yum-mirror/slang

Making it easier to work with shaders

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

jsmall-nvidiaWAR for double/host-callable/x86 (#2314)9261c7a23

master
branchesmasterall refs →
2.1 KiB95 linesraw
1// It doesn't look like fxc, dxc, vk support double versions of many of the intrinsics, so they are disabled here.
2
3// TODO(JS):
4// NOTE! war-double-host-callable category is a work around for issues with x86 & host-callable
5
6//TEST(compute, war-double-host-callable):COMPARE_COMPUTE_EX:-cpu -compute 
7//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute
8//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12
9//DISABLE_TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -render-feature double
10//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute
11
12//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name outputBuffer
13RWStructuredBuffer<int> outputBuffer;
14
15[numthreads(4, 1, 1)]
16void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
17{
18    int idx = int(dispatchThreadID.x);
19    
20    double f = idx * (1.0f / (4.0f));
21
22    int it = 0;
23    double ft = 0.0l;
24    
25    // fmod
26    ft += int(((f % 0.11l) * 100) + 0.5l);
27    
28    ft += sin(f);
29    ft += cos(f);
30    ft += tan(f);
31    
32    ft += asin(f);
33    ft += acos(f);
34    ft += atan(f);
35    
36    ft += atan2(f, 2.0l);
37    
38    {
39        double sf, cf;
40        sincos(f, sf, cf);
41        
42        ft += sf;
43        ft += cf;
44    }
45    
46    ft += rcp(1.0l + f);
47    ft += sign(f - 0.5l);
48    
49    ft += saturate(f * 4 - 2.0l);
50    
51    ft += sqrt(f);
52    ft += rsqrt(1.0l + f);
53           
54    ft += exp2(f);
55    ft += exp(f);
56        
57        
58    ft += frac(f * 3);
59    ft += ceil(f * 5 - 3);
60    
61    ft += floor(f * 10 - 7);
62    ft += trunc(f * 7);
63    
64    ft += log(f + 10.0l);
65    ft += log2(f * 3 + 2);
66
67    {
68        double v[] = { 1, 10, 100, 1000 };
69        ft += int(log10(v[idx]) + 0.5l);
70    }
71       
72    ft += abs(f * 4 - 2.0l);
73           
74    ft += min(0.5l, f);
75    ft += max(f, 0.75l);
76
77    ft += pow(0.5l, f);
78
79    ft += smoothstep(0.2l, 0.7l, f);
80    ft += lerp(-100.0l, 100.0l, f);
81
82    ft += clamp(f, 0.1l, 0.3l);
83
84    ft += step(f, 0.5l);
85
86    {
87        uint low, high;
88        asuint(f * 2.0l, low, high);
89        ft += asdouble(low, high);
90    }
91    
92    ft += ldexp(23.2, f);
93    
94    outputBuffer[idx] = int(ft * 16);
95}