@woosh/meep-engine 3.3.0 → 3.4.0

package/src/shade/renderer/gpu_primitive/sort/ffx/defs.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"defs.d.ts","sourceRoot":"","sources":["../../../../../../../src/shade/renderer/gpu_primitive/sort/ffx/defs.js"],"names":[],"mappings":"AAIA;;;;GAIG;AACH,yCAFU,MAAM,CAE4B;AAE5C;;;GAGG;AACH,2BAFU,MAAM,CAEe;AAC/B;;;GAGG;AACH,4BAFU,MAAM,CAEe;AAC/B;;;GAGG;AACH,uBAFU,MAAM,CAE2B;AAE3C;;;GAGG;AACH,oCAFU,MAAM,CAEyB;AAEzC;;;GAGG;AACH,mCAFU,WAAW,CAE+B;AACpD,6DAGE;4BAxC0B,qCAAqC;6BACpC,sCAAsC"}

package/src/shade/renderer/gpu_primitive/sort/ffx/defs.js

@@ -1,42 +0,0 @@
-import { u32 } from "../../../shader/type/WEBGPU_WGSL_PRIMITIVE_TYPES.js";
-import { WebGPUArray } from "../../../shader/type/WebGPUArray.js";
-import { WebGPUStruct } from "../../../shader/type/WebGPUStruct.js";
-
-/**
- * Good value is probably ~16
- * @see https://raw.githubusercontent.com/MircoWerner/VkRadixSort/refs/heads/main/timings/radixsort_multi_1000000.png
- * @type {number}
- */
-export const NUMBER_OF_BLOCKS_PER_GROUP = 1;
-
-/**
- * How many bits are there in a key
- * @type {number}
- */
-export const KEY_BIT_SIZE = 32;
-/**
- * How many bits do we handle per pass?
- * @type {number}
- */
-export const BITS_PER_PASS = 4;
-/**
- * How many histogram bins do we need?
- * @type {number}
- */
-export const NUM_BINS = 1 << BITS_PER_PASS;
-
-/**
- * compute shader work-group size
- * @type {number}
- */
-export const RADIX_SORT_GROUP_SIZE = 256;
-
-/**
- *
- * @type {WebGPUArray}
- */
-export const RADIX_SORT_KEYS_TYPE = WebGPUArray.u32;
-export const RADIX_SORT_PASS_PARAMETERS_STRUCT = WebGPUStruct.from({
-    element_count: u32,
-    bit_shift: u32 // current shift
-})

package/src/shade/renderer/gpu_primitive/sort/ffx/graph_radix_sort.d.ts

@@ -1,13 +0,0 @@
-/**
- * Keys are 32bit unsigned integers
- * Values are 32bit unsigned integers as well
- * @param {FrameGraph} graph
- * @param {number} keys
- * @param {number} values
- * @param {number} [size] number of input elements
- */
-export function graph_radix_sort({ graph, keys, values, size, }: FrameGraph): {
-    keys: any;
-    value: any;
-};
-//# sourceMappingURL=graph_radix_sort.d.ts.map

package/src/shade/renderer/gpu_primitive/sort/ffx/graph_radix_sort.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"graph_radix_sort.d.ts","sourceRoot":"","sources":["../../../../../../../src/shade/renderer/gpu_primitive/sort/ffx/graph_radix_sort.js"],"names":[],"mappings":"AAsBA;;;;;;;GAOG;AACH;;;EAqFC"}

