Merge pull request #100 from floooh/sgcompute

sokol-gfx compute support

build.zig

@@ -101,6 +101,7 @@ pub fn build(b: *Build) !void {
         "debugtext-userfont",
         "shapes",
         "vertexpull",
+        "instancing-compute",
     };
     inline for (examples) |example| {
         try buildExample(b, example, .{
@@ -113,7 +114,7 @@ pub fn build(b: *Build) !void {
     }
 
     // a manually invoked build step to recompile shaders via sokol-shdc
-    buildShaders(b, target);
+    buildShaders(b);
     // a manually invoked build step to build auto-docs
     buildDocs(b, target);
 }
@@ -508,22 +509,23 @@ fn emSdkSetupStep(b: *Build, emsdk: *Build.Dependency) !?*Build.Step.Run {
 
 // a separate step to compile shaders, expects the shader compiler in ../sokol-tools-bin/
 // TODO: install sokol-shdc via package manager
-fn buildShaders(b: *Build, target: Build.ResolvedTarget) void {
+fn buildShaders(b: *Build) void {
     const sokol_tools_bin_dir = "../sokol-tools-bin/bin/";
     const shaders_dir = "src/examples/shaders/";
     const shaders = .{
-        "bufferoffsets.glsl",
-        "cube.glsl",
-        "instancing.glsl",
-        "mrt.glsl",
-        "noninterleaved.glsl",
-        "offscreen.glsl",
-        "quad.glsl",
-        "shapes.glsl",
-        "texcube.glsl",
-        "blend.glsl",
-        "vertexpull.glsl",
-        "triangle.glsl",
+        .{ .src = "bufferoffsets.glsl", .needs_compute = false },
+        .{ .src = "cube.glsl", .needs_compute = false },
+        .{ .src = "instancing.glsl", .needs_compute = false },
+        .{ .src = "mrt.glsl", .needs_compute = false },
+        .{ .src = "noninterleaved.glsl", .needs_compute = false },
+        .{ .src = "offscreen.glsl", .needs_compute = false },
+        .{ .src = "quad.glsl", .needs_compute = false },
+        .{ .src = "shapes.glsl", .needs_compute = false },
+        .{ .src = "texcube.glsl", .needs_compute = false },
+        .{ .src = "blend.glsl", .needs_compute = false },
+        .{ .src = "triangle.glsl", .needs_compute = false },
+        .{ .src = "vertexpull.glsl", .needs_compute = true },
+        .{ .src = "instancing-compute.glsl", .needs_compute = true },
     };
     const optional_shdc: ?[:0]const u8 = comptime switch (builtin.os.tag) {
         .windows => "win32/sokol-shdc.exe",
@@ -537,15 +539,17 @@ fn buildShaders(b: *Build, target: Build.ResolvedTarget) void {
     }
     const shdc_path = sokol_tools_bin_dir ++ optional_shdc.?;
     const shdc_step = b.step("shaders", "Compile shaders (needs ../sokol-tools-bin)");
-    const glsl = if (isPlatform(target.result, .darwin)) "glsl410" else "glsl430";
-    const slang = glsl ++ ":metal_macos:hlsl5:glsl300es:wgsl";
     inline for (shaders) |shader| {
+        const slang = if (shader.needs_compute)
+            "glsl430:glsl310es:metal_macos:hlsl5:wgsl"
+        else
+            "glsl410:glsl300es:metal_macos:hlsl5:wgsl";
         const cmd = b.addSystemCommand(&.{
             shdc_path,
             "-i",
-            shaders_dir ++ shader,
+            shaders_dir ++ shader.src,
             "-o",
-            shaders_dir ++ shader ++ ".zig",
+            shaders_dir ++ shader.src ++ ".zig",
             "-l",
             slang,
             "-f",

