1 /*
2 * Copyright © 2020 Valve Corporation
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 * SOFTWARE.
22 */
23
24 #ifndef __FREEDRENO_GUARDBAND_H__
25 #define __FREEDRENO_GUARDBAND_H__
26
27 #include <assert.h>
28 #include <math.h>
29 #include <stdbool.h>
30
31 /* All 1's but don't overflow the GUARDBAND_CLIP_ADJ bitfields: */
32 #define MAX_GB 0x1ff
33
34 static inline unsigned
fd_calc_guardband(float offset,float scale,bool is_a3xx)35 fd_calc_guardband(float offset, float scale, bool is_a3xx)
36 {
37 /* On a3xx, the viewport max is 4k and the docs say the max guardband
38 * width is 8k. That is, GRAS cannot handle triangle coordinates more than
39 * 8k, positive or negative. On a4xx+ the viewport width was bumped to
40 * 16k, and so the guardband width was necessarily also bumped. Note that
41 * the numbers here should correspond to
42 * VkPhysicalDeviceLimits::viewportBoundsRange in Vulkan.
43 */
44 const float gb_min = is_a3xx ? -8192. : -32768.;
45 const float gb_max = is_a3xx ? 8191. : 32767.;
46
47 /* Clipping happens in normalized device coordinates, so we have to
48 * transform gb_min and gb_max to ndc using the inverse of the viewport
49 * transform. Avoid flipping min and max by using the absolute value of
50 * the scale.
51 */
52 const float gb_min_ndc = (gb_min - offset) / fabsf(scale);
53 const float gb_max_ndc = (gb_max - offset) / fabsf(scale);
54
55 /* There's only one GB_ADJ field, so presumably the guardband is
56 * [-GB_ADJ, GB_ADJ] like on Radeon. It's always safe to make the
57 * guardband smaller, so we have to take the min to get the largest range
58 * contained in [gb_min_ndc, gb_max_ndc].
59 */
60 const float gb_adj = fminf(-gb_min_ndc, gb_max_ndc);
61
62 /* The viewport should always be contained in the guardband. */
63 if (gb_adj < 1.0)
64 return MAX_GB;
65
66 /* frexp returns an unspecified value if given an infinite value, which
67 * can happen if scale == 0.
68 */
69 if (isinf(gb_adj))
70 return MAX_GB;
71
72 /* Convert gb_adj to 3.6 floating point, rounding down since it's always
73 * safe to make the guard band smaller (but not the other way around!).
74 *
75 * Note: After converting back to a float, the value the blob returns here
76 * is sometimes a little smaller than the value we return. This seems to
77 * happen around the boundary between two different rounded values. For
78 * example, using the a6xx blob:
79 *
80 * min | width | unrounded gb_adj | blob result | mesa result
81 * ------------------------------------------------------------
82 * 0 | 510 | 127.498 | 127. | 127.
83 * 0 | 511 | 127.247 | 126. | 127.
84 * 0 | 512 | 126.996 | 126. | 126.
85 *
86 * The guardband must be 32767 wide, since that's what the blob reports
87 * for viewportBoundsRange, so I'm guessing that they're rounding slightly
88 * more conservatively somehow.
89 */
90 int gb_adj_exp;
91 float gb_adj_mantissa = frexpf(gb_adj, &gb_adj_exp);
92 if (gb_adj_exp <= 0)
93 return MAX_GB;
94
95 /* Round non-representable numbers down to the largest possible number. */
96 if (gb_adj_exp > 8)
97 return MAX_GB;
98
99 return ((gb_adj_exp - 1) << 6) |
100 ((unsigned)truncf(gb_adj_mantissa * (1 << 7)) - (1 << 6));
101 }
102
103 #endif /* __FREEDRENO_GUARDBAND_H__ */
104