diff options
Diffstat (limited to 'chromium/third_party/ffmpeg/libavutil')
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/Makefile | 2 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/avstring.c | 5 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/dynarray.h | 2 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/frame.h | 61 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/hwcontext_qsv.c | 18 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/hwcontext_vaapi.c | 137 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/hwcontext_vdpau.c | 18 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/mips/generic_macros_msa.h | 42 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/pixdesc.c | 24 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/pixfmt.h | 3 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/time_internal.h | 6 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/tx.c | 803 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/tx.h | 81 | ||||
-rw-r--r-- | chromium/third_party/ffmpeg/libavutil/version.h | 2 |
14 files changed, 1145 insertions, 59 deletions
diff --git a/chromium/third_party/ffmpeg/libavutil/Makefile b/chromium/third_party/ffmpeg/libavutil/Makefile index 53208fc5877..8a7a44e4b5e 100644 --- a/chromium/third_party/ffmpeg/libavutil/Makefile +++ b/chromium/third_party/ffmpeg/libavutil/Makefile @@ -79,6 +79,7 @@ HEADERS = adler32.h \ version.h \ xtea.h \ tea.h \ + tx.h \ HEADERS-$(CONFIG_LZO) += lzo.h @@ -159,6 +160,7 @@ OBJS = adler32.o \ xga_font_data.o \ xtea.o \ tea.o \ + tx.o \ OBJS-$(CONFIG_CUDA) += hwcontext_cuda.o OBJS-$(CONFIG_D3D11VA) += hwcontext_d3d11va.o diff --git a/chromium/third_party/ffmpeg/libavutil/avstring.c b/chromium/third_party/ffmpeg/libavutil/avstring.c index f03dd251417..4c068f5bc57 100644 --- a/chromium/third_party/ffmpeg/libavutil/avstring.c +++ b/chromium/third_party/ffmpeg/libavutil/avstring.c @@ -222,12 +222,13 @@ int av_strcasecmp(const char *a, const char *b) int av_strncasecmp(const char *a, const char *b, size_t n) { - const char *end = a + n; uint8_t c1, c2; + if (n <= 0) + return 0; do { c1 = av_tolower(*a++); c2 = av_tolower(*b++); - } while (a < end && c1 && c1 == c2); + } while (--n && c1 && c1 == c2); return c1 - c2; } diff --git a/chromium/third_party/ffmpeg/libavutil/dynarray.h b/chromium/third_party/ffmpeg/libavutil/dynarray.h index 034a9fedfef..3a7e146422a 100644 --- a/chromium/third_party/ffmpeg/libavutil/dynarray.h +++ b/chromium/third_party/ffmpeg/libavutil/dynarray.h @@ -23,7 +23,7 @@ #include "mem.h" /** - * Add an element of to a dynamic array. + * Add an element to a dynamic array. * * The array is reallocated when its number of elements reaches powers of 2. * Therefore, the amortized cost of adding an element is constant. diff --git a/chromium/third_party/ffmpeg/libavutil/frame.h b/chromium/third_party/ffmpeg/libavutil/frame.h index 8aa3e88367a..5d3231e7bb6 100644 --- a/chromium/third_party/ffmpeg/libavutil/frame.h +++ b/chromium/third_party/ffmpeg/libavutil/frame.h @@ -207,31 +207,58 @@ typedef struct AVFrameSideData { } AVFrameSideData; /** - * Structure to hold Region Of Interest. + * Structure describing a single Region Of Interest. * - * self_size specifies the size of this data structure. This value - * should be set to sizeof(AVRegionOfInterest). EINVAL is returned if self_size is zero. + * When multiple regions are defined in a single side-data block, they + * should be ordered from most to least important - some encoders are only + * capable of supporting a limited number of distinct regions, so will have + * to truncate the list. * - * Number of pixels to discard from the top/bottom/left/right border of - * the frame to obtain the region of interest of the frame. - * They are encoder dependent and will be extended internally - * if the codec requires an alignment. - * If the regions overlap, the last value in the list will be used. - * - * qoffset is quant offset, and base rule here: - * returns EINVAL if AVRational.den is zero. - * the value (num/den) range is [-1.0, 1.0], clamp to +-1.0 if out of range. - * 0 means no picture quality change, - * negative offset asks for better quality (and the best with value -1.0), - * positive offset asks for worse quality (and the worst with value 1.0). - * How to explain/implement the different quilaity requirement is encoder dependent. + * When overlapping regions are defined, the first region containing a given + * area of the frame applies. */ typedef struct AVRegionOfInterest { + /** + * Must be set to the size of this data structure (that is, + * sizeof(AVRegionOfInterest)). + */ uint32_t self_size; + /** + * Distance in pixels from the top edge of the frame to the top and + * bottom edges and from the left edge of the frame to the left and + * right edges of the rectangle defining this region of interest. + * + * The constraints on a region are encoder dependent, so the region + * actually affected may be slightly larger for alignment or other + * reasons. + */ int top; int bottom; int left; int right; + /** + * Quantisation offset. + * + * Must be in the range -1 to +1. A value of zero indicates no quality + * change. A negative value asks for better quality (less quantisation), + * while a positive value asks for worse quality (greater quantisation). + * + * The range is calibrated so that the extreme values indicate the + * largest possible offset - if the rest of the frame is encoded with the + * worst possible quality, an offset of -1 indicates that this region + * should be encoded with the best possible quality anyway. Intermediate + * values are then interpolated in some codec-dependent way. + * + * For example, in 10-bit H.264 the quantisation parameter varies between + * -12 and 51. A typical qoffset value of -1/10 therefore indicates that + * this region should be encoded with a QP around one-tenth of the full + * range better than the rest of the frame. So, if most of the frame + * were to be encoded with a QP of around 30, this region would get a QP + * of around 24 (an offset of approximately -1/10 * (51 - -12) = -6.3). + * An extreme value of -1 would indicate that this region should be + * encoded with the best possible quality regardless of the treatment of + * the rest of the frame - that is, should be encoded at a QP of -12. + */ AVRational qoffset; } AVRegionOfInterest; @@ -563,6 +590,8 @@ typedef struct AVFrame { int decode_error_flags; #define FF_DECODE_ERROR_INVALID_BITSTREAM 1 #define FF_DECODE_ERROR_MISSING_REFERENCE 2 +#define FF_DECODE_ERROR_CONCEALMENT_ACTIVE 4 +#define FF_DECODE_ERROR_DECODE_SLICES 8 /** * number of audio channels, only used for audio. diff --git a/chromium/third_party/ffmpeg/libavutil/hwcontext_qsv.c b/chromium/third_party/ffmpeg/libavutil/hwcontext_qsv.c index 8f9838d7d81..59e4ed91571 100644 --- a/chromium/third_party/ffmpeg/libavutil/hwcontext_qsv.c +++ b/chromium/third_party/ffmpeg/libavutil/hwcontext_qsv.c @@ -892,8 +892,7 @@ static int qsv_transfer_data_to(AVHWFramesContext *ctx, AVFrame *dst, if (ret < 0) return ret; - - if (src->height & 16 || src->linesize[0] & 16) { + if (src->height & 15 || src->linesize[0] & 15) { realigned = 1; memset(&tmp_frame, 0, sizeof(tmp_frame)); tmp_frame.format = src->format; @@ -1207,6 +1206,7 @@ static int qsv_device_create(AVHWDeviceContext *ctx, const char *device, QSVDevicePriv *priv; enum AVHWDeviceType child_device_type; AVHWDeviceContext *child_device; + AVDictionary *child_device_opts; AVDictionaryEntry *e; mfxIMPL impl; @@ -1221,9 +1221,17 @@ static int qsv_device_create(AVHWDeviceContext *ctx, const char *device, e = av_dict_get(opts, "child_device", NULL, 0); - if (CONFIG_VAAPI) + child_device_opts = NULL; + if (CONFIG_VAAPI) { child_device_type = AV_HWDEVICE_TYPE_VAAPI; - else if (CONFIG_DXVA2) + // libmfx does not actually implement VAAPI properly, rather it + // depends on the specific behaviour of a matching iHD driver when + // used on recent Intel hardware. Set options to the VAAPI device + // creation so that we should pick a usable setup by default if + // possible, even when multiple devices and drivers are available. + av_dict_set(&child_device_opts, "kernel_driver", "i915", 0); + av_dict_set(&child_device_opts, "driver", "iHD", 0); + } else if (CONFIG_DXVA2) child_device_type = AV_HWDEVICE_TYPE_DXVA2; else { av_log(ctx, AV_LOG_ERROR, "No supported child device type is enabled\n"); @@ -1231,7 +1239,7 @@ static int qsv_device_create(AVHWDeviceContext *ctx, const char *device, } ret = av_hwdevice_ctx_create(&priv->child_device_ctx, child_device_type, - e ? e->value : NULL, NULL, 0); + e ? e->value : NULL, child_device_opts, 0); if (ret < 0) return ret; diff --git a/chromium/third_party/ffmpeg/libavutil/hwcontext_vaapi.c b/chromium/third_party/ffmpeg/libavutil/hwcontext_vaapi.c index 8624369bb9f..cf117640f28 100644 --- a/chromium/third_party/ffmpeg/libavutil/hwcontext_vaapi.c +++ b/chromium/third_party/ffmpeg/libavutil/hwcontext_vaapi.c @@ -27,6 +27,7 @@ #if CONFIG_LIBDRM # include <va/va_drmcommon.h> +# include <xf86drm.h> # include <drm_fourcc.h> # ifndef DRM_FORMAT_MOD_INVALID # define DRM_FORMAT_MOD_INVALID ((1ULL << 56) - 1) @@ -1469,6 +1470,8 @@ static int vaapi_device_create(AVHWDeviceContext *ctx, const char *device, { VAAPIDevicePriv *priv; VADisplay display = NULL; + const AVDictionaryEntry *ent; + int try_drm, try_x11, try_all; priv = av_mallocz(sizeof(*priv)); if (!priv) @@ -1479,8 +1482,95 @@ static int vaapi_device_create(AVHWDeviceContext *ctx, const char *device, ctx->user_opaque = priv; ctx->free = vaapi_device_free; + ent = av_dict_get(opts, "connection_type", NULL, 0); + if (ent) { + try_all = try_drm = try_x11 = 0; + if (!strcmp(ent->value, "drm")) { + try_drm = 1; + } else if (!strcmp(ent->value, "x11")) { + try_x11 = 1; + } else { + av_log(ctx, AV_LOG_ERROR, "Invalid connection type %s.\n", + ent->value); + return AVERROR(EINVAL); + } + } else { + try_all = 1; + try_drm = HAVE_VAAPI_DRM; + try_x11 = HAVE_VAAPI_X11; + } + +#if HAVE_VAAPI_DRM + while (!display && try_drm) { + // If the device is specified, try to open it as a DRM device node. + // If not, look for a usable render node, possibly restricted to those + // using a specified kernel driver. + int loglevel = try_all ? AV_LOG_VERBOSE : AV_LOG_ERROR; + if (device) { + priv->drm_fd = open(device, O_RDWR); + if (priv->drm_fd < 0) { + av_log(ctx, loglevel, "Failed to open %s as " + "DRM device node.\n", device); + break; + } + } else { + char path[64]; + int n, max_devices = 8; +#if CONFIG_LIBDRM + const AVDictionaryEntry *kernel_driver; + kernel_driver = av_dict_get(opts, "kernel_driver", NULL, 0); +#endif + for (n = 0; n < max_devices; n++) { + snprintf(path, sizeof(path), + "/dev/dri/renderD%d", 128 + n); + priv->drm_fd = open(path, O_RDWR); + if (priv->drm_fd < 0) { + av_log(ctx, AV_LOG_VERBOSE, "Cannot open " + "DRM render node for device %d.\n", n); + break; + } +#if CONFIG_LIBDRM + if (kernel_driver) { + drmVersion *info; + info = drmGetVersion(priv->drm_fd); + if (strcmp(kernel_driver->value, info->name)) { + av_log(ctx, AV_LOG_VERBOSE, "Ignoring device %d " + "with non-matching kernel driver (%s).\n", + n, info->name); + drmFreeVersion(info); + close(priv->drm_fd); + priv->drm_fd = -1; + continue; + } + av_log(ctx, AV_LOG_VERBOSE, "Trying to use " + "DRM render node for device %d, " + "with matching kernel driver (%s).\n", + n, info->name); + drmFreeVersion(info); + } else +#endif + { + av_log(ctx, AV_LOG_VERBOSE, "Trying to use " + "DRM render node for device %d.