/
SDL_render_metal.m
1603 lines (1371 loc) · 59.3 KB
1
2
/*
Simple DirectMedia Layer
3
Copyright (C) 1997-2018 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
158
159
@end
@implementation METAL_TextureData
160
#if !__has_feature(objc_arc)
161
162
163
- (void)dealloc
{
[_mtltexture release];
164
[_mtltexture_uv release];
165
166
167
168
[_mtlsampler release];
[super dealloc];
}
#endif
169
170
@end
171
172
173
static int
IsMetalAvailable(const SDL_SysWMinfo *syswm)
{
174
175
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.");
176
177
178
}
// this checks a weak symbol.
179
#if (defined(__MACOSX__) && (MAC_OS_X_VERSION_MIN_REQUIRED < 101100))
180
181
182
183
184
185
186
187
if (MTLCreateSystemDefaultDevice == NULL) { // probably on 10.10 or lower.
return SDL_SetError("Metal framework not available on this system");
}
#endif
return 0;
}
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
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";
237
case SDL_METAL_FRAGMENT_COPY: return @"SDL_Copy_fragment";
238
239
240
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";
241
242
243
244
default: return nil;
}
}
245
static id<MTLRenderPipelineState>
246
247
MakePipelineState(METAL_RenderData *data, METAL_PipelineCache *cache,
NSString *blendlabel, SDL_BlendMode blendmode)
248
{
249
250
id<MTLFunction> mtlvertfn = [data.mtllibrary newFunctionWithName:GetVertexFunctionName(cache->vertexFunction)];
id<MTLFunction> mtlfragfn = [data.mtllibrary newFunctionWithName:GetFragmentFunctionName(cache->fragmentFunction)];
251
252
253
254
255
256
SDL_assert(mtlvertfn != nil);
SDL_assert(mtlfragfn != nil);
MTLRenderPipelineDescriptor *mtlpipedesc = [[MTLRenderPipelineDescriptor alloc] init];
mtlpipedesc.vertexFunction = mtlvertfn;
mtlpipedesc.fragmentFunction = mtlfragfn;
257
258
259
MTLRenderPipelineColorAttachmentDescriptor *rtdesc = mtlpipedesc.colorAttachments[0];
260
rtdesc.pixelFormat = cache->renderTargetFormat;
261
262
263
264
265
266
267
268
269
270
271
272
273
274
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];
275
276
NSError *err = nil;
277
id<MTLRenderPipelineState> state = [data.mtldevice newRenderPipelineStateWithDescriptor:mtlpipedesc error:&err];
278
SDL_assert(err == nil);
279
280
281
282
283
284
285
METAL_PipelineState pipeline;
pipeline.blendMode = blendmode;
pipeline.pipe = (void *)CFBridgingRetain(state);
METAL_PipelineState *states = SDL_realloc(cache->states, (cache->count + 1) * sizeof(pipeline));
286
#if !__has_feature(objc_arc)
287
288
289
[mtlpipedesc release]; // !!! FIXME: can these be reused for each creation, or does the pipeline obtain it?
[mtlvertfn release];
[mtlfragfn release];
290
[state release];
291
#endif
292
293
294
295
296
297
298
299
300
301
302
303
if (states) {
states[cache->count++] = pipeline;
cache->states = states;
return (__bridge id<MTLRenderPipelineState>)pipeline.pipe;
} else {
CFBridgingRelease(pipeline.pipe);
SDL_OutOfMemory();
return NULL;
}
}
304
static void
305
306
MakePipelineCache(METAL_RenderData *data, METAL_PipelineCache *cache, const char *label,
MTLPixelFormat rtformat, SDL_MetalVertexFunction vertfn, SDL_MetalFragmentFunction fragfn)
307
308
309
310
311
{
SDL_zerop(cache);
cache->vertexFunction = vertfn;
cache->fragmentFunction = fragfn;
312
cache->renderTargetFormat = rtformat;
313
314
315
316
cache->label = label;
/* Create pipeline states for the default blend modes. Custom blend modes
* will be added to the cache on-demand. */
317
318
319
320
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);
321
322
323
}
static void
324
DestroyPipelineCache(METAL_PipelineCache *cache)
325
{
326
327
328
329
330
331
if (cache != NULL) {
for (int i = 0; i < cache->count; i++) {
CFBridgingRelease(cache->states[i].pipe);
}
SDL_free(cache->states);
332
333
334
}
}
335
336
337
338
339
340
341
342
343
344
345
346
347
348
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);
}
349
static METAL_ShaderPipelines *
350
ChooseShaderPipelines(METAL_RenderData *data, MTLPixelFormat rtformat)
