2 * © Copyright 2018 Alyssa Rosenzweig
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:
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
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
28 #define _LARGEFILE64_SOURCE 1
29 #define CACHE_LINE_SIZE 1024 /* TODO */
32 #include "pan_resource.h"
35 #include "pipe/p_compiler.h"
36 #include "pipe/p_config.h"
37 #include "pipe/p_context.h"
38 #include "pipe/p_defines.h"
39 #include "pipe/p_format.h"
40 #include "pipe/p_screen.h"
41 #include "pipe/p_state.h"
42 #include "util/u_blitter.h"
43 #include "util/hash_table.h"
45 /* Forward declare to avoid extra header dep */
46 struct prim_convert_context
;
48 #define MAX_DRAW_CALLS 4096
49 #define MAX_VARYINGS 4096
51 //#define PAN_DIRTY_CLEAR (1 << 0)
52 #define PAN_DIRTY_RASTERIZER (1 << 2)
53 #define PAN_DIRTY_FS (1 << 3)
54 #define PAN_DIRTY_FRAG_CORE (PAN_DIRTY_FS) /* Dirty writes are tied */
55 #define PAN_DIRTY_VS (1 << 4)
56 #define PAN_DIRTY_VERTEX (1 << 5)
57 #define PAN_DIRTY_VERT_BUF (1 << 6)
58 //#define PAN_DIRTY_VIEWPORT (1 << 7)
59 #define PAN_DIRTY_SAMPLERS (1 << 8)
60 #define PAN_DIRTY_TEXTURES (1 << 9)
62 #define SET_BIT(lval, bit, cond) \
68 struct panfrost_constant_buffer
{
74 struct panfrost_query
{
75 /* Passthrough from Gallium */
79 /* Memory for the GPU to writeback the value of the query */
80 struct panfrost_transfer transfer
;
83 struct panfrost_fence
{
84 struct pipe_reference reference
;
88 #define PANFROST_MAX_TRANSIENT_ENTRIES 64
90 struct panfrost_transient_pool
{
91 /* Memory blocks in the pool */
92 struct panfrost_memory_entry
*entries
[PANFROST_MAX_TRANSIENT_ENTRIES
];
94 /* Number of entries we own */
97 /* Current entry that we are writing to, zero-indexed, strictly less than entry_count */
100 /* Number of bytes into the current entry we are */
103 /* Entry size (all entries must be homogenous) */
107 struct panfrost_context
{
108 /* Gallium context */
109 struct pipe_context base
;
111 /* Bound job and map of panfrost_job_key to jobs */
112 struct panfrost_job
*job
;
113 struct hash_table
*jobs
;
115 /* Bit mask for supported PIPE_DRAW for this hardware */
118 struct pipe_framebuffer_state pipe_framebuffer
;
120 /* The number of concurrent FBOs allowed depends on the number of pools
121 * used; pools are ringed for parallelism opportunities */
123 struct panfrost_transient_pool transient_pools
[2];
126 struct panfrost_memory cmdstream_persistent
;
127 struct panfrost_memory shaders
;
128 struct panfrost_memory scratchpad
;
129 struct panfrost_memory tiler_heap
;
130 struct panfrost_memory varying_mem
;
131 struct panfrost_memory misc_0
;
132 struct panfrost_memory misc_1
;
133 struct panfrost_memory depth_stencil_buffer
;
135 struct panfrost_query
*occlusion_query
;
137 /* Each draw has corresponding vertex and tiler payloads */
138 struct midgard_payload_vertex_tiler payload_vertex
;
139 struct midgard_payload_vertex_tiler payload_tiler
;
141 /* The fragment shader binary itself is pointed here (for the tripipe) but
142 * also everything else in the shader core, including blending, the
143 * stencil/depth tests, etc. Refer to the presentations. */
145 struct mali_shader_meta fragment_shader_core
;
147 /* A frame is composed of a starting set value job, a number of vertex
