1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
|
/*
* Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
* Copyright 2014,2015 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* on the rights to use, copy, modify, merge, publish, distribute, sub
* license, and/or sell copies of the Software, and to permit persons to whom
* the Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
* THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
* USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* Authors:
* Jerome Glisse
*/
#include "sid.h"
#include "si_pipe.h"
#include "../radeon/r600_cs.h"
#include "util/u_format.h"
static uint32_t cik_micro_tile_mode(struct si_screen *sscreen, unsigned tile_mode)
{
if (sscreen->b.info.si_tile_mode_array_valid) {
uint32_t gb_tile_mode = sscreen->b.info.si_tile_mode_array[tile_mode];
return G_009910_MICRO_TILE_MODE_NEW(gb_tile_mode);
}
/* The kernel cannod return the tile mode array. Guess? */
return V_009910_ADDR_SURF_THIN_MICRO_TILING;
}
static void cik_sdma_do_copy_buffer(struct si_context *ctx,
struct pipe_resource *dst,
struct pipe_resource *src,
uint64_t dst_offset,
uint64_t src_offset,
uint64_t size)
{
struct radeon_winsys_cs *cs = ctx->b.rings.dma.cs;
unsigned i, ncopy, csize;
struct r600_resource *rdst = (struct r600_resource*)dst;
struct r600_resource *rsrc = (struct r600_resource*)src;
dst_offset += r600_resource(dst)->gpu_address;
src_offset += r600_resource(src)->gpu_address;
ncopy = (size + CIK_SDMA_COPY_MAX_SIZE - 1) / CIK_SDMA_COPY_MAX_SIZE;
r600_need_dma_space(&ctx->b, ncopy * 7);
r600_context_bo_reloc(&ctx->b, &ctx->b.rings.dma, rsrc, RADEON_USAGE_READ,
RADEON_PRIO_MIN);
r600_context_bo_reloc(&ctx->b, &ctx->b.rings.dma, rdst, RADEON_USAGE_WRITE,
RADEON_PRIO_MIN);
for (i = 0; i < ncopy; i++) {
csize = size < CIK_SDMA_COPY_MAX_SIZE ? size : CIK_SDMA_COPY_MAX_SIZE;
cs->buf[cs->cdw++] = CIK_SDMA_PACKET(CIK_SDMA_OPCODE_COPY,
CIK_SDMA_COPY_SUB_OPCODE_LINEAR,
0);
cs->buf[cs->cdw++] = csize;
cs->buf[cs->cdw++] = 0; /* src/dst endian swap */
cs->buf[cs->cdw++] = src_offset;
cs->buf[cs->cdw++] = src_offset >> 32;
cs->buf[cs->cdw++] = dst_offset;
cs->buf[cs->cdw++] = dst_offset >> 32;
dst_offset += csize;
src_offset += csize;
size -= csize;
}
}
static void cik_sdma_copy_buffer(struct si_context *ctx,
struct pipe_resource *dst,
struct pipe_resource *src,
uint64_t dst_offset,
uint64_t src_offset,
uint64_t size)
{
struct r600_resource *rdst = (struct r600_resource*)dst;
/* Mark the buffer range of destination as valid (initialized),
* so that transfer_map knows it should wait for the GPU when mapping
* that range. */
util_range_add(&rdst->valid_buffer_range, dst_offset,
dst_offset + size);
cik_sdma_do_copy_buffer(ctx, dst, src, dst_offset, src_offset, size);
}
static void cik_sdma_copy_tile(struct si_context *ctx,
struct pipe_resource *dst,
unsigned dst_level,
struct pipe_resource *src,
unsigned src_level,
unsigned y,
unsigned copy_height,
unsigned y_align,
unsigned pitch,
unsigned bpe)
{
struct radeon_winsys_cs *cs = ctx->b.rings.dma.cs;
struct si_screen *sscreen = ctx->screen;
struct r600_texture *rsrc = (struct r600_texture*)src;
struct r600_texture *rdst = (struct r600_texture*)dst;
struct r600_texture *rlinear, *rtiled;
unsigned linear_lvl, tiled_lvl;
unsigned array_mode, lbpe, pitch_tile_max, slice_tile_max, size;
unsigned ncopy, height, cheight, detile, i, src_mode, dst_mode;
unsigned sub_op, bank_h, bank_w, mt_aspect, nbanks, tile_split, mt;
uint64_t base, addr;
unsigned pipe_config, tile_mode_index;
dst_mode = rdst->surface.level[dst_level].mode;
src_mode = rsrc->surface.level[src_level].mode;
/* downcast linear aligned to linear to simplify test */
src_mode = src_mode == RADEON_SURF_MODE_LINEAR_ALIGNED ? RADEON_SURF_MODE_LINEAR : src_mode;
