@aardworx/wombat.rendering 0.9.19 → 0.9.21
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/runtime/derivedModes/gpuDispatcher.d.ts +4 -0
- package/dist/runtime/derivedModes/gpuDispatcher.d.ts.map +1 -1
- package/dist/runtime/derivedModes/gpuDispatcher.js +25 -6
- package/dist/runtime/derivedModes/gpuDispatcher.js.map +1 -1
- package/dist/runtime/derivedModes/gpuKernel.d.ts +1 -1
- package/dist/runtime/derivedModes/gpuKernel.d.ts.map +1 -1
- package/dist/runtime/derivedModes/gpuKernel.js +3 -3
- package/package.json +1 -1
- package/src/runtime/derivedModes/gpuDispatcher.ts +23 -6
- package/src/runtime/derivedModes/gpuKernel.ts +3 -3
|
@@ -18,6 +18,10 @@ export declare class GpuDerivedModesScene {
|
|
|
18
18
|
/** Last-known declared (for kernel uniform). Common case: all RO
|
|
19
19
|
* rules share the same declared (= the SG scope's value). */
|
|
20
20
|
private declared;
|
|
21
|
+
/** True while the staging buffer's mapAsync is in flight. We must
|
|
22
|
+
* not enqueue a new copyBufferToBuffer into it until it's been
|
|
23
|
+
* unmapped (mapAsync resolved + unmap called). */
|
|
24
|
+
private readbackInFlight;
|
|
21
25
|
/** Per-RO last-known output. CPU mirror, populated by finish(). */
|
|
22
26
|
readonly lastOutput: Uint32Array;
|
|
23
27
|
constructor(device: GPUDevice);
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"gpuDispatcher.d.ts","sourceRoot":"","sources":["../../../src/runtime/derivedModes/gpuDispatcher.ts"],"names":[],"mappings":"AAwBA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,2BAA2B,CAAC;AAc1D,qBAAa,oBAAoB;IAC/B,QAAQ,CAAC,MAAM,EAAE,SAAS,CAAC;IAE3B,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA+B;IACvD,OAAO,CAAC,SAAS,CAAK;IACtB,OAAO,CAAC,KAAK,CAAS;IAGtB,OAAO,CAAC,QAAQ,CAA4B;IAC5C,OAAO,CAAC,SAAS,CAA2B;IAC5C,OAAO,CAAC,UAAU,CAA0B;IAC5C,OAAO,CAAC,SAAS,CAA2B;IAC5C,OAAO,CAAC,QAAQ,CAAqC;IACrD,OAAO,CAAC,SAAS,CAA8B;IAC/C,OAAO,CAAC,MAAM,CAAuC;IACrD,yEAAyE;IACzE,OAAO,CAAC,QAAQ,CAAK;IACrB,uEAAuE;IACvE,OAAO,CAAC,QAAQ,CAA0B;IAC1C;kEAC8D;IAC9D,OAAO,CAAC,QAAQ,CAA4B;
|
|
1
|
+
{"version":3,"file":"gpuDispatcher.d.ts","sourceRoot":"","sources":["../../../src/runtime/derivedModes/gpuDispatcher.ts"],"names":[],"mappings":"AAwBA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,2BAA2B,CAAC;AAc1D,qBAAa,oBAAoB;IAC/B,QAAQ,CAAC,MAAM,EAAE,SAAS,CAAC;IAE3B,OAAO,CAAC,QAAQ,CAAC,OAAO,CAA+B;IACvD,OAAO,CAAC,SAAS,CAAK;IACtB,OAAO,CAAC,KAAK,CAAS;IAGtB,OAAO,CAAC,QAAQ,CAA4B;IAC5C,OAAO,CAAC,SAAS,CAA2B;IAC5C,OAAO,CAAC,UAAU,CAA0B;IAC5C,OAAO,CAAC,SAAS,CAA2B;IAC5C,OAAO,CAAC,QAAQ,CAAqC;IACrD,OAAO,CAAC,SAAS,CAA8B;IAC/C,OAAO,CAAC,MAAM,CAAuC;IACrD,yEAAyE;IACzE,OAAO,CAAC,QAAQ,CAAK;IACrB,uEAAuE;IACvE,OAAO,CAAC,QAAQ,CAA0B;IAC1C;kEAC8D;IAC9D,OAAO,CAAC,QAAQ,CAA4B;IAC5C;;uDAEmD;IACnD,OAAO,CAAC,gBAAgB,CAAS;IAEjC,mEAAmE;IACnE,QAAQ,CAAC,UAAU,EAAE,WAAW,CAAC;gBAErB,MAAM,EAAE,SAAS;IAK7B,IAAI,UAAU,IAAI,MAAM,CAA2B;IAEnD,UAAU,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ,GAAG,IAAI;IAOtE,YAAY,CAAC,MAAM,EAAE,MAAM,GAAG,IAAI;IAOlC,SAAS,IAAI,IAAI;IACjB,YAAY,IAAI,OAAO;IAEvB,OAAO,CAAC,eAAe;IAmDvB,oEAAoE;IACpE,OAAO,CAAC,YAAY;IAWpB;;;;OAIG;IACH,QAAQ,CAAC,QAAQ,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,EAAE,iBAAiB,GAAG,MAAM;IAqB7E;;;;OAIG;IACG,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAc3C,OAAO,IAAI,IAAI;CAMhB"}
