/
SDL_render_metal.m
1794 lines (1524 loc) · 65.3 KB
1
2
/*
Simple DirectMedia Layer
3
Copyright (C) 1997-2019 Sam Lantinga <slouken@libsdl.org>
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
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
#include "../../SDL_internal.h"
#if SDL_VIDEO_RENDER_METAL && !SDL_RENDER_DISABLED
#include "SDL_hints.h"
#include "SDL_log.h"
#include "SDL_assert.h"
#include "SDL_syswm.h"
#include "../SDL_sysrender.h"
31
#ifdef __MACOSX__
32
#include "../../video/cocoa/SDL_cocoametalview.h"
33
34
35
#else
#include "../../video/uikit/SDL_uikitmetalview.h"
#endif
36
#include <Availability.h>
37
38
#import <Metal/Metal.h>
#import <QuartzCore/CAMetalLayer.h>
39
40
41
42
43
44
45
/* Regenerate these with build-metal-shaders.sh */
#ifdef __MACOSX__
#include "SDL_shaders_metal_osx.h"
#else
#include "SDL_shaders_metal_ios.h"
#endif
46
47
48
/* Apple Metal renderer implementation */
49
50
51
52
53
54
55
56
57
/* macOS requires constants in a buffer to have a 256 byte alignment. */
#ifdef __MACOSX__
#define CONSTANT_ALIGN 256
#else
#define CONSTANT_ALIGN 4
#endif
#define ALIGN_CONSTANTS(size) ((size + CONSTANT_ALIGN - 1) & (~(CONSTANT_ALIGN - 1)))
58
static const size_t CONSTANTS_OFFSET_INVALID = 0xFFFFFFFF;
59
60
static const size_t CONSTANTS_OFFSET_IDENTITY = 0;
static const size_t CONSTANTS_OFFSET_HALF_PIXEL_TRANSFORM = ALIGN_CONSTANTS(CONSTANTS_OFFSET_IDENTITY + sizeof(float) * 16);
61
62
63
static const size_t CONSTANTS_OFFSET_DECODE_JPEG = ALIGN_CONSTANTS(CONSTANTS_OFFSET_HALF_PIXEL_TRANSFORM + sizeof(float) * 16);
static const size_t CONSTANTS_OFFSET_DECODE_BT601 = ALIGN_CONSTANTS(CONSTANTS_OFFSET_DECODE_JPEG + sizeof(float) * 4 * 4);
static const size_t CONSTANTS_OFFSET_DECODE_BT709 = ALIGN_CONSTANTS(CONSTANTS_OFFSET_DECODE_BT601 + sizeof(float) * 4 * 4);
64
static const size_t CONSTANTS_LENGTH = CONSTANTS_OFFSET_DECODE_BT709 + sizeof(float) * 4 * 4;
65
66
67
68
69
70
71
72
73
typedef enum SDL_MetalVertexFunction
{
SDL_METAL_VERTEX_SOLID,
SDL_METAL_VERTEX_COPY,
} SDL_MetalVertexFunction;
typedef enum SDL_MetalFragmentFunction
{
74
SDL_METAL_FRAGMENT_SOLID = 0,
75
SDL_METAL_FRAGMENT_COPY,
76
77
78
79
SDL_METAL_FRAGMENT_YUV,
SDL_METAL_FRAGMENT_NV12,
SDL_METAL_FRAGMENT_NV21,
SDL_METAL_FRAGMENT_COUNT,
80
81
82
83
84
85
86
87
88
89
90
91
92
93
} SDL_MetalFragmentFunction;
typedef struct METAL_PipelineState
{
SDL_BlendMode blendMode;
void *pipe;
} METAL_PipelineState;
typedef struct METAL_PipelineCache
{
METAL_PipelineState *states;
int count;
SDL_MetalVertexFunction vertexFunction;
SDL_MetalFragmentFunction fragmentFunction;
94
MTLPixelFormat renderTargetFormat;
95
96
97
const char *label;
} METAL_PipelineCache;
98
/* Each shader combination used by drawing functions has a separate pipeline
99
100
101
102
103
* cache, and we have a separate list of caches for each render target pixel
* format. This is more efficient than iterating over a global cache to find
* the pipeline based on the specified shader combination and RT pixel format,
* since we know what the RT pixel format is when we set the render target, and
* we know what the shader combination is inside each drawing function's code. */
104
105
typedef struct METAL_ShaderPipelines
{
106
MTLPixelFormat renderTargetFormat;
107
108
109
METAL_PipelineCache caches[SDL_METAL_FRAGMENT_COUNT];
} METAL_ShaderPipelines;
110
@interface METAL_RenderData : NSObject
111
112
113
114
115
116
@property (nonatomic, retain) id<MTLDevice> mtldevice;
@property (nonatomic, retain) id<MTLCommandQueue> mtlcmdqueue;
@property (nonatomic, retain) id<MTLCommandBuffer> mtlcmdbuffer;
@property (nonatomic, retain) id<MTLRenderCommandEncoder> mtlcmdencoder;
@property (nonatomic, retain) id<MTLLibrary> mtllibrary;
@property (nonatomic, retain) id<CAMetalDrawable> mtlbackbuffer;
117
118
@property (nonatomic, retain) id<MTLSamplerState> mtlsamplernearest;
@property (nonatomic, retain) id<MTLSamplerState> mtlsamplerlinear;
119
@property (nonatomic, retain) id<MTLBuffer> mtlbufconstants;
120
@property (nonatomic, retain) id<MTLBuffer> mtlbufquadindices;
121
122
@property (nonatomic, retain) CAMetalLayer *mtllayer;
@property (nonatomic, retain) MTLRenderPassDescriptor *mtlpassdesc;