148 * and tiler jobs, linked to the fragment job at the end. See the
149 * presentations for more information how this works */
153 mali_ptr set_value_job
;
154 mali_ptr vertex_jobs
[MAX_DRAW_CALLS
];
155 mali_ptr tiler_jobs
[MAX_DRAW_CALLS
];
157 struct mali_job_descriptor_header
*u_set_value_job
;
158 struct mali_job_descriptor_header
*u_vertex_jobs
[MAX_DRAW_CALLS
];
159 struct mali_job_descriptor_header
*u_tiler_jobs
[MAX_DRAW_CALLS
];
161 unsigned vertex_job_count
;
162 unsigned tiler_job_count
;
164 /* Per-draw Dirty flags are setup like any other driver */
167 unsigned vertex_count
;
169 union mali_attr attributes
[PIPE_MAX_ATTRIBS
];
171 unsigned varying_height
;
173 struct mali_single_framebuffer vt_framebuffer_sfbd
;
174 struct bifrost_framebuffer vt_framebuffer_mfbd
;
176 /* TODO: Multiple uniform buffers (index =/= 0), finer updates? */
178 struct panfrost_constant_buffer constant_buffer
[PIPE_SHADER_TYPES
];
181 struct panfrost_rasterizer
*rasterizer
;
183 struct panfrost_shader_variants
*vs
;
184 struct panfrost_shader_variants
*fs
;
186 struct panfrost_vertex_state
*vertex
;
188 struct pipe_vertex_buffer
*vertex_buffers
;
189 unsigned vertex_buffer_count
;
191 struct panfrost_sampler_state
*samplers
[PIPE_SHADER_TYPES
][PIPE_MAX_SAMPLERS
];
192 unsigned sampler_count
[PIPE_SHADER_TYPES
];
194 struct panfrost_sampler_view
*sampler_views
[PIPE_SHADER_TYPES
][PIPE_MAX_SHADER_SAMPLER_VIEWS
];
195 unsigned sampler_view_count
[PIPE_SHADER_TYPES
];
197 struct primconvert_context
*primconvert
;
198 struct blitter_context
*blitter
;
200 struct panfrost_blend_state
*blend
;
202 struct pipe_viewport_state pipe_viewport
;
203 struct pipe_scissor_state scissor
;
204 struct pipe_blend_color blend_color
;
205 struct pipe_depth_stencil_alpha_state
*depth_stencil
;
206 struct pipe_stencil_ref stencil_ref
;
208 /* True for t6XX, false for t8xx. */
211 /* If set, we'll require the use of single render-target framebuffer
212 * descriptors (SFBD), for older hardware -- specifically, <T760 hardware, If
213 * false, we'll use the MFBD no matter what. New hardware -does- retain support
214 * for SFBD, and in theory we could flip between them on a per-RT basis, but
215 * there's no real advantage to doing so */
221 /* Corresponds to the CSO */
223 struct panfrost_rasterizer
{
224 struct pipe_rasterizer_state base
;
226 /* Bitmask of front face, etc */
227 unsigned tiler_gl_enables
;
230 struct panfrost_blend_state
{
231 struct pipe_blend_state base
;
233 /* Whether a blend shader is in use */
234 bool has_blend_shader
;
236 /* Compiled fixed function command */
237 struct mali_blend_equation equation
;
239 /* Compiled blend shader */
240 mali_ptr blend_shader
;
241 int blend_work_count
;
244 /* Internal varyings descriptor */
245 struct panfrost_varyings
{
246 /* Varyings information: stride of each chunk of memory used for
247 * varyings (similar structure with attributes). Count is just the
248 * number of vec4's. Buffer count is the number of varying chunks (<=
249 * count). Height is used to calculate gl_Position's position ("it's
250 * not a pun, Alyssa!"). Vertex-only varyings == descriptor for
251 * gl_Position and something else apparently occupying the same space.