|
|
@@ -45,6 +45,10 @@ export class GpuDerivedModesScene {
|
|
|
45
45
|
/** Last-known declared (for kernel uniform). Common case: all RO
|
|
46
46
|
* rules share the same declared (= the SG scope's value). */
|
|
47
47
|
declared = CULL_TO_U32.back;
|
|
48
|
+
/** True while the staging buffer's mapAsync is in flight. We must
|
|
49
|
+
* not enqueue a new copyBufferToBuffer into it until it's been
|
|
50
|
+
* unmapped (mapAsync resolved + unmap called). */
|
|
51
|
+
readbackInFlight = false;
|
|
48
52
|
/** Per-RO last-known output. CPU mirror, populated by finish(). */
|
|
49
53
|
lastOutput;
|
|
50
54
|
constructor(device) {
|
|
@@ -136,14 +140,22 @@ export class GpuDerivedModesScene {
|
|
|
136
140
|
*/
|
|
137
141
|
dispatch(arenaBuf, numROs, enc) {
|
|
138
142
|
this.ensureResources(arenaBuf, numROs);
|
|
143
|
+
// If the previous frame's readback hasn't resolved yet, we can
|
|
144
|
+
// still dispatch the compute (writes to outputBuf), but we must
|
|
145
|
+
// skip the staging copy — otherwise WebGPU rejects the submit
|
|
146
|
+
// with "buffer used in submit while mapped". Cheap to skip: the
|
|
147
|
+
// next dispatch will re-copy when staging is unmapped.
|
|
139
148
|
this.uploadInputs(numROs);
|
|
140
149
|
const pass = enc.beginComputePass({ label: "gpuDerivedModes/pass" });
|
|
141
150
|
pass.setPipeline(this.pipeline);
|
|
142
151
|
pass.setBindGroup(0, this.bindGroup);
|
|
143
152
|
pass.dispatchWorkgroups(Math.ceil(numROs / WG_SIZE));
|
|
144
153
|
pass.end();
|
|
145
|
-
|
|
146
|
-
|
|
154
|
+
if (!this.readbackInFlight) {
|
|
155
|
+
const bytes = numROs * 4;
|
|
156
|
+
enc.copyBufferToBuffer(this.outputBuf, 0, this.stagingBuf, 0, bytes);
|
|
157
|
+
this.readbackInFlight = true;
|
|
158
|
+
}
|
|
147
159
|
return numROs;
|
|
148
160
|
}
|
|
149
161
|
/**
|
|
@@ -154,11 +166,18 @@ export class GpuDerivedModesScene {
|
|
|
154
166
|
async finish(numROs) {
|
|
155
167
|
if (this.stagingBuf === null || numROs === 0)
|
|
156
168
|
return;
|
|
169
|
+
if (!this.readbackInFlight)
|
|
170
|
+
return; // dispatch skipped the copy; nothing to read
|
|
157
171
|
const bytes = numROs * 4;
|
|
158
|
-
|
|
159
|
-
|
|
160
|
-
|
|
161
|
-
|
|
172
|
+
try {
|
|
173
|
+
await this.stagingBuf.mapAsync(GPUMapMode.READ, 0, bytes);
|
|
174
|
+
const view = new Uint32Array(this.stagingBuf.getMappedRange(0, bytes).slice(0));
|
|
175
|
+
this.stagingBuf.unmap();
|
|
176
|
+
this.lastOutput.set(view, 0);
|
|
177
|
+
}
|
|
178
|
+
finally {
|
|
179
|
+
this.readbackInFlight = false;
|
|
180
|
+
}
|
|
162
181
|
}
|
|
163
182
|
dispose() {
|
|
164
183
|
this.inputBuf?.destroy();