src/examples/instancing-compute.zig

@@ -0,0 +1,198 @@
+//------------------------------------------------------------------------------
+//  instancing-compute.zig
+//
+//  Like instancing.zig, but update particle positions via compute shader.
+//------------------------------------------------------------------------------
+const sokol = @import("sokol");
+const slog = sokol.log;
+const sg = sokol.gfx;
+const sapp = sokol.app;
+const sglue = sokol.glue;
+const vec3 = @import("math.zig").Vec3;
+const mat4 = @import("math.zig").Mat4;
+const shd = @import("shaders/instancing-compute.glsl.zig");
+
+const max_particles: usize = 512 * 1024;
+const num_particles_emitted_per_frame: usize = 10;
+
+const state = struct {
+    var num_particles: usize = 0;
+    var ry: f32 = 0.0;
+    const compute = struct {
+        var pip: sg.Pipeline = .{};
+        var bind: sg.Bindings = .{};
+    };
+    const display = struct {
+        var pip: sg.Pipeline = .{};
+        var bind: sg.Bindings = .{};
+        var pass_action: sg.PassAction = .{};
+        const view: mat4 = mat4.lookat(.{ .x = 0, .y = 1.5, .z = 12.0 }, vec3.zero(), vec3.up());
+    };
+};
+
+export fn init() void {
+    sg.setup(.{
+        .environment = sglue.environment(),
+        .logger = .{ .func = slog.func },
+    });
+
+    // if compute shaders not supported, clear to red color and early out
+    if (!sg.queryFeatures().compute) {
+        state.display.pass_action.colors[0] = .{
+            .load_action = .CLEAR,
+            .clear_value = .{ .r = 1, .g = 0, .b = 0, .a = 1 },
+        };
+        return;
+    }
+
+    // regular clear color
+    state.display.pass_action.colors[0] = .{
+        .load_action = .CLEAR,
+        .clear_value = .{ .r = 0, .g = 0.1, .b = 0.2, .a = 1 },
+    };
+
+    // a zero-initialized storage buffer for the particle state
+    const sbuf = sg.makeBuffer(.{
+        .type = .STORAGEBUFFER,
+        .size = max_particles * @sizeOf(shd.Particle),
+        .label = "particle-buffer",
+    });
+    state.compute.bind.storage_buffers[shd.SBUF_cs_ssbo] = sbuf;
+    state.display.bind.storage_buffers[shd.SBUF_vs_ssbo] = sbuf;
+
+    // a compute shader and pipeline object for updating the particle state
+    state.compute.pip = sg.makePipeline(.{
+        .compute = true,
+        .shader = sg.makeShader(shd.updateShaderDesc(sg.queryBackend())),
+        .label = "update-pipeline",
+    });
+
+    // vertex and index buffer for particle geometry
+    const r = 0.05;
+    state.display.bind.vertex_buffers[0] = sg.makeBuffer(.{
+        .data = sg.asRange(&[_]f32{
+            0.0, -r,  0.0, 1.0, 0.0, 0.0, 1.0,
+            r,   0.0, r,   0.0, 1.0, 0.0, 1.0,
+            r,   0.0, -r,  0.0, 0.0, 1.0, 1.0,
+            -r,  0.0, -r,  1.0, 1.0, 0.0, 1.0,
+            -r,  0.0, r,   0.0, 1.0, 1.0, 1.0,
+            0.0, r,   0.0, 1.0, 0.0, 1.0, 1.0,
+        }),
+        .label = "geometry-vbuf",
+    });
+    state.display.bind.index_buffer = sg.makeBuffer(.{
+        .type = .INDEXBUFFER,
+        .data = sg.asRange(&[_]u16{
+            2, 1, 0, 3, 2, 0,
+            4, 3, 0, 1, 4, 0,
+            5, 1, 2, 5, 2, 3,
+            5, 3, 4, 5, 4, 1,
+        }),
+        .label = "geometry-ibuf",
+    });
+
+    // shader and pipeline for rendering the particles, this uses
+    // the compute-updated storage buffer to provide the particle positions