123
124
125
@property (nonatomic, assign) METAL_ShaderPipelines *activepipelines;
@property (nonatomic, assign) METAL_ShaderPipelines *allpipelines;
@property (nonatomic, assign) int pipelinescount;
126
127
128
@end
@implementation METAL_RenderData
129
#if !__has_feature(objc_arc)
130
131
132
133
134
135
136
137
138
139
- (void)dealloc
{
[_mtldevice release];
[_mtlcmdqueue release];
[_mtlcmdbuffer release];
[_mtlcmdencoder release];
[_mtllibrary release];
[_mtlbackbuffer release];
[_mtlsamplernearest release];
[_mtlsamplerlinear release];
140
[_mtlbufconstants release];
141
[_mtlbufquadindices release];
142
143
144
145
146
[_mtllayer release];
[_mtlpassdesc release];
[super dealloc];
}
#endif
147
@end
148
149
150
@interface METAL_TextureData : NSObject
@property (nonatomic, retain) id<MTLTexture> mtltexture;
151
@property (nonatomic, retain) id<MTLTexture> mtltexture_uv;
152
@property (nonatomic, retain) id<MTLSamplerState> mtlsampler;
153
154
155
156
@property (nonatomic, assign) SDL_MetalFragmentFunction fragmentFunction;
@property (nonatomic, assign) BOOL yuv;
@property (nonatomic, assign) BOOL nv12;
@property (nonatomic, assign) size_t conversionBufferOffset;
157
@property (nonatomic, assign) BOOL hasdata;
158
159
160
@property (nonatomic, retain) id<MTLBuffer> lockedbuffer;
@property (nonatomic, assign) SDL_Rect lockedrect;
161
162
163
@end
@implementation METAL_TextureData
164
#if !__has_feature(objc_arc)
165
166
167
- (void)dealloc
{
[_mtltexture release];
168
[_mtltexture_uv release];
169
170
171
172
[_mtlsampler release];
[super dealloc];
}
#endif
173
174
@end
175
176
177
static int
IsMetalAvailable(const SDL_SysWMinfo *syswm)
{
178
179
if (syswm->subsystem != SDL_SYSWM_COCOA && syswm->subsystem != SDL_SYSWM_UIKIT) {
return SDL_SetError("Metal render target only supports Cocoa and UIKit video targets at the moment.");
180
181
182
}
// this checks a weak symbol.
183
#if (defined(__MACOSX__) && (MAC_OS_X_VERSION_MIN_REQUIRED < 101100))
184
185
186
187
188
189
190
191
if (MTLCreateSystemDefaultDevice == NULL) { // probably on 10.10 or lower.
return SDL_SetError("Metal framework not available on this system");
}
#endif
return 0;
}
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
static const MTLBlendOperation invalidBlendOperation = (MTLBlendOperation)0xFFFFFFFF;
static const MTLBlendFactor invalidBlendFactor = (MTLBlendFactor)0xFFFFFFFF;
static MTLBlendOperation
GetBlendOperation(SDL_BlendOperation operation)
{
switch (operation) {
case SDL_BLENDOPERATION_ADD: return MTLBlendOperationAdd;
case SDL_BLENDOPERATION_SUBTRACT: return MTLBlendOperationSubtract;
case SDL_BLENDOPERATION_REV_SUBTRACT: return MTLBlendOperationReverseSubtract;
case SDL_BLENDOPERATION_MINIMUM: return MTLBlendOperationMin;
case SDL_BLENDOPERATION_MAXIMUM: return MTLBlendOperationMax;
default: return invalidBlendOperation;
}
}
static MTLBlendFactor
GetBlendFactor(SDL_BlendFactor factor)
{
switch (factor) {
case SDL_BLENDFACTOR_ZERO: return MTLBlendFactorZero;
case SDL_BLENDFACTOR_ONE: return MTLBlendFactorOne;
case SDL_BLENDFACTOR_SRC_COLOR: return MTLBlendFactorSourceColor;
case SDL_BLENDFACTOR_ONE_MINUS_SRC_COLOR: return MTLBlendFactorOneMinusSourceColor;
case SDL_BLENDFACTOR_SRC_ALPHA: return MTLBlendFactorSourceAlpha;
case SDL_BLENDFACTOR_ONE_MINUS_SRC_ALPHA: return MTLBlendFactorOneMinusSourceAlpha;
case SDL_BLENDFACTOR_DST_COLOR: return MTLBlendFactorDestinationColor;
case SDL_BLENDFACTOR_ONE_MINUS_DST_COLOR: return MTLBlendFactorOneMinusDestinationColor;
case SDL_BLENDFACTOR_DST_ALPHA: return MTLBlendFactorDestinationAlpha;
case SDL_BLENDFACTOR_ONE_MINUS_DST_ALPHA: return MTLBlendFactorOneMinusDestinationAlpha;
default: return invalidBlendFactor;
}
}
static NSString *
GetVertexFunctionName(SDL_MetalVertexFunction function)
{
switch (function) {
case SDL_METAL_VERTEX_SOLID: return @"SDL_Solid_vertex";
case SDL_METAL_VERTEX_COPY: return @"SDL_Copy_vertex";
default: return nil;
}
}
static NSString *
GetFragmentFunctionName(SDL_MetalFragmentFunction function)
{
switch (function) {
case SDL_METAL_FRAGMENT_SOLID: return @"SDL_Solid_fragment";
241
case SDL_METAL_FRAGMENT_COPY: return @"SDL_Copy_fragment";
242
243
244
