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
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
|
/*
* Copyright 2018-2019 Alyssa Rosenzweig
*
* 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
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* 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 NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS 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.
*
*/
#include "pan_context.h"
#include "pan_util.h"
#include "pan_format.h"
#include "util/u_format.h"
static void
panfrost_invert_swizzle(const unsigned char *in, unsigned char *out)
{
/* First, default to all zeroes to prevent uninitialized junk */
for (unsigned c = 0; c < 4; ++c)
out[c] = PIPE_SWIZZLE_0;
/* Now "do" what the swizzle says */
for (unsigned c = 0; c < 4; ++c) {
unsigned char i = in[c];
/* Who cares? */
if (i < PIPE_SWIZZLE_X || i > PIPE_SWIZZLE_W)
continue;
/* Invert */
unsigned idx = i - PIPE_SWIZZLE_X;
out[idx] = PIPE_SWIZZLE_X + c;
}
}
static struct mali_rt_format
panfrost_mfbd_format(struct pipe_surface *surf)
{
/* Explode details on the format */
const struct util_format_description *desc =
util_format_description(surf->format);
/* The swizzle for rendering is inverted from texturing */
unsigned char swizzle[4];
panfrost_invert_swizzle(desc->swizzle, swizzle);
/* Fill in accordingly, defaulting to 8-bit UNORM */
struct mali_rt_format fmt = {
.unk1 = 0x4000000,
.unk2 = 0x1,
.nr_channels = MALI_POSITIVE(desc->nr_channels),
.unk3 = 0x4,
.flags = 0x8,
.swizzle = panfrost_translate_swizzle_4(swizzle),
.unk4 = 0x8
};
if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB)
fmt.flags |= MALI_MFBD_FORMAT_SRGB;
/* sRGB handled as a dedicated flag */
enum pipe_format linearized = util_format_linear(surf->format);
/* If RGB, we're good to go */
if (util_format_is_unorm8(desc))
return fmt;
/* Set flags for alternative formats */
switch (linearized) {
case PIPE_FORMAT_B5G6R5_UNORM:
fmt.unk1 = 0x14000000;
fmt.nr_channels = MALI_POSITIVE(2);
fmt.unk3 |= 0x1;
break;
case PIPE_FORMAT_A4B4G4R4_UNORM:
case PIPE_FORMAT_B4G4R4A4_UNORM:
fmt.unk1 = 0x10000000;
fmt.unk3 = 0x5;
fmt.nr_channels = MALI_POSITIVE(1);
break;
case PIPE_FORMAT_R10G10B10A2_UNORM:
case PIPE_FORMAT_B10G10R10A2_UNORM:
case PIPE_FORMAT_R10G10B10X2_UNORM:
case PIPE_FORMAT_B10G10R10X2_UNORM:
fmt.unk1 = 0x08000000;
fmt.unk3 = 0x6;
fmt.nr_channels = MALI_POSITIVE(1);
break;
/* Generic 8-bit */
case PIPE_FORMAT_R8_UINT:
case PIPE_FORMAT_R8_SINT:
fmt.unk1 = 0x80000000;
fmt.unk3 = 0x0;
fmt.nr_channels = MALI_POSITIVE(1);
break;
/* Generic 32-bit */
case PIPE_FORMAT_R11G11B10_FLOAT:
case PIPE_FORMAT_R8G8B8A8_UINT:
case PIPE_FORMAT_R8G8B8A8_SINT:
case PIPE_FORMAT_R16G16_FLOAT:
case PIPE_FORMAT_R16G16_UINT:
case PIPE_FORMAT_R16G16_SINT:
case PIPE_FORMAT_R32_FLOAT:
case PIPE_FORMAT_R32_UINT:
case PIPE_FORMAT_R32_SINT:
case PIPE_FORMAT_R10G10B10A2_UINT:
fmt.unk1 = 0x88000000;
fmt.unk3 = 0x0;
fmt.nr_channels = MALI_POSITIVE(4);
break;
/* Generic 16-bit */
case PIPE_FORMAT_R8G8_UINT:
case PIPE_FORMAT_R8G8_SINT:
case PIPE_FORMAT_R16_FLOAT:
case PIPE_FORMAT_R16_UINT:
case PIPE_FORMAT_R16_SINT:
case PIPE_FORMAT_B5G5R5A1_UNORM:
fmt.unk1 = 0x84000000;
fmt.unk3 = 0x0;
fmt.nr_channels = MALI_POSITIVE(2);
break;
/* Generic 64-bit */
case PIPE_FORMAT_R32G32_FLOAT:
case PIPE_FORMAT_R32G32_SINT:
case PIPE_FORMAT_R32G32_UINT:
case PIPE_FORMAT_R16G16B16A16_FLOAT:
case PIPE_FORMAT_R16G16B16A16_SINT:
case PIPE_FORMAT_R16G16B16A16_UINT:
fmt.unk1 = 0x8c000000;
fmt.unk3 = 0x1;
fmt.nr_channels = MALI_POSITIVE(2);
break;
/* Generic 128-bit */