package/src/shade/renderer/gpu_primitive/sort/ffx/graph_radix_sort.js

@@ -1,116 +0,0 @@
-import { assert } from "../../../../../core/assert.js";
-import { BITS_PER_PASS, KEY_BIT_SIZE, RADIX_SORT_GROUP_SIZE } from "./defs.js";
-import { shader_bit_histogram } from "./shader_bit_histogram.js";
-import { shader_scatter_keys_and_values } from "./shader_scatter_keys_and_values.js";
-
-/**
- *
- * @see https://github.com/GPUOpen-Drivers/gpurt/blob/f734985ebc31f471c376ed0cb217f43bdd40ee17/src/shaders/RadixSort/RadixSortParallel.hlsl#L88
- * @param {number} pass_index
- */
-function scan_exclusive_add({ pass_index }) {
-
-    // NOTE: only implements Dynamic Load Balancing (DLB) scan
-
-    assert.isNonNegativeInteger(pass_index, 'pass_index');
-
-    if (pass_index === 0) {
-
-    }
-
-}
-
-/**
- * Keys are 32bit unsigned integers
- * Values are 32bit unsigned integers as well
- * @param {FrameGraph} graph
- * @param {number} keys
- * @param {number} values
- * @param {number} [size] number of input elements
- */
-export function graph_radix_sort({
-                                     graph,
-                                     keys,
-                                     values,
-                                     size,
-                                 }) {
-
-
-    const data = {};
-
-    /**
-     *
-     * @param data
-     * @param {FramePassResources} resources
-     * @param {GraphicsExecutionContext} context
-     */
-    function execute(data, resources, context) {
-
-        const encoder = context.encoder;
-        const NUM_PASSES = KEY_BIT_SIZE / BITS_PER_PASS;
-
-        const group_block_size = (RADIX_SORT_GROUP_SIZE * BITS_PER_PASS)
-        const total_num_groups = Math.ceil(size / group_block_size)
-
-        let output_keys;
-        let output_values;
-
-        let current_from_keys;
-        let current_from_values;
-        let current_to_keys;
-        let current_to_values;
-
-        // see https://github.com/GPUOpen-Drivers/gpurt/blob/f734985ebc31f471c376ed0cb217f43bdd40ee17/src/shaders/RadixSort/RadixSortParallel.hlsl#L294
-
-        for (let pass_index = 0; pass_index < NUM_PASSES; pass_index++) {
-
-            const bit_shift_count = pass_index * BITS_PER_PASS;
-
-            // build histogram
-            shader_bit_histogram.dispatch({
-                encoder,
-                group_count_x: total_num_groups,
-            })
-
-            // scan
-            scan_exclusive_add({
-                pass_index: pass_index
-            });
-
-            // scatter
-            shader_scatter_keys_and_values.dispatch({
-                encoder,
-                group_count_x: total_num_groups,
-                bindings: {}
-            })
-
-            if (pass_index === 0) {
-                // The first pass uses the input buffer. Swap between output and temp resources for later passes.
-                current_from_keys = output_keys;
-                current_from_values = output_values;
-            }
-
-            // swap buffers
-            {
-                const t = current_from_keys;
-                current_from_keys = current_to_keys;
-                current_to_keys = t;
-            }
-            {
-                const t = current_from_values;
-                current_from_values = current_to_values;
-                current_to_values = t;
-            }
-        }
-
-    }
-
-    const graph_pass = graph.add(`radix sort`, data, execute);
-
-
-    // outputs sorted key/value pairs
-    return {
-        keys: sorted_keys,
-        value: sorted_values,
-    }
-}

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_bit_histogram.d.ts

@@ -1,3 +0,0 @@
-export const shader_bit_histogram: ComputeShader;
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-//# sourceMappingURL=shader_bit_histogram.d.ts.map

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_bit_histogram.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"shader_bit_histogram.d.ts","sourceRoot":"","sources":["../../../../../../../src/shade/renderer/gpu_primitive/sort/ffx/shader_bit_histogram.js"],"names":[],"mappings":"AAmHA,iDAIE;8BAnH4B,kCAAkC"}

