yum-mirror/slang

Making it easier to work with shaders

git clone https://git.yummers.dev/yum-mirror/slang

Jay KwakSmoke test WASM as a part of CI (#6969)15fce7c8f

master
branchesmasterall refs →
1.8 KiB63 linesraw
1// Hybrid Tausworthe PRNG
2//
3// Code adapted from: https://indico.cern.ch/event/93877/papers/2118070/files/4416-acat3.pdf (See document for license)
4//
5
6uniform float seed;
7RWStructuredBuffer<float> outputBuffer;
8
9uint seedPerThread(uint idx)
10{
11    return ((uint)idx + (uint)(seed * 1000000)) * 1099087573UL;
12}
13
14uint tauStep(uint z, uint s1, uint s2, uint s3, uint M)
15{
16    uint b = (((z << s1) ^ z) >> s2);
17    return (((z & M) << s3) ^ b);
18}
19
20[shader("compute")]
21[numthreads(64, 1, 1)]
22void computeMain(uint2 dispatchThreadId : SV_DispatchThreadID)
23{
24    uint idx = dispatchThreadId.x;
25    uint val = ((uint)idx) * 1099087573UL + ((uint)seed) * 12003927;
26
27    uint z = tauStep(val, 13, 19, 12, 4294967294);
28    z = tauStep(z, 2, 25, 4, 4294967288);
29    z = tauStep(z, 3, 11, 17, 4294967280);
30
31    uint z1, z2, z3, z4;
32    uint r0, r1, r2, r3;
33
34    // STEP 1
35    uint _seed = seedPerThread(idx);
36    z1 = tauStep(_seed, 13, 19, 12, 429496729UL);
37    z2 = tauStep(_seed, 2, 25, 4, 4294967288UL);
38    z3 = tauStep(_seed, 3, 11, 17, 429496280UL);
39    z4 = (1664525 * _seed + 1013904223UL);
40    r0 = (z1 ^ z2 ^ z3 ^ z4);
41    // STEP 2
42    z1 = tauStep(r0, 13, 19, 12, 429496729UL);
43    z2 = tauStep(r0, 2, 25, 4, 4294967288UL);
44    z3 = tauStep(r0, 3, 11, 17, 429496280UL);
45    z4 = (1664525 * r0 + 1013904223UL);
46    r1 = (z1 ^ z2 ^ z3 ^ z4);
47    // STEP 3
48    z1 = tauStep(r1, 13, 19, 12, 429496729UL);
49    z2 = tauStep(r1, 2, 25, 4, 4294967288UL);
50    z3 = tauStep(r1, 3, 11, 17, 429496280UL);
51    z4 = (1664525 * r1 + 1013904223UL);
52    r2 = (z1 ^ z2 ^ z3 ^ z4);
53    // STEP 4
54    z1 = tauStep(r2, 13, 19, 12, 429496729UL);
55    z2 = tauStep(r2, 2, 25, 4, 4294967288UL);
56    z3 = tauStep(r2, 3, 11, 17, 429496280UL);
57    z4 = (1664525 * r2 + 1013904223UL);
58    r3 = (z1 ^ z2 ^ z3 ^ z4);
59
60    float u4 = r3 * 2.3283064365387e-10;
61
62    outputBuffer[idx] = u4;
63}