case PIPE_FORMAT_R32G32B32A32_FLOAT:
case PIPE_FORMAT_R32G32B32A32_SINT:
case PIPE_FORMAT_R32G32B32A32_UINT:
fmt.unk1 = 0x90000000;
fmt.unk3 = 0x1;
fmt.nr_channels = MALI_POSITIVE(4);
break;
default:
unreachable("Invalid format rendering");
}
return fmt;
}
static void
panfrost_mfbd_clear(
struct panfrost_job *job,
struct bifrost_framebuffer *fb,
struct bifrost_fb_extra *fbx,
struct bifrost_render_target *rts,
unsigned rt_count)
{
for (unsigned i = 0; i < rt_count; ++i) {
if (!(job->clear & (PIPE_CLEAR_COLOR0 << i)))
continue;
rts[i].clear_color_1 = job->clear_color[i][0];
rts[i].clear_color_2 = job->clear_color[i][1];
rts[i].clear_color_3 = job->clear_color[i][2];
rts[i].clear_color_4 = job->clear_color[i][3];
}
if (job->clear & PIPE_CLEAR_DEPTH) {
fb->clear_depth = job->clear_depth;
}
if (job->clear & PIPE_CLEAR_STENCIL) {
fb->clear_stencil = job->clear_stencil;
}
}
static void
panfrost_mfbd_set_cbuf(
struct bifrost_render_target *rt,
struct pipe_surface *surf)
{
struct panfrost_resource *rsrc = pan_resource(surf->texture);
unsigned level = surf->u.tex.level;
unsigned first_layer = surf->u.tex.first_layer;
assert(surf->u.tex.last_layer == first_layer);
int stride = rsrc->slices[level].stride;
mali_ptr base = panfrost_get_texture_address(rsrc, level, first_layer);
rt->format = panfrost_mfbd_format(surf);
/* Now, we set the layout specific pieces */
if (rsrc->layout == PAN_LINEAR) {
rt->format.block = MALI_MFBD_BLOCK_LINEAR;
rt->framebuffer = base;
rt->framebuffer_stride = stride / 16;
} else if (rsrc->layout == PAN_TILED) {
rt->format.block = MALI_MFBD_BLOCK_TILED;
rt->framebuffer = base;
rt->framebuffer_stride = stride;
} else if (rsrc->layout == PAN_AFBC) {
rt->format.block = MALI_MFBD_BLOCK_AFBC;
unsigned header_size = rsrc->slices[level].header_size;
rt->framebuffer = base + header_size;
rt->afbc.metadata = base;
rt->afbc.stride = 0;
rt->afbc.unk = 0x30009;
/* TODO: Investigate shift */
rt->framebuffer_stride = stride << 1;
} else {
fprintf(stderr, "Invalid render layout (cbuf)");
assert(0);
}
}
static void
panfrost_mfbd_set_zsbuf(
struct bifrost_framebuffer *fb,
struct bifrost_fb_extra *fbx,
struct pipe_surface *surf)
{
struct panfrost_resource *rsrc = pan_resource(surf->texture);
unsigned level = surf->u.tex.level;
assert(surf->u.tex.first_layer == 0);
unsigned offset = rsrc->slices[level].offset;
if (rsrc->layout == PAN_AFBC) {
/* The only Z/S format we can compress is Z24S8 or variants
* thereof (handled by the state tracker) */
assert(surf->format == PIPE_FORMAT_Z24_UNORM_S8_UINT);
mali_ptr base = rsrc->bo->gpu + offset;
unsigned header_size = rsrc->slices[level].header_size;
fb->mfbd_flags |= MALI_MFBD_EXTRA;
fbx->flags =
MALI_EXTRA_PRESENT |
MALI_EXTRA_AFBC |
MALI_EXTRA_AFBC_ZS |
MALI_EXTRA_ZS |
0x1; /* unknown */
fbx->ds_afbc.depth_stencil = base + header_size;
fbx->ds_afbc.depth_stencil_afbc_metadata = base;
fbx->ds_afbc.depth_stencil_afbc_stride = 0;
fbx->ds_afbc.zero1 = 0x10009;
fbx->ds_afbc.padding = 0x1000;
} else if (rsrc->layout == PAN_LINEAR) {
/* TODO: Z32F(S8) support, which is always linear */
assert(surf->format == PIPE_FORMAT_Z24_UNORM_S8_UINT);
int stride = rsrc->slices[level].stride;
fb->mfbd_flags |= MALI_MFBD_EXTRA;
fbx->flags |= MALI_EXTRA_PRESENT | MALI_EXTRA_ZS | 0x1;
fbx->ds_linear.depth = rsrc->bo->gpu + offset;
fbx->ds_linear.depth_stride = stride;
} else {
assert(0);
}
}
/* Helper for sequential uploads used for MFBD */
#define UPLOAD(dest, offset, src, max) { \
size_t sz = sizeof(*src); \
memcpy(dest.cpu + offset, src, sz); \
assert((offset + sz) <= max); \
offset += sz; \
}
