summaryrefslogtreecommitdiffstats
path: root/src/video_core/host1x/ffmpeg/ffmpeg.cpp
blob: dcd07e6d2526d14a1e476f3e5644486aeb3d511a (plain) (blame)
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
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later

#include "common/assert.h"
#include "common/logging/log.h"
#include "common/scope_exit.h"
#include "common/settings.h"
#include "video_core/host1x/ffmpeg/ffmpeg.h"

extern "C" {
#ifdef LIBVA_FOUND
// for querying VAAPI driver information
#include <libavutil/hwcontext_vaapi.h>
#endif
}

namespace FFmpeg {

namespace {

constexpr AVPixelFormat PreferredGpuFormat = AV_PIX_FMT_NV12;
constexpr AVPixelFormat PreferredCpuFormat = AV_PIX_FMT_YUV420P;
constexpr std::array PreferredGpuDecoders = {
    AV_HWDEVICE_TYPE_CUDA,
#ifdef _WIN32
    AV_HWDEVICE_TYPE_D3D11VA,
    AV_HWDEVICE_TYPE_DXVA2,
#elif defined(__unix__)
    AV_HWDEVICE_TYPE_VAAPI,
    AV_HWDEVICE_TYPE_VDPAU,
#endif
    // last resort for Linux Flatpak (w/ NVIDIA)
    AV_HWDEVICE_TYPE_VULKAN,
};

AVPixelFormat GetGpuFormat(AVCodecContext* codec_context, const AVPixelFormat* pix_fmts) {
    for (const AVPixelFormat* p = pix_fmts; *p != AV_PIX_FMT_NONE; ++p) {
        if (*p == codec_context->pix_fmt) {
            return codec_context->pix_fmt;
        }
    }

    LOG_INFO(HW_GPU, "Could not find compatible GPU AV format, falling back to CPU");
    av_buffer_unref(&codec_context->hw_device_ctx);

    codec_context->pix_fmt = PreferredCpuFormat;
    return codec_context->pix_fmt;
}

std::string AVError(int errnum) {
    char errbuf[AV_ERROR_MAX_STRING_SIZE] = {};
    av_make_error_string(errbuf, sizeof(errbuf) - 1, errnum);
    return errbuf;
}

} // namespace

Packet::Packet(std::span<const u8> data) {
    m_packet = av_packet_alloc();
    m_packet->data = const_cast<u8*>(data.data());
    m_packet->size = static_cast<s32>(data.size());
}

Packet::~Packet() {
    av_packet_free(&m_packet);
}

Frame::Frame() {
    m_frame = av_frame_alloc();
}

Frame::~Frame() {
    av_frame_free(&m_frame);
}

Decoder::Decoder(Tegra::Host1x::NvdecCommon::VideoCodec codec) {
    const AVCodecID av_codec = [&] {
        switch (codec) {
        case Tegra::Host1x::NvdecCommon::VideoCodec::H264:
            return AV_CODEC_ID_H264;
        case Tegra::Host1x::NvdecCommon::VideoCodec::VP8:
            return AV_CODEC_ID_VP8;
        case Tegra::Host1x::NvdecCommon::VideoCodec::VP9:
            return AV_CODEC_ID_VP9;
        default:
            UNIMPLEMENTED_MSG("Unknown codec {}", codec);
            return AV_CODEC_ID_NONE;
        }
    }();

    m_codec = avcodec_find_decoder(av_codec);
}

bool Decoder::SupportsDecodingOnDevice(AVPixelFormat* out_pix_fmt, AVHWDeviceType type) const {
    for (int i = 0;; i++) {
        const AVCodecHWConfig* config = avcodec_get_hw_config(m_codec, i);
        if (!config) {
            LOG_DEBUG(HW_GPU, "{} decoder does not support device type {}", m_codec->name,
                      av_hwdevice_get_type_name(type));
            break;
        }
        if ((config->methods & AV_CODEC_HW_CONFIG_METHOD_HW_DEVICE_CTX) != 0 &&
            config->device_type == type) {
            LOG_INFO(HW_GPU, "Using {} GPU decoder", av_hwdevice_get_type_name(type));
            *out_pix_fmt = config->pix_fmt;
            return true;
        }
    }

