1/*
2 * © Copyright 2018 Alyssa Rosenzweig
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
25#ifndef __BUILDER_H__
26#define __BUILDER_H__
27
28#define _LARGEFILE64_SOURCE 1
29#include <sys/mman.h>
30#include <assert.h>
31#include "pan_resource.h"
32#include "pan_job.h"
33#include "pan_blend_cso.h"
34#include "pan_encoder.h"
35#include "pan_texture.h"
36
37#include "pipe/p_compiler.h"
38#include "pipe/p_config.h"
39#include "pipe/p_context.h"
40#include "pipe/p_defines.h"
41#include "pipe/p_format.h"
42#include "pipe/p_screen.h"
43#include "pipe/p_state.h"
44#include "util/u_blitter.h"
45#include "util/hash_table.h"
46#include "util/simple_mtx.h"
47
48#include "midgard/midgard_compile.h"
49#include "compiler/shader_enums.h"
50
51/* Forward declare to avoid extra header dep */
52struct prim_convert_context;
53
54#define SET_BIT(lval, bit, cond) \
55	if (cond) \
56		lval |= (bit); \
57	else \
58		lval &= ~(bit);
59
60/* Dirty tracking flags. 3D is for general 3D state. Shader flags are
61 * per-stage. Renderer refers to Renderer State Descriptors. Vertex refers to
62 * vertex attributes/elements. */
63
64enum pan_dirty_3d {
65        PAN_DIRTY_VIEWPORT       = BITFIELD_BIT(0),
66        PAN_DIRTY_SCISSOR        = BITFIELD_BIT(1),
67        PAN_DIRTY_VERTEX         = BITFIELD_BIT(2),
68        PAN_DIRTY_PARAMS         = BITFIELD_BIT(3),
69        PAN_DIRTY_DRAWID         = BITFIELD_BIT(4),
70        PAN_DIRTY_TLS_SIZE       = BITFIELD_BIT(5),
71};
72
73enum pan_dirty_shader {
74        PAN_DIRTY_STAGE_RENDERER = BITFIELD_BIT(0),
75        PAN_DIRTY_STAGE_TEXTURE  = BITFIELD_BIT(1),
76        PAN_DIRTY_STAGE_SAMPLER  = BITFIELD_BIT(2),
77        PAN_DIRTY_STAGE_IMAGE    = BITFIELD_BIT(3),
78        PAN_DIRTY_STAGE_CONST    = BITFIELD_BIT(4),
79        PAN_DIRTY_STAGE_SSBO     = BITFIELD_BIT(5),
80};
81
82struct panfrost_constant_buffer {
83        struct pipe_constant_buffer cb[PIPE_MAX_CONSTANT_BUFFERS];
84        uint32_t enabled_mask;
85};
86
87struct panfrost_query {
88        /* Passthrough from Gallium */
89        unsigned type;
90        unsigned index;
91
92        /* For computed queries. 64-bit to prevent overflow */
93        struct {
94                uint64_t start;
95                uint64_t end;
96        };
97
98        /* Memory for the GPU to writeback the value of the query */
99        struct pipe_resource *rsrc;
100
101        /* Whether an occlusion query is for a MSAA framebuffer */
102        bool msaa;
103};
104
105struct pipe_fence_handle {
106        struct pipe_reference reference;
107        uint32_t syncobj;
108        bool signaled;
109};
110
111struct panfrost_streamout_target {
112        struct pipe_stream_output_target base;
113        uint32_t offset;
114};
115
116struct panfrost_streamout {
117        struct pipe_stream_output_target *targets[PIPE_MAX_SO_BUFFERS];
118        unsigned num_targets;
119};
120
121struct panfrost_context {
122        /* Gallium context */
123        struct pipe_context base;
124
125        /* Dirty global state */
126        enum pan_dirty_3d dirty;
127
128        /* Per shader stage dirty state */
129        enum pan_dirty_shader dirty_shader[PIPE_SHADER_TYPES];
130
131        /* Unowned pools, so manage yourself. */
132        struct panfrost_pool descs, shaders;
133
134        /* Sync obj used to keep track of in-flight jobs. */
135        uint32_t syncobj;
136
137        /* Set of 32 batches. When the set is full, the LRU entry (the batch
138         * with the smallest seqnum) is flushed to free a slot.