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_bit_histogram.js

@@ -1,120 +0,0 @@
-import { assert } from "../../../../../core/assert.js";
-import { WebGPUArray } from "../../../shader/type/WebGPUArray.js";
-import { chunk_round_up_u32 } from "../../../shader/chunk/math/chunk_round_up_u32.js";
-import { CodeChunk } from "../../../shader/compiler/CodeChunk.js";
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-import { ShaderResourceSetDescriptor } from "../../../shader/resource/ShaderResourceSetDescriptor.js";
-import { chunk_safe_load_uint4_umax } from "./chunk_safe_load_uint4_umax.js";
-import {
-    BITS_PER_PASS,
-    NUM_BINS,
-    NUMBER_OF_BLOCKS_PER_GROUP,
-    RADIX_SORT_GROUP_SIZE,
-    RADIX_SORT_KEYS_TYPE,
-    RADIX_SORT_PASS_PARAMETERS_STRUCT
-} from "./defs.js";
-
-const resources = new ShaderResourceSetDescriptor();
-resources.createGroup()
-    .addUniform("settings", RADIX_SORT_PASS_PARAMETERS_STRUCT)
-    .addStorageBuffer("input_keys", RADIX_SORT_KEYS_TYPE);
-
-resources.createGroup()
-    .addStorageBuffer("histogram", WebGPUArray.u32, true);
-
-
-assert.isNonNegativeInteger(NUM_BINS, 'NUM_BINS'); // must be divisible
-
-
-/**
- * see https://github.com/GPUOpen-Drivers/gpurt/blob/f734985ebc31f471c376ed0cb217f43bdd40ee17/src/shaders/RadixSort/BitHistogram.hlsl
- * @type {CodeChunk}
- */
-const body = CodeChunk.from(
-    //language=WGSL
-    `
-const GROUP_SIZE = ${RADIX_SORT_GROUP_SIZE};
-const BITS_PER_PASS = ${BITS_PER_PASS};
-const NUM_BINS = ${NUM_BINS};
-const NUMBER_OF_BLOCKS_PER_GROUP = ${NUMBER_OF_BLOCKS_PER_GROUP};
-
-var<workgroup> SharedMem:array<atomic<u32>, ${RADIX_SORT_GROUP_SIZE * NUM_BINS} >;
-
-@compute @workgroup_size(${RADIX_SORT_GROUP_SIZE}, 1, 1)
-fn main(
-    @builtin(local_invocation_index) localId: u32,
-    @builtin(workgroup_id) workgroup_id: vec3<u32>,
-    @builtin(num_workgroups) workgroup_count: vec3<u32>,
-){
-      
-    let groupId = workgroup_id.x;
-    let totalGroups = workgroup_count.x;
-    let numElements = settings.element_count;
-    let bitShiftCount = settings.bit_shift;
-    
-    // Clear local histogram
-    for (var i = 0; i < NUM_BINS; i++)
-    {
-        SharedMem[i * GROUP_SIZE + localId] = 0;
-    }
-
-    // Make sure everything is up to date
-    workgroupBarrier();
-
-    const numblocks_per_group = NUMBER_OF_BLOCKS_PER_GROUP;
-    const numelems_per_group = numblocks_per_group * GROUP_SIZE;
-
-    let numblocks_total = round_up_u32(numElements, GROUP_SIZE * 4);
-    let maxblocks = numblocks_total - groupId * numblocks_per_group;
-
-    var loadidx = groupId * numelems_per_group + localId;
-    
-    let block_count = min(numblocks_per_group, maxblocks);
-    
-    for (let block = 0; block < block_count; block++){
-    
-        // Load single int4 value
-        let key: vec4u = safe_load_uint4_umax(
-                    &input_keys,
-                    loadidx,
-                    numElements
-        );
-
-        // Handle value adding histogram bins
-        // for all 4 elements
-        let bin = ((key >> bitShiftCount) & 0xF);
-
-        atomicAdd(SharedMem[bin.x * GROUP_SIZE + localId], 1);
-        atomicAdd(SharedMem[bin.y * GROUP_SIZE + localId], 1);
-        atomicAdd(SharedMem[bin.z * GROUP_SIZE + localId], 1);
-        atomicAdd(SharedMem[bin.w * GROUP_SIZE + localId], 1);
-        
-        loadidx += GROUP_SIZE;
-    }
-
-    // Wait for all threads to finish
-    workgroupBarrier();
-
-    var sum = 0;
-    if (localId < NUM_BINS)
-    {
-        for (var i = 0; i < GROUP_SIZE; i++)
-        {
-            sum += SharedMem[localId * GROUP_SIZE + i];
-        }
-
-        let outputIndex = totalGroups * localId + groupId;
-        histogram[outputOffset + outputIndex ] = sum;
-    }
-}
-    `, [
-        chunk_round_up_u32,
-        chunk_safe_load_uint4_umax,
-    ]
-)
-
-export const shader_bit_histogram = ComputeShader.from({
-    label: "Bit Histogram",
-    resources,
-    body,
-})

