r600g: reduce memory usage from range/block hash table.
[mesa.git] / src / gallium / winsys / r600 / drm / r600_priv.h
1 /*
2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
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 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the 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 NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
22 *
23 * Authors:
24 * Jerome Glisse
25 */
26 #ifndef R600_PRIV_H
27 #define R600_PRIV_H
28
29 #include <errno.h>
30 #include <stdint.h>
31 #include <stdlib.h>
32 #include <assert.h>
33 #include <util/u_double_list.h>
34 #include <util/u_inlines.h>
35 #include "util/u_hash_table.h"
36 #include <os/os_thread.h>
37 #include "r600.h"
38
39 #define PKT_COUNT_C 0xC000FFFF
40 #define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
41
42 struct r600_bomgr;
43 struct r600_bo;
44
45 struct radeon {
46 int fd;
47 int refcount;
48 unsigned device;
49 unsigned family;
50 enum chip_class chip_class;
51 struct r600_tiling_info tiling_info;
52 struct r600_bomgr *bomgr;
53 unsigned fence;
54 unsigned *cfence;
55 struct r600_bo *fence_bo;
56 unsigned clock_crystal_freq;
57 unsigned num_backends;
58 unsigned minor_version;
59
60 /* List of buffer handles and its mutex. */
61 struct util_hash_table *bo_handles;
62 pipe_mutex bo_handles_mutex;
63 };
64
65 #define REG_FLAG_NEED_BO 1
66 #define REG_FLAG_DIRTY_ALWAYS 2
67 #define REG_FLAG_RV6XX_SBU 4
68
69 struct r600_reg {
70 unsigned opcode;
71 unsigned offset_base;
72 unsigned offset;
73 unsigned flags;
74 unsigned flush_flags;
75 unsigned flush_mask;
76 };
77
78 struct radeon_bo {
79 struct pipe_reference reference;
80 unsigned handle;
81 unsigned size;
82 unsigned alignment;
83 int map_count;
84 void *data;
85 struct list_head fencedlist;
86 unsigned fence;
87 struct r600_context *ctx;
88 boolean shared;
89 struct r600_reloc *reloc;
90 unsigned reloc_id;
91 unsigned last_flush;
92 unsigned name;
93 };
94
95 struct r600_bo {
96 struct pipe_reference reference;
97 unsigned size;
98 unsigned tiling_flags;
99 unsigned kernel_pitch;
100 unsigned domains;
101 struct radeon_bo *bo;
102 unsigned fence;
103 /* manager data */
104 struct list_head list;
105 unsigned manager_id;
106 unsigned alignment;
107 unsigned offset;
108 int64_t start;
109 int64_t end;
110 };
111
112 struct r600_bomgr {
113 struct radeon *radeon;
114 unsigned usecs;
115 pipe_mutex mutex;
116 struct list_head delayed;
117 unsigned num_delayed;
118 };
119
120 /*
121 * r600_drm.c
122 */
123 struct radeon *r600_new(int fd, unsigned device);
124 void r600_delete(struct radeon *r600);
125
126 /*
127 * radeon_pciid.c
128 */
129 unsigned radeon_family_from_device(unsigned device);
130
131 /*
132 * radeon_bo.c
133 */
134 struct radeon_bo *radeon_bo(struct radeon *radeon, unsigned handle,
135 unsigned size, unsigned alignment);
136 void radeon_bo_reference(struct radeon *radeon, struct radeon_bo **dst,
137 struct radeon_bo *src);
138 int radeon_bo_wait(struct radeon *radeon, struct radeon_bo *bo);
139 int radeon_bo_busy(struct radeon *radeon, struct radeon_bo *bo, uint32_t *domain);
140 int radeon_bo_fencelist(struct radeon *radeon, struct radeon_bo **bolist, uint32_t num_bo);