+    state.display.pip = sg.makePipeline(.{
+        .shader = sg.makeShader(shd.displayShaderDesc(sg.queryBackend())),
+        .layout = brk: {
+            var layout: sg.VertexLayoutState = .{};
+            layout.attrs[0] = .{ .format = .FLOAT3 };
+            layout.attrs[1] = .{ .format = .FLOAT4 };
+            break :brk layout;
+        },
+        .index_type = .UINT16,
+        .depth = .{
+            .compare = .LESS_EQUAL,
+            .write_enabled = true,
+        },
+        .cull_mode = .BACK,
+        .label = "render-pipeline",
+    });
+
+    // one-time init of particle velocities via a compute shader
+    const pip = sg.makePipeline(.{
+        .compute = true,
+        .shader = sg.makeShader(shd.initShaderDesc(sg.queryBackend())),
+    });
+    sg.beginPass(.{ .compute = true });
+    sg.applyPipeline(pip);
+    sg.applyBindings(state.compute.bind);
+    sg.dispatch(max_particles / 64, 1, 1);
+    sg.endPass();
+    sg.destroyPipeline(pip);
+}
+
+export fn frame() void {
+    if (!sg.queryFeatures().compute) {
+        drawFallback();
+        return;
+    }
+
+    state.num_particles += num_particles_emitted_per_frame;
+    if (state.num_particles > max_particles) {
+        state.num_particles = max_particles;
+    }
+    const dt: f32 = @floatCast(sapp.frameDuration());
+
+    // compute pass to update particle positions
+    const cs_params: shd.CsParams = .{
+        .dt = dt,
+        .num_particles = @intCast(state.num_particles),
+    };
+    sg.beginPass(.{ .compute = true, .label = "compute-pass" });
+    sg.applyPipeline(state.compute.pip);
+    sg.applyBindings(state.compute.bind);
+    sg.applyUniforms(shd.UB_cs_params, sg.asRange(&cs_params));
+    sg.dispatch(@intCast((state.num_particles + 63) / 64), 1, 1);
+    sg.endPass();
+
+    // render pass to render the particles via instancing, with the
+    // instance positions coming from the storage buffer
+    state.ry += 60.0 * dt;
+    const vs_params = computeVsParams(1.0, state.ry);
+    sg.beginPass(.{
+        .action = state.display.pass_action,
+        .swapchain = sglue.swapchain(),
+        .label = "render-pass",
+    });
+    sg.applyPipeline(state.display.pip);
+    sg.applyBindings(state.display.bind);
+    sg.applyUniforms(shd.UB_vs_params, sg.asRange(&vs_params));
+    sg.draw(0, 24, @intCast(state.num_particles));
+    sg.endPass();
+    sg.commit();
+}
+
+export fn cleanup() void {
+    sg.shutdown();
+}
+
+pub fn main() void {
+    sapp.run(.{
+        .init_cb = init,
+        .frame_cb = frame,
+        .cleanup_cb = cleanup,
+        .width = 800,
+        .height = 600,
+        .sample_count = 4,
+        .icon = .{ .sokol_default = true },
+        .window_title = "instancing-compute.zig",
+        .logger = .{ .func = slog.func },
+    });
+}
+
+fn computeVsParams(rx: f32, ry: f32) shd.VsParams {
+    const rxm = mat4.rotate(rx, .{ .x = 1.0, .y = 0.0, .z = 0.0 });
+    const rym = mat4.rotate(ry, .{ .x = 0.0, .y = 1.0, .z = 0.0 });
+    const model = mat4.mul(rxm, rym);
+    const aspect = sapp.widthf() / sapp.heightf();
+    const proj = mat4.persp(60.0, aspect, 0.01, 50.0);
+    return shd.VsParams{ .mvp = mat4.mul(mat4.mul(proj, state.display.view), model) };
+}
+
+fn drawFallback() void {
+    sg.beginPass(.{ .action = state.display.pass_action, .swapchain = sglue.swapchain(), .label = "render-pass" });
+    sg.endPass();
+    sg.commit();
+}