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_distribute_part_sum_int4.d.ts

@@ -1,3 +0,0 @@
-export const shader_distribute_part_sum_int4: ComputeShader;
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-//# sourceMappingURL=shader_distribute_part_sum_int4.d.ts.map

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_distribute_part_sum_int4.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"shader_distribute_part_sum_int4.d.ts","sourceRoot":"","sources":["../../../../../../../src/shade/renderer/gpu_primitive/sort/ffx/shader_distribute_part_sum_int4.js"],"names":[],"mappings":"AAuEA,4DAIG;8BAvE2B,kCAAkC"}

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_distribute_part_sum_int4.js

@@ -1,76 +0,0 @@
-import { u32 } from "../../../shader/type/WEBGPU_WGSL_PRIMITIVE_TYPES.js";
-import { WebGPUArray } from "../../../shader/type/WebGPUArray.js";
-import { WebGPUStruct } from "../../../shader/type/WebGPUStruct.js";
-import { CodeChunk } from "../../../shader/compiler/CodeChunk.js";
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-import { ShaderResourceSetDescriptor } from "../../../shader/resource/ShaderResourceSetDescriptor.js";
-import { chunk_safe_load_uint4, chunk_safe_store_uint4 } from "./chunk_safe_load_uint4_umax.js";
-import { RADIX_SORT_GROUP_SIZE, RADIX_SORT_KEYS_TYPE } from "./defs.js";
-
-const ROOT_CONSTANTS = WebGPUStruct.from({
-    num_elements: u32,
-});
-
-const resources = new ShaderResourceSetDescriptor();
-
-resources.createGroup()
-    .addUniform('root_constants', ROOT_CONSTANTS)
-    .addStorageBuffer('partial_sums', WebGPUArray.u32, true)
-    .addStorageBuffer('in_out_array', RADIX_SORT_KEYS_TYPE, true)
-
-
-/**
- * see https://github.com/GPUOpen-Drivers/gpurt/blob/f734985ebc31f471c376ed0cb217f43bdd40ee17/src/shaders/RadixSort/DistributePartSumInt4.hlsl
- * @type {CodeChunk}
- */
-const body = CodeChunk.from(
-    //language=WGSL
-    `
-
-// This corresponds to the [numthreads] attribute in HLSL.
-// It can be a constant or a pipeline-overridable constant.
-const GROUP_SIZE = ${RADIX_SORT_GROUP_SIZE};
-
-@compute
-@workgroup_size(GROUP_SIZE, 1, 1)
-fn distribute_part_sum_int4(
-    @builtin(global_invocation_id) global_id: vec3<u32>,
-    @builtin(group_id) group_id: vec3<u32>
-) {
-    // The logic from the DistributePartSumInt4Impl function is inlined here.
-
-    // 1. Load the data vector (vec4<i32>) for the current thread.
-    var v1 = safe_load_uint4(
-        &in_out_array,
-        global_id.x,
-        root_constants.num_elements
-    );
-
-    // 2. Calculate the index to load the partial sum for this workgroup.
-    // The original HLSL is: partialSumsOffset + ((groupId >> 1) * sizeof(int)).
-    // In WGSL, we convert the byte offset to an element index by dividing by 4.
-    let sum_index = (group_id.x >> 1u);
-    let sum = partial_sums[sum_index];
-
-    // 3. Add the group's partial sum to all four components of the vector.
-    v1 += vec4(sum);
-
-    // 4. Store the modified vector back into the buffer.
-    safe_store_uint4(
-        v1,
-        &in_out_array,
-        global_id.x,
-        root_constants.num_elements
-    );
-}
-    `, [
-        chunk_safe_load_uint4,
-        chunk_safe_store_uint4,
-    ]
-)
-
-export const shader_distribute_part_sum_int4 = ComputeShader.from({
-    label: "Distribute Part Sum Int4",
-    resources,
-    body,
-});