dst_mode = dst_mode == RADEON_SURF_MODE_LINEAR_ALIGNED ? RADEON_SURF_MODE_LINEAR : dst_mode;
assert(dst_mode != src_mode);
assert(src_mode == RADEON_SURF_MODE_LINEAR || dst_mode == RADEON_SURF_MODE_LINEAR);
sub_op = CIK_SDMA_COPY_SUB_OPCODE_TILED;
lbpe = util_logbase2(bpe);
pitch_tile_max = ((pitch / bpe) / 8) - 1;
detile = dst_mode == RADEON_SURF_MODE_LINEAR;
rlinear = detile ? rdst : rsrc;
rtiled = detile ? rsrc : rdst;
linear_lvl = detile ? dst_level : src_level;
tiled_lvl = detile ? src_level : dst_level;
assert(!util_format_is_depth_and_stencil(rtiled->resource.b.b.format));
array_mode = si_array_mode(rtiled->surface.level[tiled_lvl].mode);
slice_tile_max = (rtiled->surface.level[tiled_lvl].nblk_x *
rtiled->surface.level[tiled_lvl].nblk_y) / (8*8) - 1;
height = rlinear->surface.level[linear_lvl].nblk_y;
base = rtiled->surface.level[tiled_lvl].offset;
addr = rlinear->surface.level[linear_lvl].offset;
bank_h = cik_bank_wh(rtiled->surface.bankh);
bank_w = cik_bank_wh(rtiled->surface.bankw);
mt_aspect = cik_macro_tile_aspect(rtiled->surface.mtilea);
tile_split = cik_tile_split(rtiled->surface.tile_split);
tile_mode_index = si_tile_mode_index(rtiled, tiled_lvl, false);
nbanks = si_num_banks(sscreen, rtiled);
base += rtiled->resource.gpu_address;
addr += rlinear->resource.gpu_address;
pipe_config = cik_db_pipe_config(sscreen, tile_mode_index);
mt = cik_micro_tile_mode(sscreen, tile_mode_index);
size = (copy_height * pitch) / 4;
cheight = copy_height;
if (((cheight * pitch) / 4) > CIK_SDMA_COPY_MAX_SIZE) {
cheight = (CIK_SDMA_COPY_MAX_SIZE * 4) / pitch;
cheight &= ~(y_align - 1);
}
ncopy = (copy_height + cheight - 1) / cheight;
r600_need_dma_space(&ctx->b, ncopy * 12);
r600_context_bo_reloc(&ctx->b, &ctx->b.rings.dma, &rsrc->resource,
RADEON_USAGE_READ, RADEON_PRIO_MIN);
r600_context_bo_reloc(&ctx->b, &ctx->b.rings.dma, &rdst->resource,
RADEON_USAGE_WRITE, RADEON_PRIO_MIN);
copy_height = size * 4 / pitch;
for (i = 0; i < ncopy; i++) {
cheight = copy_height;
if (((cheight * pitch) / 4) > CIK_SDMA_COPY_MAX_SIZE) {
cheight = (CIK_SDMA_COPY_MAX_SIZE * 4) / pitch;
cheight &= ~(y_align - 1);
}
size = (cheight * pitch) / 4;
cs->buf[cs->cdw++] = CIK_SDMA_PACKET(CIK_SDMA_OPCODE_COPY,
sub_op, detile << 15);
cs->buf[cs->cdw++] = base;
cs->buf[cs->cdw++] = base >> 32;
cs->buf[cs->cdw++] = ((height - 1) << 16) | pitch_tile_max;
cs->buf[cs->cdw++] = slice_tile_max;
cs->buf[cs->cdw++] = (pipe_config << 26) | (mt_aspect << 24) |
(nbanks << 21) | (bank_h << 18) | (bank_w << 15) |
(tile_split << 11) | (mt << 8) | (array_mode << 3) |
lbpe;
cs->buf[cs->cdw++] = y << 16; /* | x */
cs->buf[cs->cdw++] = 0; /* z */;
cs->buf[cs->cdw++] = addr & 0xfffffffc;
cs->buf[cs->cdw++] = addr >> 32;
cs->buf[cs->cdw++] = (pitch / bpe) - 1;
cs->buf[cs->cdw++] = size;
copy_height -= cheight;
y += cheight;
}
}
void cik_sdma_copy(struct pipe_context *ctx,
struct pipe_resource *dst,
unsigned dst_level,
unsigned dstx, unsigned dsty, unsigned dstz,
struct pipe_resource *src,
unsigned src_level,
const struct pipe_box *src_box)
{
struct si_context *sctx = (struct si_context *)ctx;
struct r600_texture *rsrc = (struct r600_texture*)src;
struct r600_texture *rdst = (struct r600_texture*)dst;
unsigned dst_pitch, src_pitch, bpe, dst_mode, src_mode;
unsigned src_w, dst_w;
unsigned src_x, src_y;
unsigned copy_height, y_align;
unsigned dst_x = dstx, dst_y = dsty, dst_z = dstz;
if (sctx->b.rings.dma.cs == NULL) {
goto fallback;
}
if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) {
cik_sdma_copy_buffer(sctx, dst, src, dst_x, src_box->x, src_box->width);
return;
}
/* Before re-enabling this, please make sure you can hit all newly
* enabled paths in your testing, preferably with both piglit (in
* particular the streaming-texture-leak test) and real world apps
* (e.g. the UE4 Elemental demo).