\n", n); + } + break; + } + if (n >= max_devices) + break; + } + + display = vaGetDisplayDRM(priv->drm_fd); + if (!display) { + av_log(ctx, AV_LOG_VERBOSE, "Cannot open a VA display " + "from DRM device %s.\n", device); + return AVERROR_EXTERNAL; + } + break; + } +#endif + #if HAVE_VAAPI_X11 - if (!display && !(device && device[0] == '/')) { + if (!display && try_x11) { // Try to open the device as an X11 display. priv->x11_display = XOpenDisplay(device); if (!priv->x11_display) { @@ -1500,34 +1590,31 @@ static int vaapi_device_create(AVHWDeviceContext *ctx, const char *device, } #endif -#if HAVE_VAAPI_DRM if (!display) { - // Try to open the device as a DRM path. - // Default to using the first render node if the user did not - // supply a path. - const char *path = device ? device : "/dev/dri/renderD128"; - priv->drm_fd = open(path, O_RDWR); - if (priv->drm_fd < 0) { - av_log(ctx, AV_LOG_VERBOSE, "Cannot open DRM device %s.\n", - path); - } else { - display = vaGetDisplayDRM(priv->drm_fd); - if (!display) { - av_log(ctx, AV_LOG_ERROR, "Cannot open a VA display " - "from DRM device %s.\n", path); - return AVERROR_UNKNOWN; - } + if (device) + av_log(ctx, AV_LOG_ERROR, "No VA display found for " + "device %s.\n", device); + else + av_log(ctx, AV_LOG_ERROR, "No VA display found for " + "any default device.\n"); + return AVERROR(EINVAL); + } - av_log(ctx, AV_LOG_VERBOSE, "Opened VA display via " - "DRM device %s.\n", path); + ent = av_dict_get(opts, "driver", NULL, 0); + if (ent) { +#if VA_CHECK_VERSION(0, 38, 0) + VAStatus vas; + vas = vaSetDriverName(display, ent->value); + if (vas != VA_STATUS_SUCCESS) { + av_log(ctx, AV_LOG_ERROR, "Failed to set driver name to " + "%s: %d (%s).\n", ent->value, vas, vaErrorStr(vas)); + vaTerminate(display); + return AVERROR_EXTERNAL; } - } +#else + av_log(ctx, AV_LOG_WARNING, "Driver name setting is not " + "supported with this VAAPI version.\n"); #endif - - if (!display) { - av_log(ctx, AV_LOG_ERROR, "No VA display found for " - "device: %s.\n", device ? device : ""); - return AVERROR(EINVAL); } return vaapi_device_connect(ctx, display); diff --git a/chromium/third_party/ffmpeg/libavutil/hwcontext_vdpau.c b/chromium/third_party/ffmpeg/libavutil/hwcontext_vdpau.c index c11c3cfdab2..6b8c1d5f767 100644 --- a/chromium/third_party/ffmpeg/libavutil/hwcontext_vdpau.c +++ b/chromium/third_party/ffmpeg/libavutil/hwcontext_vdpau.c @@ -73,8 +73,10 @@ static const VDPAUPixFmtMap pix_fmts_422[] = { }; static const VDPAUPixFmtMap pix_fmts_444[] = { - { VDP_YCBCR_FORMAT_YV12, AV_PIX_FMT_YUV444P }, - { 0, AV_PIX_FMT_NONE, }, +#ifdef VDP_YCBCR_FORMAT_Y_U_V_444 + { VDP_YCBCR_FORMAT_Y_U_V_444, AV_PIX_FMT_YUV444P }, +#endif + { 0, AV_PIX_FMT_NONE, }, }; static const struct { @@ -349,7 +351,11 @@ static int vdpau_transfer_data_from(AVHWFramesContext *ctx, AVFrame *dst, return AVERROR(EINVAL); } - if (vdpau_format == VDP_YCBCR_FORMAT_YV12) + if ((vdpau_format == VDP_YCBCR_FORMAT_YV12) +#ifdef VDP_YCBCR_FORMAT_Y_U_V_444 + || (vdpau_format == VDP_YCBCR_FORMAT_Y_U_V_444) +#endif + ) FFSWAP(void*, data[1], data[2]); err = priv->get_data(surf, vdpau_format, data, linesize); @@ -400,7 +406,11 @@ static int vdpau_transfer_data_to(AVHWFramesContext *ctx, AVFrame *dst, return AVERROR(EINVAL); } - if (vdpau_format == VDP_YCBCR_FORMAT_YV12) + if ((vdpau_format == VDP_YCBCR_FORMAT_YV12) +#ifdef VDP_YCBCR_FORMAT_Y_U_V_444 + || (vdpau_format == VDP_YCBCR_FORMAT_Y_U_V_444) +#endif + ) FFSWAP(const void*, data[1], data[2]); err = priv->put_data(surf, vdpau_format, data, linesize); diff --git a/chromium/third_party/ffmpeg/libavutil/mips/generic_macros_msa.h b/chromium/third_party/ffmpeg/libavutil/mips/generic_macros_msa.h index 6a467046631..a3774281f9e 100644 --- a/chromium/third_party/ffmpeg/libavutil/mips/generic_macros_msa.h +++ b/chromium/third_party/ffmpeg/libavutil/mips/generic_macros_msa.h @@ -23,6 +23,11 @@ #include <stdint.h> #include <msa.h> +#include <config.h> + +#if HAVE_MSA2 +#include <msa2.h> +#endif #define ALIGNMENT 16 #define ALLOC_ALIGNED(align) __attribute__ ((aligned((align) << 1))) @@ -1234,6 +1239,15 @@ unsigned absolute diff values, even-odd pairs are added together to generate 8 halfword results. */ +#if HAVE_MSA2 +#define SAD_UB2_UH(in0, in1, ref0, ref1) \ +( { \ + v8u16 sad_m = { 0 }; \ + sad_m += __builtin_msa2_sad_adj2_u_w2x_b((v16u8) in0, (v16u8) ref0); \ + sad_m += __builtin_msa2_sad_adj2_u_w2x_b((v16u8) in1, (v16u8) ref1); \ + sad_m; \ +} ) +#else #define SAD_UB2_UH(in0, in1, ref0, ref1) \ ( { \ v16u8 diff0_m, diff1_m; \ @@ -1247,6 +1261,7 @@ \ sad_m; \ } ) +#endif // #if HAVE_MSA2 /* Description : Insert specified word elements from input vectors to 1 destination vector @@ -2287,6 +2302,12 @@ extracted and interleaved with same vector 'in0' to generate 4 word elements keeping sign intact */ +#if HAVE_MSA2 +#define UNPCK_R_SH_SW(in, out) \ +{ \ + out = (v4i32) __builtin_msa2_w2x_lo_s_h((v8i16) in); \ +} +#else #define UNPCK_R_SH_SW(in, out) \ { \ v8i16 sign_m; \ @@ -2294,6 +2315,7 @@ sign_m = __msa_clti_s_h((v8i16) in, 0); \ out = (v4i32) __msa_ilvr_h(sign_m, (v8i16) in); \ } +#endif // #if HAVE_MSA2 /* Description : Sign extend byte elements from input vector and return halfword results in pair of vectors @@ -2306,6 +2328,13 @@ Then interleaved left with same vector 'in0' to generate 8 signed halfword elements in 'out1' */ +#if HAVE_MSA2 +#define UNPCK_SB_SH(in, out0, out1) \ +{ \ + out0 = (v4i32) __builtin_msa2_w2x_lo_s_b((v16i8) in); \ + out1 = (v4i32) __builtin_msa2_w2x_hi_s_b((v16i8) in); \ +} +#else #define UNPCK_SB_SH(in, out0, out1) \ { \ v16i8 tmp_m; \ @@ -2313,6 +2342,7 @@ tmp_m = __msa_clti_s_b((v16i8) in, 0); \ ILVRL_B2_SH(tmp_m, in, out0, out1); \ } +#endif // #if HAVE_MSA2 /* Description : Zero extend unsigned byte elements to halfword elements Arguments : Inputs - in (1 input unsigned byte vector) @@ -2339,6 +2369,13 @@ Then interleaved left with same vector 'in0' to generate 4 signed word elements in 'out1' */ +#if HAVE_MSA2 +#define UNPCK_SH_SW(in, out0, out1) \ +{ \ + out0 = (v4i32) __builtin_msa2_w2x_lo_s_h((v8i16) in); \ + out1 = (v4i32) __builtin_msa2_w2x_hi_s_h((v8i16) in); \ +} +#else #define UNPCK_SH_SW(in, out0, out1) \ { \ v8i16 tmp_m; \ @@ -2346,6 +2383,7 @@ tmp_m = __msa_clti_s_h((v8i16) in, 0); \ ILVRL_H2_SW(tmp_m, in, out0, out1); \ } +#endif // #if HAVE_MSA2 /* Description : Swap two variables Arguments : Inputs - in0, in1 @@ -2850,13 +2888,11 @@ */ #define DPADD_SH3_SH(in0, in1, in2, coeff0, coeff1, coeff2) \ ( { \ - v8i16 tmp1_m; \ v8i16 out0_m; \ \ out0_m = __msa_dotp_s_h((v16i8) in0, (v16i8) coeff0); \ out0_m = __msa_dpadd_s_h(out0_m, (v16i8) in1, (v16i8) coeff1); \ - tmp1_m = __msa_dotp_s_h((v16i8) in2, (v16i8) coeff2); \ - out0_m = __msa_adds_s_h(out0_m, tmp1_m); \ + out0_m = __msa_dpadd_s_h(out0_m, (v16i8) in2, (v16i8) coeff2); \ \ out0_m; \ } ) diff --git a/chromium/third_party/ffmpeg/libavutil/pixdesc.c b/chromium/third_party/ffmpeg/libavutil/pixdesc.c index fe38344d732..b97b0665b0a 100644 --- a/chromium/third_party/ffmpeg/libavutil/pixdesc.c +++ b/chromium/third_party/ffmpeg/libavutil/pixdesc.c @@ -2320,6 +2320,30 @@ static const AVPixFmtDescriptor av_pix_fmt_descriptors[AV_PIX_FMT_NB] = { }, .flags = AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_ALPHA, }, + [AV_PIX_FMT_NV24] = { + .name = "nv24", + .nb_components = 3, + .log2_chroma_w = 0, + .log2_chroma_h = 0, + .comp = { + { 0, 1, 0, 0, 8, 0, 7, 1 }, /* Y */ + { 1, 2, 0, 0, 8, 1, 7, 1 }, /* U */ + { 1, 2, 1, 0, 8, 1, 7, 2 }, /* V */ + }, + .flags = AV_PIX_FMT_FLAG_PLANAR, + }, + [AV_PIX_FMT_NV42] = { + .name = "nv42", + .nb_components = 3, + .log2_chroma_w = 0, + .log2_chroma_h = 0, + .comp = { + { 0, 1, 0, 0, 8, 0, 7, 1 }, /* Y */ + { 1, 2, 1, 0, 8, 1, 7, 2 }, /* U */ + { 1, 2, 0, 0, 8, 1, 7, 1 }, /* V */ + }, + .flags = AV_PIX_FMT_FLAG_PLANAR, + }, }; #if FF_API_PLUS1_MINUS1 FF_ENABLE_DEPRECATION_WARNINGS diff --git a/chromium/third_party/ffmpeg/libavutil/pixfmt.h b/chromium/third_party/ffmpeg/libavutil/pixfmt.h index 24d1b7e4153..8b54c9415bb 100644 --- a/chromium/third_party/ffmpeg/libavutil/pixfmt.h +++ b/chromium/third_party/ffmpeg/libavutil/pixfmt.h @@ -345,6 +345,9 @@ enum AVPixelFormat { AV_PIX_FMT_YUVA444P12BE, ///< planar YUV 4:4:4,36bpp, (1 Cr & Cb sample per 1x1 Y samples), 12b alpha, big-endian AV_PIX_FMT_YUVA444P12LE, ///< planar YUV 4:4:4,36bpp, (1 Cr & Cb sample per 1x1 Y samples), 12b alpha, little-endian + AV_PIX_FMT_NV24, ///< planar YUV 4:4:4, 24bpp, 1 plane for Y and 1 plane for the UV components, which are interleaved (first byte U and the following byte V) + AV_PIX_FMT_NV42, ///< as above, but U and V bytes are swapped + AV_PIX_FMT_NB ///< number of pixel formats, DO NOT USE THIS if you want to link with shared libav* because the number of formats might differ between versions }; diff --git a/chromium/third_party/ffmpeg/libavutil/time_internal.h b/chromium/third_party/ffmpeg/libavutil/time_internal.h index 612a75a0411..d0f007ab1c4 100644 --- a/chromium/third_party/ffmpeg/libavutil/time_internal.h +++ b/chromium/third_party/ffmpeg/libavutil/time_internal.h @@ -23,7 +23,7 @@ #include "config.h" #if !HAVE_GMTIME_R && !defined(gmtime_r) -static inline struct tm *gmtime_r(const time_t* clock, struct tm *result) +static inline struct tm *ff_gmtime_r(const time_t* clock, struct tm *result) { struct tm *ptr = gmtime(clock); if (!ptr) @@ -31,10 +31,11 @@ static inline struct tm *gmtime_r(const time_t* clock, struct tm *result) *result = *ptr; return result; } +#define gmtime_r ff_gmtime_r #endif #if !HAVE_LOCALTIME_R && !defined(localtime_r) -static inline struct tm *localtime_r(const time_t* clock, struct tm *result) +static inline struct tm *ff_localtime_r(const time_t* clock, struct tm *result) { struct tm *ptr = localtime(clock); if (!ptr) @@ -42,6 +43,7 @@ static inline struct tm *localtime_r(const time_t* clock, struct tm *result) *result = *ptr; return result; } +#define localtime_r ff_localtime_r #endif #endif /* AVUTIL_TIME_INTERNAL_H */ diff --git a/chromium/third_party/ffmpeg/libavutil/tx.c b/chromium/third_party/ffmpeg/libavutil/tx.c new file mode 100644 index 00000000000..93f6e489d36 --- /dev/null +++ b/chromium/third_party/ffmpeg/libavutil/tx.c @@ -0,0 +1,803 @@ +/* + * Copyright (c) 2019 Lynne <dev@lynne.ee> + * Power of two FFT: + * Copyright (c) 2008 Loren Merritt + * Copyright (c) 2002 Fabrice Bellard + * Partly based on libdjbfft by D. J. Bernstein + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stddef.h> +#include "tx.h" +#include "thread.h" +#include "mem.h" +#include "avassert.h" + +typedef float FFTSample; +typedef AVComplexFloat FFTComplex; + +struct AVTXContext { + int n; /* Nptwo part */ + int m; /* Ptwo part */ + + FFTComplex *exptab; /* MDCT exptab */ + FFTComplex *tmp; /* Temporary buffer needed for all compound transforms */ + int *pfatab; /* Input/Output mapping for compound transforms */ + int *revtab; /* Input mapping for power of two transforms */ +}; + +#define FFT_NAME(x) x + +#define COSTABLE(size) \ + static DECLARE_ALIGNED(32, FFTSample, FFT_NAME(ff_cos_##size))[size/2] + +static FFTSample * const FFT_NAME(ff_cos_tabs)[18]; + +COSTABLE(16); +COSTABLE(32); +COSTABLE(64); +COSTABLE(128); +COSTABLE(256); +COSTABLE(512); +COSTABLE(1024); +COSTABLE(2048); +COSTABLE(4096); +COSTABLE(8192); +COSTABLE(16384); +COSTABLE(32768); +COSTABLE(65536); +COSTABLE(131072); + +static av_cold void init_ff_cos_tabs(int index) +{ + int m = 1 << index; + double freq = 2*M_PI/m; + FFTSample *tab = FFT_NAME(ff_cos_tabs)[index]; + for(int i = 0; i <= m/4; i++) + tab[i] = cos(i*freq); + for(int i = 1; i < m/4; i++) + tab[m/2 - i] = tab[i]; +} + +typedef struct CosTabsInitOnce { + void (*func)(void); + AVOnce control; +} CosTabsInitOnce; + +#define INIT_FF_COS_TABS_FUNC(index, size) \ +static av_cold void init_ff_cos_tabs_ ## size (void) \ +{ \ + init_ff_cos_tabs(index); \ +} + +INIT_FF_COS_TABS_FUNC(4, 16) +INIT_FF_COS_TABS_FUNC(5, 32) +INIT_FF_COS_TABS_FUNC(6, 64) +INIT_FF_COS_TABS_FUNC(7, 128) +INIT_FF_COS_TABS_FUNC(8, 256) +INIT_FF_COS_TABS_FUNC(9, 512) +INIT_FF_COS_TABS_FUNC(10, 1024) +INIT_FF_COS_TABS_FUNC(11, 2048) +INIT_FF_COS_TABS_FUNC(12, 4096) +INIT_FF_COS_TABS_FUNC(13, 8192) +INIT_FF_COS_TABS_FUNC(14, 16384) +INIT_FF_COS_TABS_FUNC(15, 32768) +INIT_FF_COS_TABS_FUNC(16, 65536) +INIT_FF_COS_TABS_FUNC(17, 131072) + +static CosTabsInitOnce cos_tabs_init_once[] = { + { NULL }, + { NULL }, + { NULL }, + { NULL }, + { init_ff_cos_tabs_16, AV_ONCE_INIT }, + { init_ff_cos_tabs_32, AV_ONCE_INIT }, + { init_ff_cos_tabs_64, AV_ONCE_INIT }, + { init_ff_cos_tabs_128, AV_ONCE_INIT }, + { init_ff_cos_tabs_256, AV_ONCE_INIT }, + { init_ff_cos_tabs_512, AV_ONCE_INIT }, + { init_ff_cos_tabs_1024, AV_ONCE_INIT }, + { init_ff_cos_tabs_2048, AV_ONCE_INIT }, + { init_ff_cos_tabs_4096, AV_ONCE_INIT }, + { init_ff_cos_tabs_8192, AV_ONCE_INIT }, + { init_ff_cos_tabs_16384, AV_ONCE_INIT }, + { init_ff_cos_tabs_32768, AV_ONCE_INIT }, + { init_ff_cos_tabs_65536, AV_ONCE_INIT }, + { init_ff_cos_tabs_131072, AV_ONCE_INIT }, +}; + +static FFTSample * const FFT_NAME(ff_cos_tabs)[] = { + NULL, NULL, NULL, NULL, + FFT_NAME(ff_cos_16), + FFT_NAME(ff_cos_32), + FFT_NAME(ff_cos_64), + FFT_NAME(ff_cos_128), + FFT_NAME(ff_cos_256), + FFT_NAME(ff_cos_512), + FFT_NAME(ff_cos_1024), + FFT_NAME(ff_cos_2048), + FFT_NAME(ff_cos_4096), + FFT_NAME(ff_cos_8192), + FFT_NAME(ff_cos_16384), + FFT_NAME(ff_cos_32768), + FFT_NAME(ff_cos_65536), + FFT_NAME(ff_cos_131072), +}; + +static av_cold void ff_init_ff_cos_tabs(int index) +{ + ff_thread_once(&cos_tabs_init_once[index].control, + cos_tabs_init_once[index].func); +} + +static AVOnce tabs_53_once = AV_ONCE_INIT; +static DECLARE_ALIGNED(32, FFTComplex, FFT_NAME(ff_53_tabs))[4]; + +static av_cold void ff_init_53_tabs(void) +{ + ff_53_tabs[0] = (FFTComplex){ cos(2 * M_PI / 12), cos(2 * M_PI / 12) }; + ff_53_tabs[1] = (FFTComplex){ 0.5, 0.5 }; + ff_53_tabs[2] = (FFTComplex){ cos(2 * M_PI / 5), sin(2 * M_PI / 5) }; + ff_53_tabs[3] = (FFTComplex){ cos(2 * M_PI / 10), sin(2 * M_PI / 10) }; +} + +#define BF(x, y, a, b) do { \ + x = (a) - (b); \ + y = (a) + (b); \ + } while (0) + +#define CMUL(dre, dim, are, aim, bre, bim) do { \ + (dre) = (are) * (bre) - (aim) * (bim); \ + (dim) = (are) * (bim) + (aim) * (bre); \ + } while (0) + +#define CMUL3(c, a, b) CMUL((c).re, (c).im, (a).re, (a).im, (b).re, (b).im) + +static av_always_inline void fft3(FFTComplex *out, FFTComplex *in, + ptrdiff_t stride) +{ + FFTComplex tmp[2]; + + tmp[0].re = in[1].im - in[2].im; + tmp[0].im = in[1].re - in[2].re; + tmp[1].re = in[1].re + in[2].re; + tmp[1].im = in[1].im + in[2].im; + + out[0*stride].re = in[0].re + tmp[1].re; + out[0*stride].im = in[0].im + tmp[1].im; + + tmp[0].re *= ff_53_tabs[0].re; + tmp[0].im *= ff_53_tabs[0].im; + tmp[1].re *= ff_53_tabs[1].re; + tmp[1].im *= ff_53_tabs[1].re; + + out[1*stride].re = in[0].re - tmp[1].re + tmp[0].re; + out[1*stride].im = in[0].im - tmp[1].im - tmp[0].im; + out[2*stride].re = in[0].re - tmp[1].re - tmp[0].re; + out[2*stride].im = in[0].im - tmp[1].im + tmp[0].im; +} + +#define DECL_FFT5(NAME, D0, D1, D2, D3, D4) \ +static av_always_inline void NAME(FFTComplex *out, FFTComplex *in, \ + ptrdiff_t stride) \ +{ \ + FFTComplex z0[4], t[6]; \ + \ + t[0].re = in[1].re + in[4].re; \ + t[0].im = in[1].im + in[4].im; \ + t[1].im = in[1].re - in[4].re; \ + t[1].re = in[1].im - in[4].im; \ + t[2].re = in[2].re + in[3].re; \ + t[2].im = in[2].im + in[3].im; \ + t[3].im = in[2].re - in[3].re; \ + t[3].re = in[2].im - in[3].im; \ + \ + out[D0*stride].re = in[0].re + in[1].re + in[2].re + \ + in[3].re + in[4].re; \ + out[D0*stride].im = in[0].im + in[1].im + in[2].im + \ + in[3].im + in[4].im; \ + \ + t[4].re = ff_53_tabs[2].re * t[2].re - ff_53_tabs[3].re * t[0].re; \ + t[4].im = ff_53_tabs[2].re * t[2].im - ff_53_tabs[3].re * t[0].im; \ + t[0].re = ff_53_tabs[2].re * t[0].re - ff_53_tabs[3].re * t[2].re; \ + t[0].im = ff_53_tabs[2].re * t[0].im - ff_53_tabs[3].re * t[2].im; \ + t[5].re = ff_53_tabs[2].im * t[3].re - ff_53_tabs[3].im * t[1].re; \ + t[5].im = ff_53_tabs[2].im * t[3].im - ff_53_tabs[3].im * t[1].im; \ + t[1].re = ff_53_tabs[2].im * t[1].re + ff_53_tabs[3].im * t[3].re; \ + t[1].im = ff_53_tabs[2].im * t[1].im + ff_53_tabs[3].im * t[3].im; \ + \ + z0[0].re = t[0].re - t[1].re; \ + z0[0].im = t[0].im - t[1].im; \ + z0[1].re = t[4].re + t[5].re; \ + z0[1].im = t[4].im + t[5].im; \ + \ + z0[2].re = t[4].re - t[5].re; \ + z0[2].im = t[4].im - t[5].im; \ + z0[3].re = t[0].re + t[1].re; \ + z0[3].im = t[0].im + t[1].im; \ + \ + out[D1*stride].re = in[0].re + z0[3].re; \ + out[D1*stride].im = in[0].im + z0[0].im; \ + out[D2*stride].re = in[0].re + z0[2].re; \ + out[D2*stride].im = in[0].im + z0[1].im; \ + out[D3*stride].re = in[0].re + z0[1].re; \ + out[D3*stride].im = in[0].im + z0[2].im; \ + out[D4*stride].re = in[0].re + z0[0].re; \ + out[D4*stride].im = in[0].im + z0[3].im; \ +} + +DECL_FFT5(fft5, 0, 1, 2, 3, 4) +DECL_FFT5(fft5_m1, 0, 6, 12, 3, 9) +DECL_FFT5(fft5_m2, 10, 1, 7, 13, 4) +DECL_FFT5(fft5_m3, 5, 11, 2, 8, 14) + +static av_always_inline void fft15(FFTComplex *out, FFTComplex *in, + ptrdiff_t stride) +{ + FFTComplex tmp[15]; + + for (int i = 0; i < 5; i++) + fft3(tmp + i, in + i*3, 