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"
34 #include "pan_blend.h"
36 #include "pipe/p_compiler.h"
37 #include "pipe/p_config.h"
38 #include "pipe/p_context.h"
39 #include "pipe/p_defines.h"
40 #include "pipe/p_format.h"
41 #include "pipe/p_screen.h"
42 #include "pipe/p_state.h"
43 #include "util/u_blitter.h"
44 #include "util/hash_table.h"
46 #include "midgard/midgard_compile.h"
48 /* Forward declare to avoid extra header dep */
49 struct prim_convert_context
;
51 #define MAX_VARYINGS 4096
53 //#define PAN_DIRTY_CLEAR (1 << 0)
54 #define PAN_DIRTY_RASTERIZER (1 << 2)
55 #define PAN_DIRTY_FS (1 << 3)
56 #define PAN_DIRTY_FRAG_CORE (PAN_DIRTY_FS) /* Dirty writes are tied */
57 #define PAN_DIRTY_VS (1 << 4)
58 #define PAN_DIRTY_VERTEX (1 << 5)
59 #define PAN_DIRTY_VERT_BUF (1 << 6)
60 //#define PAN_DIRTY_VIEWPORT (1 << 7)
61 #define PAN_DIRTY_SAMPLERS (1 << 8)
62 #define PAN_DIRTY_TEXTURES (1 << 9)
64 #define SET_BIT(lval, bit, cond) \
70 struct panfrost_constant_buffer
{
71 struct pipe_constant_buffer cb
[PIPE_MAX_CONSTANT_BUFFERS
];
72 uint32_t enabled_mask
;
76 struct panfrost_query
{
77 /* Passthrough from Gallium */
81 /* Memory for the GPU to writeback the value of the query */
82 struct panfrost_transfer transfer
;
85 struct panfrost_fence
{
86 struct pipe_reference reference
;
90 #define PANFROST_MAX_TRANSIENT_ENTRIES 64
92 struct panfrost_transient_pool
{
93 /* Memory blocks in the pool */
94 struct panfrost_memory_entry
*entries
[PANFROST_MAX_TRANSIENT_ENTRIES
];
96 /* Number of entries we own */
99 /* Current entry that we are writing to, zero-indexed, strictly less than entry_count */
100 unsigned entry_index
;
102 /* Number of bytes into the current entry we are */
105 /* Entry size (all entries must be homogenous) */
109 struct panfrost_context
{
110 /* Gallium context */
111 struct pipe_context base
;
113 /* Bound job and map of panfrost_job_key to jobs */
114 struct panfrost_job
*job
;
115 struct hash_table
*jobs
;
117 /* panfrost_resource -> panfrost_job */
118 struct hash_table
*write_jobs
;
120 /* Bit mask for supported PIPE_DRAW for this hardware */
123 struct pipe_framebuffer_state pipe_framebuffer
;
125 /* The number of concurrent FBOs allowed depends on the number of pools
126 * used; pools are ringed for parallelism opportunities */
128 struct panfrost_transient_pool transient_pools
[2];
131 struct panfrost_memory cmdstream_persistent
;
132 struct panfrost_memory shaders
;
133 struct panfrost_memory scratchpad
;
134 struct panfrost_memory tiler_heap
;
135 struct panfrost_memory varying_mem
;
136 struct panfrost_memory tiler_polygon_list
;
137 struct panfrost_memory tiler_dummy
;
138 struct panfrost_memory depth_stencil_buffer
;
140 struct panfrost_query
*occlusion_query
;
142 /* Each draw has corresponding vertex and tiler payloads */
143 struct midgard_payload_vertex_tiler payload_vertex
;
144 struct midgard_payload_vertex_tiler payload_tiler
;
146 /* The fragment shader binary itself is pointed here (for the tripipe) but
147 * also everything else in the shader core, including blending, the
148 * stencil/depth tests, etc. Refer to the presentations. */
150 struct mali_shader_meta fragment_shader_core
;
152 /* Per-draw Dirty flags are setup like any other driver */
155 unsigned vertex_count
;
156 unsigned instance_count
;
158 /* If instancing is enabled, vertex count padded for instance; if
159 * it is disabled, just equal to plain vertex count */
160 unsigned padded_count
;
162 union mali_attr attributes
[PIPE_MAX_ATTRIBS
];
164 unsigned varying_height
;
166 struct mali_single_framebuffer vt_framebuffer_sfbd
;
167 struct bifrost_framebuffer vt_framebuffer_mfbd
;
169 /* TODO: Multiple uniform buffers (index =/= 0), finer updates? */
171 struct panfrost_constant_buffer constant_buffer
[PIPE_SHADER_TYPES
];
174 struct panfrost_rasterizer
*rasterizer
;
176 struct panfrost_shader_variants
*vs
;
177 struct panfrost_shader_variants
*fs
;
179 struct panfrost_vertex_state
*vertex
;
181 struct pipe_vertex_buffer vertex_buffers
[PIPE_MAX_ATTRIBS
];
184 struct panfrost_sampler_state
*samplers
[PIPE_SHADER_TYPES
][PIPE_MAX_SAMPLERS
];
185 unsigned sampler_count
[PIPE_SHADER_TYPES
];
187 struct panfrost_sampler_view
*sampler_views
[PIPE_SHADER_TYPES
][PIPE_MAX_SHADER_SAMPLER_VIEWS
];
188 unsigned sampler_view_count
[PIPE_SHADER_TYPES
];
190 struct primconvert_context
*primconvert
;
191 struct blitter_context
*blitter
;
193 /* Blitting the wallpaper (the old contents of the framebuffer back to
194 * itself) uses a dedicated u_blitter instance versus general blit()