case SDL_METAL_FRAGMENT_YUV: return @"SDL_YUV_fragment";
case SDL_METAL_FRAGMENT_NV12: return @"SDL_NV12_fragment";
case SDL_METAL_FRAGMENT_NV21: return @"SDL_NV21_fragment";
245
246
247
248
default: return nil;
}
}
249
static id<MTLRenderPipelineState>
250
251
MakePipelineState(METAL_RenderData *data, METAL_PipelineCache *cache,
NSString *blendlabel, SDL_BlendMode blendmode)
252
{
253
254
id<MTLFunction> mtlvertfn = [data.mtllibrary newFunctionWithName:GetVertexFunctionName(cache->vertexFunction)];
id<MTLFunction> mtlfragfn = [data.mtllibrary newFunctionWithName:GetFragmentFunctionName(cache->fragmentFunction)];
255
256
257
258
259
260
SDL_assert(mtlvertfn != nil);
SDL_assert(mtlfragfn != nil);
MTLRenderPipelineDescriptor *mtlpipedesc = [[MTLRenderPipelineDescriptor alloc] init];
mtlpipedesc.vertexFunction = mtlvertfn;
mtlpipedesc.fragmentFunction = mtlfragfn;
261
262
263
MTLRenderPipelineColorAttachmentDescriptor *rtdesc = mtlpipedesc.colorAttachments[0];
264
rtdesc.pixelFormat = cache->renderTargetFormat;
265
266
267
268
269
270
271
272
273
274
275
276
277
278
if (blendmode != SDL_BLENDMODE_NONE) {
rtdesc.blendingEnabled = YES;
rtdesc.sourceRGBBlendFactor = GetBlendFactor(SDL_GetBlendModeSrcColorFactor(blendmode));
rtdesc.destinationRGBBlendFactor = GetBlendFactor(SDL_GetBlendModeDstColorFactor(blendmode));
rtdesc.rgbBlendOperation = GetBlendOperation(SDL_GetBlendModeColorOperation(blendmode));
rtdesc.sourceAlphaBlendFactor = GetBlendFactor(SDL_GetBlendModeSrcAlphaFactor(blendmode));
rtdesc.destinationAlphaBlendFactor = GetBlendFactor(SDL_GetBlendModeDstAlphaFactor(blendmode));
rtdesc.alphaBlendOperation = GetBlendOperation(SDL_GetBlendModeAlphaOperation(blendmode));
} else {
rtdesc.blendingEnabled = NO;
}
mtlpipedesc.label = [@(cache->label) stringByAppendingString:blendlabel];
279
280
NSError *err = nil;
281
id<MTLRenderPipelineState> state = [data.mtldevice newRenderPipelineStateWithDescriptor:mtlpipedesc error:&err];
282
SDL_assert(err == nil);
283
284
285
286
287
288
289
METAL_PipelineState pipeline;
pipeline.blendMode = blendmode;
pipeline.pipe = (void *)CFBridgingRetain(state);
METAL_PipelineState *states = SDL_realloc(cache->states, (cache->count + 1) * sizeof(pipeline));
290
#if !__has_feature(objc_arc)
291
292
293
[mtlpipedesc release]; // !!! FIXME: can these be reused for each creation, or does the pipeline obtain it?
[mtlvertfn release];
[mtlfragfn release];
294
[state release];
295
#endif
296
297
298
299
300
301
302
303
304
305
306
307
if (states) {
states[cache->count++] = pipeline;
cache->states = states;
return (__bridge id<MTLRenderPipelineState>)pipeline.pipe;
} else {
CFBridgingRelease(pipeline.pipe);
SDL_OutOfMemory();
return NULL;
}
}
308
static void
309
310
MakePipelineCache(METAL_RenderData *data, METAL_PipelineCache *cache, const char *label,
MTLPixelFormat rtformat, SDL_MetalVertexFunction vertfn, SDL_MetalFragmentFunction fragfn)
311
312
313
314
315
{
SDL_zerop(cache);
cache->vertexFunction = vertfn;
cache->fragmentFunction = fragfn;
316
cache->renderTargetFormat = rtformat;
317
318
319
320
cache->label = label;
/* Create pipeline states for the default blend modes. Custom blend modes
* will be added to the cache on-demand. */
321
322
323
324
MakePipelineState(data, cache, @" (blend=none)", SDL_BLENDMODE_NONE);
MakePipelineState(data, cache, @" (blend=blend)", SDL_BLENDMODE_BLEND);
MakePipelineState(data, cache, @" (blend=add)", SDL_BLENDMODE_ADD);
MakePipelineState(data, cache, @" (blend=mod)", SDL_BLENDMODE_MOD);
325
326
327
}
static void
328
DestroyPipelineCache(METAL_PipelineCache *cache)
329
{
330
331
332
333
334
335
if (cache != NULL) {
for (int i = 0; i < cache->count; i++) {
CFBridgingRelease(cache->states[i].pipe);
}
SDL_free(cache->states);
336
337
338
}
}
339
340
341
342
343
344
345
346
347
348
349
350
351
352
void
MakeShaderPipelines(METAL_RenderData *data, METAL_ShaderPipelines *pipelines, MTLPixelFormat rtformat)
{
SDL_zerop(pipelines);
pipelines->renderTargetFormat = rtformat;
MakePipelineCache(data, &pipelines->caches[SDL_METAL_FRAGMENT_SOLID], "SDL primitives pipeline", rtformat, SDL_METAL_VERTEX_SOLID, SDL_METAL_FRAGMENT_SOLID);