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_partial.d.ts

@@ -1,3 +0,0 @@
-export const shader_scan_exclusive_partial: ComputeShader;
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-//# sourceMappingURL=shader_scan_exclusive_partial.d.ts.map

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_partial.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"shader_scan_exclusive_partial.d.ts","sourceRoot":"","sources":["../../../../../../../src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_partial.js"],"names":[],"mappings":"AAwHA,0DAIG;8BA1H2B,kCAAkC"}

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_partial.js

@@ -1,125 +0,0 @@
-import { WebGPUArray } from "../../../shader/type/WebGPUArray.js";
-import { CodeChunk } from "../../../shader/compiler/CodeChunk.js";
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-import { ShaderResourceSetDescriptor } from "../../../shader/resource/ShaderResourceSetDescriptor.js";
-import { chunk_safe_load_uint4, chunk_safe_store_uint4 } from "./chunk_safe_load_uint4_umax.js";
-import { RADIX_SORT_GROUP_SIZE, RADIX_SORT_KEYS_TYPE, RADIX_SORT_PASS_PARAMETERS_STRUCT } from "./defs.js";
-
-
-const resources = new ShaderResourceSetDescriptor();
-resources.createGroup()
-    .addUniform("settings", RADIX_SORT_PASS_PARAMETERS_STRUCT)
-
-resources.createGroup()
-    .addStorageBuffer("keys", RADIX_SORT_KEYS_TYPE, true)
-    .addStorageBuffer("partial_sums", WebGPUArray.u32, true);
-
-/**
- * @see https://github.com/GPUOpen-Drivers/gpurt/blob/f734985ebc31f471c376ed0cb217f43bdd40ee17/src/shaders/RadixSort/ScanExclusivePartInt4.hlsl
- * @type {CodeChunk}
- */
-const body = CodeChunk.from(
-    //language=WGSL
-    `
-var<workgroup> SharedMem:array<atomic<u32>, ${RADIX_SORT_GROUP_SIZE} >;
-
-fn group_scan_exclusive_part_int4(
-    localId: u32, 
-    groupSize: u32,
-) -> u32 {
-    var sum = 0u;
-    
-    // 1. Up-sweep phase (reduction)
-    for (var stride = 1u; stride <= (groupSize >> 1); stride <<= 1)
-    {
-        if (localId < groupSize / (2 * stride)){
-        
-            let a = 2*(localId + 1)*stride - 1;
-            let b = (2*localId + 1)*stride - 1;
-            
-            SharedMem[a] += SharedMem[b];
-        }
-
-        // Wait for all threads to finish
-        workgroupBarrier();
-    }
-
-    if (localId == 0)
-    {
-        sum = SharedMem[groupSize - 1];
-        SharedMem[groupSize - 1] = 0;
-    }
-
-    workgroupBarrier();
-    
-    // 2. Down-sweep phase
-    for (var stride = (groupSize >> 1); stride > 0; stride >>= 1)
-    {
-        if (localId < groupSize / (2 * stride))
-        {
-            let a = (2 * localId + 1)*stride - 1;
-            let b = 2 * (localId + 1)*stride - 1;
-            
-            let temp = SharedMem[a];
-            
-            SharedMem[a] = SharedMem[b]; // swap
-            SharedMem[b] += temp;
-        }
-
-        workgroupBarrier();
-    }
-
-    return sum;
-}
-
-@compute @workgroup_size(${RADIX_SORT_GROUP_SIZE}, 1, 1)
-fn main(
-    @builtin(local_invocation_index) localId: u32,
-    @builtin(workgroup_id) workgroup_id: vec3<u32>,
-){
-    let groupId = workgroup_id.x;
-    let numElements = settings.element_count;
-
-    var v1 = safe_load_uint4(&keys, 2 * globalId, numElements);
-    var v2 = safe_load_uint4(&keys, 2 * globalId + 1, numElements);
-    
-    v1.y += v1.x; v1.w += v1.z; v1.w += v1.y;
-    v2.y += v2.x; v2.w += v2.z; v2.w += v2.y;
-    v2.w += v1.w;
-    
-    SharedMem[localId] = v2.w;
-    
-    // Sync shared memory writes
-    workgroupBarrier();
-    
-    let sum = group_scan_exclusive_part_int4(localId, ${RADIX_SORT_GROUP_SIZE});
-    
-    if (localId == 0)
-    {
-        partial_sums[groupId] = sum;
-    }
-    
-    v2.w = SharedMem[localId];
-    
-    var t = v1.w; v1.w = v2.w; v2.w += t;
-    t = v1.y; v1.y = v1.w; v1.w += t;
-    t = v2.y; v2.y = v2.w; v2.w += t;
-    t = v1.x; v1.x = v1.y; v1.y += t;
-    t = v2.x; v2.x = v2.y; v2.y += t;
-    t = v1.z; v1.z = v1.w; v1.w += t;
-    t = v2.z; v2.z = v2.w; v2.w += t;
-    
-    safe_store_uint4(&keys, v2, 2 * globalId + 1, numElements);
-    safe_store_uint4(&keys, v1, 2 * globalId, numElements);
-}
-    `, [
-        chunk_safe_load_uint4,
-        chunk_safe_store_uint4
-    ]
-);
-
-export const shader_scan_exclusive_partial = ComputeShader.from({
-    label: "Scan Exclusive Partial",
-    body,
-    resources,
-});

