Merge branch '7.8'
[mesa.git] / src / gallium / auxiliary / gallivm / lp_bld_sample.c
1 /**************************************************************************
2 *
3 * Copyright 2009 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 above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 /**
29 * @file
30 * Texture sampling -- common code.
31 *
32 * @author Jose Fonseca <jfonseca@vmware.com>
33 */
34
35 #include "pipe/p_defines.h"
36 #include "pipe/p_state.h"
37 #include "util/u_format.h"
38 #include "util/u_math.h"
39 #include "lp_bld_debug.h"
40 #include "lp_bld_const.h"
41 #include "lp_bld_arit.h"
42 #include "lp_bld_type.h"
43 #include "lp_bld_format.h"
44 #include "lp_bld_sample.h"
45
46
47 /**
48 * Initialize lp_sampler_static_state object with the gallium sampler
49 * and texture state.
50 * The former is considered to be static and the later dynamic.
51 */
52 void
53 lp_sampler_static_state(struct lp_sampler_static_state *state,
54 const struct pipe_sampler_view *view,
55 const struct pipe_sampler_state *sampler)
56 {
57 const struct pipe_resource *texture = view->texture;
58
59 memset(state, 0, sizeof *state);
60
61 if(!texture)
62 return;
63
64 if(!sampler)
65 return;
66
67 /*
68 * We don't copy sampler state over unless it is actually enabled, to avoid
69 * spurious recompiles, as the sampler static state is part of the shader
70 * key.
71 *
72 * Ideally the state tracker or cso_cache module would make all state
73 * canonical, but until that happens it's better to be safe than sorry here.
74 *
75 * XXX: Actually there's much more than can be done here, especially
76 * regarding 1D/2D/3D/CUBE textures, wrap modes, etc.
77 */
78
79 state->format = view->format;
80 state->target = texture->target;
81 state->pot_width = util_is_pot(texture->width0);
82 state->pot_height = util_is_pot(texture->height0);
83 state->pot_depth = util_is_pot(texture->depth0);
84
85 state->wrap_s = sampler->wrap_s;
86 state->wrap_t = sampler->wrap_t;
87 state->wrap_r = sampler->wrap_r;
88 state->min_img_filter = sampler->min_img_filter;
89 state->mag_img_filter = sampler->mag_img_filter;
90 if (texture->last_level) {
91 state->min_mip_filter = sampler->min_mip_filter;
92 } else {
93 state->min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
94 }
95
96 state->compare_mode = sampler->compare_mode;
97 if (sampler->compare_mode != PIPE_TEX_COMPARE_NONE) {
98 state->compare_func = sampler->compare_func;
99 }
100
101 state->normalized_coords = sampler->normalized_coords;
102 state->lod_bias = sampler->lod_bias;
103 state->min_lod = sampler->min_lod;
104 state->max_lod = sampler->max_lod;
105 state->border_color[0] = sampler->border_color[0];
106 state->border_color[1] = sampler->border_color[1];
107 state->border_color[2] = sampler->border_color[2];
108 state->border_color[3] = sampler->border_color[3];
109
110 /*
111 * FIXME: Handle the remainder of pipe_sampler_view.
112 */
113 }
114
115
116 /**
117 * Gather elements from scatter positions in memory into a single vector.
118 *
119 * @param src_width src element width
120 * @param dst_width result element width (source will be expanded to fit)
121 * @param length length of the offsets,
122 * @param base_ptr base pointer, should be a i8 pointer type.