MakePipelineCache(data, &pipelines->caches[SDL_METAL_FRAGMENT_COPY], "SDL copy pipeline", rtformat, SDL_METAL_VERTEX_COPY, SDL_METAL_FRAGMENT_COPY);
MakePipelineCache(data, &pipelines->caches[SDL_METAL_FRAGMENT_YUV], "SDL YUV pipeline", rtformat, SDL_METAL_VERTEX_COPY, SDL_METAL_FRAGMENT_YUV);
MakePipelineCache(data, &pipelines->caches[SDL_METAL_FRAGMENT_NV12], "SDL NV12 pipeline", rtformat, SDL_METAL_VERTEX_COPY, SDL_METAL_FRAGMENT_NV12);
MakePipelineCache(data, &pipelines->caches[SDL_METAL_FRAGMENT_NV21], "SDL NV21 pipeline", rtformat, SDL_METAL_VERTEX_COPY, SDL_METAL_FRAGMENT_NV21);
}
353
static METAL_ShaderPipelines *
354
ChooseShaderPipelines(METAL_RenderData *data, MTLPixelFormat rtformat)
355
{
356
357
358
359
360
361
362
363
364
365
366
367
METAL_ShaderPipelines *allpipelines = data.allpipelines;
int count = data.pipelinescount;
for (int i = 0; i < count; i++) {
if (allpipelines[i].renderTargetFormat == rtformat) {
return &allpipelines[i];
}
}
allpipelines = SDL_realloc(allpipelines, (count + 1) * sizeof(METAL_ShaderPipelines));
if (allpipelines == NULL) {
368
369
370
371
SDL_OutOfMemory();
return NULL;
}
372
373
374
375
MakeShaderPipelines(data, &allpipelines[count], rtformat);
data.allpipelines = allpipelines;
data.pipelinescount = count + 1;
376
377
return &data.allpipelines[count];
378
379
380
}
static void
381
DestroyAllPipelines(METAL_ShaderPipelines *allpipelines, int count)
382
{
383
384
385
386
387
if (allpipelines != NULL) {
for (int i = 0; i < count; i++) {
for (int cache = 0; cache < SDL_METAL_FRAGMENT_COUNT; cache++) {
DestroyPipelineCache(&allpipelines[i].caches[cache]);
}
388
389
}
390
SDL_free(allpipelines);
391
}
392
393
394
}
static inline id<MTLRenderPipelineState>
395
ChoosePipelineState(METAL_RenderData *data, METAL_ShaderPipelines *pipelines, SDL_MetalFragmentFunction fragfn, SDL_BlendMode blendmode)
396
{
397
398
METAL_PipelineCache *cache = &pipelines->caches[fragfn];
399
400
401
402
for (int i = 0; i < cache->count; i++) {
if (cache->states[i].blendMode == blendmode) {
return (__bridge id<MTLRenderPipelineState>)cache->states[i].pipe;
}
403
}
404
405
return MakePipelineState(data, cache, [NSString stringWithFormat:@" (blend=custom 0x%x)", blendmode], blendmode);
406
407
}
408
409
410
411
static void
METAL_ActivateRenderCommandEncoder(SDL_Renderer * renderer, MTLLoadAction load, MTLClearColor *clear_color)
{
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
412
413
414
415
416
/* Our SetRenderTarget just signals that the next render operation should
* set up a new render pass. This is where that work happens. */
if (data.mtlcmdencoder == nil) {
id<MTLTexture> mtltexture = nil;
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
if (renderer->target != NULL) {
METAL_TextureData *texdata = (__bridge METAL_TextureData *)renderer->target->driverdata;
mtltexture = texdata.mtltexture;
} else {
if (data.mtlbackbuffer == nil) {
/* The backbuffer's contents aren't guaranteed to persist after
* presenting, so we can leave it undefined when loading it. */
data.mtlbackbuffer = [data.mtllayer nextDrawable];
if (load == MTLLoadActionLoad) {
load = MTLLoadActionDontCare;
}
}
mtltexture = data.mtlbackbuffer.texture;
}
432
433
SDL_assert(mtltexture);
434
435
436
437
438
if (load == MTLLoadActionClear) {
SDL_assert(clear_color != NULL);
data.mtlpassdesc.colorAttachments[0].clearColor = *clear_color;
}
439
440
441
data.mtlpassdesc.colorAttachments[0].loadAction = load;
data.mtlpassdesc.colorAttachments[0].texture = mtltexture;
442
443
444
data.mtlcmdbuffer = [data.mtlcmdqueue commandBuffer];
data.mtlcmdencoder = [data.mtlcmdbuffer renderCommandEncoderWithDescriptor:data.mtlpassdesc];
445
446
447
448
449
450
if (data.mtlbackbuffer != nil && mtltexture == data.mtlbackbuffer.texture) {
data.mtlcmdencoder.label = @"SDL metal renderer backbuffer";
} else {
data.mtlcmdencoder.label = @"SDL metal renderer render target";
}
451
452
data.activepipelines = ChooseShaderPipelines(data, mtltexture.pixelFormat);
453
454
455
456
457
458
459
// make sure this has a definite place in the queue. This way it will
// execute reliably whether the app tries to make its own command buffers
// or whatever. This means we can _always_ batch rendering commands!
