1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
|
#version 450 core
#extension GL_AMD_gpu_shader_half_float: enable
#extension GL_ARB_gpu_shader_int64: enable
void main()
{
}
// Half float literals
void literal()
{
const float16_t f16c = 0.000001hf;
const f16vec2 f16cv = f16vec2(-0.25HF, 0.03HF);
f16vec2 f16v;
f16v.x = f16c;
f16v += f16cv;
}
// Block memory layout
struct S
{
float16_t x; // rule 1: align = 2, takes offsets 0-1
f16vec2 y; // rule 2: align = 4, takes offsets 4-7
f16vec3 z; // rule 3: align = 8, takes offsets 8-13
};
layout(column_major, std140) uniform B1
{
float16_t a; // rule 1: align = 2, takes offsets 0-1
f16vec2 b; // rule 2: align = 4, takes offsets 4-7
f16vec3 c; // rule 3: align = 8, takes offsets 8-15
float16_t d[2]; // rule 4: align = 16, array stride = 16,
// takes offsets 16-47
f16mat2x3 e; // rule 5: align = 16, matrix stride = 16,
// takes offsets 48-79
f16mat2x3 f[2]; // rule 6: align = 16, matrix stride = 16,
// array stride = 32, f[0] takes
// offsets 80-111, f[1] takes offsets
// 112-143
S g; // rule 9: align = 16, g.x takes offsets
// 144-145, g.y takes offsets 148-151,
// g.z takes offsets 152-159
S h[2]; // rule 10: align = 16, array stride = 16, h[0]
// takes offsets 160-175, h[1] takes
// offsets 176-191
};
layout(row_major, std430) buffer B2
{
float16_t o; // rule 1: align = 2, takes offsets 0-1
f16vec2 p; // rule 2: align = 4, takes offsets 4-7
f16vec3 q; // rule 3: align = 8, takes offsets 8-13
float16_t r[2]; // rule 4: align = 2, array stride = 2, takes
// offsets 14-17
f16mat2x3 s; // rule 7: align = 4, matrix stride = 4, takes
// offsets 20-31
f16mat2x3 t[2]; // rule 8: align = 4, matrix stride = 4, array
// stride = 12, t[0] takes offsets
// 32-43, t[1] takes offsets 44-55
S u; // rule 9: align = 8, u.x takes offsets
// 56-57, u.y takes offsets 60-63, u.z
// takes offsets 64-69
S v[2]; // rule 10: align = 8, array stride = 16, v[0]
// takes offsets 72-87, v[1] takes
// offsets 88-103
};
// Specialization constant
layout(constant_id = 100) const float16_t sf16 = 0.125hf;
layout(constant_id = 101) const float sf = 0.25;
layout(constant_id = 102) const double sd = 0.5lf;
const float f16_to_f = float(sf16);
const double f16_to_d = float(sf16);
const float16_t f_to_f16 = float16_t(sf);
const float16_t d_to_f16 = float16_t(sd);
void operators()
{
float16_t f16;
f16vec2 f16v;
f16mat2x2 f16m;
bool b;
// Arithmetic
f16v += f16v;
f16v -= f16v;
f16v *= f16v;
f16v /= f16v;
f16v++;
f16v--;
++f16m;
--f16m;
f16v = -f16v;
f16m = -f16m;
f16 = f16v.x + f16v.y;
f16 = f16v.x - f16v.y;
f16 = f16v.x * f16v.y;
f16 = f16v.x / f16v.y;
// Relational
b = (f16v.x != f16);
b = (f16v.y == f16);
b = (f16v.x > f16);
b = (f16v.y < f16);
b = (f16v.x >= f16);
b = (f16v.y <= f16);
// Vector/matrix operations
f16v = f16v * f16;
f16m = f16m * f16;
f16v = f16m * f16v;
f16v = f16v * f16m;
f16m = f16m * f16m;
}
void typeCast()
{
bvec3 bv;
vec3 fv;
dvec3 dv;
ivec3 iv;
uvec3 uv;
i64vec3 i64v;
u64vec3 u64v;
f16vec3 f16v;
f16v = f16vec3(bv); // bool -> float16
bv = bvec3(f16v); // float16 -> bool
f16v = f16vec3(fv); // float -> float16
fv = vec3(f16v); // float16 -> float
f16v = f16vec3(dv); // double -> float16
dv = dvec3(dv); // float16 -> double
f16v = f16vec3(iv); // int -> float16
