@graphty/webgpu-graph-algorithms 0.6.20 → 0.6.22

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 10 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 11 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   |
 | ---- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- | ---------- | -------- |

package/dist/browser.js

@@ -1,4 +1,4 @@
-import { G as GpuContext, W as WebGpuGraphError } from "./chunks/context-VIvatQOo.js";
+import { G as GpuContext, W as WebGpuGraphError } from "./chunks/context-B40Z6lV_.js";
 function navigatorGpu() {
   if (typeof navigator === "undefined") {
     return void 0;

package/dist/chunks/{context-VIvatQOo.js → context-B40Z6lV_.js}

@@ -109,6 +109,8 @@ 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 BORUVKA_ROUNDS_PER_SUBMIT = 4;
+const F32_SIGN_BIT = 2147483648;
 const GROUP_ROW_THREAD_MAX = 32;
 const GROUP_ROW_THREAD_LIMIT = 128;
 const PARALLEL_MERGE_LIMIT = 65e3;
@@ -920,6 +922,7 @@ const PRELUDE_WGSL = (
 const INVALID_INDEX: u32 = ${INVALID_INDEX}u;
 const U32_MAX: u32 = ${U32_MAX}u;
 const F32_INF_BITS: u32 = ${F32_INF_BITS}u;
+const F32_SIGN_BIT: u32 = ${F32_SIGN_BIT}u;
 const MAX_WORKGROUPS_PER_DIM: u32 = ${MAX_WORKGROUPS_PER_DIM}u;
 const EXACT_TILES_PER_PASS: u32 = ${EXACT_TILES_PER_PASS}u;
 const FA2_DIST_FLOOR: f32 = ${wgslF32Literal(FA2_DISTANCE_FLOOR)};
@@ -958,6 +961,11 @@ fn lowbias32(x0: u32) -> u32 {
 fn mask_bit(w: u32, i: u32) -> bool { return ((w >> (i & 31u)) & 1u) == 1u; }
 fn unpack_u8(w: u32, i: u32) -> u32 { return (w >> (8u * (i & 3u))) & 0xFFu; }
 fn pair_hash(i: u32, j: u32) -> u32 { return lowbias32((min(i, j) * 0x9E3779B9u) ^ max(i, j)); }
+fn order_key(w: f32) -> u32 {
+    let raw = bitcast<u32>(w);
+    let bits = select(raw, 0u, raw == F32_SIGN_BIT);
+    return select(bits | F32_SIGN_BIT, ~bits, (bits & F32_SIGN_BIT) != 0u);
+}
 fn hash_unit(h: u32) -> f32 { return f32(h >> 8u) * (1.0 / 16777216.0); }
 fn hash_dir(h: u32, dim: u32) -> vec3f {
     let phi = 6.283185307179586 * hash_unit(h);
@@ -3537,40 +3545,41 @@ class GpuContext {
   }
 }
 export {
-  SE_SCALE_REFERENCE_NODES as $,
+  SETTLE_FLOOR_REFERENCE_NODES as $,
   APSP_MAX_DISPATCHES_PER_SUBMIT as A,
   BufferUsage as B,
-  LAYOUT_TUNING_DEFAULTS as C,
-  EXACT_MAX_NODES as D,
+  GRID_HUB_CELL as C,
+  LAYOUT_TUNING_DEFAULTS as D,
   EXACT_TILES_PER_PASS as E,
   FUSED_FRONTIER_MAX as F,
   GpuContext as G,
-  SETTLE_FLOOR_UNBOUNDED as H,
+  EXACT_MAX_NODES as H,
   INDIRECT_ARGS_STRIDE as I,
-  GRID_BBOX_MARGIN as J,
-  GRID_EXTENT_FLOOR as K,
+  SETTLE_FLOOR_UNBOUNDED as J,
+  GRID_BBOX_MARGIN as K,
   LABEL_PROP_PASSES_PER_SUBMIT as L,
   MAX_WORKGROUPS_PER_DIM as M,
-  FR_ADAPTIVE_MAX_ITERATIONS as N,
-  FR_START_TEMPERATURE as O,
+  GRID_EXTENT_FLOOR as N,
+  FR_ADAPTIVE_MAX_ITERATIONS as O,
   PARALLEL_MERGE_LIMIT as P,
-  FA2_FLAG_ADAPTIVE as Q,
+  FR_START_TEMPERATURE as Q,
   RADIX_BINS as R,
   STATE_HEADER_BYTES as S,
   TRACE_RECORD_BYTES as T,
   UNIFORM_SLOT_BYTES as U,
-  SETTLE_FLOOR_FRACTION as V,
+  FA2_FLAG_ADAPTIVE as V,
   WebGpuGraphError as W,
-  FR_REHEAT_FRACTION as X,
-  FR_DEFAULTS as Y,
-  SE_DEFAULTS as Z,
-  SETTLE_FLOOR_REFERENCE_NODES as _,
+  SETTLE_FLOOR_FRACTION as X,
+  FR_REHEAT_FRACTION as Y,
+  FR_DEFAULTS as Z,
+  SE_DEFAULTS 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,
+  SE_SCALE_REFERENCE_NODES as a0,
+  ARC_WINDOW_ALIGN as a1,
+  PASSTHROUGH_FORMAT_CODES as a2,
+  STORAGE_ALIGN as a3,
+  WORKGROUP_SIZE as a4,
+  isSoftwareAdapter as a5,
   U32_MAX as b,
   MAX_LEVELS_PER_SUBMIT as c,
   deviceLostError as d,
@@ -3586,15 +3595,15 @@ export {
   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
+  BORUVKA_ROUNDS_PER_SUBMIT as q,
+  GRID_COARSEST_SIDE as r,
+  GRID_MIN_SIDE as s,
+  GRID_SORT_BITS as t,
+  FA2_DEFAULTS as u,
+  MAX_ITERATIONS_PER_STEP as v,
+  MAX_1D_ITEMS as w,
+  hasErrorCode as x,
+  FA2_FLAG_FIRST as y,
+  PARTIAL_BYTES as z
 };
-//# sourceMappingURL=context-VIvatQOo.js.map
+//# sourceMappingURL=context-B40Z6lV_.js.map