139         */
140        struct {
141                uint64_t seqnum;
142                struct panfrost_batch slots[PAN_MAX_BATCHES];
143
144                /** Set of active batches for faster traversal */
145                BITSET_DECLARE(active, PAN_MAX_BATCHES);
146        } batches;
147
148        /* Map from resources to panfrost_batches */
149        struct hash_table *writers;
150
151        /* Bound job batch */
152        struct panfrost_batch *batch;
153
154        /* Within a launch_grid call.. */
155        const struct pipe_grid_info *compute_grid;
156
157        struct pipe_framebuffer_state pipe_framebuffer;
158        struct panfrost_streamout streamout;
159
160        bool active_queries;
161        uint64_t prims_generated;
162        uint64_t tf_prims_generated;
163        struct panfrost_query *occlusion_query;
164
165        bool indirect_draw;
166        unsigned drawid;
167        unsigned vertex_count;
168        unsigned instance_count;
169        unsigned offset_start;
170        unsigned base_vertex;
171        unsigned base_instance;
172        mali_ptr first_vertex_sysval_ptr;
173        mali_ptr base_vertex_sysval_ptr;
174        mali_ptr base_instance_sysval_ptr;
175        enum pipe_prim_type active_prim;
176
177        /* If instancing is enabled, vertex count padded for instance; if
178         * it is disabled, just equal to plain vertex count */
179        unsigned padded_count;
180
181        struct panfrost_constant_buffer constant_buffer[PIPE_SHADER_TYPES];
182        struct panfrost_rasterizer *rasterizer;
183        struct panfrost_shader_variants *shader[PIPE_SHADER_TYPES];
184        struct panfrost_vertex_state *vertex;
185
186        struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
187        uint32_t vb_mask;
188
189        struct pipe_shader_buffer ssbo[PIPE_SHADER_TYPES][PIPE_MAX_SHADER_BUFFERS];
190        uint32_t ssbo_mask[PIPE_SHADER_TYPES];
191
192        struct pipe_image_view images[PIPE_SHADER_TYPES][PIPE_MAX_SHADER_IMAGES];
193        uint32_t image_mask[PIPE_SHADER_TYPES];
194
195        struct panfrost_sampler_state *samplers[PIPE_SHADER_TYPES][PIPE_MAX_SAMPLERS];
196        unsigned sampler_count[PIPE_SHADER_TYPES];
197
198        struct panfrost_sampler_view *sampler_views[PIPE_SHADER_TYPES][PIPE_MAX_SHADER_SAMPLER_VIEWS];
199        unsigned sampler_view_count[PIPE_SHADER_TYPES];
200
201        struct blitter_context *blitter;
202
203        struct panfrost_blend_state *blend;
204
205        struct pipe_viewport_state pipe_viewport;
206        struct pipe_scissor_state scissor;
207        struct pipe_blend_color blend_color;
208        struct panfrost_zsa_state *depth_stencil;
209        struct pipe_stencil_ref stencil_ref;
210        uint16_t sample_mask;
211        unsigned min_samples;
212
213        struct panfrost_query *cond_query;
214        bool cond_cond;
215        enum pipe_render_cond_flag cond_mode;
216
217        bool is_noop;
218
219        /* Mask of active render targets */
220        uint8_t fb_rt_mask;
221};
222
223/* Corresponds to the CSO */
224
225struct panfrost_rasterizer;
226
227/* Linked varyings */
228struct pan_linkage {
229        /* If the upload is owned by the CSO instead
230         * of the pool, the referenced BO. Else,
231         * NULL. */
232        struct panfrost_bo *bo;
233
234        /* Uploaded attribute descriptors */
235        mali_ptr producer, consumer;
236
237        /* Varyings buffers required */
238        uint32_t present;
239
240        /* Per-vertex stride for general varying buffer */
241        uint32_t stride;
242};
243
244#define RSD_WORDS 16
245
246/* Variants bundle together to form the backing CSO, bundling multiple
247 * shaders with varying emulated features baked in */
248
249/* A shader state corresponds to the actual, current variant of the shader */
250struct panfrost_shader_state {
251        /* Compiled, mapped descriptor, ready for the hardware */
252        bool compiled;
253
254        /* Respectively, shader binary and Renderer State Descriptor */
255        struct panfrost_pool_ref bin, state;
256
257        /* For fragment shaders, a prepared (but not uploaded RSD) */
258        uint32_t partial_rsd[RSD_WORDS];
259
260        struct pan_shader_info info;
261
262        /* Linked varyings, for non-separable programs */
263        struct pan_linkage linkage;
264