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"gpuDispatcher.js","sourceRoot":"","sources":["../../../src/runtime/derivedModes/gpuDispatcher.ts"],"names":[],"mappings":"AAAA,oEAAoE;AACpE,kDAAkD;AAClD,EAAE;AACF,YAAY;AACZ,YAAY;AACZ,EAAE;AACF,qDAAqD;AACrD,sEAAsE;AACtE,kEAAkE;AAClE,kDAAkD;AAClD,sDAAsD;AACtD,gEAAgE;AAChE,wCAAwC;AACxC,sEAAsE;AACtE,uEAAuE;AACvE,oBAAoB;AACpB,iEAAiE;AACjE,iEAAiE;AACjE,kEAAkE;AAClE,kDAAkD;AAClD,EAAE;AACF,mDAAmD;AAEnD,OAAO,EAAE,yBAAyB,EAAE,WAAW,EAAE,MAAM,gBAAgB,CAAC;AAGxE,MAAM,OAAO,GAAG,UAAU,CAAC;AAS3B,MAAM,OAAO,GAAG,EAAE,CAAC;AACnB,MAAM,IAAI,GAAG,CAAC,CAAS,EAAU,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC;IAAE,CAAC,KAAK,CAAC,CAAC,CAAC,OAAO,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAElG,MAAM,OAAO,oBAAoB;IACtB,MAAM,CAAY;IAEV,OAAO,GAA4B,EAAE,CAAC;IAC/C,SAAS,GAAG,CAAC,CAAC;IACd,KAAK,GAAG,KAAK,CAAC;IAEtB,wDAAwD;IAChD,QAAQ,GAAuB,IAAI,CAAC;IACpC,SAAS,GAAsB,IAAI,CAAC;IACpC,UAAU,GAAqB,IAAI,CAAC;IACpC,SAAS,GAAsB,IAAI,CAAC;IACpC,QAAQ,GAAgC,IAAI,CAAC;IAC7C,SAAS,GAAyB,IAAI,CAAC;IACvC,MAAM,GAAkC,IAAI,CAAC;IACrD,yEAAyE;IACjE,QAAQ,GAAG,CAAC,CAAC;IACrB,uEAAuE;IAC/D,QAAQ,GAAqB,IAAI,CAAC;IAC1C;kEAC8D;IACtD,QAAQ,GAAW,WAAW,CAAC,IAAI,CAAC;
|
|
1
|
+
{"version":3,"file":"gpuDispatcher.js","sourceRoot":"","sources":["../../../src/runtime/derivedModes/gpuDispatcher.ts"],"names":[],"mappings":"AAAA,oEAAoE;AACpE,kDAAkD;AAClD,EAAE;AACF,YAAY;AACZ,YAAY;AACZ,EAAE;AACF,qDAAqD;AACrD,sEAAsE;AACtE,kEAAkE;AAClE,kDAAkD;AAClD,sDAAsD;AACtD,gEAAgE;AAChE,wCAAwC;AACxC,sEAAsE;AACtE,uEAAuE;AACvE,oBAAoB;AACpB,iEAAiE;AACjE,iEAAiE;AACjE,kEAAkE;AAClE,kDAAkD;AAClD,EAAE;AACF,mDAAmD;AAEnD,OAAO,EAAE,yBAAyB,EAAE,WAAW,EAAE,MAAM,gBAAgB,CAAC;AAGxE,MAAM,OAAO,GAAG,UAAU,CAAC;AAS3B,MAAM,OAAO,GAAG,EAAE,CAAC;AACnB,MAAM,IAAI,GAAG,CAAC,CAAS,EAAU,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC;IAAE,CAAC,KAAK,CAAC,CAAC,CAAC,OAAO,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAElG,MAAM,OAAO,oBAAoB;IACtB,MAAM,CAAY;IAEV,OAAO,GAA4B,EAAE,CAAC;IAC/C,SAAS,GAAG,CAAC,CAAC;IACd,KAAK,GAAG,KAAK,CAAC;IAEtB,wDAAwD;IAChD,QAAQ,GAAuB,IAAI,CAAC;IACpC,SAAS,GAAsB,IAAI,CAAC;IACpC,UAAU,GAAqB,IAAI,CAAC;IACpC,SAAS,GAAsB,IAAI,CAAC;IACpC,QAAQ,GAAgC,IAAI,CAAC;IAC7C,SAAS,GAAyB,IAAI,CAAC;IACvC,MAAM,GAAkC,IAAI,CAAC;IACrD,yEAAyE;IACjE,QAAQ,GAAG,CAAC,CAAC;IACrB,uEAAuE;IAC/D,QAAQ,GAAqB,IAAI,CAAC;IAC1C;kEAC8D;IACtD,QAAQ,GAAW,WAAW,CAAC,IAAI,CAAC;IAC5C;;uDAEmD;IAC3C,gBAAgB,GAAG,KAAK,CAAC;IAEjC,mEAAmE;IAC1D,UAAU,CAAc;IAEjC,YAAY,MAAiB;QAC3B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,UAAU,GAAG,IAAI,WAAW,CAAC,CAAC,CAAC,CAAC;IACvC,CAAC;IAED,IAAI,UAAU,KAAa,OAAO,IAAI,CAAC,SAAS,CAAC,CAAC,CAAC;IAEnD,UAAU,CAAC,MAAc,EAAE,QAAgB,EAAE,QAAkB;QAC7D,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,GAAG,EAAE,QAAQ,EAAE,WAAW,EAAE,WAAW,CAAC,QAAQ,CAAC,EAAE,CAAC;QACxE,IAAI,CAAC,SAAS,EAAE,CAAC;QACjB,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QAClB,IAAI,CAAC,QAAQ,GAAG,WAAW,CAAC,QAAQ,CAAC,CAAC;IACxC,CAAC;IAED,YAAY,CAAC,MAAc;QACzB,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,SAAS;YAAE,OAAO;QAC/C,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,GAAG,SAAS,CAAC;QACjC,IAAI,CAAC,SAAS,EAAE,CAAC;QACjB,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;IACpB,CAAC;IAED,SAAS,KAAW,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC,CAAC,CAAC;IACxC,YAAY,KAAc,MAAM,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC,OAAO,GAAG,CAAC,CAAC,CAAC;IAE3E,eAAe,CAAC,QAAmB,EAAE,MAAc;QACzD,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,CAAC;QAC1C,IAAI,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,OAAO,GAAG,IAAI,CAAC,QAAQ,EAAE,CAAC;YACvD,IAAI,CAAC,QAAU,EAAE,OAAO,EAAE,CAAC;YAC3B,IAAI,CAAC,SAAU,EAAE,OAAO,EAAE,CAAC;YAC3B,IAAI,CAAC,UAAU,EAAE,OAAO,EAAE,CAAC;YAC3B,MAAM,KAAK,GAAG,OAAO,GAAG,CAAC,CAAC;YAC1B,IAAI,CAAC,QAAQ,GAAK,IAAI,CAAC,MAAM,CAAC,YAAY,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,EAAE,cAAc,CAAC,OAAO,GAAG,cAAc,CAAC,QAAQ,EAAE,KAAK,EAAE,oBAAoB,EAAE,CAAC,CAAC;YAClJ,IAAI,CAAC,SAAS,GAAI,IAAI,CAAC,MAAM,CAAC,YAAY,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,EAAE,cAAc,CAAC,OAAO,GAAG,cAAc,CAAC,QAAQ,GAAG,cAAc,CAAC,QAAQ,EAAE,KAAK,EAAE,qBAAqB,EAAE,CAAC,CAAC;YAC7K,IAAI,CAAC,UAAU,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,EAAE,cAAc,CAAC,QAAQ,GAAG,cAAc,CAAC,QAAQ,EAAE,KAAK,EAAE,yBAAyB,EAAE,CAAC,CAAC;YACvJ,IAAoC,CAAC,UAAU,GAAG,IAAI,WAAW,CAAC,OAAO,CAAC,CAAC;YAC5E,IAAI,CAAC,QAAQ,GAAG,OAAO,CAAC;YACxB,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;QACxB,CAAC;QACD,IAAI,IAAI,CAAC,SAAS,KAAK,IAAI,EAAE,CAAC;YAC5B,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,MAAM,CAAC,YAAY,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,KAAK,EAAE,cAAc,CAAC,OAAO,GAAG,cAAc,CAAC,QAAQ,EAAE,KAAK,EAAE,wBAAwB,EAAE,CAAC,CAAC;QACpJ,CAAC;QACD,IAAI,IAAI,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC;YACzB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,qBAAqB,CAAC;gBAC9C,KAAK,EAAE,qBAAqB;gBAC5B,OAAO,EAAE;oBACP,EAAE,OAAO,EAAE,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,IAAI,EAAE,mBAAmB,EAAE,EAAE;oBACzF,EAAE,OAAO,EAAE,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,IAAI,EAAE,mBAAmB,EAAE,EAAE;oBACzF,EAAE,OAAO,EAAE,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,EAAE;oBAC/E,EAAE,OAAO,EAAE,CAAC,EAAE,UAAU,EAAE,cAAc,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,EAAE;iBAChF;aACF,CAAC,CAAC;QACL,CAAC;QACD,IAAI,IAAI,CAAC,QAAQ,KAAK,IAAI,EAAE,CAAC;YAC3B,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,kBAAkB,CAAC,EAAE,IAAI,EAAE,yBAAyB,EAAE,KAAK,EAAE,wBAAwB,EAAE,CAAC,CAAC;YACpH,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC,MAAM,CAAC,qBAAqB,CAAC;gBAChD,KAAK,EAAE,0BAA0B;gBACjC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,oBAAoB,CAAC,EAAE,gBAAgB,EAAE,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC7E,OAAO,EAAE,EAAE,MAAM,EAAE,UAAU,EAAE,UAAU,EAAE;aAC5C,CAAC,CAAC;QACL,CAAC;QACD,IAAI,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,IAAI,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;YAC1D,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,MAAM,CAAC,eAAe,CAAC;gBAC3C,KAAK,EAAE,2BAA2B;gBAClC,MAAM,EAAE,IAAI,CAAC,MAAM;gBACnB,OAAO,EAAE;oBACP,EAAE,OAAO,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