@graphty/webgpu-graph-algorithms 0.6.15 → 0.6.16

package/README.md

@@ -695,7 +695,7 @@ never relaxed silently.
 
 <!-- targets-table:nvidia-lovelace-driver580 -->
 
-Generated by `pnpm run bench:readme` from `benchmarks/results/targets.json`. Each figure is the best median the row has recorded, the baseline `bench:compare` pins: Node rows across the 8 session(s) of `benchmarks/results/nvidia-lovelace-driver580.json`, Chromium rows across the 3 of `benchmarks/results/nvidia-lovelace-driver0.json`.
+Generated by `pnpm run bench:readme` from `benchmarks/results/targets.json`. Each figure is the best median the row has recorded, the baseline `bench:compare` pins: Node rows across the 10 session(s) of `benchmarks/results/nvidia-lovelace-driver580.json`, Chromium rows across the 3 of `benchmarks/results/nvidia-lovelace-driver0.json`.
 
 | Id   | What                                                                                                                                                                   | Target     | Measured   | Status   |
 | ---- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- | ---------- | -------- |
@@ -722,11 +722,11 @@ Generated by `pnpm run bench:readme` from `benchmarks/results/targets.json`. Eac
 | T-14 | Fruchterman-Reingold exact tier, GPU time per iteration (profiler) at 100k (`repulsion: "exact"`)                                                                      | recorded   | 16.322 ms  | recorded |
 | T-14 | Spring-electrical preset, GPU time per iteration (profiler) at 10k                                                                                                     | recorded   | 0.648 ms   | recorded |
 | T-14 | Spring-electrical preset, GPU time per iteration (profiler) at 100k (`repulsion: "exact"`)                                                                             | recorded   | 18.154 ms  | recorded |
-| T-10 | BFS direction-optimizing (`breadthFirstSearch`), wall end to end including the upload, from node 0 of the undirected RMAT at 1M / 10M (a contract on the dev box only) | <= 100 ms  | 89.706 ms  | met      |
-| T-10 | BFS top-down, the same traversal at 1M / 10M                                                                                                                           | recorded   | 97.061 ms  | recorded |
+| T-10 | BFS direction-optimizing (`breadthFirstSearch`), wall end to end including the upload, from node 0 of the undirected RMAT at 1M / 10M (a contract on the dev box only) | <= 100 ms  | 84.989 ms  | met      |
+| T-10 | BFS top-down, the same traversal at 1M / 10M                                                                                                                           | recorded   | 81.261 ms  | recorded |
 | T-10 | BFS direction-optimizing, the same traversal at 100k / 1M                                                                                                              | recorded   | 12.837 ms  | recorded |
-| T-10 | BFS on the 1000 x 1000 grid from a corner (1,999 levels), wall including the upload (a contract on the dev box only)                                                   | <= 1500 ms | 425.858 ms | met      |
-| T-10 | `sssp` (the near-far queue), wall end to end including the upload, random f32 weights in [0.1, 10), at 1M / 10M                                                        | recorded   | 266.241 ms | recorded |
+| T-10 | BFS on the 1000 x 1000 grid from a corner (1,999 levels), wall including the upload (a contract on the dev box only)                                                   | <= 1500 ms | 369.673 ms | met      |
+| T-10 | `sssp` (the near-far queue), wall end to end including the upload, random f32 weights in [0.1, 10), at 1M / 10M                                                        | recorded   | 256.320 ms | recorded |
 | T-10 | `sssp`, the same run at 100k / 1M                                                                                                                                      | recorded   | 47.421 ms  | recorded |
 
 <!-- /targets-table:nvidia-lovelace-driver580 -->

package/dist/browser.js

@@ -1,4 +1,4 @@
-import { G as GpuContext, W as WebGpuGraphError } from "./chunks/context-oXphO3yj.js";
+import { G as GpuContext, W as WebGpuGraphError } from "./chunks/context-VIvatQOo.js";
 function navigatorGpu() {
   if (typeof navigator === "undefined") {
     return void 0;