141 int radeon_bo_get_tiling_flags(struct radeon *radeon,
142 struct radeon_bo *bo,
143 uint32_t *tiling_flags,
144 uint32_t *pitch);
145 int radeon_bo_get_name(struct radeon *radeon,
146 struct radeon_bo *bo,
147 uint32_t *name);
148 int radeon_bo_fixed_map(struct radeon *radeon, struct radeon_bo *bo);
149
150 /*
151 * r600_hw_context.c
152 */
153 int r600_context_init_fence(struct r600_context *ctx);
154 void r600_context_bo_reloc(struct r600_context *ctx, u32 *pm4, struct r600_bo *rbo);
155 void r600_context_bo_flush(struct r600_context *ctx, unsigned flush_flags,
156 unsigned flush_mask, struct r600_bo *rbo);
157 struct r600_bo *r600_context_reg_bo(struct r600_context *ctx, unsigned offset);
158 int r600_context_add_block(struct r600_context *ctx, const struct r600_reg *reg, unsigned nreg);
159 void r600_context_pipe_state_set_resource(struct r600_context *ctx, struct r600_pipe_state *state, unsigned offset);
160 void r600_context_block_emit_dirty(struct r600_context *ctx, struct r600_block *block);
161 void r600_context_dirty_block(struct r600_context *ctx, struct r600_block *block,
162 int dirty, int index);
163
164 void r600_context_reg(struct r600_context *ctx,
165 unsigned offset, unsigned value,
166 unsigned mask);
167 /*
168 * r600_bo.c
169 */
170 void r600_bo_destroy(struct radeon *radeon, struct r600_bo *bo);
171
172 /*
173 * r600_bomgr.c
174 */
175 struct r600_bomgr *r600_bomgr_create(struct radeon *radeon, unsigned usecs);
176 void r600_bomgr_destroy(struct r600_bomgr *mgr);
177 bool r600_bomgr_bo_destroy(struct r600_bomgr *mgr, struct r600_bo *bo);
178 void r600_bomgr_bo_init(struct r600_bomgr *mgr, struct r600_bo *bo);
179 struct r600_bo *r600_bomgr_bo_create(struct r600_bomgr *mgr,
180 unsigned size,
181 unsigned alignment,
182 unsigned cfence);
183
184
185 /*
186 * helpers
187 */
188
189 /* each range covers 9 bits of dword space = 512 dwords = 2k bytes */
190 /* there is a block entry for each register so 512 blocks */
191 /* we have no registers to read/write below 0x8000 (0x2000 in dw space) */
192 /* we use some fake offsets at 0x40000 to do evergreen sampler borders so take 0x42000 as a max bound*/
193 #define RANGE_OFFSET_START 0x8000
194 #define HASH_SHIFT 9
195 #define NUM_RANGES (0x42000 - RANGE_OFFSET_START) / (4 << HASH_SHIFT) /* 128 << 9 = 64k */
196
197 #define CTX_RANGE_ID(ctx, offset) ((((offset - RANGE_OFFSET_START) >> 2) >> HASH_SHIFT) & 255)
198 #define CTX_BLOCK_ID(ctx, offset) (((offset - RANGE_OFFSET_START) >> 2) & ((1 << HASH_SHIFT) - 1))
199
200 /*
201 * radeon_bo.c
202 */
203 static inline int radeon_bo_map(struct radeon *radeon, struct radeon_bo *bo)
204 {
205 if (bo->map_count == 0 && !bo->data)
206 return radeon_bo_fixed_map(radeon, bo);
207 bo->map_count++;
208 return 0;
209 }
210
211 static inline void radeon_bo_unmap(struct radeon *radeon, struct radeon_bo *bo)
212 {
213 bo->map_count--;
214 assert(bo->map_count >= 0);
215 }
216
217 /*
218 * fence
219 */
220 static inline bool fence_is_after(unsigned fence, unsigned ofence)
221 {
222 /* handle wrap around */
223 if (fence < 0x80000000 && ofence > 0x80000000)
224 return TRUE;
225 if (fence > ofence)
226 return TRUE;
227 return FALSE;
228 }
229
230 #endif