static int map_avcodec_id(enum AVCodecID id)
{
switch (id) {
- case AV_CODEC_ID_H264: return cudaVideoCodec_H264;
- case AV_CODEC_ID_HEVC: return cudaVideoCodec_HEVC;
+ case AV_CODEC_ID_H264: return cudaVideoCodec_H264;
+ case AV_CODEC_ID_HEVC: return cudaVideoCodec_HEVC;
+ case AV_CODEC_ID_MPEG1VIDEO: return cudaVideoCodec_MPEG1;
+ case AV_CODEC_ID_MPEG2VIDEO: return cudaVideoCodec_MPEG2;
+ case AV_CODEC_ID_MPEG4: return cudaVideoCodec_MPEG4;
+ case AV_CODEC_ID_VC1: return cudaVideoCodec_VC1;
+ case AV_CODEC_ID_VP8: return cudaVideoCodec_VP8;
+ case AV_CODEC_ID_VP9: return cudaVideoCodec_VP9;
+ case AV_CODEC_ID_WMV3: return cudaVideoCodec_VC1;
}
return -1;
}
return -1;
}
+static int nvdec_test_capabilities(NVDECDecoder *decoder,
+ CUVIDDECODECREATEINFO *params, void *logctx)
+{
+ CUresult err;
+ CUVIDDECODECAPS caps = { 0 };
+
+ caps.eCodecType = params->CodecType;
+ caps.eChromaFormat = params->ChromaFormat;
+ caps.nBitDepthMinus8 = params->bitDepthMinus8;
+
+ if (!decoder->cvdl->cuvidGetDecoderCaps) {
+ av_log(logctx, AV_LOG_WARNING, "Used Nvidia driver is too old to perform a capability check.\n");
+ av_log(logctx, AV_LOG_WARNING, "The minimum required version is "
+#if defined(_WIN32) || defined(__CYGWIN__)
+ "378.66"
+#else
+ "378.13"
+#endif
+ ". Continuing blind.\n");
+ return 0;
+ }
+
+ err = decoder->cvdl->cuvidGetDecoderCaps(&caps);
+ if (err != CUDA_SUCCESS) {
+ av_log(logctx, AV_LOG_ERROR, "Failed querying decoder capabilities\n");
+ return AVERROR_UNKNOWN;
+ }
+
+ av_log(logctx, AV_LOG_VERBOSE, "NVDEC capabilities:\n");
+ av_log(logctx, AV_LOG_VERBOSE, "format supported: %s, max_mb_count: %d\n",
+ caps.bIsSupported ? "yes" : "no", caps.nMaxMBCount);
+ av_log(logctx, AV_LOG_VERBOSE, "min_width: %d, max_width: %d\n",
+ caps.nMinWidth, caps.nMaxWidth);
+ av_log(logctx, AV_LOG_VERBOSE, "min_height: %d, max_height: %d\n",
+ caps.nMinHeight, caps.nMaxHeight);
+
+ if (!caps.bIsSupported) {
+ av_log(logctx, AV_LOG_ERROR, "Hardware is lacking required capabilities\n");
+ return AVERROR(EINVAL);
+ }
+
+ if (params->ulWidth > caps.nMaxWidth || params->ulWidth < caps.nMinWidth) {
+ av_log(logctx, AV_LOG_ERROR, "Video width %d not within range from %d to %d\n",
+ (int)params->ulWidth, caps.nMinWidth, caps.nMaxWidth);
+ return AVERROR(EINVAL);
+ }
+
+ if (params->ulHeight > caps.nMaxHeight || params->ulHeight < caps.nMinHeight) {
+ av_log(logctx, AV_LOG_ERROR, "Video height %d not within range from %d to %d\n",
+ (int)params->ulHeight, caps.nMinHeight, caps.nMaxHeight);
+ return AVERROR(EINVAL);
+ }
+
+ if ((params->ulWidth * params->ulHeight) / 256 > caps.nMaxMBCount) {
+ av_log(logctx, AV_LOG_ERROR, "Video macroblock count %d exceeds maximum of %d\n",
+ (int)(params->ulWidth * params->ulHeight) / 256, caps.nMaxMBCount);
+ return AVERROR(EINVAL);
+ }
+
+ return 0;
+}
+
static void nvdec_decoder_free(void *opaque, uint8_t *data)
{
NVDECDecoder *decoder = (NVDECDecoder*)data;
decoder->cuda_ctx = device_hwctx->cuda_ctx;
decoder->cudl = device_hwctx->internal->cuda_dl;
- ret = cuvid_load_functions(&decoder->cvdl);