,MAAM,EAAE,QAAQ,EAAE,EAAE;oBAC9C,EAAE,OAAO,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,MAAM,EAAE,IAAI,CAAC,QAAS,EAAG,EAAE;oBACrD,EAAE,OAAO,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,MAAM,EAAE,IAAI,CAAC,SAAU,EAAE,EAAE;oBACrD,EAAE,OAAO,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,MAAM,EAAE,IAAI,CAAC,SAAU,EAAE,EAAE;iBACtD;aACF,CAAC,CAAC;YACH,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC;QAC3B,CAAC;IACH,CAAC;IAED,oEAAoE;IAC5D,YAAY,CAAC,MAAc;QACjC,MAAM,IAAI,GAAG,IAAI,WAAW,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;QAC1D,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YAChC,MAAM,CAAC,GAAG,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;YAC1B,IAAI,CAAC,KAAK,SAAS;gBAAE,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC;QAC5C,CAAC;QACD,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,WAAW,CAAC,IAAI,CAAC,QAAS,EAAE,CAAC,EAAE,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,UAAU,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC;QAChG,MAAM,MAAM,GAAG,IAAI,WAAW,CAAC,CAAC,MAAM,EAAE,IAAI,CAAC,QAAQ,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QAC9D,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,WAAW,CAAC,IAAI,CAAC,SAAU,EAAE,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,UAAU,EAAE,MAAM,CAAC,UAAU,CAAC,CAAC;IACzG,CAAC;IAED;;;;OAIG;IACH,QAAQ,CAAC,QAAmB,EAAE,MAAc,EAAE,GAAsB;QAClE,IAAI,CAAC,eAAe,CAAC,QAAQ,EAAE,MAAM,CAAC,CAAC;QACvC,+DAA+D;QAC/D,gEAAgE;QAChE,8DAA8D;QAC9D,gEAAgE;QAChE,uDAAuD;QACvD,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;QAC1B,MAAM,IAAI,GAAG,GAAG,CAAC,gBAAgB,CAAC,EAAE,KAAK,EAAE,sBAAsB,EAAE,CAAC,CAAC;QACrE,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,QAAS,CAAC,CAAC;QACjC,IAAI,CAAC,YAAY,CAAC,CAAC,EAAE,IAAI,CAAC,SAAU,CAAC,CAAC;QACtC,IAAI,CAAC,kBAAkB,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,GAAG,OAAO,CAAC,CAAC,CAAC;QACrD,IAAI,CAAC,GAAG,EAAE,CAAC;QACX,IAAI,CAAC,IAAI,CAAC,gBAAgB,EAAE,CAAC;YAC3B,MAAM,KAAK,GAAG,MAAM,GAAG,CAAC,CAAC;YACzB,GAAG,CAAC,kBAAkB,CAAC,IAAI,CAAC,SAAU,EAAE,CAAC,EAAE,IAAI,CAAC,UAAW,EAAE,CAAC,EAAE,KAAK,CAAC,CAAC;YACvE,IAAI,CAAC,gBAAgB,GAAG,IAAI,CAAC;QAC/B,CAAC;QACD,OAAO,MAAM,CAAC;IAChB,CAAC;IAED;;;;OAIG;IACH,KAAK,CAAC,MAAM,CAAC,MAAc;QACzB,IAAI,IAAI,CAAC,UAAU,KAAK,IAAI,IAAI,MAAM,KAAK,CAAC;YAAE,OAAO;QACrD,IAAI,CAAC,IAAI,CAAC,gBAAgB;YAAE,OAAO,CAAC,6CAA6C;QACjF,MAAM,KAAK,GAAG,MAAM,GAAG,CAAC,CAAC;QACzB,IAAI,CAAC;YACH,MAAM,IAAI,CAAC,UAAU,CAAC,QAAQ,CAAC,UAAU,CAAC,IAAI,EAAE,CAAC,EAAE,KAAK,CAAC,CAAC;YAC1D,MAAM,IAAI,GAAG,IAAI,WAAW,CAAC,IAAI,CAAC,UAAU,CAAC,cAAc,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;YAChF,IAAI,CAAC,UAAU,CAAC,KAAK,EAAE,CAAC;YACvB,IAAoC,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;QAChE,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC,gBAAgB,GAAG,KAAK,CAAC;QAChC,CAAC;IACH,CAAC;IAED,OAAO;QACL,IAAI,CAAC,QAAU,EAAE,OAAO,EAAE,CAAC;QAC3B,IAAI,CAAC,SAAU,EAAE,OAAO,EAAE,CAAC;QAC3B,IAAI,CAAC,UAAU,EAAE,OAAO,EAAE,CAAC;QAC3B,IAAI,CAAC,SAAU,EAAE,OAAO,EAAE,CAAC;IAC7B,CAAC;CACF"}
|