195 * callbacks from Gallium, as the blit() callback can trigger
196 * wallpapering without Gallium realising, which in turns u_blitter
197 * errors due to unsupported reucrsion */
199 struct blitter_context
*blitter_wallpaper
;
200 struct panfrost_job
*wallpaper_batch
;
202 struct panfrost_blend_state
*blend
;
204 struct pipe_viewport_state pipe_viewport
;
205 struct pipe_scissor_state scissor
;
206 struct pipe_blend_color blend_color
;
207 struct pipe_depth_stencil_alpha_state
*depth_stencil
;
208 struct pipe_stencil_ref stencil_ref
;
210 /* True for t6XX, false for t8xx. */
213 /* If set, we'll require the use of single render-target framebuffer
214 * descriptors (SFBD), for older hardware -- specifically, <T760 hardware, If
215 * false, we'll use the MFBD no matter what. New hardware -does- retain support
216 * for SFBD, and in theory we could flip between them on a per-RT basis, but
217 * there's no real advantage to doing so */
223 /* Corresponds to the CSO */
225 struct panfrost_rasterizer
{
226 struct pipe_rasterizer_state base
;
228 /* Bitmask of front face, etc */
229 unsigned tiler_gl_enables
;
232 /* Variants bundle together to form the backing CSO, bundling multiple
233 * shaders with varying emulated features baked in (alpha test
234 * parameters, etc) */
235 #define MAX_SHADER_VARIANTS 8
237 /* A shader state corresponds to the actual, current variant of the shader */
238 struct panfrost_shader_state
{
239 struct pipe_shader_state
*base
;
241 /* Compiled, mapped descriptor, ready for the hardware */
243 struct mali_shader_meta
*tripipe
;
244 mali_ptr tripipe_gpu
;
246 /* Non-descript information */
249 bool writes_point_size
;
250 bool reads_point_coord
;
252 struct mali_attr_meta varyings
[PIPE_MAX_ATTRIBS
];
253 gl_varying_slot varyings_loc
[PIPE_MAX_ATTRIBS
];
255 unsigned sysval_count
;
256 unsigned sysval
[MAX_SYSVAL_COUNT
];
258 /* Information on this particular shader variant */
259 struct pipe_alpha_state alpha_state
;
261 uint16_t point_sprite_mask
;
262 unsigned point_sprite_upper_left
: 1;
265 /* A collection of varyings (the CSO) */
266 struct panfrost_shader_variants
{
267 struct pipe_shader_state base
;
269 struct panfrost_shader_state variants
[MAX_SHADER_VARIANTS
];
270 unsigned variant_count
;
272 /* The current active variant */
273 unsigned active_variant
;
276 struct panfrost_vertex_state
{
277 unsigned num_elements
;
279 struct pipe_vertex_element pipe
[PIPE_MAX_ATTRIBS
];
280 struct mali_attr_meta hw
[PIPE_MAX_ATTRIBS
];
283 struct panfrost_sampler_state
{
284 struct pipe_sampler_state base
;
285 struct mali_sampler_descriptor hw
;
288 /* Misnomer: Sampler view corresponds to textures, not samplers */
290 struct panfrost_sampler_view
{
291 struct pipe_sampler_view base
;
292 struct mali_texture_descriptor hw
;
295 static inline struct panfrost_context
*
296 pan_context(struct pipe_context
*pcontext
)
298 return (struct panfrost_context
*) pcontext
;
301 struct pipe_context
*
302 panfrost_create_context(struct pipe_screen
*screen
, void *priv
, unsigned flags
);
305 panfrost_emit_for_draw(struct panfrost_context
*ctx
, bool with_vertex_data
);
307 struct panfrost_transfer
308 panfrost_vertex_tiler_job(struct panfrost_context
*ctx
, bool is_tiler
);
311 panfrost_get_default_swizzle(unsigned components
);
315 struct pipe_context
*pipe
,
316 struct pipe_fence_handle
**fence
,
320 panfrost_is_scanout(struct panfrost_context
*ctx
);
322 mali_ptr
panfrost_sfbd_fragment(struct panfrost_context
*ctx
, bool has_draws
);
323 mali_ptr
panfrost_mfbd_fragment(struct panfrost_context
*ctx
, bool has_draws
);
325 struct bifrost_framebuffer
326 panfrost_emit_mfbd(struct panfrost_context
*ctx
, unsigned vertex_count
);
328 struct mali_single_framebuffer
329 panfrost_emit_sfbd(struct panfrost_context
*ctx
, unsigned vertex_count
);
332 panfrost_fragment_job(struct panfrost_context
*ctx
, bool has_draws
);
335 panfrost_shader_compile(struct panfrost_context
*ctx
, struct mali_shader_meta
*meta
, const char *src
, int type
, struct panfrost_shader_state
*state
);
338 panfrost_pack_work_groups_compute(
339 struct mali_vertex_tiler_prefix
*out
,
348 panfrost_pack_work_groups_fused(
349 struct mali_vertex_tiler_prefix
*vertex
,
350 struct mali_vertex_tiler_prefix
*tiler
,
361 panfrost_vertex_buffer_address(struct panfrost_context
*ctx
, unsigned i
);
364 panfrost_emit_vertex_data(struct panfrost_job
*batch
);
366 struct pan_shift_odd
{
372 panfrost_padded_vertex_count(
373 unsigned vertex_count
,
378 pan_expand_shift_odd(struct pan_shift_odd o
);