+ ret = cuvid_load_functions(&decoder->cvdl, logctx);
if (ret < 0) {
av_log(logctx, AV_LOG_ERROR, "Failed loading nvcuvid.\n");
goto fail;
goto fail;
}
+ ret = nvdec_test_capabilities(decoder, params, logctx);
+ if (ret < 0) {
+ decoder->cudl->cuCtxPopCurrent(&dummy);
+ goto fail;
+ }
+
err = decoder->cvdl->cuvidCreateDecoder(&decoder->decoder, params);
decoder->cudl->cuCtxPopCurrent(&dummy);
params.ulNumOutputSurfaces = 1;
ret = nvdec_decoder_create(&ctx->decoder_ref, frames_ctx->device_ref, ¶ms, avctx);
- if (ret < 0)
+ if (ret < 0) {
+ if (params.ulNumDecodeSurfaces > 32) {
+ av_log(avctx, AV_LOG_WARNING, "Using more than 32 (%d) decode surfaces might cause nvdec to fail.\n",
+ (int)params.ulNumDecodeSurfaces);
+ av_log(avctx, AV_LOG_WARNING, "Try lowering the amount of threads. Using %d right now.\n",
+ avctx->thread_count);
+ }
return ret;
+ }
pool = av_mallocz(sizeof(*pool));
if (!pool) {
return AVERROR(ENOMEM);
cf->decoder_ref = av_buffer_ref(ctx->decoder_ref);
- if (!cf->decoder_ref)
+ if (!cf->decoder_ref) {
+ ret = AVERROR(ENOMEM);
goto fail;
+ }
cf->idx_ref = av_buffer_pool_get(ctx->decoder_pool);
if (!cf->idx_ref) {
return ret;
}
+int ff_nvdec_simple_end_frame(AVCodecContext *avctx)
+{
+ NVDECContext *ctx = avctx->internal->hwaccel_priv_data;
+ int ret = ff_nvdec_end_frame(avctx);
+ ctx->bitstream = NULL;
+ return ret;
+}
+
+int ff_nvdec_simple_decode_slice(AVCodecContext *avctx, const uint8_t *buffer,
+ uint32_t size)
+{
+ NVDECContext *ctx = avctx->internal->hwaccel_priv_data;
+ void *tmp;
+
+ tmp = av_fast_realloc(ctx->slice_offsets, &ctx->slice_offsets_allocated,
+ (ctx->nb_slices + 1) * sizeof(*ctx->slice_offsets));
+ if (!tmp)
+ return AVERROR(ENOMEM);
+ ctx->slice_offsets = tmp;
+
+ if (!ctx->bitstream)
+ ctx->bitstream = (uint8_t*)buffer;
+
+ ctx->slice_offsets[ctx->nb_slices] = buffer - ctx->bitstream;
+ ctx->bitstream_len += size;
+ ctx->nb_slices++;
+
+ return 0;
+}
+
int ff_nvdec_frame_params(AVCodecContext *avctx,
AVBufferRef *hw_frames_ctx,
int dpb_size)
return AVERROR(EINVAL);
}
- if (avctx->thread_type & FF_THREAD_FRAME)
- dpb_size += avctx->thread_count;
-
frames_ctx->format = AV_PIX_FMT_CUDA;
- frames_ctx->width = avctx->coded_width;
- frames_ctx->height = avctx->coded_height;
- frames_ctx->sw_format = sw_desc->comp[0].depth > 8 ?
- AV_PIX_FMT_P010 : AV_PIX_FMT_NV12;
+ frames_ctx->width = (avctx->coded_width + 1) & ~1;
+ frames_ctx->height = (avctx->coded_height + 1) & ~1;
frames_ctx->initial_pool_size = dpb_size;
+ switch (sw_desc->comp[0].depth) {
+ case 8:
+ frames_ctx->sw_format = AV_PIX_FMT_NV12;
+ break;
+ case 10:
+ frames_ctx->sw_format = AV_PIX_FMT_P010;
+ break;
+ case 12:
+ frames_ctx->sw_format = AV_PIX_FMT_P016;
+ break;
+ default:
+ return AVERROR(EINVAL);
+ }
+
return 0;
}
+
+int ff_nvdec_get_ref_idx(AVFrame *frame)
+{
+ FrameDecodeData *fdd;
+ NVDECFrame *cf;
+
+ if (!frame || !frame->private_ref)
+ return -1;
+
+ fdd = (FrameDecodeData*)frame->private_ref->data;
+ cf = (NVDECFrame*)fdd->hwaccel_priv;
+ if (!cf)
+ return -1;
+
+ return cf->idx;
+}