package/dist/chunks/{context-oXphO3yj.js → context-VIvatQOo.js}

@@ -106,6 +106,14 @@ const BC_BATCH_BUDGET_FRACTION = 0.25;
 const BC_MAX_BATCH = 64;
 const BC_EDGE_PARALLEL_GAMMA = 2;
 const BC_BACKWARD_LEVELS_PER_SUBMIT = 64;
+const APSP_TILE = 32;
+const APSP_MAX_DISPATCHES_PER_SUBMIT = 4096;
+const LABEL_PROP_PASSES_PER_SUBMIT = 8;
+const GROUP_ROW_THREAD_MAX = 32;
+const GROUP_ROW_THREAD_LIMIT = 128;
+const PARALLEL_MERGE_LIMIT = 65e3;
+const GROUP_HASH_LOAD_FACTOR = 2;
+const TRIANGLE_BINARY_SEARCH_RATIO = 32;
 const PASSTHROUGH_FORMAT_CODES = Object.freeze(["E_GPU_INELIGIBLE", "E_UNKNOWN_NODE", "E_UNKNOWN_COLUMN", "E_COLUMN_LENGTH"]);
 const EMPTY_DETAILS = Object.freeze({});
 class WebGpuGraphError extends Error {
@@ -926,6 +934,9 @@ const GRID_EXTENT_FLOOR: f32 = ${wgslF32Literal(GRID_EXTENT_FLOOR)};
 const GRID_BBOX_MARGIN: f32 = ${wgslF32Literal(GRID_BBOX_MARGIN)};
 const RADIX_BINS: u32 = ${RADIX_BINS}u;
 const RADIX_DIGIT_MASK: u32 = ${RADIX_BINS - 1}u;
+const APSP_TILE: u32 = ${APSP_TILE}u;
+const GROUP_HASH_LOAD_FACTOR: u32 = ${GROUP_HASH_LOAD_FACTOR}u;
+const TRIANGLE_BINARY_SEARCH_RATIO: u32 = ${TRIANGLE_BINARY_SEARCH_RATIO}u;
 const F32_MAX: f32 = 0x1.fffffep+127;
 override WG: u32 = ${WORKGROUP_SIZE}u;
 override USE_PERM: bool = false;
@@ -2637,7 +2648,8 @@ class GraphResidency {
   /**
    * Uploads (or finds) a view: outDegree / inDegree / degreeOrder / reverseDegreeOrder upload one array each;
    * reverse and edgeList (P7) upload their arrays perArray, never into the arena, and are memoised per record so
-   * a second call uploads nothing (spec 4.3). coo and mate -> E_UNSUPPORTED until P11. packViews concatenates a
+   * a second call uploads nothing (spec 4.3). coo uploads its per-arc `src` (P11; the rest aliases the core); mate ->
+   * E_UNSUPPORTED. packViews concatenates a
    * reverse or edgeList view into ONE buffer at STORAGE_ALIGN offsets; on any other view `true` is E_UNSUPPORTED
    * { option: "packViews" }.
    * @param s - the snapshot
@@ -2704,10 +2716,18 @@ class GraphResidency {
         break;
       }
       case "coo":
+        this.assertNotReleased(s);
+        this.assertNonEmpty(s);
+        if (s.arcCount === 0) {
+          return Object.freeze({ view: name, bindings: Object.freeze({}), scalars: Object.freeze({}) });
+        }
+        array = s.coo().src;
+        bindingName = "src";
+        break;
       case "mate":
-        throw new WebGpuGraphError("E_UNSUPPORTED", `the ${name} view is not uploaded before P11`, {
-          feature: `view:${name}`,
-          hint: "outDegree, inDegree, degreeOrder, reverseDegreeOrder, reverse and edgeList are uploaded"
+        throw new WebGpuGraphError("E_UNSUPPORTED", "the mate view is not uploaded: no kernel reads it", {
