shader_image_load_formatted : enable
|
||
#extension GL_EXT_shader_explicit_arithmetic_types_int16 : enable
|
||
#extension GL_EXT_descriptor_heap : enable
|
||
#extension GL_EXT_nonuniform_qualifier : enable
|
||
#extension GL_EXT_buffer_reference : enable
|
||
|
||
// ─── bindless heap declarations ─────────────────────────────────────────
|
||
// The same heap slot can be read as either uiImages[] (storage image) or
|
||
// uiTextures[] (sampled image) depending on which descriptor was written
|
||
// at that slot. Samplers live in a separate sampler heap.
|
||
layout(descriptor_heap) uniform image2D uiImages[];
|
||
layout(descriptor_heap) uniform texture2D uiTextures[];
|
||
layout(descriptor_heap) uniform sampler uiSamplers[];
|
||
|
||
// ─── push-constant header ───────────────────────────────────────────────
|
||
// Every UI dispatch's push-constant struct begins with this. User shaders
|
||
// MUST embed it as the first member so UIRenderer::FillHeader works.
|
||
struct UIDispatchHeader {
|
||
uint outImage; // heap slot of the swapchain image (this frame)
|
||
uint itemBuffer; // heap slot of the item SSBO
|
||
uvec2 surfaceSize; // window pixel size
|
||
vec4 clipRectPx; // (xy, wh) — every standard shader honors this
|
||
uint itemCount;
|
||
uint frameIdx;
|
||
uint flags; // user-defined feature bits
|
||
uint _pad; // reserved — keep zeroed
|
||
};
|
||
|
||
// ─── standard item structs ──────────────────────────────────────────────
|
||
// These match the C++ Crafter::QuadItem / CircleItem / ImageItem / GlyphItem
|
||
// byte-for-byte under std430.
|
||
struct QuadItem { vec4 rect; vec4 color; vec4 corners; vec4 outline; };
|
||
// rect = (x, y, w, h) in pixels
|
||
// color = filled body RGBA (premultiplied alpha not assumed)
|
||
// corners = per-corner radius in px (TL, TR, BR, BL); 0 = sharp
|
||
// outline = (thickness, R, G, B); thickness > 0 paints an outline of given color
|
||
|
||
struct CircleItem { vec4 centerRadius; vec4 color; vec4 outline; };
|
||
// centerRadius = (cx, cy, radius, _)
|
||
// outline.x = thickness (0 = filled), .yzw = outline RGB
|
||
|
||
struct ImageItem { vec4 rect; vec4 uv; vec4 tint; uvec4 slots; };
|
||
// rect = (x, y, w, h)
|
||
// uv = (u0, v0, u1, v1) into the source texture
|
||
// tint = multiplied with the sampled color
|
||
// slots = (textureHeapSlot, samplerHeapSlot, _, _)
|
||
|
||
struct GlyphItem { vec4 rect; vec4 uv; vec4 color; };
|
||
// rect = (x, y, w, h) on screen
|
||
// uv = (u0, v0, u1, v1) into the SDF font atlas
|
||
// color = glyph color (alpha modulated by SDF)
|
||
|
||
// ─── SSBO heap views ────────────────────────────────────────────────────
|
||
// One declaration per item type; each shader uses the one matching its
|
||
// dispatch. Indexed by hdr.itemBuffer.
|
||
layout(descriptor_heap, std430) readonly buffer UIQuadBuf { QuadItem items[]; } uiQuadHeap[];
|
||
layout(descriptor_heap, std430) readonly buffer UICircleBuf { CircleItem items[]; } uiCircleHeap[];
|
||
layout(descriptor_heap, std430) readonly buffer UIImageBuf { ImageItem items[]; } uiImageHeap[];
|
||
layout(descriptor_heap, std430) readonly buffer UIGlyphBuf { GlyphItem items[]; } uiGlyphHeap[];
|
||
|
||
|
||
// ──── Driver workaround: per-member SSBO load ────────────────────────────
|
||
// `UIItem it = itemHeap[idx].items[i]` emits an OpLoad of a composite type
|
||
// from a descriptor-heap'd SSBO, which crashes the GPU on the NVIDIA
|
||
// VK_EXT_descriptor_heap path (verified with a 1-float struct repro).
|
||
// Reading individual members works (each becomes OpAccessChain + scalar
|
||
// OpLoad). LoadItem reassembles the struct member-by-member into a local;
|
||
// the rest of the shader then operates on a regular local var.