    return false;
}

std::vector<AVHWDeviceType> HardwareContext::GetSupportedDeviceTypes() {
    std::vector<AVHWDeviceType> types;
    AVHWDeviceType current_device_type = AV_HWDEVICE_TYPE_NONE;

    while (true) {
        current_device_type = av_hwdevice_iterate_types(current_device_type);
        if (current_device_type == AV_HWDEVICE_TYPE_NONE) {
            return types;
        }

        types.push_back(current_device_type);
    }
}

HardwareContext::~HardwareContext() {
    av_buffer_unref(&m_gpu_decoder);
}

bool HardwareContext::InitializeForDecoder(DecoderContext& decoder_context,
                                           const Decoder& decoder) {
    const auto supported_types = GetSupportedDeviceTypes();
    for (const auto type : PreferredGpuDecoders) {
        AVPixelFormat hw_pix_fmt;

        if (std::ranges::find(supported_types, type) == supported_types.end()) {
            LOG_DEBUG(HW_GPU, "{} explicitly unsupported", av_hwdevice_get_type_name(type));
            continue;
        }

        if (!this->InitializeWithType(type)) {
            continue;
        }

        if (decoder.SupportsDecodingOnDevice(&hw_pix_fmt, type)) {
            decoder_context.InitializeHardwareDecoder(*this, hw_pix_fmt);
            return true;
        }
    }

    return false;
}

bool HardwareContext::InitializeWithType(AVHWDeviceType type) {
    av_buffer_unref(&m_gpu_decoder);

    if (const int ret = av_hwdevice_ctx_create(&m_gpu_decoder, type, nullptr, nullptr, 0);
        ret < 0) {
        LOG_DEBUG(HW_GPU, "av_hwdevice_ctx_create({}) failed: {}", av_hwdevice_get_type_name(type),
                  AVError(ret));
        return false;
    }

#ifdef LIBVA_FOUND
    if (type == AV_HWDEVICE_TYPE_VAAPI) {
        // We need to determine if this is an impersonated VAAPI driver.
        auto* hwctx = reinterpret_cast<AVHWDeviceContext*>(m_gpu_decoder->data);
        auto* vactx = static_cast<AVVAAPIDeviceContext*>(hwctx->hwctx);
        const char* vendor_name = vaQueryVendorString(vactx->display);
        if (strstr(vendor_name, "VDPAU backend")) {
            // VDPAU impersonated VAAPI impls are super buggy, we need to skip them.
            LOG_DEBUG(HW_GPU, "Skipping VDPAU impersonated VAAPI driver");
            return false;
        } else {
            // According to some user testing, certain VAAPI drivers (Intel?) could be buggy.
            // Log the driver name just in case.
            LOG_DEBUG(HW_GPU, "Using VAAPI driver: {}", vendor_name);
        }
    }
#endif

    return true;
}

DecoderContext::DecoderContext(const Decoder& decoder) {
    m_codec_context = avcodec_alloc_context3(decoder.GetCodec());
    av_opt_set(m_codec_context->priv_data, "tune", "zerolatency", 0);
    m_codec_context->thread_count = 0;
    m_codec_context->thread_type &= ~FF_THREAD_FRAME;
}