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
export declare const GPU_FLIP_CULL_BY_DET_WGSL = "\n// Per-RO record. Index i corresponds to drawId i in the scene's\n// flat RO table.\nstruct GpuRuleParams {\n numROs: u32,\n declared: u32, // 0 = none, 1 = front, 2 = back (must match CULL_MODE[] order in bitfield.ts)\n _pad0: u32,\n _pad1: u32,\n};\n\n@group(0) @binding(0) var<storage, read> arena: array<u32>;\n// One per-RO entry: byte offset within the arena of the ROs ModelTrafo\n// allocation header. Data is at offset + 16 (ALLOC_HEADER_PAD_TO).\n// 0xFFFFFFFFu means \"this RO does NOT have a flip-cull-by-det rule\".\n@group(0) @binding(1) var<storage, read> roInputRefs: array<u32>;\n@group(0) @binding(2) var<storage, read_write> roOutputs: array<u32>;\n@group(0) @binding(3) var<uniform> uIn: GpuRuleParams;\n\n// Read a row-major mat4 from the arena at byte offset `refBytes + 16`\n// (header is 16 bytes wide due to ALLOC_HEADER_PAD_TO). Returns the\n// 3\u00D73 upper-left in row-major order: [m00,m01,m02, m10,m11,m12, m20,m21,m22].\nfn loadUpper3x3(refBytes: u32) -> array<f32, 9> {\n let baseU32 = (refBytes + 16u) >> 2u;\n // The arena is a Float32 storage of mat4 in row-major, 16 floats.\n // Read 9 of them.\n return array<f32, 9>(\n bitcast<f32>(arena[baseU32 + 0u]), bitcast<f32>(arena[baseU32 + 1u]), bitcast<f32>(arena[baseU32 + 2u]),\n bitcast<f32>(arena[baseU32 + 4u]), bitcast<f32>(arena[baseU32 + 5u]), bitcast<f32>(arena[baseU32 + 6u]),\n bitcast<f32>(arena[baseU32 + 8u]), bitcast<f32>(arena[baseU32 + 9u]), bitcast<f32>(arena[baseU32 + 10u]),\n );\n}\n\nfn det3x3(m: array<f32, 9>) -> f32 {\n return m[0] * (m[4] * m[8] - m[5] * m[7])\n - m[1] * (m[3] * m[8] - m[5] * m[6])\n + m[2] * (m[3] * m[7] - m[4] * m[6]);\n}\n\n// Mirror of bitfield.ts CULL_MODE[] order: 0=none, 1=front, 2=back.\nfn flipCull(c: u32) -> u32 {\n if (c == 1u) { return 2u; } // front -> back\n if (c == 2u) { return 1u; } // back -> front\n return c; // none\n}\n\n@compute @workgroup_size(64)\nfn evaluate(@builtin(global_invocation_id) gid: vec3<u32>) {\n let i = gid.x;\n if (i >= uIn.numROs) { return; }\n let
|
|
1
|
+
export declare const GPU_FLIP_CULL_BY_DET_WGSL = "\n// Per-RO record. Index i corresponds to drawId i in the scene's\n// flat RO table.\nstruct GpuRuleParams {\n numROs: u32,\n declared: u32, // 0 = none, 1 = front, 2 = back (must match CULL_MODE[] order in bitfield.ts)\n _pad0: u32,\n _pad1: u32,\n};\n\n@group(0) @binding(0) var<storage, read> arena: array<u32>;\n// One per-RO entry: byte offset within the arena of the ROs ModelTrafo\n// allocation header. Data is at offset + 16 (ALLOC_HEADER_PAD_TO).\n// 0xFFFFFFFFu means \"this RO does NOT have a flip-cull-by-det rule\".\n@group(0) @binding(1) var<storage, read> roInputRefs: array<u32>;\n@group(0) @binding(2) var<storage, read_write> roOutputs: array<u32>;\n@group(0) @binding(3) var<uniform> uIn: GpuRuleParams;\n\n// Read a row-major mat4 from the arena at byte offset `refBytes + 16`\n// (header is 16 bytes wide due to ALLOC_HEADER_PAD_TO). Returns the\n// 3\u00D73 upper-left in row-major order: [m00,m01,m02, m10,m11,m12, m20,m21,m22].