src/examples/shaders/blend.glsl.zig

@@ -7,7 +7,7 @@ const m = @import("../math.zig");
 //    Generated by sokol-shdc (https://github.com/floooh/sokol-tools)
 //
 //    Cmdline:
-//        sokol-shdc -i src/examples/shaders/blend.glsl -o src/examples/shaders/blend.glsl.zig -l glsl410:metal_macos:hlsl5:glsl300es:wgsl -f sokol_zig --reflection
+//        sokol-shdc -i src/examples/shaders/blend.glsl -o src/examples/shaders/blend.glsl.zig -l glsl410:glsl300es:metal_macos:hlsl5:wgsl -f sokol_zig --reflection
 //
 //    Overview:
 //    =========
@@ -1269,12 +1269,11 @@ pub fn bgUniformOffset(ub_name: []const u8, u_name: []const u8) ?usize {
     }
     return null;
 }
-pub fn bgUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniformDesc {
+pub fn bgUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniform {
     if (std.mem.eql(u8, ub_name, "bg_fs_params")) {
         if (std.mem.eql(u8, u_name, "tick")) {
-            var desc: sg.ShaderUniformDesc = .{};
+            var desc: sg.GlslShaderUniform = .{};
             desc.type = .FLOAT;
-            desc.offset = 0;
             desc.array_count = 0;
             desc.glsl_name = "tick";
             return desc;
@@ -1323,12 +1322,11 @@ pub fn quadUniformOffset(ub_name: []const u8, u_name: []const u8) ?usize {
     }
     return null;
 }
-pub fn quadUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniformDesc {
+pub fn quadUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniform {
     if (std.mem.eql(u8, ub_name, "quad_vs_params")) {
         if (std.mem.eql(u8, u_name, "mvp")) {
-            var desc: sg.ShaderUniformDesc = .{};
+            var desc: sg.GlslShaderUniform = .{};
             desc.type = .MAT4;
-            desc.offset = 0;
             desc.array_count = 0;
             desc.glsl_name = "mvp";
             return desc;

src/examples/shaders/bufferoffsets.glsl.zig

@@ -6,7 +6,7 @@ const std = @import("std");
 //    Generated by sokol-shdc (https://github.com/floooh/sokol-tools)
 //
 //    Cmdline:
-//        sokol-shdc -i src/examples/shaders/bufferoffsets.glsl -o src/examples/shaders/bufferoffsets.glsl.zig -l glsl410:metal_macos:hlsl5:glsl300es:wgsl -f sokol_zig --reflection
+//        sokol-shdc -i src/examples/shaders/bufferoffsets.glsl -o src/examples/shaders/bufferoffsets.glsl.zig -l glsl410:glsl300es:metal_macos:hlsl5:wgsl -f sokol_zig --reflection
 //
 //    Overview:
 //    =========
@@ -572,7 +572,7 @@ pub fn bufferoffsetsUniformOffset(ub_name: []const u8, u_name: []const u8) ?usiz
     _ = u_name;
     return null;
 }
-pub fn bufferoffsetsUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniformDesc {
+pub fn bufferoffsetsUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniform {
     _ = ub_name;
     _ = u_name;
     return null;

src/examples/shaders/cube.glsl.zig

@@ -7,7 +7,7 @@ const m = @import("../math.zig");
 //    Generated by sokol-shdc (https://github.com/floooh/sokol-tools)
 //
 //    Cmdline:
-//        sokol-shdc -i src/examples/shaders/cube.glsl -o src/examples/shaders/cube.glsl.zig -l glsl410:metal_macos:hlsl5:glsl300es:wgsl -f sokol_zig --reflection
+//        sokol-shdc -i src/examples/shaders/cube.glsl -o src/examples/shaders/cube.glsl.zig -l glsl410:glsl300es:metal_macos:hlsl5:wgsl -f sokol_zig --reflection
 //
 //    Overview:
 //    =========
@@ -662,12 +662,11 @@ pub fn cubeUniformOffset(ub_name: []const u8, u_name: []const u8) ?usize {
     }
     return null;
 }
-pub fn cubeUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniformDesc {
+pub fn cubeUniformDesc(ub_name: []const u8, u_name: []const u8) ?sg.GlslShaderUniform {
     if (std.mem.eql(u8, ub_name, "vs_params")) {
         if (std.mem.eql(u8, u_name, "mvp")) {
-            var desc: sg.ShaderUniformDesc = .{};
+            var desc: sg.GlslShaderUniform = .{};
             desc.type = .MAT4;
-            desc.offset = 0;
             desc.array_count = 0;
             desc.glsl_name = "mvp";
             return desc;