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_uint4.d.ts

@@ -1,3 +0,0 @@
-export const shader_scan_exclusive_uint4: ComputeShader;
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-//# sourceMappingURL=shader_scan_exclusive_uint4.d.ts.map

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_uint4.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"shader_scan_exclusive_uint4.d.ts","sourceRoot":"","sources":["../../../../../../../src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_uint4.js"],"names":[],"mappings":"AAuIA,wDAIE;8BAxI4B,kCAAkC"}

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scan_exclusive_uint4.js

@@ -1,140 +0,0 @@
-import { u32 } from "../../../shader/type/WEBGPU_WGSL_PRIMITIVE_TYPES.js";
-import { WebGPUStruct } from "../../../shader/type/WebGPUStruct.js";
-import { CodeChunk } from "../../../shader/compiler/CodeChunk.js";
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-import { ShaderResourceSetDescriptor } from "../../../shader/resource/ShaderResourceSetDescriptor.js";
-import { chunk_safe_load_uint4, chunk_safe_store_uint4 } from "./chunk_safe_load_uint4_umax.js";
-import { RADIX_SORT_GROUP_SIZE, RADIX_SORT_KEYS_TYPE } from "./defs.js";
-
-const resources = new ShaderResourceSetDescriptor();
-resources.createGroup()
-    .addUniform('root_constants', WebGPUStruct.from({
-        num_elements: u32,
-    }))
-    .addStorageBuffer('in_out_array', RADIX_SORT_KEYS_TYPE, true)
-
-/**
- * @see https://github.com/GPUOpen-Drivers/gpurt/blob/f734985ebc31f471c376ed0cb217f43bdd40ee17/src/shaders/RadixSort/ScanExclusiveInt4.hlsl
- * @type {CodeChunk}
- */
-const body = CodeChunk.from(
-    //language=WGSL
-    `
-var<workgroup> shared_mem: array<u32, ${RADIX_SORT_GROUP_SIZE} >;
-
-
-// --- Helper & Implementation Functions ---
-
-// This is a direct translation of the GroupScanExclusiveInt4 helper function.
-// It performs a two-pass exclusive scan (up-sweep and down-sweep) on the data
-// already present in the workgroup's shared memory.
-fn group_scan_exclusive(local_id: u32, group_size: u32) {
-
-    // 1. Up-sweep phase (reduction)
-    for (var stride = 1u; stride <= (group_size >> 1u); stride <<= 1u) {
-    
-        if (local_id < group_size / (2u * stride)) {
-        
-            let idx1 = (2u * (local_id + 1u) * stride) - 1u;
-            let idx2 = ((2u * local_id + 1u) * stride) - 1u;
-            
-            shared_mem[idx1] = shared_mem[idx1] + shared_mem[idx2];
-        }
-        
-        workgroupBarrier();
-    }
-
-    // Set the last element to 0 for an exclusive scan
-    if (local_id == 0u) {
-        shared_mem[group_size - 1u] = 0;
-    }
-    
-    workgroupBarrier();
-
-    // 2. Down-sweep phase
-    for (var stide = group_size >> 1u; stride > 0u; stride = stride >> 1u) {
-       
-        if (local_id < group_size / (2u * stride)) {
-        
-            let idx1 = ((2u * local_id + 1u) * stride) - 1u;