\nfn loadUpper3x3(refBytes: u32) -> array<f32, 9> {\n let baseU32 = (refBytes + 16u) >> 2u;\n // The arena is a Float32 storage of mat4 in row-major, 16 floats.\n // Read 9 of them.\n return array<f32, 9>(\n bitcast<f32>(arena[baseU32 + 0u]), bitcast<f32>(arena[baseU32 + 1u]), bitcast<f32>(arena[baseU32 + 2u]),\n bitcast<f32>(arena[baseU32 + 4u]), bitcast<f32>(arena[baseU32 + 5u]), bitcast<f32>(arena[baseU32 + 6u]),\n bitcast<f32>(arena[baseU32 + 8u]), bitcast<f32>(arena[baseU32 + 9u]), bitcast<f32>(arena[baseU32 + 10u]),\n );\n}\n\nfn det3x3(m: array<f32, 9>) -> f32 {\n return m[0] * (m[4] * m[8] - m[5] * m[7])\n - m[1] * (m[3] * m[8] - m[5] * m[6])\n + m[2] * (m[3] * m[7] - m[4] * m[6]);\n}\n\n// Mirror of bitfield.ts CULL_MODE[] order: 0=none, 1=front, 2=back.\nfn flipCull(c: u32) -> u32 {\n if (c == 1u) { return 2u; } // front -> back\n if (c == 2u) { return 1u; } // back -> front\n return c; // none\n}\n\n@compute @workgroup_size(64)\nfn evaluate(@builtin(global_invocation_id) gid: vec3<u32>) {\n let i = gid.x;\n if (i >= uIn.numROs) { return; }\n let refByte = roInputRefs[i];\n if (refByte == 0xFFFFFFFFu) {\n // No rule on this RO; leave existing output untouched. Allows a\n // single kernel dispatch to skip non-participating ROs.\n return;\n }\n let m = loadUpper3x3(refByte);\n let d = det3x3(m);\n let declared = uIn.declared;\n let outVal = select(declared, flipCull(declared), d < 0.0);\n roOutputs[i] = outVal;\n}\n";
|
|
2
2
|
/** Map between bitfield.ts's CULL_MODE order and the kernel's u32 enum. */
|
|
3
3
|
export declare const CULL_TO_U32: {
|
|
4
4
|
readonly none: 0;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"gpuKernel.d.ts","sourceRoot":"","sources":["../../../src/runtime/derivedModes/gpuKernel.ts"],"names":[],"mappings":"AAgCA,eAAO,MAAM,yBAAyB,
|
|
1
|
+
{"version":3,"file":"gpuKernel.d.ts","sourceRoot":"","sources":["../../../src/runtime/derivedModes/gpuKernel.ts"],"names":[],"mappings":"AAgCA,eAAO,MAAM,yBAAyB,k/EA6DrC,CAAC;AAEF,2EAA2E;AAC3E,eAAO,MAAM,WAAW;;;;CAAgD,CAAC;AACzE,eAAO,MAAM,WAAW,oCAAqC,CAAC"}
|
|
@@ -78,13 +78,13 @@ fn flipCull(c: u32) -> u32 {
|
|
|
78
78
|
fn evaluate(@builtin(global_invocation_id) gid: vec3<u32>) {
|
|
79
79
|
let i = gid.x;
|
|
80
80
|
if (i >= uIn.numROs) { return; }
|
|
81
|
-
let
|
|
82
|
-
if (
|
|
81
|
+
let refByte = roInputRefs[i];
|
|
82
|
+
if (refByte == 0xFFFFFFFFu) {
|
|
83
83
|
// No rule on this RO; leave existing output untouched. Allows a
|
|
84
84
|
// single kernel dispatch to skip non-participating ROs.
|
|
85
85
|
return;
|
|
86
86
|
}
|
|
87
|
-
let m = loadUpper3x3(
|
|
87
|
+
let m = loadUpper3x3(refByte);
|
|
88
88
|
let d = det3x3(m);
|
|
89
89
|
let declared = uIn.declared;
|
|
90
90
|
let outVal = select(declared, flipCull(declared), d < 0.0);
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@aardworx/wombat.rendering",
|
|
3
|
-
"version": "0.9.