|
||
|
||
ImageItem LoadImageItem(uint heap, uint i) {
|
||
ImageItem it;
|
||
it.rect = uiImageHeap[heap].items[i].rect;
|
||
it.uv = uiImageHeap[heap].items[i].uv;
|
||
it.tint = uiImageHeap[heap].items[i].tint;
|
||
it.slots = uiImageHeap[heap].items[i].slots;
|
||
return it;
|
||
}
|
||
|
||
GlyphItem LoadGlpyhtem(uint heap, uint i) {
|
||
GlyphItem it;
|
||
it.rect = uiGlyphHeap[heap].items[i].rect;
|
||
it.uv = uiGlyphHeap[heap].items[i].uv;
|
||
it.color = uiGlyphHeap[heap].items[i].color;
|
||
return it;
|
||
}
|
||
|
||
CircleItem LoadCircleItem(uint heap, uint i) {
|
||
CircleItem it;
|
||
it.centerRadius = uiCircleHeap[heap].items[i].centerRadius;
|
||
it.color = uiCircleHeap[heap].items[i].color;
|
||
it.outline = uiCircleHeap[heap].items[i].outline;
|
||
return it;
|
||
}
|
||
|
||
QuadItem LoadQuadItem(uint heap, uint i) {
|
||
QuadItem it;
|
||
it.rect = uiQuadHeap[heap].items[i].rect;
|
||
it.color = uiQuadHeap[heap].items[i].color;
|
||
it.corners = uiQuadHeap[heap].items[i].corners;
|
||
it.outline = uiQuadHeap[heap].items[i].outline;
|
||
return it;
|
||
}
|
||
|
||
// ─── pixel-tile dispatch model ─────────────────────────────────────────
|
||
// Standard shaders dispatch one workgroup per 8×8 screen tile. Each thread
|
||
// owns ONE pixel and iterates ALL items in order, accumulating the result
|
||
// in a local register, then stores once at the end. This guarantees correct
|
||
// z-order within a single dispatch (no inter-workgroup race on imageLoad/
|
||
// imageStore) and gives the user simple semantics: "items render in array
|
||
// order, later items overdraw earlier ones".
|
||
//
|
||
// Caller dispatches `(ceil(W/8), ceil(H/8), 1)` — no need to know the max
|
||
// item size.
|
||
|
||
// Returns the screen pixel and validates against the surface and clip rect.
|
||
bool uiResolveScreenPixel(UIDispatchHeader hdr, out ivec2 screenPx) {
|
||
uvec2 px = gl_GlobalInvocationID.xy;
|
||
if (px.x >= hdr.surfaceSize.x || px.y >= hdr.surfaceSize.y) return false;
|
||
if (float(px.x) < hdr.clipRectPx.x || float(px.y) < hdr.clipRectPx.y) return false;
|
||
if (float(px.x) >= hdr.clipRectPx.x + hdr.clipRectPx.z) return false;
|
||
if (float(px.y) >= hdr.clipRectPx.y + hdr.clipRectPx.w) return false;
|
||
screenPx = ivec2(px);
|
||
return true;
|
||
}
|
||
|
||
// Non-premultiplied "src over dst" blend. Both operands and result are
|
||
// straight-alpha vec4. Use this when iterating items in a loop with a local
|
||
// accumulator.
|
||
vec4 uiBlendOver(vec4 dst, vec4 src) {
|
||
float a = clamp(src.a, 0.0, 1.0);
|
||
vec3 outRGB = mix(dst.rgb, src.rgb, a);
|
||
float outA = a + dst.a * (1.0 - a);
|
||
return vec4(outRGB, outA);
|
||
}
|
||
|
||
// SDF for a rounded rect with per-corner radius. p is the point relative to
|
||
// the rect's center; halfSize is the rect half-extents; r is per-corner
|
||
// (TL, TR, BR, BL). Returns signed distance (negative inside).
|
||
float uiSdRoundRect(vec2 p, vec2 halfSize, vec4 r) {
|
||
// Pick the radius for the quadrant p is in.
|
||
r.xy = (p.x > 0.0) ? r.zy : r.wx; // pick TR/BR vs TL/BL
|
||
r.x = (p.y > 0.0) ? r.x : r.y;
|
||
vec2 q = abs(p) - halfSize + r.x;
|
||
return min(max(q.x, q.y), 0.0) + length(max(q, 0.0)) - r.x;
|
||
}
|