351
{
352
353
354
355
356
357
358
359
360
361
362
363
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) {
364
365
366
367
SDL_OutOfMemory();
return NULL;
}
368
369
370
371
MakeShaderPipelines(data, &allpipelines[count], rtformat);
data.allpipelines = allpipelines;
data.pipelinescount = count + 1;
372
373
return &data.allpipelines[count];
374
375
376
}
static void
377
DestroyAllPipelines(METAL_ShaderPipelines *allpipelines, int count)
378
{
379
380
381
382
383
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]);
}
384
385
}
386
SDL_free(allpipelines);
387
}
388
389
390
}
static inline id<MTLRenderPipelineState>
391
ChoosePipelineState(METAL_RenderData *data, METAL_ShaderPipelines *pipelines, SDL_MetalFragmentFunction fragfn, SDL_BlendMode blendmode)
392
{
393
394
METAL_PipelineCache *cache = &pipelines->caches[fragfn];
395
396
397
398
for (int i = 0; i < cache->count; i++) {
if (cache->states[i].blendMode == blendmode) {
return (__bridge id<MTLRenderPipelineState>)cache->states[i].pipe;
}
399
}
400
401
return MakePipelineState(data, cache, [NSString stringWithFormat:@" (blend=custom 0x%x)", blendmode], blendmode);
402
403
}
404
405
406
407
static void
METAL_ActivateRenderCommandEncoder(SDL_Renderer * renderer, MTLLoadAction load, MTLClearColor *clear_color)
{
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
408
409
410
411
412
/* 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;
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
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;
}
428
429
SDL_assert(mtltexture);
430
431
432
433
434
if (load == MTLLoadActionClear) {
SDL_assert(clear_color != NULL);
data.mtlpassdesc.colorAttachments[0].clearColor = *clear_color;
}
435
436
437
data.mtlpassdesc.colorAttachments[0].loadAction = load;
data.mtlpassdesc.colorAttachments[0].texture = mtltexture;
438
439
440
data.mtlcmdbuffer = [data.mtlcmdqueue commandBuffer];
data.mtlcmdencoder = [data.mtlcmdbuffer renderCommandEncoderWithDescriptor:data.mtlpassdesc];
441
442
443
444
445
446
if (data.mtlbackbuffer != nil && mtltexture == data.mtlbackbuffer.texture) {
data.mtlcmdencoder.label = @"SDL metal renderer backbuffer";
} else {
data.mtlcmdencoder.label = @"SDL metal renderer render target";
}
447
448
data.activepipelines = ChooseShaderPipelines(data, mtltexture.pixelFormat);
449
450
451
452
453
454
455
// 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];
}
}
456
457
458
459
460
461
462
463
464
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
}
}
465
466
467
468
469
470
471
472
473
474
475
476
477
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;
}}
478
479
480
481
482
483
484
485
486
487
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);
488
489
490
491
492
493
494
495
496
497
498
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;
}
499
500
501
502
503
504
static int
METAL_CreateTexture(SDL_Renderer * renderer, SDL_Texture * texture)
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
MTLPixelFormat pixfmt;
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
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));
}
522
523
524
MTLTextureDescriptor *mtltexdesc = [MTLTextureDescriptor texture2DDescriptorWithPixelFormat:pixfmt
width:(NSUInteger)texture->w height:(NSUInteger)texture->h mipmapped:NO];
525
526
527
528
529
530
531
532
533
534
535
536
537
538
/* 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");
}
539
540
id<MTLTexture> mtltexture_uv = nil;
541
542
543
BOOL yuv = (texture->format == SDL_PIXELFORMAT_IYUV) || (texture->format == SDL_PIXELFORMAT_YV12);
BOOL nv12 = (texture->format == SDL_PIXELFORMAT_NV12) || (texture->format == SDL_PIXELFORMAT_NV21);
544
545
546
547
548
549
550
551
552
553
554
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;
555
556
}
557
if (yuv || nv12) {
558
mtltexture_uv = [data.mtldevice newTextureWithDescriptor:mtltexdesc];
559
560
561
562
563
564
if (mtltexture_uv == nil) {
#if !__has_feature(objc_arc)
[mtltexture release];
#endif