iv = ivec3(f16v); // float16 -> int
f16v = f16vec3(uv); // uint -> float16
uv = uvec3(f16v); // float16 -> uint
f16v = f16vec3(i64v); // int64 -> float16
i64v = i64vec3(f16v); // float16 -> int64
f16v = f16vec3(u64v); // uint64 -> float16
u64v = u64vec3(f16v); // float16 -> uint64
}
void builtinAngleTrigFuncs()
{
f16vec4 f16v1, f16v2;
f16v2 = radians(f16v1);
f16v2 = degrees(f16v1);
f16v2 = sin(f16v1);
f16v2 = cos(f16v1);
f16v2 = tan(f16v1);
f16v2 = asin(f16v1);
f16v2 = acos(f16v1);
f16v2 = atan(f16v1, f16v2);
f16v2 = atan(f16v1);
f16v2 = sinh(f16v1);
f16v2 = cosh(f16v1);
f16v2 = tanh(f16v1);
f16v2 = asinh(f16v1);
f16v2 = acosh(f16v1);
f16v2 = atanh(f16v1);
}
void builtinExpFuncs()
{
f16vec2 f16v1, f16v2;
f16v2 = pow(f16v1, f16v2);
f16v2 = exp(f16v1);
f16v2 = log(f16v1);
f16v2 = exp2(f16v1);
f16v2 = log2(f16v1);
f16v2 = sqrt(f16v1);
f16v2 = inversesqrt(f16v1);
}
void builtinCommonFuncs()
{
f16vec3 f16v1, f16v2, f16v3;
float16_t f16;
bool b;
bvec3 bv;
ivec3 iv;
f16v2 = abs(f16v1);
f16v2 = sign(f16v1);
f16v2 = floor(f16v1);
f16v2 = trunc(f16v1);
f16v2 = round(f16v1);
f16v2 = roundEven(f16v1);
f16v2 = ceil(f16v1);
f16v2 = fract(f16v1);
f16v2 = mod(f16v1, f16v2);
f16v2 = mod(f16v1, f16);
f16v3 = modf(f16v1, f16v2);
f16v3 = min(f16v1, f16v2);
f16v3 = min(f16v1, f16);
f16v3 = max(f16v1, f16v2);
f16v3 = max(f16v1, f16);
f16v3 = clamp(f16v1, f16, f16v2.x);
f16v3 = clamp(f16v1, f16v2, f16vec3(f16));
f16v3 = mix(f16v1, f16v2, f16);
f16v3 = mix(f16v1, f16v2, f16v3);
f16v3 = mix(f16v1, f16v2, bv);
f16v3 = step(f16v1, f16v2);
f16v3 = step(f16, f16v3);
f16v3 = smoothstep(f16v1, f16v2, f16v3);
f16v3 = smoothstep(f16, f16v1.x, f16v2);
b = isnan(f16);
bv = isinf(f16v1);
f16v3 = fma(f16v1, f16v2, f16v3);
f16v2 = frexp(f16v1, iv);
f16v2 = ldexp(f16v1, iv);
}
void builtinPackUnpackFuncs()
{
uint u;
f16vec2 f16v;
u = packFloat2x16(f16v);
f16v = unpackFloat2x16(u);
}
void builtinGeometryFuncs()
{
float16_t f16;
f16vec3 f16v1, f16v2, f16v3;
f16 = length(f16v1);
f16 = distance(f16v1, f16v2);
f16 = dot(f16v1, f16v2);
f16v3 = cross(f16v1, f16v2);
f16v2 = normalize(f16v1);
f16v3 = faceforward(f16v1, f16v2, f16v3);
f16v3 = reflect(f16v1, f16v2);
f16v3 = refract(f16v1, f16v2, f16);
}
void builtinMatrixFuncs()
{
f16mat2x3 f16m1, f16m2, f16m3;
f16mat3x2 f16m4;
f16mat3 f16m5;
f16mat4 f16m6, f16m7;
f16vec3 f16v1;
f16vec2 f16v2;
float16_t f16;
f16m3 = matrixCompMult(f16m1, f16m2);
f16m1 = outerProduct(f16v1, f16v2);
f16m4 = transpose(f16m1);
f16 = determinant(f16m5);
f16m6 = inverse(f16m7);
}
void builtinVecRelFuncs()
{
f16vec3 f16v1, f16v2;
bvec3 bv;
bv = lessThan(f16v1, f16v2);
bv = lessThanEqual(f16v1, f16v2);
bv = greaterThan(f16v1, f16v2);
bv = greaterThanEqual(f16v1, f16v2);
bv = equal(f16v1, f16v2);
bv = notEqual(f16v1, f16v2);
}
in f16vec3 if16v;
void builtinFragProcFuncs()
{
f16vec3 f16v;
// Derivative
f16v.x = dFdx(if16v.x);
f16v.y = dFdy(if16v.y);
f16v.xy = dFdxFine(if16v.xy);
f16v.xy = dFdyFine(if16v.xy);
f16v = dFdxCoarse(if16v);
f16v = dFdxCoarse(if16v);
f16v.x = fwidth(if16v.x);
f16v.xy = fwidthFine(if16v.xy);
f16v = fwidthCoarse(if16v);
// Interpolation
f16v.x = interpolateAtCentroid(if16v.x);
f16v.xy = interpolateAtSample(if16v.xy, 1);
f16v = interpolateAtOffset(if16v, f16vec2(0.5hf));
}
|