+          feature: "view:mate",
+          hint: "outDegree, inDegree, degreeOrder, reverseDegreeOrder, coo, reverse and edgeList are uploaded"
         });
       default:
         throw invalid("name", name, "a view name");
@@ -3517,33 +3537,40 @@ class GpuContext {
   }
 }
 export {
-  FR_START_TEMPERATURE as A,
+  SE_SCALE_REFERENCE_NODES as $,
+  APSP_MAX_DISPATCHES_PER_SUBMIT as A,
   BufferUsage as B,
-  FA2_FLAG_ADAPTIVE as C,
-  SETTLE_FLOOR_FRACTION as D,
+  LAYOUT_TUNING_DEFAULTS as C,
+  EXACT_MAX_NODES as D,
   EXACT_TILES_PER_PASS as E,
   FUSED_FRONTIER_MAX as F,
   GpuContext as G,
-  FR_REHEAT_FRACTION as H,
+  SETTLE_FLOOR_UNBOUNDED as H,
   INDIRECT_ARGS_STRIDE as I,
-  FR_DEFAULTS as J,
-  SE_DEFAULTS as K,
-  LAYOUT_TUNING_DEFAULTS as L,
+  GRID_BBOX_MARGIN as J,
+  GRID_EXTENT_FLOOR as K,
+  LABEL_PROP_PASSES_PER_SUBMIT as L,
   MAX_WORKGROUPS_PER_DIM as M,
-  SETTLE_FLOOR_REFERENCE_NODES as N,
-  SE_SCALE_REFERENCE_NODES as O,
-  PARTIAL_BYTES as P,
-  ARC_WINDOW_ALIGN as Q,
+  FR_ADAPTIVE_MAX_ITERATIONS as N,
+  FR_START_TEMPERATURE as O,
+  PARALLEL_MERGE_LIMIT as P,
+  FA2_FLAG_ADAPTIVE as Q,
   RADIX_BINS as R,
   STATE_HEADER_BYTES as S,
   TRACE_RECORD_BYTES as T,
   UNIFORM_SLOT_BYTES as U,
-  PASSTHROUGH_FORMAT_CODES as V,
+  SETTLE_FLOOR_FRACTION as V,
   WebGpuGraphError as W,
-  STORAGE_ALIGN as X,
-  WORKGROUP_SIZE as Y,
-  isSoftwareAdapter as Z,
+  FR_REHEAT_FRACTION as X,
+  FR_DEFAULTS as Y,
+  SE_DEFAULTS as Z,
+  SETTLE_FLOOR_REFERENCE_NODES as _,
   WGSL_RESERVED_WORDS as a,
+  ARC_WINDOW_ALIGN as a0,
+  PASSTHROUGH_FORMAT_CODES as a1,
+  STORAGE_ALIGN as a2,
+  WORKGROUP_SIZE as a3,
+  isSoftwareAdapter as a4,
   U32_MAX as b,
   MAX_LEVELS_PER_SUBMIT as c,
   deviceLostError as d,
@@ -3555,19 +3582,19 @@ export {
   BC_BATCH_BUDGET_FRACTION as j,
   BC_MAX_BATCH as k,
   BC_BACKWARD_LEVELS_PER_SUBMIT as l,
-  GRID_COARSEST_SIDE as m,
-  GRID_MIN_SIDE as n,
-  GRID_SORT_BITS as o,
-  FA2_DEFAULTS as p,
-  MAX_ITERATIONS_PER_STEP as q,
-  MAX_1D_ITEMS as r,
-  hasErrorCode as s,
-  FA2_FLAG_FIRST as t,
-  GRID_HUB_CELL as u,
-  EXACT_MAX_NODES as v,
-  SETTLE_FLOOR_UNBOUNDED as w,
-  GRID_BBOX_MARGIN as x,
-  GRID_EXTENT_FLOOR as y,
-  FR_ADAPTIVE_MAX_ITERATIONS as z
+  APSP_TILE as m,
+  GROUP_ROW_THREAD_MAX as n,
+  GROUP_ROW_THREAD_LIMIT as o,
+  GROUP_HASH_LOAD_FACTOR as p,
+  GRID_COARSEST_SIDE as q,
+  GRID_MIN_SIDE as r,
+  GRID_SORT_BITS as s,
+  FA2_DEFAULTS as t,
+  MAX_ITERATIONS_PER_STEP as u,
+  MAX_1D_ITEMS as v,
+  hasErrorCode as w,
+  FA2_FLAG_FIRST as x,
+  PARTIAL_BYTES as y,
+  GRID_HUB_CELL as z
 };
-//# sourceMappingURL=context-oXphO3yj.js.map
+//# sourceMappingURL=context-VIvatQOo.js.map