static mali_ptr
panfrost_mfbd_upload(
struct panfrost_context *ctx,
struct bifrost_framebuffer *fb,
struct bifrost_fb_extra *fbx,
struct bifrost_render_target *rts,
unsigned cbufs)
{
off_t offset = 0;
/* There may be extra data stuck in the middle */
bool has_extra = fb->mfbd_flags & MALI_MFBD_EXTRA;
/* Compute total size for transfer */
size_t total_sz =
sizeof(struct bifrost_framebuffer) +
(has_extra ? sizeof(struct bifrost_fb_extra) : 0) +
sizeof(struct bifrost_render_target) * cbufs;
struct panfrost_transfer m_f_trans =
panfrost_allocate_transient(ctx, total_sz);
/* Do the transfer */
UPLOAD(m_f_trans, offset, fb, total_sz);
if (has_extra)
UPLOAD(m_f_trans, offset, fbx, total_sz);
for (unsigned c = 0; c < cbufs; ++c) {
UPLOAD(m_f_trans, offset, &rts[c], total_sz);
}
/* Return pointer suitable for the fragment section */
return m_f_trans.gpu | MALI_MFBD | (has_extra ? 2 : 0);
}
#undef UPLOAD
/* Creates an MFBD for the FRAGMENT section of the bound framebuffer */
mali_ptr
panfrost_mfbd_fragment(struct panfrost_context *ctx, bool has_draws)
{
struct panfrost_job *job = panfrost_get_job_for_fbo(ctx);
struct bifrost_framebuffer fb = panfrost_emit_mfbd(ctx, has_draws);
struct bifrost_fb_extra fbx = {};
struct bifrost_render_target rts[4] = {};
/* XXX: MRT case */
fb.rt_count_2 = 1;
fb.mfbd_flags = 0x100;
/* TODO: MRT clear */
panfrost_mfbd_clear(job, &fb, &fbx, rts, fb.rt_count_2);
for (int cb = 0; cb < ctx->pipe_framebuffer.nr_cbufs; ++cb) {
struct pipe_surface *surf = ctx->pipe_framebuffer.cbufs[cb];
unsigned bpp = util_format_get_blocksize(surf->format);
panfrost_mfbd_set_cbuf(&rts[cb], surf);
/* What is this? Looks like some extension of the bpp field.
* Maybe it establishes how much internal tilebuffer space is
* reserved? */
fb.rt_count_2 = MAX2(fb.rt_count_2, ALIGN_POT(bpp, 4) / 4);
}
if (ctx->pipe_framebuffer.zsbuf) {
panfrost_mfbd_set_zsbuf(&fb, &fbx, ctx->pipe_framebuffer.zsbuf);
}
/* For the special case of a depth-only FBO, we need to attach a dummy render target */
if (ctx->pipe_framebuffer.nr_cbufs == 0) {
struct mali_rt_format null_rt = {
.unk1 = 0x4000000,
.unk4 = 0x8
};
rts[0].format = null_rt;
rts[0].framebuffer = 0;
rts[0].framebuffer_stride = 0;
}
/* When scanning out, the depth buffer is immediately invalidated, so
* we don't need to waste bandwidth writing it out. This can improve
* performance substantially (Z32_UNORM 1080p @ 60fps is 475 MB/s of
* memory bandwidth!).
*
* The exception is ReadPixels, but this is not supported on GLES so we
* can safely ignore it. */
if (panfrost_is_scanout(ctx)) {
job->requirements &= ~PAN_REQ_DEPTH_WRITE;
}
/* Actualize the requirements */
if (job->requirements & PAN_REQ_MSAA) {
rts[0].format.flags |= MALI_MFBD_FORMAT_MSAA;
/* XXX */
fb.unk1 |= (1 << 4) | (1 << 1);
fb.rt_count_2 = 4;
}
if (job->requirements & PAN_REQ_DEPTH_WRITE)
fb.mfbd_flags |= MALI_MFBD_DEPTH_WRITE;
/* Checksumming only works with a single render target */
if (ctx->pipe_framebuffer.nr_cbufs == 1) {
struct pipe_surface *surf = ctx->pipe_framebuffer.cbufs[0];
struct panfrost_resource *rsrc = pan_resource(surf->texture);
struct panfrost_bo *bo = rsrc->bo;
if (rsrc->checksummed) {
unsigned level = surf->u.tex.level;
struct panfrost_slice *slice = &rsrc->slices[level];
fb.mfbd_flags |= MALI_MFBD_EXTRA;
fbx.flags |= MALI_EXTRA_PRESENT;
fbx.checksum_stride = slice->checksum_stride;
fbx.checksum = bo->gpu + slice->checksum_offset;
}
}
/* We always upload at least one (dummy) cbuf */
unsigned cbufs = MAX2(ctx->pipe_framebuffer.nr_cbufs, 1);
return panfrost_mfbd_upload(ctx, &fb, &fbx, rts, cbufs);
}
|