265        struct pipe_stream_output_info stream_output;
266        uint64_t so_mask;
267
268        /* Variants */
269        enum pipe_format rt_formats[8];
270        unsigned nr_cbufs;
271
272        /* Mask of state that dirties the sysvals */
273        unsigned dirty_3d, dirty_shader;
274};
275
276/* A collection of varyings (the CSO) */
277struct panfrost_shader_variants {
278        /* A panfrost_shader_variants can represent a shader for
279         * either graphics or compute */
280
281        bool is_compute;
282
283        union {
284                struct pipe_shader_state base;
285                struct pipe_compute_state cbase;
286        };
287
288        /** Lock for the variants array */
289        simple_mtx_t lock;
290
291        struct panfrost_shader_state *variants;
292        unsigned variant_space;
293
294        unsigned variant_count;
295
296        /* The current active variant */
297        unsigned active_variant;
298};
299
300struct pan_vertex_buffer {
301        unsigned vbi;
302        unsigned divisor;
303};
304
305struct panfrost_vertex_state {
306        unsigned num_elements;
307
308        /* buffers corresponds to attribute buffer, element_buffers corresponds
309         * to an index in buffers for each vertex element */
310        struct pan_vertex_buffer buffers[PIPE_MAX_ATTRIBS];
311        unsigned element_buffer[PIPE_MAX_ATTRIBS];
312        unsigned nr_bufs;
313
314        struct pipe_vertex_element pipe[PIPE_MAX_ATTRIBS];
315        unsigned formats[PIPE_MAX_ATTRIBS];
316};
317
318struct panfrost_zsa_state;
319struct panfrost_sampler_state;
320struct panfrost_sampler_view;
321
322static inline struct panfrost_context *
323pan_context(struct pipe_context *pcontext)
324{
325        return (struct panfrost_context *) pcontext;
326}
327
328static inline struct panfrost_streamout_target *
329pan_so_target(struct pipe_stream_output_target *target)
330{
331        return (struct panfrost_streamout_target *)target;
332}
333
334static inline struct panfrost_shader_state *
335panfrost_get_shader_state(struct panfrost_context *ctx,
336                          enum pipe_shader_type st)
337{
338        struct panfrost_shader_variants *all = ctx->shader[st];
339
340        if (!all)
341                return NULL;
342
343        return &all->variants[all->active_variant];
344}
345
346struct pipe_context *
347panfrost_create_context(struct pipe_screen *screen, void *priv, unsigned flags);
348
349bool
350panfrost_writes_point_size(struct panfrost_context *ctx);
351
352struct panfrost_ptr
353panfrost_vertex_tiler_job(struct panfrost_context *ctx, bool is_tiler);
354
355void
356panfrost_flush(
357        struct pipe_context *pipe,
358        struct pipe_fence_handle **fence,
359        unsigned flags);
360
361bool
362panfrost_render_condition_check(struct panfrost_context *ctx);
363
364void
365panfrost_shader_compile(struct pipe_screen *pscreen,
366                        struct panfrost_pool *shader_pool,
367                        struct panfrost_pool *desc_pool,
368                        enum pipe_shader_ir ir_type,
369                        const void *ir,
370                        gl_shader_stage stage,
371                        struct panfrost_shader_state *state);
372
373void
374panfrost_analyze_sysvals(struct panfrost_shader_state *ss);
375
376mali_ptr
377panfrost_get_index_buffer_bounded(struct panfrost_batch *batch,
378                                  const struct pipe_draw_info *info,
379                                  const struct pipe_draw_start_count_bias *draw,
380                                  unsigned *min_index, unsigned *max_index);
381
382/* Instancing */
383
384mali_ptr
385panfrost_vertex_buffer_address(struct panfrost_context *ctx, unsigned i);
386
387/* Compute */
388
389void
390panfrost_compute_context_init(struct pipe_context *pctx);
391
392static inline void
393panfrost_dirty_state_all(struct panfrost_context *ctx)
394{
395        ctx->dirty = ~0;
396
397        for (unsigned i = 0; i < PIPE_SHADER_TYPES; ++i)
398                ctx->dirty_shader[i] = ~0;
399}
400
401static inline void
402panfrost_clean_state_3d(struct panfrost_context *ctx)
403{
404        ctx->dirty = 0;
405
406        for (unsigned i = 0; i < PIPE_SHADER_TYPES; ++i) {
407                if (i != PIPE_SHADER_COMPUTE)
408                        ctx->dirty_shader[i] = 0;
409        }
410}
411
412#endif
413