123 * @param offsets vector with offsets
124 */
125 LLVMValueRef
126 lp_build_gather(LLVMBuilderRef builder,
127 unsigned length,
128 unsigned src_width,
129 unsigned dst_width,
130 LLVMValueRef base_ptr,
131 LLVMValueRef offsets)
132 {
133 LLVMTypeRef src_type = LLVMIntType(src_width);
134 LLVMTypeRef src_ptr_type = LLVMPointerType(src_type, 0);
135 LLVMTypeRef dst_elem_type = LLVMIntType(dst_width);
136 LLVMTypeRef dst_vec_type = LLVMVectorType(dst_elem_type, length);
137 LLVMValueRef res;
138 unsigned i;
139
140 res = LLVMGetUndef(dst_vec_type);
141 for(i = 0; i < length; ++i) {
142 LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
143 LLVMValueRef elem_offset;
144 LLVMValueRef elem_ptr;
145 LLVMValueRef elem;
146
147 elem_offset = LLVMBuildExtractElement(builder, offsets, index, "");
148 elem_ptr = LLVMBuildGEP(builder, base_ptr, &elem_offset, 1, "");
149 elem_ptr = LLVMBuildBitCast(builder, elem_ptr, src_ptr_type, "");
150 elem = LLVMBuildLoad(builder, elem_ptr, "");
151
152 assert(src_width <= dst_width);
153 if(src_width > dst_width)
154 elem = LLVMBuildTrunc(builder, elem, dst_elem_type, "");
155 if(src_width < dst_width)
156 elem = LLVMBuildZExt(builder, elem, dst_elem_type, "");
157
158 res = LLVMBuildInsertElement(builder, res, elem, index, "");
159 }
160
161 return res;
162 }
163
164
165 /**
166 * Compute the offset of a pixel block.
167 *
168 * x, y, z, y_stride, z_stride are vectors, and they refer to pixel blocks, as
169 * per format description, and not individual pixels.
170 */
171 LLVMValueRef
172 lp_build_sample_offset(struct lp_build_context *bld,
173 const struct util_format_description *format_desc,
174 LLVMValueRef x,
175 LLVMValueRef y,
176 LLVMValueRef z,
177 LLVMValueRef y_stride,
178 LLVMValueRef z_stride)
179 {
180 LLVMValueRef x_stride;
181 LLVMValueRef offset;
182
183 x_stride = lp_build_const_vec(bld->type, format_desc->block.bits/8);
184
185 if(format_desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS) {
186 LLVMValueRef x_lo, x_hi;
187 LLVMValueRef y_lo, y_hi;
188 LLVMValueRef x_stride_lo, x_stride_hi;
189 LLVMValueRef y_stride_lo, y_stride_hi;
190 LLVMValueRef x_offset_lo, x_offset_hi;
191 LLVMValueRef y_offset_lo, y_offset_hi;
192 LLVMValueRef offset_lo, offset_hi;
193
194 /* XXX 1D & 3D addressing not done yet */
195 assert(!z);
196 assert(!z_stride);
197
198 x_lo = LLVMBuildAnd(bld->builder, x, bld->one, "");
199 y_lo = LLVMBuildAnd(bld->builder, y, bld->one, "");
200
201 x_hi = LLVMBuildLShr(bld->builder, x, bld->one, "");
202 y_hi = LLVMBuildLShr(bld->builder, y, bld->one, "");
203
204 x_stride_lo = x_stride;
205 y_stride_lo = lp_build_const_vec(bld->type, 2*format_desc->block.bits/8);
206
207 x_stride_hi = lp_build_const_vec(bld->type, 4*format_desc->block.bits/8);
208 y_stride_hi = LLVMBuildShl(bld->builder, y_stride, bld->one, "");
209
210 x_offset_lo = lp_build_mul(bld, x_lo, x_stride_lo);
211 y_offset_lo = lp_build_mul(bld, y_lo, y_stride_lo);
212 offset_lo = lp_build_add(bld, x_offset_lo, y_offset_lo);
213
214 x_offset_hi = lp_build_mul(bld, x_hi, x_stride_hi);
215 y_offset_hi = lp_build_mul(bld, y_hi, y_stride_hi);
216 offset_hi = lp_build_add(bld, x_offset_hi, y_offset_hi);
217
218 offset = lp_build_add(bld, offset_hi, offset_lo);
219 }
220 else {
221 offset = lp_build_mul(bld, x, x_stride);
222
223 if (y && y_stride) {
224 LLVMValueRef y_offset = lp_build_mul(bld, y, y_stride);
225 offset = lp_build_add(bld, offset, y_offset);
226 }
227
228 if (z && z_stride) {
229 LLVMValueRef z_offset = lp_build_mul(bld, z, z_stride);
230 offset = lp_build_add(bld, offset, z_offset);
231 }
232 }
233
234 return offset;
235 }