return SDL_SetError("Texture allocation failed");
}
565
}
566
567
568
569
570
571
572
573
574
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;
575
576
577
texturedata.yuv = yuv;
texturedata.nv12 = nv12;
578
579
580
581
582
583
584
585
586
587
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;
}
588
589
590
591
592
593
594
595
596
597
598
599
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;
}
600
601
texture->driverdata = (void*)CFBridgingRetain(texturedata);
602
603
604
605
606
607
#if !__has_feature(objc_arc)
[texturedata release];
[mtltexture release];
[mtltexture_uv release];
#endif
608
609
610
return 0;
}}
611
612
613
614
615
616
static int
METAL_UpdateTexture(SDL_Renderer * renderer, SDL_Texture * texture,
const SDL_Rect * rect, const void *pixels, int pitch)
{ @autoreleasepool {
METAL_TextureData *texturedata = (__bridge METAL_TextureData *)texture->driverdata;
617
618
619
620
621
622
623
624
/* !!! FIXME: replaceRegion does not do any synchronization, so it might
* !!! FIXME: stomp on a previous frame's data that's currently being read
* !!! FIXME: by the GPU. */
[texturedata.mtltexture replaceRegion:MTLRegionMake2D(rect->x, rect->y, rect->w, rect->h)
mipmapLevel:0
withBytes:pixels
bytesPerRow:pitch];
625
626
627
628
if (texturedata.yuv) {
int Uslice = texture->format == SDL_PIXELFORMAT_YV12 ? 1 : 0;
int Vslice = texture->format == SDL_PIXELFORMAT_YV12 ? 0 : 1;
629
630
631
632
633
634
635
636
637
/* Skip to the correct offset into the next texture */
pixels = (const void*)((const Uint8*)pixels + rect->h * pitch);
[texturedata.mtltexture_uv replaceRegion:MTLRegionMake2D(rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2)
mipmapLevel:0
slice:Uslice
withBytes:pixels
bytesPerRow:(pitch + 1) / 2
bytesPerImage:0];
638
639
640
641
642
643
644
645
646
647
/* Skip to the correct offset into the next texture */
pixels = (const void*)((const Uint8*)pixels + ((rect->h + 1) / 2) * ((pitch + 1)/2));
[texturedata.mtltexture_uv replaceRegion:MTLRegionMake2D(rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2)
mipmapLevel:0
slice:Vslice
withBytes:pixels
bytesPerRow:(pitch + 1) / 2
bytesPerImage:0];
}
648
649
650
651
652
653
654
655
656
657
658
if (texturedata.nv12) {
/* Skip to the correct offset into the next texture */
pixels = (const void*)((const Uint8*)pixels + rect->h * pitch);
[texturedata.mtltexture_uv replaceRegion:MTLRegionMake2D(rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2)
mipmapLevel:0
slice:0
withBytes:pixels
bytesPerRow:2 * ((pitch + 1) / 2)
bytesPerImage:0];
}
659
660
661
return 0;
}}
662
663
664
665
666
667
668
669
670
671
672
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;
673
674
675
676
/* Bail out if we're supposed to update an empty rectangle */
if (rect->w <= 0 || rect->h <= 0) {
return 0;
677
678
}
679
680
681
682
[texturedata.mtltexture replaceRegion:MTLRegionMake2D(rect->x, rect->y, rect->w, rect->h)
mipmapLevel:0
withBytes:Yplane
bytesPerRow:Ypitch];
683
684
685
686
687
688
689
[texturedata.mtltexture_uv replaceRegion:MTLRegionMake2D(rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2)
mipmapLevel:0
slice:Uslice
withBytes:Uplane
bytesPerRow:Upitch
bytesPerImage:0];
690
691
692
693
694
695
696
697
698
[texturedata.mtltexture_uv replaceRegion:MTLRegionMake2D(rect->x / 2, rect->y / 2, (rect->w + 1) / 2, (rect->h + 1) / 2)
mipmapLevel:0
slice:Vslice
withBytes:Vplane
bytesPerRow:Vpitch
bytesPerImage:0];
return 0;
699
}}
700
701
702
703
704
705
706
707
static int
METAL_LockTexture(SDL_Renderer * renderer, SDL_Texture * texture,
const SDL_Rect * rect, void **pixels, int *pitch)
{
return SDL_Unsupported(); // !!! FIXME: write me
}
708
static void
709
METAL_UnlockTexture(SDL_Renderer * renderer, SDL_Texture * texture)
710
{
711
712
// !!! FIXME: write me
}
713
714
715
716
717
static int
METAL_SetRenderTarget(SDL_Renderer * renderer, SDL_Texture * texture)
{ @autoreleasepool {
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
718
719
720
721
722
723
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];
724
725
726
727
data.mtlcmdencoder = nil;
data.mtlcmdbuffer = nil;
}
728
729
730
731
732
733
/* 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;
}}
734
735
736
737
738
739
740
// 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;
741
742
}
743
744
static int
METAL_QueueSetViewport(SDL_Renderer * renderer, SDL_RenderCommand *cmd)
745
{
746
747
748
749
750
751
752
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;
753
754
}
755
756
757
758
759
760
761
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;
762
}
763
764
765
SDL_memcpy(matrix, projection, matrixlen);
return 0;
766
767
768
}
static int
769
770
771
772
773
774
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;
775
}
776
777
778
779
*(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;
780
return 0;
781
}
782
783
784
static int
METAL_QueueDrawPoints(SDL_Renderer * renderer, SDL_RenderCommand *cmd, const SDL_FPoint * points, int count)
785
{
786
787
788
789
const size_t vertlen = (sizeof (float) * 2) * count;
float *verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, 0, &cmd->data.draw.first);
if (!verts) {
return -1;
790
}
791
792
793
cmd->data.draw.count = count;
SDL_memcpy(verts, points, vertlen);
return 0;
794
795
}
796
static int
797
798
799
800
801
802
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;
803
804
}
805
cmd->data.draw.count = count;
806
807
808
809
810
811
/* Quads in the following vertex order (matches the quad index buffer):
* 1---3
* | \ |
* 0---2
*/
812
813
814
for (int i = 0; i < count; i++, rects++) {
if ((rects->w <= 0.0f) || (rects->h <= 0.0f)) {
cmd->data.draw.count--;
815
} else {
816
817
818
819
820
821
822
823
*(verts++) = rects->x;
*(verts++) = rects->y + rects->h;
*(verts++) = rects->x;
*(verts++) = rects->y;
*(verts++) = rects->x + rects->w;
*(verts++) = rects->y + rects->h;
*(verts++) = rects->x + rects->w;
*(verts++) = rects->y;
824
}
825
}
826
827
828
if (cmd->data.draw.count == 0) {
cmd->command = SDL_RENDERCMD_NO_OP; // nothing to do, just skip this one later.
829
}
830
831
832
return 0;
}
833
834
835
836
837
static int
METAL_QueueCopy(SDL_Renderer * renderer, SDL_RenderCommand *cmd, SDL_Texture * texture,
const SDL_Rect * srcrect, const SDL_FRect * dstrect)
{
838
839
const float texw = (float) texture->w;
const float texh = (float) texture->h;
840
841
842
843
844
// !!! FIXME: use an index buffer
const size_t vertlen = (sizeof (float) * 16);
float *verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, 0, &cmd->data.draw.first);
if (!verts) {
return -1;
845
846
}
847
cmd->data.draw.count = 1;
848
849
850
851
852
853
854
855
856
*(verts++) = dstrect->x;
*(verts++) = dstrect->y + dstrect->h;
*(verts++) = dstrect->x;
*(verts++) = dstrect->y;
*(verts++) = dstrect->x + dstrect->w;
*(verts++) = dstrect->y + dstrect->h;
*(verts++) = dstrect->x + dstrect->w;
*(verts++) = dstrect->y;
857
858
859
860
861
862
863
864
865
*(verts++) = normtex(srcrect->x, texw);
*(verts++) = normtex(srcrect->y + srcrect->h, texh);
*(verts++) = normtex(srcrect->x, texw);
*(verts++) = normtex(srcrect->y, texh);
*(verts++) = normtex(srcrect->x + srcrect->w, texw);
*(verts++) = normtex(srcrect->y + srcrect->h, texh);
*(verts++) = normtex(srcrect->x + srcrect->w, texw);
*(verts++) = normtex(srcrect->y, texh);
866
867
return 0;
868
}
869
870
static int
871
872
873
874
METAL_QueueCopyEx(SDL_Renderer * renderer, SDL_RenderCommand *cmd, SDL_Texture * texture,
const SDL_Rect * srcquad, const SDL_FRect * dstrect,
const double angle, const SDL_FPoint *center, const SDL_RendererFlip flip)
{
875
876
const float texw = (float) texture->w;
const float texh = (float) texture->h;
877
878
879
880
const float rads = (float)(M_PI * (float) angle / 180.0f);
const float c = cosf(rads), s = sinf(rads);
float minu, maxu, minv, maxv;
const size_t vertlen = (sizeof (float) * 32);
881
882
883
884
float *verts;
// cheat and store this offset in (count) because it needs to be aligned in ways other fields don't and we aren't using count otherwise.
verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, CONSTANT_ALIGN, &cmd->data.draw.count);
885
886
887
if (!verts) {
return -1;
}
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
// transform matrix
SDL_memset(verts, '\0', sizeof (*verts) * 16);
verts[10] = verts[15] = 1.0f;
// rotation
verts[0] = c;
verts[1] = s;
verts[4] = -s;
verts[5] = c;
// translation
verts[12] = dstrect->x + center->x;
verts[13] = dstrect->y + center->y;
// rest of the vertices don't need the aggressive alignment. Pack them in.
verts = (float *) SDL_AllocateRenderVertices(renderer, vertlen, 0, &cmd->data.draw.first);
if (!verts) {
return -1;
}
907
908
909
910
911
minu = normtex(srcquad->x, texw);
maxu = normtex(srcquad->x + srcquad->w, texw);
minv = normtex(srcquad->y, texh);
maxv = normtex(srcquad->y + srcquad->h, texh);
912
913
914
915
916
917
918
919
920
921
922
if (flip & SDL_FLIP_HORIZONTAL) {
float tmp = maxu;
maxu = minu;
minu = tmp;
}
if (flip & SDL_FLIP_VERTICAL) {
float tmp = maxv;
maxv = minv;
minv = tmp;
}
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
// vertices
*(verts++) = -center->x;
*(verts++) = dstrect->h - center->y;
*(verts++) = -center->x;
*(verts++) = -center->y;
*(verts++) = dstrect->w - center->x;
*(verts++) = dstrect->h - center->y;
*(verts++) = dstrect->w - center->x;
*(verts++) = -center->y;
// texcoords
*(verts++) = minu;
*(verts++) = maxv;
*(verts++) = minu;
*(verts++) = minv;
*(verts++) = maxu;
*(verts++) = maxv;
*(verts++) = maxu;
*(verts++) = minv;
944
945
946
return 0;
}
947
948
typedef struct
949
{
950
951
952
#if __has_feature(objc_arc)
__unsafe_unretained id<MTLRenderPipelineState> pipeline;
#else
953
id<MTLRenderPipelineState> pipeline;
954
#endif
955
956
957
958
size_t constants_offset;
SDL_Texture *texture;
SDL_bool cliprect_dirty;
SDL_bool cliprect_enabled;
959
SDL_Rect cliprect;
960
961
SDL_bool viewport_dirty;
SDL_Rect viewport;
962
963
964
size_t projection_offset;
SDL_bool color_dirty;
size_t color_offset;
965
} METAL_DrawStateCache;
966
967
968
969
970
static void
SetDrawState(SDL_Renderer *renderer, const SDL_RenderCommand *cmd, const SDL_MetalFragmentFunction shader,
const size_t constants_offset, id<MTLBuffer> mtlbufvertex, METAL_DrawStateCache *statecache)
{
971
METAL_RenderData *data = (__bridge METAL_RenderData *) renderer->driverdata;
972
const SDL_BlendMode blend = cmd->data.draw.blend;
973
size_t first = cmd->data.draw.first;
974
id<MTLRenderPipelineState> newpipeline;
975
976
METAL_ActivateRenderCommandEncoder(renderer, MTLLoadActionLoad, NULL);
977
978
979
980
981
982
983
984
985
986
if (statecache->viewport_dirty) {
MTLViewport viewport;
viewport.originX = statecache->viewport.x;
viewport.originY = statecache->viewport.y;
viewport.width = statecache->viewport.w;
viewport.height = statecache->viewport.h;
viewport.znear = 0.0;
viewport.zfar = 1.0;
[data.mtlcmdencoder setViewport:viewport];
987
[data.mtlcmdencoder setVertexBuffer:mtlbufvertex offset:statecache->projection_offset atIndex:2]; // projection
988
statecache->viewport_dirty = SDL_FALSE;
989
990
}
991
992
993
994
995
996
997
998
999
1000
if (statecache->cliprect_dirty) {
MTLScissorRect mtlrect;
if (statecache->cliprect_enabled) {
const SDL_Rect *rect = &statecache->cliprect;
mtlrect.x = statecache->viewport.x + rect->x;
mtlrect.y = statecache->viewport.y + rect->y;
mtlrect.width = rect->w;
mtlrect.height = rect->h;
} else {
mtlrect.x = statecache->viewport.x;