*/
goto fallback;
if (src->format != dst->format ||
rdst->surface.nsamples > 1 || rsrc->surface.nsamples > 1 ||
rdst->dirty_level_mask & (1 << dst_level)) {
goto fallback;
}
if (rsrc->dirty_level_mask & (1 << src_level)) {
if (rsrc->htile_buffer)
goto fallback;
ctx->flush_resource(ctx, src);
}
src_x = util_format_get_nblocksx(src->format, src_box->x);
dst_x = util_format_get_nblocksx(src->format, dst_x);
src_y = util_format_get_nblocksy(src->format, src_box->y);
dst_y = util_format_get_nblocksy(src->format, dst_y);
dst_pitch = rdst->surface.level[dst_level].pitch_bytes;
src_pitch = rsrc->surface.level[src_level].pitch_bytes;
src_w = rsrc->surface.level[src_level].npix_x;
dst_w = rdst->surface.level[dst_level].npix_x;
if (src_pitch != dst_pitch || src_box->x || dst_x || src_w != dst_w ||
src_box->width != src_w ||
rsrc->surface.level[src_level].nblk_y !=
rdst->surface.level[dst_level].nblk_y) {
/* FIXME CIK can do partial blit */
goto fallback;
}
bpe = rdst->surface.bpe;
copy_height = src_box->height / rsrc->surface.blk_h;
dst_mode = rdst->surface.level[dst_level].mode;
src_mode = rsrc->surface.level[src_level].mode;
/* downcast linear aligned to linear to simplify test */
src_mode = src_mode == RADEON_SURF_MODE_LINEAR_ALIGNED ? RADEON_SURF_MODE_LINEAR : src_mode;
dst_mode = dst_mode == RADEON_SURF_MODE_LINEAR_ALIGNED ? RADEON_SURF_MODE_LINEAR : dst_mode;
/* Dimensions must be aligned to (macro)tiles */
switch (src_mode == RADEON_SURF_MODE_LINEAR ? dst_mode : src_mode) {
case RADEON_SURF_MODE_1D:
if ((src_x % 8) || (src_y % 8) || (dst_x % 8) || (dst_y % 8) ||
(copy_height % 8))
goto fallback;
y_align = 8;
break;
case RADEON_SURF_MODE_2D: {
unsigned mtilew, mtileh, num_banks;
switch (si_num_banks(sctx->screen, rsrc)) {
case V_02803C_ADDR_SURF_2_BANK:
default:
num_banks = 2;
break;
case V_02803C_ADDR_SURF_4_BANK:
num_banks = 4;
break;
case V_02803C_ADDR_SURF_8_BANK:
num_banks = 8;
break;
case V_02803C_ADDR_SURF_16_BANK:
num_banks = 16;
break;
}
mtilew = (8 * rsrc->surface.bankw *
sctx->screen->b.tiling_info.num_channels) *
rsrc->surface.mtilea;
assert(!(mtilew & (mtilew - 1)));
mtileh = (8 * rsrc->surface.bankh * num_banks) /
rsrc->surface.mtilea;
assert(!(mtileh & (mtileh - 1)));
if ((src_x & (mtilew - 1)) || (src_y & (mtileh - 1)) ||
(dst_x & (mtilew - 1)) || (dst_y & (mtileh - 1)) ||
(copy_height & (mtileh - 1)))
goto fallback;
y_align = mtileh;
break;
}
default:
y_align = 1;
}
if (src_mode == dst_mode) {
uint64_t dst_offset, src_offset;
unsigned src_h, dst_h;
src_h = rsrc->surface.level[src_level].npix_y;
dst_h = rdst->surface.level[dst_level].npix_y;
if (src_box->depth > 1 &&
(src_y || dst_y || src_h != dst_h || src_box->height != src_h))
goto fallback;
/* simple dma blit would do NOTE code here assume :
* dst_pitch == src_pitch
*/
src_offset= rsrc->surface.level[src_level].offset;
src_offset += rsrc->surface.level[src_level].slice_size * src_box->z;
src_offset += src_y * src_pitch + src_x * bpe;
dst_offset = rdst->surface.level[dst_level].offset;
dst_offset += rdst->surface.level[dst_level].slice_size * dst_z;
dst_offset += dst_y * dst_pitch + dst_x * bpe;
cik_sdma_do_copy_buffer(sctx, dst, src, dst_offset, src_offset,
src_box->depth *
rsrc->surface.level[src_level].slice_size);
} else {
if (dst_y != src_y || src_box->depth > 1 || src_box->z || dst_z)
goto fallback;
cik_sdma_copy_tile(sctx, dst, dst_level, src, src_level,
src_y, copy_height, y_align, dst_pitch, bpe);
}
return;
fallback:
si_resource_copy_region(ctx, dst, dst_level, dstx, dsty, dstz,
src, src_level, src_box);
}
|