[data.mtlcmdbuffer enqueue];
}
}
460
461
462
463
464
465
466
467
468
static void
METAL_WindowEvent(SDL_Renderer * renderer, const SDL_WindowEvent *event)
{
if (event->event == SDL_WINDOWEVENT_SHOWN ||
event->event == SDL_WINDOWEVENT_HIDDEN) {
// !!! FIXME: write me
}
}
469
470
471
472
473
474
475
476
477
478
479
480
481
static int
METAL_GetOutputSize(SDL_Renderer * renderer, int *w, int *h)
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
if (w) {
*w = (int)data.mtllayer.drawableSize.width;
}
if (h) {
*h = (int)data.mtllayer.drawableSize.height;
}
return 0;
}}
482
483
484
485
486
487
488
489
490
491
static SDL_bool
METAL_SupportsBlendMode(SDL_Renderer * renderer, SDL_BlendMode blendMode)
{
SDL_BlendFactor srcColorFactor = SDL_GetBlendModeSrcColorFactor(blendMode);
SDL_BlendFactor srcAlphaFactor = SDL_GetBlendModeSrcAlphaFactor(blendMode);
SDL_BlendOperation colorOperation = SDL_GetBlendModeColorOperation(blendMode);
SDL_BlendFactor dstColorFactor = SDL_GetBlendModeDstColorFactor(blendMode);
SDL_BlendFactor dstAlphaFactor = SDL_GetBlendModeDstAlphaFactor(blendMode);
SDL_BlendOperation alphaOperation = SDL_GetBlendModeAlphaOperation(blendMode);
492
493
494
495
496
497
498
499
500
501
502
if (GetBlendFactor(srcColorFactor) == invalidBlendFactor ||
GetBlendFactor(srcAlphaFactor) == invalidBlendFactor ||
GetBlendOperation(colorOperation) == invalidBlendOperation ||
GetBlendFactor(dstColorFactor) == invalidBlendFactor ||
GetBlendFactor(dstAlphaFactor) == invalidBlendFactor ||
GetBlendOperation(alphaOperation) == invalidBlendOperation) {
return SDL_FALSE;
}
return SDL_TRUE;
}
503
504
505
506
507
508
static int
METAL_CreateTexture(SDL_Renderer * renderer, SDL_Texture * texture)
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
MTLPixelFormat pixfmt;
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
switch (texture->format) {
case SDL_PIXELFORMAT_ABGR8888:
pixfmt = MTLPixelFormatRGBA8Unorm;
break;
case SDL_PIXELFORMAT_ARGB8888:
pixfmt = MTLPixelFormatBGRA8Unorm;
break;
case SDL_PIXELFORMAT_IYUV:
case SDL_PIXELFORMAT_YV12:
case SDL_PIXELFORMAT_NV12:
case SDL_PIXELFORMAT_NV21:
pixfmt = MTLPixelFormatR8Unorm;
break;
default:
return SDL_SetError("Texture format %s not supported by Metal", SDL_GetPixelFormatName(texture->format));
}
526
527
528
MTLTextureDescriptor *mtltexdesc = [MTLTextureDescriptor texture2DDescriptorWithPixelFormat:pixfmt
width:(NSUInteger)texture->w height:(NSUInteger)texture->h mipmapped:NO];
529
530
531
532
533
534
535
536
537
538
539
540
541
542
/* Not available in iOS 8. */
if ([mtltexdesc respondsToSelector:@selector(usage)]) {
if (texture->access == SDL_TEXTUREACCESS_TARGET) {
mtltexdesc.usage = MTLTextureUsageShaderRead | MTLTextureUsageRenderTarget;
} else {
mtltexdesc.usage = MTLTextureUsageShaderRead;
}
}
id<MTLTexture> mtltexture = [data.mtldevice newTextureWithDescriptor:mtltexdesc];
if (mtltexture == nil) {
return SDL_SetError("Texture allocation failed");
}
543
544
id<MTLTexture> mtltexture_uv = nil;
545
546
547
BOOL yuv = (texture->format == SDL_PIXELFORMAT_IYUV) || (texture->format == SDL_PIXELFORMAT_YV12);
BOOL nv12 = (texture->format == SDL_PIXELFORMAT_NV12) || (texture->format == SDL_PIXELFORMAT_NV21);
548
549
550
551
552
553
554
555
556
557
558
if (yuv) {
mtltexdesc.pixelFormat = MTLPixelFormatR8Unorm;
mtltexdesc.width = (texture->w + 1) / 2;
mtltexdesc.height = (texture->h + 1) / 2;
mtltexdesc.textureType = MTLTextureType2DArray;
mtltexdesc.arrayLength = 2;
} else if (nv12) {
mtltexdesc.pixelFormat = MTLPixelFormatRG8Unorm;
mtltexdesc.width = (texture->w + 1) / 2;
mtltexdesc.height = (texture->h + 1) / 2;
559
560
}
561
if (yuv || nv12) {
562
mtltexture_uv = [data.mtldevice newTextureWithDescriptor:mtltexdesc];
563
564
565
566
567
568
if (mtltexture_uv == nil) {
#if !__has_feature(objc_arc)
[mtltexture release];
#endif
return SDL_SetError("Texture allocation failed");
}
569
}
570
571
572
573
574
575
576
577
578
METAL_TextureData *texturedata = [[METAL_TextureData alloc] init];
if (texture->scaleMode == SDL_ScaleModeNearest) {
texturedata.mtlsampler = data.mtlsamplernearest;
} else {
texturedata.mtlsampler = data.mtlsamplerlinear;
}
texturedata.mtltexture = mtltexture;
texturedata.mtltexture_uv = mtltexture_uv;
579
580
581
texturedata.yuv = yuv;
texturedata.nv12 = nv12;
582
583
584
585
586
587
588
589
590
591
if (yuv) {
