1 /**************************************************************************
2 *
3 * Copyright 2010-2021 VMware, Inc.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
18 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
19 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
20 * USE OR OTHER DEALINGS IN THE SOFTWARE.
21 *
22 * The above copyright notice and this permission notice (including the
23 * next paragraph) shall be included in all copies or substantial portions
24 * of the Software.
25 *
26 **************************************************************************/
27
28
29 #include "pipe/p_config.h"
30
31 #include "util/u_math.h"
32 #include "util/u_cpu_detect.h"
33 #include "util/u_pack_color.h"
34 #include "util/u_surface.h"
35 #include "util/u_sse.h"
36
37 #include "lp_jit.h"
38 #include "lp_rast.h"
39 #include "lp_debug.h"
40 #include "lp_state_fs.h"
41 #include "lp_linear_priv.h"
42
43
44 #if defined(PIPE_ARCH_SSE)
45
46
47 /* This file contains various special-case fastpaths which implement
48 * the entire linear pipeline in a single funciton.
49 *
50 * These include simple blits and some debug code.
51 *
52 * These functions fully implement the linear path and do not need to
53 * be combined with blending, interpolation or sampling routines.
54 */
55
56 /* Linear shader which implements the BLIT_RGBA shader with the
57 * additional constraints imposed by lp_setup_is_blit().
58 */
59 static boolean
lp_linear_blit_rgba_blit(const struct lp_rast_state * state,unsigned x,unsigned y,unsigned width,unsigned height,const float (* a0)[4],const float (* dadx)[4],const float (* dady)[4],uint8_t * color,unsigned stride)60 lp_linear_blit_rgba_blit(const struct lp_rast_state *state,
61 unsigned x, unsigned y,
62 unsigned width, unsigned height,
63 const float (*a0)[4],
64 const float (*dadx)[4],
65 const float (*dady)[4],
66 uint8_t *color,
67 unsigned stride)
68 {
69 const struct lp_jit_context *context = &state->jit_context;
70 const struct lp_jit_texture *texture = &context->textures[0];
71 const uint8_t *src;
72 unsigned src_stride;
73 int src_x, src_y;
74
75 LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);
76
77 /* Require w==1.0:
78 */
79 if (a0[0][3] != 1.0 ||
80 dadx[0][3] != 0.0 ||
81 dady[0][3] != 0.0)
82 return FALSE;
83
84 src_x = x + util_iround(a0[1][0]*texture->width - 0.5f);
85 src_y = y + util_iround(a0[1][1]*texture->height - 0.5f);
86
87 src = texture->base;
88 src_stride = texture->row_stride[0];
89
90 /* Fall back to blit_rgba() if clamping required:
91 */
92 if (src_x < 0 ||
93 src_y < 0 ||
94 src_x + width > texture->width ||
95 src_y + height > texture->height)
96 return FALSE;
97
98 util_copy_rect(color, PIPE_FORMAT_B8G8R8A8_UNORM, stride,
99 x, y,
100 width, height,
101 src, src_stride,
102 src_x, src_y);
103
104 return TRUE;
105 }
106
107
108 /* Linear shader which implements the BLIT_RGB1 shader, with the
109 * additional constraints imposed by lp_setup_is_blit().