|
|
3
|
+
"version": "0.9.21",
|
|
4
4
|
"description": "WebGPU rendering layer for the Wombat TypeScript stack — RenderObject/RenderTask/runtime + window glue, port of Aardvark.Rendering's lower layers on top of WebGPU and wombat.shader.",
|
|
5
5
|
"license": "MIT",
|
|
6
6
|
"author": "krauthaufen",
|
|
@@ -58,6 +58,10 @@ export class GpuDerivedModesScene {
|
|
|
58
58
|
/** Last-known declared (for kernel uniform). Common case: all RO
|
|
59
59
|
* rules share the same declared (= the SG scope's value). */
|
|
60
60
|
private declared: number = CULL_TO_U32.back;
|
|
61
|
+
/** True while the staging buffer's mapAsync is in flight. We must
|
|
62
|
+
* not enqueue a new copyBufferToBuffer into it until it's been
|
|
63
|
+
* unmapped (mapAsync resolved + unmap called). */
|
|
64
|
+
private readbackInFlight = false;
|
|
61
65
|
|
|
62
66
|
/** Per-RO last-known output. CPU mirror, populated by finish(). */
|
|
63
67
|
readonly lastOutput: Uint32Array;
|
|
@@ -156,14 +160,22 @@ export class GpuDerivedModesScene {
|
|
|
156
160
|
*/
|
|
157
161
|
dispatch(arenaBuf: GPUBuffer, numROs: number, enc: GPUCommandEncoder): number {
|
|
158
162
|
this.ensureResources(arenaBuf, numROs);
|
|
163
|
+
// If the previous frame's readback hasn't resolved yet, we can
|
|
164
|
+
// still dispatch the compute (writes to outputBuf), but we must
|
|
165
|
+
// skip the staging copy — otherwise WebGPU rejects the submit
|
|
166
|
+
// with "buffer used in submit while mapped". Cheap to skip: the
|
|
167
|
+
// next dispatch will re-copy when staging is unmapped.
|
|
159
168
|
this.uploadInputs(numROs);
|
|
160
169
|
const pass = enc.beginComputePass({ label: "gpuDerivedModes/pass" });
|
|
161
170
|
pass.setPipeline(this.pipeline!);
|
|
162
171
|
pass.setBindGroup(0, this.bindGroup!);
|
|
163
172
|
pass.dispatchWorkgroups(Math.ceil(numROs / WG_SIZE));
|
|
164
173
|
pass.end();
|
|
165
|
-
|
|
166
|
-
|
|
174
|
+
if (!this.readbackInFlight) {
|
|
175
|
+
const bytes = numROs * 4;
|
|
176
|
+
enc.copyBufferToBuffer(this.outputBuf!, 0, this.stagingBuf!, 0, bytes);
|
|
177
|
+
this.readbackInFlight = true;
|
|
178
|
+
}
|
|
167
179
|
return numROs;
|
|
168
180
|
}
|
|
169
181
|
|
|
@@ -174,11 +186,16 @@ export class GpuDerivedModesScene {
|
|
|
174
186
|
*/
|
|
175
187
|
async finish(numROs: number): Promise<void> {
|
|
176
188
|
if (this.stagingBuf === null || numROs === 0) return;
|
|
189
|
+
if (!this.readbackInFlight) return; // dispatch skipped the copy; nothing to read
|
|
177
190
|
const bytes = numROs * 4;
|
|
178
|
-
|
|
179
|
-
|
|
180
|
-
|
|
181
|
-
|
|
191
|
+
try {
|
|
192
|
+
await this.stagingBuf.mapAsync(GPUMapMode.READ, 0, bytes);
|
|
193
|
+
const view = new Uint32Array(this.stagingBuf.getMappedRange(0, bytes).slice(0));
|
|
194
|
+
this.stagingBuf.unmap();
|
|
195
|
+
(this as { lastOutput: Uint32Array }).lastOutput.set(view, 0);
|
|
196
|
+
} finally {
|
|
197
|
+
this.readbackInFlight = false;
|
|
198
|
+
}
|
|
182
199
|
}
|
|
183
200
|
|
|
184
201
|
dispose(): void {
|
|
@@ -79,13 +79,13 @@ fn flipCull(c: u32) -> u32 {
|
|
|
79
79
|
fn evaluate(@builtin(global_invocation_id) gid: vec3<u32>) {
|
|
80
80
|
let i = gid.x;
|
|
81
81
|
if (i >= uIn.numROs) { return; }
|
|
82
|
-
let
|
|
83
|
-
if (
|
|
82
|
+
let refByte = roInputRefs[i];
|
|
83
|
+
if (refByte == 0xFFFFFFFFu) {
|
|
84
84
|
// No rule on this RO; leave existing output untouched. Allows a
|
|
85
85
|
// single kernel dispatch to skip non-participating ROs.
|
|
86
86
|
return;
|
|
87
87
|
}
|
|
88
|
-
let m = loadUpper3x3(
|
|
88
|
+
let m = loadUpper3x3(refByte);
|
|
89
89
|
let d = det3x3(m);
|
|
90
90
|
let declared = uIn.declared;
|
|
91
91
|
let outVal = select(declared, flipCull(declared), d < 0.0);
|