DecoderContext::~DecoderContext() {
    av_buffer_unref(&m_codec_context->hw_device_ctx);
    avcodec_free_context(&m_codec_context);
}

void DecoderContext::InitializeHardwareDecoder(const HardwareContext& context,
                                               AVPixelFormat hw_pix_fmt) {
    m_codec_context->hw_device_ctx = av_buffer_ref(context.GetBufferRef());
    m_codec_context->get_format = GetGpuFormat;
    m_codec_context->pix_fmt = hw_pix_fmt;
}

bool DecoderContext::OpenContext(const Decoder& decoder) {
    if (const int ret = avcodec_open2(m_codec_context, decoder.GetCodec(), nullptr); ret < 0) {
        LOG_ERROR(HW_GPU, "avcodec_open2 error: {}", AVError(ret));
        return false;
    }

    if (!m_codec_context->hw_device_ctx) {
        LOG_INFO(HW_GPU, "Using FFmpeg software decoding");
    }

    return true;
}

bool DecoderContext::SendPacket(const Packet& packet) {
    if (const int ret = avcodec_send_packet(m_codec_context, packet.GetPacket()); ret < 0) {
        LOG_ERROR(HW_GPU, "avcodec_send_packet error: {}", AVError(ret));
        return false;
    }

    return true;
}

std::unique_ptr<Frame> DecoderContext::ReceiveFrame(bool* out_is_interlaced) {
    auto dst_frame = std::make_unique<Frame>();

    const auto ReceiveImpl = [&](AVFrame* frame) {
        if (const int ret = avcodec_receive_frame(m_codec_context, frame); ret < 0) {
            LOG_ERROR(HW_GPU, "avcodec_receive_frame error: {}", AVError(ret));
            return false;
        }

        *out_is_interlaced = frame->interlaced_frame != 0;
        return true;
    };

    if (m_codec_context->hw_device_ctx) {
        // If we have a hardware context, make a separate frame here to receive the
        // hardware result before sending it to the output.
        Frame intermediate_frame;

        if (!ReceiveImpl(intermediate_frame.GetFrame())) {
            return {};
        }

        dst_frame->SetFormat(PreferredGpuFormat);
        if (const int ret =
                av_hwframe_transfer_data(dst_frame->GetFrame(), intermediate_frame.GetFrame(), 0);
            ret < 0) {
            LOG_ERROR(HW_GPU, "av_hwframe_transfer_data error: {}", AVError(ret));
            return {};
        }
    } else {
        // Otherwise, decode the frame as normal.
        if (!ReceiveImpl(dst_frame->GetFrame())) {
            return {};
        }
    }

    return dst_frame;
}

DeinterlaceFilter::DeinterlaceFilter(const Frame& frame) {
    const AVFilter* buffer_src = avfilter_get_by_name("buffer");
    const AVFilter* buffer_sink = avfilter_get_by_name("buffersink");
    AVFilterInOut* inputs = avfilter_inout_alloc();
    AVFilterInOut* outputs = avfilter_inout_alloc();
    SCOPE_EXIT({
        avfilter_inout_free(&inputs);
        avfilter_inout_free(&outputs);
    });

    // Don't know how to get the accurate time_base but it doesn't matter for yadif filter
    // so just use 1/1 to make buffer filter happy
    std::string args = fmt::format("video_size={}x{}:pix_fmt={}:time_base=1/1", frame.GetWidth(),
                                   frame.GetHeight(), static_cast<int>(frame.GetPixelFormat()));

    m_filter_graph = avfilter_graph_alloc();
    int ret = avfilter_graph_create_filter(&m_source_context, buffer_src, "in", args.c_str(),
                                           nullptr, m_filter_graph);
    if (ret < 0) {
        LOG_ERROR(HW_GPU, "avfilter_graph_create_filter source error: {}", AVError(ret));
        return;
    }

    ret = avfilter_graph_create_filter(&m_sink_context, buffer_sink, "out", nullptr, nullptr,
                                       m_filter_graph);
    if (ret < 0) {
        LOG_ERROR(HW_GPU, "avfilter_graph_create_filter sink error: {}", AVError(ret));
        return;
    }

    inputs->name = av_strdup("out");
    inputs->filter_ctx = m_sink_context;
    inputs->pad_idx = 0;
    inputs->next = nullptr;

    outputs->name = av_strdup("in");
    outputs->filter_ctx = m_source_context;
    outputs->pad_idx = 0;
    outputs->next = nullptr;

    const char* description = "yadif=1:-1:0";
    ret = avfilter_graph_parse_ptr(m_filter_graph, description, &inputs, &outputs, nullptr);
    if (ret < 0) {
        LOG_ERROR(HW_GPU, "avfilter_graph_parse_ptr error: {}", AVError(ret));
        return;
    }

    ret = avfilter_graph_config(m_filter_graph, nullptr);
    if (ret < 0) {
        LOG_ERROR(HW_GPU, "avfilter_graph_config error: {}", AVError(ret));
        return;
    }

    m_initialized = true;
}

bool DeinterlaceFilter::AddSourceFrame(const Frame& frame) {
    if (const int ret = av_buffersrc_add_frame_flags(m_source_context, frame.GetFrame(),
                                                     AV_BUFFERSRC_FLAG_KEEP_REF);
        ret < 0) {
        LOG_ERROR(HW_GPU, "av_buffersrc_add_frame_flags error: {}", AVError(ret));
        return false;
    }

    return true;
}

std::unique_ptr<Frame> DeinterlaceFilter::DrainSinkFrame() {
    auto dst_frame = std::make_unique<Frame>();
    const int ret = av_buffersink_get_frame(m_sink_context, dst_frame->GetFrame());

    if (ret == AVERROR(EAGAIN) || ret == AVERROR(AVERROR_EOF)) {
        return {};
    }

    if (ret < 0) {
        LOG_ERROR(HW_GPU, "av_buffersink_get_frame error: {}", AVError(ret));
        return {};
    }

    return dst_frame;
}

DeinterlaceFilter::~DeinterlaceFilter() {
    avfilter_graph_free(&m_filter_graph);
}

void DecodeApi::Reset() {
    m_deinterlace_filter.reset();
    m_hardware_context.reset();
    m_decoder_context.reset();
    m_decoder.reset();
}

bool DecodeApi::Initialize(Tegra::Host1x::NvdecCommon::VideoCodec codec) {
    this->Reset();
    m_decoder.emplace(codec);
    m_decoder_context.emplace(*m_decoder);

    // Enable GPU decoding if requested.
    if (Settings::values.nvdec_emulation.GetValue() == Settings::NvdecEmulation::Gpu) {
        m_hardware_context.emplace();
        m_hardware_context->InitializeForDecoder(*m_decoder_context, *m_decoder);
    }

    // Open the decoder context.
    if (!m_decoder_context->OpenContext(*m_decoder)) {
        this->Reset();
        return false;
    }

    return true;
}

bool DecodeApi::SendPacket(std::span<const u8> packet_data, size_t configuration_size) {
    FFmpeg::Packet packet(packet_data);
    return m_decoder_context->SendPacket(packet);
}

void DecodeApi::ReceiveFrames(std::queue<std::unique_ptr<Frame>>& frame_queue) {
    // Receive raw frame from decoder.
    bool is_interlaced;
    auto frame = m_decoder_context->ReceiveFrame(&is_interlaced);
    if (!frame) {
        return;
    }

    if (!is_interlaced) {
        // If the frame is not interlaced, we can pend it now.
        frame_queue.push(std::move(frame));
    } else {
        // Create the deinterlacer if needed.
        if (!m_deinterlace_filter) {
            m_deinterlace_filter.emplace(*frame);
        }

        // Add the frame we just received.
        if (!m_deinterlace_filter->AddSourceFrame(*frame)) {
            return;
        }

        // Pend output fields.
        while (true) {
            auto filter_frame = m_deinterlace_filter->DrainSinkFrame();
            if (!filter_frame) {
                break;
            }

            frame_queue.push(std::move(filter_frame));
        }
    }
}

} // namespace FFmpeg