5); + + fft5_m1(out, tmp + 0, stride); + fft5_m2(out, tmp + 5, stride); + fft5_m3(out, tmp + 10, stride); +} + +#define BUTTERFLIES(a0,a1,a2,a3) {\ + BF(t3, t5, t5, t1);\ + BF(a2.re, a0.re, a0.re, t5);\ + BF(a3.im, a1.im, a1.im, t3);\ + BF(t4, t6, t2, t6);\ + BF(a3.re, a1.re, a1.re, t4);\ + BF(a2.im, a0.im, a0.im, t6);\ +} + +// force loading all the inputs before storing any. +// this is slightly slower for small data, but avoids store->load aliasing +// for addresses separated by large powers of 2. +#define BUTTERFLIES_BIG(a0,a1,a2,a3) {\ + FFTSample r0=a0.re, i0=a0.im, r1=a1.re, i1=a1.im;\ + BF(t3, t5, t5, t1);\ + BF(a2.re, a0.re, r0, t5);\ + BF(a3.im, a1.im, i1, t3);\ + BF(t4, t6, t2, t6);\ + BF(a3.re, a1.re, r1, t4);\ + BF(a2.im, a0.im, i0, t6);\ +} + +#define TRANSFORM(a0,a1,a2,a3,wre,wim) {\ + CMUL(t1, t2, a2.re, a2.im, wre, -wim);\ + CMUL(t5, t6, a3.re, a3.im, wre, wim);\ + BUTTERFLIES(a0,a1,a2,a3)\ +} + +#define TRANSFORM_ZERO(a0,a1,a2,a3) {\ + t1 = a2.re;\ + t2 = a2.im;\ + t5 = a3.re;\ + t6 = a3.im;\ + BUTTERFLIES(a0,a1,a2,a3)\ +} + +/* z[0...8n-1], w[1...2n-1] */ +#define PASS(name)\ +static void name(FFTComplex *z, const FFTSample *wre, unsigned int n)\ +{\ + FFTSample t1, t2, t3, t4, t5, t6;\ + int o1 = 2*n;\ + int o2 = 4*n;\ + int o3 = 6*n;\ + const FFTSample *wim = wre+o1;\ + n--;\ +\ + TRANSFORM_ZERO(z[0],z[o1],z[o2],z[o3]);\ + TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\ + do {\ + z += 2;\ + wre += 2;\ + wim -= 2;\ + TRANSFORM(z[0],z[o1],z[o2],z[o3],wre[0],wim[0]);\ + TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\ + } while(--n);\ +} + +PASS(pass) +#undef BUTTERFLIES +#define BUTTERFLIES BUTTERFLIES_BIG +PASS(pass_big) + +#define DECL_FFT(n,n2,n4)\ +static void fft##n(FFTComplex *z)\ +{\ + fft##n2(z);\ + fft##n4(z+n4*2);\ + fft##n4(z+n4*3);\ + pass(z,FFT_NAME(ff_cos_##n),n4/2);\ +} + +static void fft4(FFTComplex *z) +{ + FFTSample t1, t2, t3, t4, t5, t6, t7, t8; + + BF(t3, t1, z[0].re, z[1].re); + BF(t8, t6, z[3].re, z[2].re); + BF(z[2].re, z[0].re, t1, t6); + BF(t4, t2, z[0].im, z[1].im); + BF(t7, t5, z[2].im, z[3].im); + BF(z[3].im, z[1].im, t4, t8); + BF(z[3].re, z[1].re, t3, t7); + BF(z[2].im, z[0].im, t2, t5); +} + +static void fft8(FFTComplex *z) +{ + FFTSample t1, t2, t3, t4, t5, t6; + + fft4(z); + + BF(t1, z[5].re, z[4].re, -z[5].re); + BF(t2, z[5].im, z[4].im, -z[5].im); + BF(t5, z[7].re, z[6].re, -z[7].re); + BF(t6, z[7].im, z[6].im, -z[7].im); + + BUTTERFLIES(z[0],z[2],z[4],z[6]); + TRANSFORM(z[1],z[3],z[5],z[7],M_SQRT1_2,M_SQRT1_2); +} + +static void fft16(FFTComplex *z) +{ + FFTSample t1, t2, t3, t4, t5, t6; + FFTSample cos_16_1 = FFT_NAME(ff_cos_16)[1]; + FFTSample cos_16_3 = FFT_NAME(ff_cos_16)[3]; + + fft8(z); + fft4(z+8); + fft4(z+12); + + TRANSFORM_ZERO(z[0],z[4],z[8],z[12]); + TRANSFORM(z[2],z[6],z[10],z[14],M_SQRT1_2,M_SQRT1_2); + TRANSFORM(z[1],z[5],z[9],z[13],cos_16_1,cos_16_3); + TRANSFORM(z[3],z[7],z[11],z[15],cos_16_3,cos_16_1); +} + +DECL_FFT(32,16,8) +DECL_FFT(64,32,16) +DECL_FFT(128,64,32) +DECL_FFT(256,128,64) +DECL_FFT(512,256,128) +#define pass pass_big +DECL_FFT(1024,512,256) +DECL_FFT(2048,1024,512) +DECL_FFT(4096,2048,1024) +DECL_FFT(8192,4096,2048) +DECL_FFT(16384,8192,4096) +DECL_FFT(32768,16384,8192) +DECL_FFT(65536,32768,16384) +DECL_FFT(131072,65536,32768) + +static void (* const fft_dispatch[])(FFTComplex*) = { + fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024, + fft2048, fft4096, fft8192, fft16384, fft32768, fft65536, fft131072 +}; + +#define DECL_COMP_FFT(N) \ +static void compound_fft_##N##xM(AVTXContext *s, void *_out, \ + void *_in, ptrdiff_t stride) \ +{ \ + const int m = s->m, *in_map = s->pfatab, *out_map = in_map + N*m; \ + FFTComplex *in = _in; \ + FFTComplex *out = _out; \ + FFTComplex fft##N##in[N]; \ + void (*fftp)(FFTComplex *z) = fft_dispatch[av_log2(m) - 2]; \ + \ + for (int i = 0; i < m; i++) { \ + for (int j = 0; j < N; j++) \ + fft##N##in[j] = in[in_map[i*N + j]]; \ + fft##N(s->tmp + s->revtab[i], fft##N##in, m); \ + } \ + \ + for (int i = 0; i < N; i++) \ + fftp(s->tmp + m*i); \ + \ + for (int i = 0; i < N*m; i++) \ + out[i] = s->tmp[out_map[i]]; \ +} + +DECL_COMP_FFT(3) +DECL_COMP_FFT(5) +DECL_COMP_FFT(15) + +static void monolithic_fft(AVTXContext *s, void *_out, void *_in, + ptrdiff_t stride) +{ + FFTComplex *in = _in; + FFTComplex *out = _out; + int m = s->m, mb = av_log2(m) - 2; + for (int i = 0; i < m; i++) + out[s->revtab[i]] = in[i]; + fft_dispatch[mb](out); +} + +#define DECL_COMP_IMDCT(N) \ +static void compound_imdct_##N##xM(AVTXContext *s, void *_dst, void *_src, \ + ptrdiff_t stride) \ +{ \ + FFTComplex fft##N##in[N]; \ + FFTComplex *z = _dst, *exp = s->exptab; \ + const int m = s->m, len8 = N*m >> 1; \ + const int *in_map = s->pfatab, *out_map = in_map + N*m; \ + const float *src = _src, *in1, *in2; \ + void (*fftp)(FFTComplex *) = fft_dispatch[av_log2(m) - 2]; \ + \ + stride /= sizeof(*src); /* To convert it from bytes */ \ + in1 = src; \ + in2 = src + ((N*m*2) - 1) * stride; \ + \ + for (int i = 0; i < m; i++) { \ + for (int j = 0; j < N; j++) { \ + const int k = in_map[i*N + j]; \ + FFTComplex tmp = { in2[-k*stride], in1[k*stride] }; \ + CMUL3(fft##N##in[j], tmp, exp[k >> 1]); \ + } \ + fft##N(s->tmp + s->revtab[i], fft##N##in, m); \ + } \ + \ + for (int i = 0; i < N; i++) \ + fftp(s->tmp + m*i); \ + \ + for (int i = 0; i < len8; i++) { \ + const int i0 = len8 + i, i1 = len8 - i - 1; \ + const int s0 = out_map[i0], s1 = out_map[i1]; \ + FFTComplex src1 = { s->tmp[s1].im, s->tmp[s1].re }; \ + FFTComplex src0 = { s->tmp[s0].im, s->tmp[s0].re }; \ + \ + CMUL(z[i1].re, z[i0].im, src1.re, src1.im, exp[i1].im, exp[i1].re); \ + CMUL(z[i0].re, z[i1].im, src0.re, src0.im, exp[i0].im, exp[i0].re); \ + } \ +} + +DECL_COMP_IMDCT(3) +DECL_COMP_IMDCT(5) +DECL_COMP_IMDCT(15) + +#define