110 */
111 static boolean
lp_linear_blit_rgb1_blit(const struct lp_rast_state * state,unsigned x,unsigned y,unsigned width,unsigned height,const float (* a0)[4],const float (* dadx)[4],const float (* dady)[4],uint8_t * color,unsigned stride)112 lp_linear_blit_rgb1_blit(const struct lp_rast_state *state,
113 unsigned x, unsigned y,
114 unsigned width, unsigned height,
115 const float (*a0)[4],
116 const float (*dadx)[4],
117 const float (*dady)[4],
118 uint8_t *color,
119 unsigned stride)
120 {
121 const struct lp_jit_context *context = &state->jit_context;
122 const struct lp_jit_texture *texture = &context->textures[0];
123 const uint8_t *src;
124 unsigned src_stride;
125 int src_x, src_y;
126
127 LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);
128
129 /* Require w==1.0:
130 */
131 if (a0[0][3] != 1.0 ||
132 dadx[0][3] != 0.0 ||
133 dady[0][3] != 0.0)
134 return FALSE;
135
136 color += x * 4 + y * stride;
137
138 src_x = x + util_iround(a0[1][0]*texture->width - 0.5f);
139 src_y = y + util_iround(a0[1][1]*texture->height - 0.5f);
140
141 src = texture->base;
142 src_stride = texture->row_stride[0];
143 src += src_x * 4;
144 src += src_y * src_stride;
145
146 if (src_x < 0 ||
147 src_y < 0 ||
148 src_x + width > texture->width ||
149 src_y + height > texture->height)
150 return FALSE;
151
152 for (y = 0; y < height; y++) {
153 const uint32_t *src_row = (const uint32_t *)src;
154 uint32_t *dst_row = (uint32_t *)color;
155
156 for (x = 0; x < width; x++) {
157 *dst_row++ = *src_row++ | 0xff000000;
158 }
159
160 color += stride;
161 src += src_stride;
162 }
163
164 return TRUE;
165 }
166
167 /* Linear shader which always emits purple. Used for debugging.
168 */
169 static boolean
lp_linear_purple(const struct lp_rast_state * state,unsigned x,unsigned y,unsigned width,unsigned height,const float (* a0)[4],const float (* dadx)[4],const float (* dady)[4],uint8_t * color,unsigned stride)170 lp_linear_purple(const struct lp_rast_state *state,
171 unsigned x, unsigned y,
172 unsigned width, unsigned height,
173 const float (*a0)[4],
174 const float (*dadx)[4],
175 const float (*dady)[4],
176 uint8_t *color,
177 unsigned stride)
178 {
179 union util_color uc;
180
181 util_pack_color_ub(0xff, 0, 0xff, 0xff,
182 PIPE_FORMAT_B8G8R8A8_UNORM, &uc);
183
184 util_fill_rect(color,
185 PIPE_FORMAT_B8G8R8A8_UNORM,
186 stride,
187 x,
188 y,
189 width,
190 height,
191 &uc);
192
193 return TRUE;
194 }
195
196 /* Examine the fragment shader varient and determine whether we can
197 * substitute a fastpath linear shader implementation.
198 */
199 boolean
lp_linear_check_fastpath(struct lp_fragment_shader_variant * variant)200 lp_linear_check_fastpath(struct lp_fragment_shader_variant *variant)
201 {
202 struct lp_sampler_static_state *samp0 = lp_fs_variant_key_sampler_idx(&variant->key, 0);
203
204 if (!samp0)
205 return false;
206
207 enum pipe_format tex_format = samp0->texture_state.format;
208 if (variant->shader->kind == LP_FS_KIND_BLIT_RGBA &&
209 tex_format == PIPE_FORMAT_B8G8R8A8_UNORM &&
210 is_nearest_clamp_sampler(samp0) &&
211 variant->opaque) {
212 variant->jit_linear_blit = lp_linear_blit_rgba_blit;
213 }
214
215 if (variant->shader->kind == LP_FS_KIND_BLIT_RGB1 &&
216 variant->opaque &&
217 (tex_format == PIPE_FORMAT_B8G8R8A8_UNORM ||
218 tex_format == PIPE_FORMAT_B8G8R8X8_UNORM) &&
219 is_nearest_clamp_sampler(samp0)) {
220 variant->jit_linear_blit = lp_linear_blit_rgb1_blit;
221 }
222
223 if (0) {
224 variant->jit_linear = lp_linear_purple;
225 }
226
227
228 /* Stop now if jit_linear has been initialized. Otherwise keep
229 * searching - even if jit_linear_blit has been instantiated.
230 */
231 return variant->jit_linear != NULL;
232 }
233 #else
234 boolean
lp_linear_check_fastpath(struct lp_fragment_shader_variant * variant)235 lp_linear_check_fastpath(struct lp_fragment_shader_variant *variant)
236 {
237 return FALSE;
238 }
239 #endif
240
241