texturedata.fragmentFunction = SDL_METAL_FRAGMENT_YUV;
} else if (texture->format == SDL_PIXELFORMAT_NV12) {
texturedata.fragmentFunction = SDL_METAL_FRAGMENT_NV12;
} else if (texture->format == SDL_PIXELFORMAT_NV21) {
texturedata.fragmentFunction = SDL_METAL_FRAGMENT_NV21;
} else {
texturedata.fragmentFunction = SDL_METAL_FRAGMENT_COPY;
}
592
593
594
595
596
597
598
599
600
601
602
603
if (yuv || nv12) {
size_t offset = 0;
SDL_YUV_CONVERSION_MODE mode = SDL_GetYUVConversionModeForResolution(texture->w, texture->h);
switch (mode) {
case SDL_YUV_CONVERSION_JPEG: offset = CONSTANTS_OFFSET_DECODE_JPEG; break;
case SDL_YUV_CONVERSION_BT601: offset = CONSTANTS_OFFSET_DECODE_BT601; break;
case SDL_YUV_CONVERSION_BT709: offset = CONSTANTS_OFFSET_DECODE_BT709; break;
default: offset = 0; break;
}
texturedata.conversionBufferOffset = offset;
}
604
605
texture->driverdata = (void*)CFBridgingRetain(texturedata);
606
607
608
609
610
611
#if !__has_feature(objc_arc)
[texturedata release];
[mtltexture release];
[mtltexture_uv release];
#endif
612
613
614
return 0;
}}
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
static void
METAL_UploadTextureData(id<MTLTexture> texture, SDL_Rect rect, int slice,
const void * pixels, int pitch)
{
[texture replaceRegion:MTLRegionMake2D(rect.x, rect.y, rect.w, rect.h)
mipmapLevel:0
slice:slice
withBytes:pixels
bytesPerRow:pitch
bytesPerImage:0];
}
static MTLStorageMode
METAL_GetStorageMode(id<MTLResource> resource)
{
/* iOS 8 does not have this method. */
if ([resource respondsToSelector:@selector(storageMode)]) {
return resource.storageMode;
}
return MTLStorageModeShared;
}
static int
METAL_UpdateTextureInternal(SDL_Renderer * renderer, METAL_TextureData *texturedata,
id<MTLTexture> texture, SDL_Rect rect, int slice,
const void * pixels, int pitch)
{
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
SDL_Rect stagingrect = {0, 0, rect.w, rect.h};
MTLTextureDescriptor *desc;
/* If the texture is managed or shared and this is the first upload, we can
* use replaceRegion to upload to it directly. Otherwise we upload the data
* to a staging texture and copy that over. */
if (!texturedata.hasdata && METAL_GetStorageMode(texture) != MTLStorageModePrivate) {
METAL_UploadTextureData(texture, rect, slice, pixels, pitch);
return 0;
}
desc = [MTLTextureDescriptor texture2DDescriptorWithPixelFormat:texture.pixelFormat
width:rect.w
height:rect.h
mipmapped:NO];
if (desc == nil) {
return SDL_OutOfMemory();
}
/* TODO: We could have a pool of textures or a MTLHeap we allocate from,
* and release a staging texture back to the pool in the command buffer's
* completion handler. */
id<MTLTexture> stagingtex = [data.mtldevice newTextureWithDescriptor:desc];
if (stagingtex == nil) {
return SDL_OutOfMemory();
}
#if !__has_feature(objc_arc)
[stagingtex autorelease];
#endif
METAL_UploadTextureData(stagingtex, stagingrect, 0, pixels, pitch);
if (data.mtlcmdencoder != nil) {
[data.mtlcmdencoder endEncoding];
data.mtlcmdencoder = nil;
}
if (data.mtlcmdbuffer == nil) {
data.mtlcmdbuffer = [data.mtlcmdqueue commandBuffer];
}
id<MTLBlitCommandEncoder> blitcmd = [data.mtlcmdbuffer blitCommandEncoder];
[blitcmd copyFromTexture:stagingtex
sourceSlice:0
sourceLevel:0
sourceOrigin:MTLOriginMake(0, 0, 0)
sourceSize:MTLSizeMake(rect.w, rect.h, 1)
toTexture:texture
destinationSlice:slice
destinationLevel:0
destinationOrigin:MTLOriginMake(rect.x, rect.y, 0)];
[blitcmd endEncoding];
/* TODO: This isn't very efficient for the YUV formats, which call
* UpdateTextureInternal multiple times in a row. */
[data.mtlcmdbuffer commit];
data.mtlcmdbuffer = nil;
return 0;
}
709
710
static int
METAL_UpdateTexture(SDL_Renderer * renderer, SDL_Texture * texture,
711
const SDL_Rect * rect, const void *pixels, int pitch)
712
713
{ @autoreleasepool {
METAL_TextureData *texturedata = (__bridge METAL_TextureData *)texture->driverdata;
714
715
716
717
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture, *rect, 0, pixels, pitch) < 0) {
return -1;
}
718
719
720
721
if (texturedata.yuv) {
int Uslice = texture->format == SDL_PIXELFORMAT_YV12 ? 1 : 0;
int Vslice = texture->format == SDL_PIXELFORMAT_YV12 ? 0 : 1;
722
723
int UVpitch = (pitch + 1) / 2;
SDL_Rect UVrect = {rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2};
724
725
726
/* Skip to the correct offset into the next texture */
pixels = (const void*)((const Uint8*)pixels + rect->h * pitch);
727
728
729
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture_uv, UVrect, Uslice, pixels, UVpitch) < 0) {
return -1;
}
730
731
/* Skip to the correct offset into the next texture */
732
733
734
735
pixels = (const void*)((const Uint8*)pixels + UVrect.h * UVpitch);
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture_uv, UVrect, Vslice, pixels, UVpitch) < 0) {
return -1;
}
736
}
737
738
if (texturedata.nv12) {
739
740
741
SDL_Rect UVrect = {rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2};
int UVpitch = 2 * ((pitch + 1) / 2);
742
743
/* Skip to the correct offset into the next texture */
pixels = (const void*)((const Uint8*)pixels + rect->h * pitch);
744
745
746
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture_uv, UVrect, 0, pixels, UVpitch) < 0) {
return -1;
}
747
}
748
749
750
texturedata.hasdata = YES;
751
752
return 0;
}}
753
754
755
756
757
758
759
760
761
762
763
static int
METAL_UpdateTextureYUV(SDL_Renderer * renderer, SDL_Texture * texture,
const SDL_Rect * rect,
const Uint8 *Yplane, int Ypitch,
const Uint8 *Uplane, int Upitch,
const Uint8 *Vplane, int Vpitch)
{ @autoreleasepool {
METAL_TextureData *texturedata = (__bridge METAL_TextureData *)texture->driverdata;
const int Uslice = 0;
const int Vslice = 1;
764
SDL_Rect UVrect = {rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2};
765
766
767
768
/* Bail out if we're supposed to update an empty rectangle */
if (rect->w <= 0 || rect->h <= 0) {
return 0;
769
770
}
771
772
773
774
775
776
777
778
779
780
781
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture, *rect, 0, Yplane, Ypitch) < 0) {
return -1;
}
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture_uv, UVrect, Uslice, Uplane, Upitch)) {
return -1;
}
if (METAL_UpdateTextureInternal(renderer, texturedata, texturedata.mtltexture_uv, UVrect, Vslice, Vplane, Vpitch)) {
return -1;
}
texturedata.hasdata = YES;
782
783
return 0;
784
}}
785
786
787
788
static int
METAL_LockTexture(SDL_Renderer * renderer, SDL_Texture * texture,
const SDL_Rect * rect, void **pixels, int *pitch)
789
790
791
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
METAL_TextureData *texturedata = (__bridge METAL_TextureData *)texture->driverdata;
792
int buffersize = 0;
793
794
795
if (rect->w <= 0 || rect->h <= 0) {
return SDL_SetError("Invalid rectangle dimensions for LockTexture.");
796
797
798
799
}
*pitch = SDL_BYTESPERPIXEL(texture->format) * rect->w;
800
801
802
803
804
805
if (texturedata.yuv || texturedata.nv12) {
buffersize = ((*pitch) * rect->h) + (2 * (*pitch + 1) / 2) * ((rect->h + 1) / 2);
} else {
buffersize = (*pitch) * rect->h;
}
806
texturedata.lockedrect = *rect;
807
texturedata.lockedbuffer = [data.mtldevice newBufferWithLength:buffersize options:MTLResourceStorageModeShared];
808
809
810
811
812
813
814
815
if (texturedata.lockedbuffer == nil) {
return SDL_OutOfMemory();
}
*pixels = [texturedata.lockedbuffer contents];
return 0;
}}
816
817
static void
818
METAL_UnlockTexture(SDL_Renderer * renderer, SDL_Texture * texture)
819
820
821
822
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
METAL_TextureData *texturedata = (__bridge METAL_TextureData *)texture->driverdata;
SDL_Rect rect = texturedata.lockedrect;
823
824
int pitch = SDL_BYTESPERPIXEL(texture->format) * rect.w;
SDL_Rect UVrect = {rect.x / 2, rect.y / 2, (rect.w + 1) / 2, (rect.h + 1) / 2};
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
if (texturedata.lockedbuffer == nil) {
return;
}
if (data.mtlcmdencoder != nil) {
[data.mtlcmdencoder endEncoding];
data.mtlcmdencoder = nil;
}
if (data.mtlcmdbuffer == nil) {
data.mtlcmdbuffer = [data.mtlcmdqueue commandBuffer];
}
id<MTLBlitCommandEncoder> blitcmd = [data.mtlcmdbuffer blitCommandEncoder];
[blitcmd copyFromBuffer:texturedata.lockedbuffer
sourceOffset:0
843
sourceBytesPerRow:pitch
844
845
846
847
848
849
850
sourceBytesPerImage:0
sourceSize:MTLSizeMake(rect.w, rect.h, 1)
toTexture:texturedata.mtltexture
destinationSlice:0
destinationLevel:0
destinationOrigin:MTLOriginMake(rect.x, rect.y, 0)];
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
if (texturedata.yuv) {
int Uslice = texture->format == SDL_PIXELFORMAT_YV12 ? 1 : 0;
int Vslice = texture->format == SDL_PIXELFORMAT_YV12 ? 0 : 1;
int UVpitch = (pitch + 1) / 2;
[blitcmd copyFromBuffer:texturedata.lockedbuffer
sourceOffset:rect.h * pitch
sourceBytesPerRow:UVpitch
sourceBytesPerImage:UVpitch * UVrect.h
sourceSize:MTLSizeMake(UVrect.w, UVrect.h, 1)
toTexture:texturedata.mtltexture_uv
destinationSlice:Uslice
destinationLevel:0
destinationOrigin:MTLOriginMake(UVrect.x, UVrect.y, 0)];
[blitcmd copyFromBuffer:texturedata.lockedbuffer
sourceOffset:(rect.h * pitch) + UVrect.h * UVpitch
sourceBytesPerRow:UVpitch
sourceBytesPerImage:UVpitch * UVrect.h
sourceSize:MTLSizeMake(UVrect.w, UVrect.h, 1)
toTexture:texturedata.mtltexture_uv
destinationSlice:Vslice
destinationLevel:0
destinationOrigin:MTLOriginMake(UVrect.x, UVrect.y, 0)];
}
if (texturedata.nv12) {
int UVpitch = 2 * ((pitch + 1) / 2);
[blitcmd copyFromBuffer:texturedata.lockedbuffer
sourceOffset:rect.h * pitch
sourceBytesPerRow:UVpitch
sourceBytesPerImage:0
sourceSize:MTLSizeMake(UVrect.w, UVrect.h, 1)
toTexture:texturedata.mtltexture_uv
destinationSlice:0
destinationLevel:0
destinationOrigin:MTLOriginMake(UVrect.x, UVrect.y, 0)];
}
891
892
893
894
895
896
897
898
899
900
901
902
[blitcmd endEncoding];
[data.mtlcmdbuffer commit];
data.mtlcmdbuffer = nil;
#if !__has_feature(objc_arc)
[texturedata.lockedbuffer release];
#endif
texturedata.lockedbuffer = nil;
texturedata.hasdata = YES;
}}
903
904
905
906
907
static int
METAL_SetRenderTarget(SDL_Renderer * renderer, SDL_Texture * texture)
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
908
909
910
911
912
913
if (data.mtlcmdencoder) {
/* End encoding for the previous render target so we can set up a new
* render pass for this one. */
[data.mtlcmdencoder endEncoding];
[data.mtlcmdbuffer commit];
914
915
916
917
data.mtlcmdencoder = nil;
data.mtlcmdbuffer = nil;
}
918
919
920
921
922
923
/* We don't begin a new render pass right away - we delay it until an actual
* draw or clear happens. That way we can use hardware clears when possible,
* which are only available when beginning a new render pass. */
return 0;
}}
924
925
926
927
928
929
930
// normalize a value from 0.0f to len into 0.0f to 1.0f.
static inline float
normtex(const float _val, const float len)
{
return _val / len;
931
932
}
933
934
static int
METAL_QueueSetViewport(SDL_Renderer * renderer, SDL_RenderCommand *cmd)
935
{
936
937
938
939
940
941
942
float projection[4][4]; /* Prepare an orthographic projection */
const int w = cmd->data.viewport.rect.w;
const int h = cmd->data.viewport.rect.h;
const size_t matrixlen = sizeof (projection);
float *matrix = (float *) SDL_AllocateRenderVertices(renderer, matrixlen, CONSTANT_ALIGN, &cmd->data.viewport.first);
if (!matrix) {
return -1;
943
944
}
945
946
947
948
949
950
951
SDL_memset(projection, '\0', matrixlen);
if (w && h) {
projection[0][0] = 2.0f / w;
projection[1][1] = -2.0f / h;
projection[3][0] = -1.0f;
projection[3][1] = 1.0f;
projection[3][3] = 1.0f;
952
}
953
954
955
SDL_memcpy(matrix, projection, matrixlen);
return 0;
956
957
958
}
static int
959
960
961
962
963
964
METAL_QueueSetDrawColor(SDL_Renderer *renderer, SDL_RenderCommand *cmd)
{
const size_t vertlen = sizeof (float) * 4;
float *verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, CONSTANT_ALIGN, &cmd->data.color.first);
if (!verts) {
return -1;
965
}
966
967
968
969
*(verts++) = ((float)cmd->data.color.r) / 255.0f;
*(verts++) = ((float)cmd->data.color.g) / 255.0f;
*(verts++) = ((float)cmd->data.color.b) / 255.0f;
*(verts++) = ((float)cmd->data.color.a) / 255.0f;
970
return 0;
971
}
972
973
974
static int
METAL_QueueDrawPoints(SDL_Renderer * renderer, SDL_RenderCommand *cmd, const SDL_FPoint * points, int count)
975
{
976
977
978
979
const size_t vertlen = (sizeof (float) * 2) * count;
float *verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, 0, &cmd->data.draw.first);
if (!verts) {
return -1;
980
}
981
982
983
cmd->data.draw.count = count;
SDL_memcpy(verts, points, vertlen);
return 0;
984
985
}
986
static int
987
988
989
990
991
992
METAL_QueueFillRects(SDL_Renderer * renderer, SDL_RenderCommand *cmd, const SDL_FRect * rects, int count)
{
const size_t vertlen = (sizeof (float) * 8) * count;
float *verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, 0, &cmd->data.draw.first);
if (!verts) {
return -1;
993
994
}
995
cmd->data.draw.count = count;
996
997
998
999
1000
/* Quads in the following vertex order (matches the quad index buffer):
* 1---3
* | \ |
* 0---2