DECL_COMP_MDCT(N) \ +static void compound_mdct_##N##xM(AVTXContext *s, void *_dst, void *_src, \ + ptrdiff_t stride) \ +{ \ + float *src = _src, *dst = _dst; \ + FFTComplex *exp = s->exptab, tmp, fft##N##in[N]; \ + const int m = s->m, len4 = N*m, len3 = len4 * 3, len8 = len4 >> 1; \ + const int *in_map = s->pfatab, *out_map = in_map + N*m; \ + void (*fftp)(FFTComplex *) = fft_dispatch[av_log2(m) - 2]; \ + \ + stride /= sizeof(*dst); \ + \ + for (int i = 0; i < m; i++) { /* Folding and pre-reindexing */ \ + for (int j = 0; j < N; j++) { \ + const int k = in_map[i*N + j]; \ + if (k < len4) { \ + tmp.re = -src[ len4 + k] + src[1*len4 - 1 - k]; \ + tmp.im = -src[ len3 + k] - src[1*len3 - 1 - k]; \ + } else { \ + tmp.re = -src[ len4 + k] - src[5*len4 - 1 - k]; \ + tmp.im = src[-len4 + k] - src[1*len3 - 1 - k]; \ + } \ + CMUL(fft##N##in[j].im, fft##N##in[j].re, tmp.re, tmp.im, \ + exp[k >> 1].re, exp[k >> 1].im); \ + } \ + fft##N(s->tmp + s->revtab[i], fft##N##in, m); \ + } \ + \ + for (int i = 0; i < N; i++) \ + fftp(s->tmp + m*i); \ + \ + for (int i = 0; i < len8; i++) { \ + const int i0 = len8 + i, i1 = len8 - i - 1; \ + const int s0 = out_map[i0], s1 = out_map[i1]; \ + FFTComplex src1 = { s->tmp[s1].re, s->tmp[s1].im }; \ + FFTComplex src0 = { s->tmp[s0].re, s->tmp[s0].im }; \ + \ + CMUL(dst[2*i1*stride + stride], dst[2*i0*stride], src0.re, src0.im, \ + exp[i0].im, exp[i0].re); \ + CMUL(dst[2*i0*stride + stride], dst[2*i1*stride], src1.re, src1.im, \ + exp[i1].im, exp[i1].re); \ + } \ +} + +DECL_COMP_MDCT(3) +DECL_COMP_MDCT(5) +DECL_COMP_MDCT(15) + +static void monolithic_imdct(AVTXContext *s, void *_dst, void *_src, + ptrdiff_t stride) +{ + FFTComplex *z = _dst, *exp = s->exptab; + const int m = s->m, len8 = m >> 1; + const float *src = _src, *in1, *in2; + void (*fftp)(FFTComplex *) = fft_dispatch[av_log2(m) - 2]; + + stride /= sizeof(*src); + in1 = src; + in2 = src + ((m*2) - 1) * stride; + + for (int i = 0; i < m; i++) { + FFTComplex tmp = { in2[-2*i*stride], in1[2*i*stride] }; + CMUL3(z[s->revtab[i]], tmp, exp[i]); + } + + fftp(z); + + for (int i = 0; i < len8; i++) { + const int i0 = len8 + i, i1 = len8 - i - 1; + FFTComplex src1 = { z[i1].im, z[i1].re }; + FFTComplex src0 = { z[i0].im, z[i0].re }; + + CMUL(z[i1].re, z[i0].im, src1.re, src1.im, exp[i1].im, exp[i1].re); + CMUL(z[i0].re, z[i1].im, src0.re, src0.im, exp[i0].im, exp[i0].re); + } +} + +static void monolithic_mdct(AVTXContext *s, void *_dst, void *_src, + ptrdiff_t stride) +{ + float *src = _src, *dst = _dst; + FFTComplex *exp = s->exptab, tmp, *z = _dst; + const int m = s->m, len4 = m, len3 = len4 * 3, len8 = len4 >> 1; + void (*fftp)(FFTComplex *) = fft_dispatch[av_log2(m) - 2]; + + stride /= sizeof(*dst); + + for (int i = 0; i < m; i++) { /* Folding and pre-reindexing */ + const int k = 2*i; + if (k < len4) { + tmp.re = -src[ len4 + k] + src[1*len4 - 1 - k]; + tmp.im = -src[ len3 + k] - src[1*len3 - 1 - k]; + } else { + tmp.re = -src[ len4 + k] - src[5*len4 - 1 - k]; + tmp.im = src[-len4 + k] - src[1*len3 - 1 - k]; + } + CMUL(z[s->revtab[i]].im, z[s->revtab[i]].re, tmp.re, tmp.im, + exp[i].re, exp[i].im); + } + + fftp(z); + + for (int i = 0; i < len8; i++) { + const int i0 = len8 + i, i1 = len8 - i - 1; + FFTComplex src1 = { z[i1].re, z[i1].im }; + FFTComplex src0 = { z[i0].re, z[i0].im }; + + CMUL(dst[2*i1*stride + stride], dst[2*i0*stride], src0.re, src0.im, + exp[i0].im, exp[i0].re); + CMUL(dst[2*i0*stride + stride], dst[2*i1*stride], src1.re, src1.im, + exp[i1].im, exp[i1].re); + } +} + +/* Calculates the modular multiplicative inverse, not fast, replace */ +static int mulinv(int n, int m) +{ + n = n % m; + for (int x = 1; x < m; x++) + if (((n * x) % m) == 1) + return x; + av_assert0(0); /* Never reached */ +} + +/* Guaranteed to work for any n, m where gcd(n, m) == 1 */ +static int gen_compound_mapping(AVTXContext *s, int n, int m, int inv, + enum AVTXType type) +{ + int *in_map, *out_map; + const int len = n*m; + const int m_inv = mulinv(m, n); + const int n_inv = mulinv(n, m); + const int mdct = type == AV_TX_FLOAT_MDCT; + + if (!(s->pfatab = av_malloc(2*len*sizeof(*s->pfatab)))) + return AVERROR(ENOMEM); + + in_map = s->pfatab; + out_map = s->pfatab + n*m; + + /* Ruritanian map for input, CRT map for output, can be swapped */ + for (int j = 0; j < m; j++) { + for (int i = 0; i < n; i++) { + /* Shifted by 1 to simplify forward MDCTs */ + in_map[j*n + i] = ((i*m + j*n) % len) << mdct; + out_map[(i*m*m_inv + j*n*n_inv) % len] = i*m + j; + } + } + + /* Change transform direction by reversing all ACs */ + if (inv) { + for (int i = 0; i < m; i++) { + int *in = &in_map[i*n + 1]; /* Skip the DC */ + for (int j = 0; j < ((n - 1) >> 1); j++) + FFSWAP(int, in[j], in[n - j - 2]); + } + } + + /* Our 15-point transform is also a compound one, so embed its input map */ + if (n == 15) { + for (int k = 0; k < m; k++) { + int tmp[15]; + memcpy(tmp, &in_map[k*15], 15*sizeof(*tmp)); + for (int i = 0; i < 5; i++) { + for (int j = 0; j < 3; j++) + in_map[k*15 + i*3 + j] = tmp[(i*3 + j*5) % 15]; + } + } + } + + return 0; +} + +static int split_radix_permutation(int i, int n, int inverse) +{ + int m; + if (n <= 2) + return i & 1; + m = n >> 1; + if (!(i & m)) + return split_radix_permutation(i, m, inverse)*2; + m >>= 1; + if (inverse == !(i & m)) + return split_radix_permutation(i, m, inverse)*4 + 1; + else + return split_radix_permutation(i, m, inverse)*4 - 1; +} + +static int get_ptwo_revtab(AVTXContext *s, int m, int inv) +{ + if (!(s->revtab = av_malloc(m*sizeof(*s->revtab)))) + return AVERROR(ENOMEM); + + /* Default */ + for (int i = 0; i < m; i++) { + int k = -split_radix_permutation(i, m, inv) & (m - 1); + s->revtab[k] = i; + } + + return 0; +} + +static int gen_mdct_exptab(AVTXContext *s, int len4, double scale) +{ + const double theta = (scale < 0 ? len4 : 0) + 1.0/8.0; + + if (!