-            let idx2 = (2u * (local_id + 1u) * stride) - 1u;
-            
-            let temp = shared_mem[idx1];
-            
-            shared_mem[idx1] = shared_mem[idx2];
-            shared_mem[idx2] = shared_mem[idx2] + temp;
-            
-        }
-        
-        workgroupBarrier();
-    }
-}
-
-// This function contains the main logic, equivalent to ScanExclusiveInt4Impl.
-fn scan_exclusive_int4_impl(global_id: u32, local_id: u32) {
-    // Load two vectors per thread
-    var v1 = safe_load_uint4(
-        &in_out_array, 2u * global_id, root_constants.num_elements
-    );
-    var v2 = safe_load_uint4(
-        &in_out_array, 2u * global_id + 1u, root_constants.num_elements
-    );
-
-    // Perform a local exclusive scan within the 8 integers (2x vec4)
-    v1.y = v1.y + v1.x;
-    v1.w = v1.w + v1.z;
-    v1.w = v1.w + v1.y;
-
-    v2.y = v2.y + v2.x;
-    v2.w = v2.w + v2.z;
-    v2.w = v2.w + v2.y;
-
-    // The sum of the first vector is added to the second
-    v2.w = v2.w + v1.w;
-
-    // Store the total sum of this thread's elements into shared memory
-    shared_mem[local_id] = v2.w;
-    workgroupBarrier();
-
-    // Perform an exclusive scan on the sums from all threads in the workgroup
-    group_scan_exclusive(local_id, ${RADIX_SORT_GROUP_SIZE});
-
-    // Read the scanned sum from shared memory and distribute it
-    v2.w = shared_mem[local_id];
-
-    // Distribute the scanned prefix sum back through the local 8 integers
-    var t = v1.w; v1.w = v2.w; v2.w = v2.w + t;
-    t = v1.y; v1.y = v1.w; v1.w = v1.w + t;
-    t = v2.y; v2.y = v2.w; v2.w = v2.w + t;
-    t = v1.x; v1.x = v1.y; v1.y = v1.y + t;
-    t = v2.x; v2.x = v2.y; v2.y = v2.y + t;
-    t = v1.z; v1.z = v1.w; v1.w = v1.w + t;
-    t = v2.z; v2.z = v2.w; v2.w = v2.w + t;
-
-    // Store the final results
-    safe_store_uint4(&in_out_array, v2, 2u * global_id + 1u, root_constants.num_elements);
-    safe_store_uint4(&in_out_array, v1, 2u * global_id, root_constants.num_elements);
-}
-
-@compute
-@workgroup_size(${RADIX_SORT_GROUP_SIZE}, 1, 1)
-fn scan_exclusive_int4(
-    @builtin(global_invocation_id) global_id_vec: vec3<u32>,
-    @builtin(local_invocation_id) local_id_vec: vec3<u32>
-) {
-    let global_id = global_id_vec.x;
-    let local_id = local_id_vec.x;
-
-    scan_exclusive_int4_impl(global_id, local_id);
-}
-    `, [
-        chunk_safe_store_uint4,
-        chunk_safe_load_uint4,
-    ]
-)
-
-export const shader_scan_exclusive_uint4 = ComputeShader.from({
-    label: "Scan Exclusive Uint4",
-    body,
-    resources,
-})

package/src/shade/renderer/gpu_primitive/sort/ffx/shader_scatter_keys_and_values.d.ts

@@ -1,3 +0,0 @@
-export const shader_scatter_keys_and_values: ComputeShader;
-import { ComputeShader } from "../../../shader/ComputeShader.js";
-//# sourceMappingURL=shader_scatter_keys_and_values.d.ts.map