252 * Varyings == main varyings descriptors following typical mali_attr
255 unsigned varyings_stride
[MAX_VARYINGS
];
256 unsigned varying_count
;
257 unsigned varying_buffer_count
;
259 /* Map of the actual varyings buffer */
260 uint8_t *varyings_buffer_cpu
;
261 mali_ptr varyings_descriptor
;
262 mali_ptr varyings_descriptor_fragment
;
265 /* Variants bundle together to form the backing CSO, bundling multiple
266 * shaders with varying emulated features baked in (alpha test
267 * parameters, etc) */
268 #define MAX_SHADER_VARIANTS 8
270 /* A shader state corresponds to the actual, current variant of the shader */
271 struct panfrost_shader_state
{
272 struct pipe_shader_state
*base
;
274 /* Compiled, mapped descriptor, ready for the hardware */
276 struct mali_shader_meta
*tripipe
;
277 mali_ptr tripipe_gpu
;
279 /* Non-descript information */
282 bool writes_point_size
;
284 /* Valid for vertex shaders only due to when this is calculated */
285 struct panfrost_varyings varyings
;
287 /* Information on this particular shader variant */
288 struct pipe_alpha_state alpha_state
;
291 /* A collection of varyings (the CSO) */
292 struct panfrost_shader_variants
{
293 struct pipe_shader_state base
;
295 struct panfrost_shader_state variants
[MAX_SHADER_VARIANTS
];
296 unsigned variant_count
;
298 /* The current active variant */
299 unsigned active_variant
;
302 struct panfrost_vertex_state
{
303 unsigned num_elements
;
305 struct pipe_vertex_element pipe
[PIPE_MAX_ATTRIBS
];
306 int nr_components
[PIPE_MAX_ATTRIBS
];
308 /* The actual attribute meta, prebaked and GPU mapped. TODO: Free memory */
309 struct mali_attr_meta
*hw
;
310 mali_ptr descriptor_ptr
;
313 struct panfrost_sampler_state
{
314 struct pipe_sampler_state base
;
315 struct mali_sampler_descriptor hw
;
318 /* Misnomer: Sampler view corresponds to textures, not samplers */
320 struct panfrost_sampler_view
{
321 struct pipe_sampler_view base
;
322 struct mali_texture_descriptor hw
;
325 static inline struct panfrost_context
*
326 pan_context(struct pipe_context
*pcontext
)
328 return (struct panfrost_context
*) pcontext
;
331 static inline struct panfrost_screen
*
332 pan_screen(struct pipe_screen
*p
)
334 return (struct panfrost_screen
*)p
;
337 struct pipe_context
*
338 panfrost_create_context(struct pipe_screen
*screen
, void *priv
, unsigned flags
);
341 panfrost_emit_for_draw(struct panfrost_context
*ctx
, bool with_vertex_data
);
343 struct panfrost_transfer
344 panfrost_vertex_tiler_job(struct panfrost_context
*ctx
, bool is_tiler
, bool is_elided_tiler
);
347 panfrost_get_default_swizzle(unsigned components
);
351 struct pipe_context
*pipe
,
352 struct pipe_fence_handle
**fence
,
356 panfrost_is_scanout(struct panfrost_context
*ctx
);
359 panfrost_sfbd_fragment(struct panfrost_context
*ctx
, bool flip_y
);
362 panfrost_mfbd_fragment(struct panfrost_context
*ctx
, bool flip_y
);
364 struct bifrost_framebuffer
365 panfrost_emit_mfbd(struct panfrost_context
*ctx
);
367 struct mali_single_framebuffer
368 panfrost_emit_sfbd(struct panfrost_context
*ctx
);
371 panfrost_fragment_job(struct panfrost_context
*ctx
);
374 panfrost_shader_compile(struct panfrost_context
*ctx
, struct mali_shader_meta
*meta
, const char *src
, int type
, struct panfrost_shader_state
*state
);