(s->exptab = av_malloc_array(len4, sizeof(*s->exptab)))) + return AVERROR(ENOMEM); + + scale = sqrt(fabs(scale)); + for (int i = 0; i < len4; i++) { + const double alpha = M_PI_2 * (i + theta) / len4; + s->exptab[i].re = cos(alpha) * scale; + s->exptab[i].im = sin(alpha) * scale; + } + + return 0; +} + +av_cold void av_tx_uninit(AVTXContext **ctx) +{ + if (!(*ctx)) + return; + + av_free((*ctx)->pfatab); + av_free((*ctx)->exptab); + av_free((*ctx)->revtab); + av_free((*ctx)->tmp); + + av_freep(ctx); +} + +static int init_mdct_fft(AVTXContext *s, av_tx_fn *tx, enum AVTXType type, + int inv, int len, const void *scale, uint64_t flags) +{ + int err, n = 1, m = 1, max_ptwo = 1 << (FF_ARRAY_ELEMS(fft_dispatch) + 1); + + if (type == AV_TX_FLOAT_MDCT) + len >>= 1; + +#define CHECK_FACTOR(DST, FACTOR, SRC) \ + if (DST == 1 && !(SRC % FACTOR)) { \ + DST = FACTOR; \ + SRC /= FACTOR; \ + } + CHECK_FACTOR(n, 15, len) + CHECK_FACTOR(n, 5, len) + CHECK_FACTOR(n, 3, len) +#undef CHECK_NPTWO_FACTOR + + /* len must be a power of two now */ + if (!(len & (len - 1)) && len >= 4 && len <= max_ptwo) { + m = len; + len = 1; + } + + /* Filter out direct 3, 5 and 15 transforms, too niche */ + if (len > 1 || m == 1) { + av_log(NULL, AV_LOG_ERROR, "Unsupported transform size: n = %i, " + "m = %i, residual = %i!\n", n, m, len); + return AVERROR(EINVAL); + } else if (n > 1 && m > 1) { /* 2D transform case */ + if ((err = gen_compound_mapping(s, n, m, inv, type))) + return err; + if (!(s->tmp = av_malloc(n*m*sizeof(*s->tmp)))) + return AVERROR(ENOMEM); + *tx = n == 3 ? compound_fft_3xM : + n == 5 ? compound_fft_5xM : + compound_fft_15xM; + if (type == AV_TX_FLOAT_MDCT) + *tx = n == 3 ? inv ? compound_imdct_3xM : compound_mdct_3xM : + n == 5 ? inv ? compound_imdct_5xM : compound_mdct_5xM : + inv ? compound_imdct_15xM : compound_mdct_15xM; + } else { /* Direct transform case */ + *tx = monolithic_fft; + if (type == AV_TX_FLOAT_MDCT) + *tx = inv ? monolithic_imdct : monolithic_mdct; + } + + if (n != 1) + ff_thread_once(&tabs_53_once, ff_init_53_tabs); + if (m != 1) { + get_ptwo_revtab(s, m, inv); + for (int i = 4; i <= av_log2(m); i++) + ff_init_ff_cos_tabs(i); + } + + if (type == AV_TX_FLOAT_MDCT) + if ((err = gen_mdct_exptab(s, n*m, *((float *)scale)))) + return err; + + s->n = n; + s->m = m; + + return 0; +} + +av_cold int av_tx_init(AVTXContext **ctx, av_tx_fn *tx, enum AVTXType type, + int inv, int len, const void *scale, uint64_t flags) +{ + int err; + AVTXContext *s = av_mallocz(sizeof(*s)); + if (!s) + return AVERROR(ENOMEM); + + switch (type) { + case AV_TX_FLOAT_FFT: + case AV_TX_FLOAT_MDCT: + if ((err = init_mdct_fft(s, tx, type, inv, len, scale, flags))) + goto fail; + break; + default: + err = AVERROR(EINVAL); + goto fail; + } + + *ctx = s; + + return 0; + +fail: + av_tx_uninit(&s); + *tx = NULL; + return err; +} diff --git a/chromium/third_party/ffmpeg/libavutil/tx.h b/chromium/third_party/ffmpeg/libavutil/tx.h new file mode 100644 index 00000000000..b1f2d963533 --- /dev/null +++ b/chromium/third_party/ffmpeg/libavutil/tx.h @@ -0,0 +1,81 @@ +/* + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVUTIL_TX_H +#define AVUTIL_TX_H + +#include <stdint.h> +#include <stddef.h> + +typedef struct AVTXContext AVTXContext; + +typedef struct AVComplexFloat { + float re, im; +} AVComplexFloat; + +enum AVTXType { + /** + * Standard complex to complex FFT with sample data type AVComplexFloat. + * Scaling currently unsupported + */ + AV_TX_FLOAT_FFT = 0, + /** + * Standard MDCT with sample data type of float and a scale type of + * float. Length is the frame size, not the window size (which is 2x frame) + */ + AV_TX_FLOAT_MDCT = 1, +}; + +/** + * Function pointer to a function to perform the transform. + * + * @note Using a different context than the one allocated during av_tx_init() + * is not allowed. + * + * @param s the transform context + * @param out the output array + * @param in the input array + * @param stride the input or output stride (depending on transform direction) + * in bytes, currently implemented for all MDCT transforms + */ +typedef void (*av_tx_fn)(AVTXContext *s, void *out, void *in, ptrdiff_t stride); + +/** + * Initialize a transform context with the given configuration + * Currently power of two lengths from 4 to 131072 are supported, along with + * any length decomposable to a power of two and either 3, 5 or 15. + * + * @param ctx the context to allocate, will be NULL on error + * @param tx pointer to the transform function pointer to set + * @param type type the type of transform + * @param inv whether to do an inverse or a forward transform + * @param len the size of the transform in samples + * @param scale pointer to the value to scale the output if supported by type + * @param flags currently unused + * + * @return 0 on success, negative error code on failure + */ +int av_tx_init(AVTXContext **ctx, av_tx_fn *tx, enum AVTXType type, + int inv, int len, const void *scale, uint64_t flags); + +/** + * Frees a context and sets ctx to NULL, does nothing when ctx == NULL + */ +void av_tx_uninit(AVTXContext **ctx); + +#endif /* AVUTIL_TX_H */ diff --git a/chromium/third_party/ffmpeg/libavutil/version.h b/chromium/third_party/ffmpeg/libavutil/version.h index 1fcdea95bf4..e16b93e877e 100644 --- a/chromium/third_party/ffmpeg/libavutil/version.h +++ b/chromium/third_party/ffmpeg/libavutil/version.h @@ -79,7 +79,7 @@ */ #define LIBAVUTIL_VERSION_MAJOR 56 -#define LIBAVUTIL_VERSION_MINOR 26 +#define LIBAVUTIL_VERSION_MINOR 30 #define LIBAVUTIL_VERSION_MICRO 100 #define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \ |