| /* |
| * Copyright 2026 Rive |
| * |
| * GM test / Phase 0 witness for the post-review fix sweep |
| * (`/Users/luigi/Projects/plan/Ore/Post-Review Fix Plan.md` Phase 0 |
| * + Phase 3). |
| * |
| * Locks in the WebGPU contract that `dynamicOffsets[i]` is consumed in |
| * BindGroupLayout-entry order (= ascending `@binding` per RFC §3.6 BGL |
| * sort), NOT in the caller's `desc.ubos[]` order. |
| * |
| * Two dynamic UBOs at @group(0) @binding(0) and @group(0) @binding(1). |
| * One buffer with two 256-byte aligned ranges: |
| * |
| * range A @ offset 0 = (1, 0, 0, 0) — should be read by u_a |
| * range B @ offset 256 = (0, 1, 0, 0) — should be read by u_b |
| * |
| * The GM constructs `BindGroupDesc::ubos` in REVERSE slot order |
| * (`{slot:1}` first, `{slot:0}` second) and passes |
| * `dynamicOffsets = [0, 256]` in BGL/ascending-slot order. The shader |
| * emits `vec4f(u_a.r, u_b.g, 0, 1)`: |
| * |
| * - Correct (BGL order): a=range A, b=range B → (1, 1, 0, 1) yellow. |
| * - Wrong (insertion order, today's Metal `m_mtlBuffers` walk): a=B, |
| * b=A → (0, 0, 0, 1) black. |
| * - Wrong (Dawn validation fail post-fix on WGPU if the BGL emits |
| * dynamic entries out of ascending order): pipeline rejection. |
| * |
| * Expected today (PRE-FIX): |
| * - Metal: BLACK (insertion-order walk is wrong). |
| * - Vulkan / D3D11 / D3D12 / GL / WGPU: YELLOW (slot-ascending walk |
| * happens to coincide with BGL order). |
| * |
| * Expected POST-FIX (Phase 3 of the fix plan, Metal sort by slot at |
| * BindGroup construction): |
| * - Every backend renders yellow. |
| */ |
| |
| #include "gm.hpp" |
| #include "gmutils.hpp" |
| #include "ore_gm_helper.hpp" |
| #if defined(ORE_BACKEND_METAL) || defined(ORE_BACKEND_GL) || \ |
| defined(ORE_BACKEND_VK) || defined(ORE_BACKEND_WGPU) || \ |
| defined(ORE_BACKEND_D3D11) || defined(ORE_BACKEND_D3D12) |
| #include "rive/renderer/render_canvas.hpp" |
| #include "rive/renderer/ore/ore_buffer.hpp" |
| #include "rive/renderer/ore/ore_bind_group.hpp" |
| #include "rive/renderer/ore/ore_shader_module.hpp" |
| #include "rive/renderer/ore/ore_pipeline.hpp" |
| #include "rive/renderer/ore/ore_render_pass.hpp" |
| #endif |
| |
| using namespace rivegm; |
| using namespace rive; |
| using namespace rive::gpu; |
| #if defined(ORE_BACKEND_METAL) || defined(ORE_BACKEND_GL) || \ |
| defined(ORE_BACKEND_VK) || defined(ORE_BACKEND_WGPU) || \ |
| defined(ORE_BACKEND_D3D11) || defined(ORE_BACKEND_D3D12) |
| using namespace rive::ore; |
| #endif |
| |
| #if defined(ORE_BACKEND_METAL) || defined(ORE_BACKEND_GL) || \ |
| defined(ORE_BACKEND_VK) || defined(ORE_BACKEND_WGPU) || \ |
| defined(ORE_BACKEND_D3D11) || defined(ORE_BACKEND_D3D12) |
| #define ORE_BINDING_DYNAMIC_UBO_ACTIVE |
| #endif |
| |
| // 256-byte stride is the strictest dynamic-offset alignment requirement |
| // across our backends (D3D12 needs 256, D3D11.1 needs 256 enforced by |
| // firstConstant×16, Vulkan minUniformBufferOffsetAlignment is at most 256 |
| // on every adapter we ship to, WebGPU spec mandates 256). |
| static constexpr uint32_t kDynamicAlign = 256; |
| |
| class OreBindingDynamicUBOGM : public GM |
| { |
| public: |
| OreBindingDynamicUBOGM() : GM(128, 128) {} |
| |
| ColorInt clearColor() const override { return 0xff000000; } |
| |
| void onDraw(rive::Renderer* originalRenderer) override |
| { |
| auto renderContext = TestingWindow::Get()->renderContext(); |
| if (!renderContext || !m_ore.ensureContext(renderContext)) |
| return; |
| |
| #ifdef ORE_BINDING_DYNAMIC_UBO_ACTIVE |
| auto& ctx = *m_ore.oreContext; |
| |
| // ── One buffer with two 256-byte ranges. Range A holds the |
| // values u_a should read; range B the values u_b should read. |
| struct Uniforms |
| { |
| float r, g, b, a; |
| }; |
| std::vector<uint8_t> bufBytes(kDynamicAlign * 2, 0); |
| const Uniforms kRangeA = {1.0f, 0.0f, 0.0f, 0.0f}; // u_a sees red |
| const Uniforms kRangeB = {0.0f, 1.0f, 0.0f, 0.0f}; // u_b sees green |
| memcpy(bufBytes.data() + 0, &kRangeA, sizeof(kRangeA)); |
| memcpy(bufBytes.data() + kDynamicAlign, &kRangeB, sizeof(kRangeB)); |
| |
| BufferDesc bufDesc{}; |
| bufDesc.usage = BufferUsage::uniform; |
| bufDesc.size = static_cast<uint32_t>(bufBytes.size()); |
| bufDesc.data = bufBytes.data(); |
| bufDesc.label = "dynamic_ubo_buffer"; |
| auto buf = ctx.makeBuffer(bufDesc); |
| |
| // ── Shader from the RSTB ── |
| auto shader = ore_gm::loadShader(ctx, ore_gm::kDynamicUBOWitness); |
| if (!shader.vsModule) |
| return; |
| |
| // Dynamic UBO bindings declared on the layout (RFC §3.6 / Dawn). |
| const uint32_t dynBindings[] = {0, 1}; |
| auto layout0 = ore_gm::makeLayoutFromShader(ctx, |
| shader.vsModule.get(), |
| 0, |
| dynBindings, |
| 2); |
| BindGroupLayout* layouts[] = {layout0.get()}; |
| |
| PipelineDesc pipeDesc{}; |
| pipeDesc.vertexModule = shader.vsModule.get(); |
| pipeDesc.fragmentModule = shader.psModule.get(); |
| pipeDesc.vertexEntryPoint = shader.vsEntryPoint; |
| pipeDesc.fragmentEntryPoint = shader.fsEntryPoint; |
| pipeDesc.vertexBufferCount = 0; |
| pipeDesc.topology = PrimitiveTopology::triangleList; |
| pipeDesc.colorTargets[0].format = TextureFormat::rgba8unorm; |
| pipeDesc.colorCount = 1; |
| pipeDesc.depthStencil.depthCompare = CompareFunction::always; |
| pipeDesc.depthStencil.depthWriteEnabled = false; |
| pipeDesc.bindGroupLayouts = layouts; |
| pipeDesc.bindGroupLayoutCount = 1; |
| pipeDesc.label = "ore_binding_dynamic_ubo_pipeline"; |
| auto pipeline = ctx.makePipeline(pipeDesc); |
| if (!pipeline) |
| { |
| fprintf(stderr, |
| "[ore_binding_dynamic_ubo] pipeline creation failed: %s\n", |
| ctx.lastError().c_str()); |
| return; |
| } |
| |
| // ── BindGroup with UBOs in REVERSE slot order. The runtime |
| // must ignore this insertion order and walk the bind group in |
| // BGL/ascending-slot order when consuming `dynamicOffsets[]`. |
| BindGroupDesc::UBOEntry uboEntries[2]{}; |
| uboEntries[0].slot = 1; // u_b |
| uboEntries[0].buffer = buf.get(); |
| uboEntries[0].offset = 0; |
| uboEntries[0].size = sizeof(Uniforms); |
| uboEntries[1].slot = 0; // u_a |
| uboEntries[1].buffer = buf.get(); |
| uboEntries[1].offset = 0; |
| uboEntries[1].size = sizeof(Uniforms); |
| |
| BindGroupDesc bgDesc{}; |
| bgDesc.layout = layout0.get(); |
| bgDesc.ubos = uboEntries; |
| bgDesc.uboCount = 2; |
| bgDesc.label = "dynamic_ubo_bg"; |
| auto bg = ctx.makeBindGroup(bgDesc); |
| if (!bg) |
| return; |
| |
| // ── Dynamic offsets in BGL/ascending-slot order: |
| // index 0 → u_a (slot 0) → range A at offset 0 |
| // index 1 → u_b (slot 1) → range B at offset kDynamicAlign |
| const uint32_t dynamicOffsets[2] = {0u, kDynamicAlign}; |
| |
| // ── Render into a RenderCanvas ── |
| auto canvas = renderContext->makeRenderCanvas(128, 128); |
| if (!canvas) |
| return; |
| |
| auto canvasTarget = ctx.wrapCanvasTexture(canvas.get()); |
| if (!canvasTarget) |
| return; |
| |
| m_ore.beginFrame(); |
| |
| ColorAttachment ca{}; |
| ca.view = canvasTarget.get(); |
| ca.loadOp = LoadOp::clear; |
| ca.storeOp = StoreOp::store; |
| ca.clearColor = {0, 0, 0, 1}; |
| |
| RenderPassDesc rpDesc{}; |
| rpDesc.colorAttachments[0] = ca; |
| rpDesc.colorCount = 1; |
| rpDesc.label = "ore_binding_dynamic_ubo_pass"; |
| |
| auto pass = ctx.beginRenderPass(rpDesc); |
| pass.setPipeline(pipeline.get()); |
| pass.setBindGroup(0, bg.get(), dynamicOffsets, 2); |
| pass.setViewport(0, 0, 128, 128); |
| pass.draw(3); // fullscreen triangle |
| pass.finish(); |
| |
| m_ore.endFrame(); |
| ore_gm::invalidateGLStateAfterOre(renderContext); |
| |
| originalRenderer->save(); |
| originalRenderer->drawImage(canvas->renderImage(), |
| {.filter = ImageFilter::nearest}, |
| BlendMode::srcOver, |
| 1); |
| originalRenderer->restore(); |
| #endif // ORE_BINDING_DYNAMIC_UBO_ACTIVE |
| } |
| |
| private: |
| ore_gm::OreGMContext m_ore; |
| }; |
| |
| GMREGISTER(ore_binding_dynamic_ubo, return new OreBindingDynamicUBOGM) |