Merge remote-tracking branch 'qatar/master'

* qatar/master:
  dsputilenc_mmx: split assignment of ff_sse16_sse2 to SSE2 section.
  dnxhdenc: add space between function argument type and comment.
  x86: fmtconvert: add special asm for float_to_int16_interleave_misc_*
  attributes: Add a definition of av_always_inline for MSVC
  cmdutils: Pass the actual chosen encoder to filter_codec_opts
  os_support: Add fallback definitions for stat flags
  os_support: Rename the poll fallback function to ff_poll
  network: Check for struct pollfd
  os_support: Don't compare a negative number against socket descriptors
  os_support: Include all the necessary headers for the win32 open function
  x86: vc1: fix and enable optimised loop filter

Conflicts:
	cmdutils.c
	cmdutils.h
	ffmpeg.c
	ffplay.c
	libavformat/os_support.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

cmdutils.c

@@ -1135,8 +1135,8 @@ int check_stream_specifier(AVFormatContext *s, AVStream *st, const char *spec)
     return AVERROR(EINVAL);
 }
 
-AVDictionary *filter_codec_opts(AVDictionary *opts, AVCodec *codec,
+AVDictionary *filter_codec_opts(AVDictionary *opts, enum CodecID codec_id,
-                                AVFormatContext *s, AVStream *st)
+                                AVFormatContext *s, AVStream *st, AVCodec *codec)
 {
     AVDictionary    *ret = NULL;
     AVDictionaryEntry *t = NULL;
@@ -1145,6 +1145,9 @@ AVDictionary *filter_codec_opts(AVDictionary *opts, AVCodec *codec,
     char          prefix = 0;
     const AVClass    *cc = avcodec_get_class();
 
+    if (!codec)
+        codec            = s->oformat ? avcodec_find_encoder(codec_id)
+                                      : avcodec_find_decoder(codec_id);
     if (!codec)
         return NULL;
 
@@ -1205,8 +1208,8 @@ AVDictionary **setup_find_stream_info_opts(AVFormatContext *s,
         return NULL;
     }
     for (i = 0; i < s->nb_streams; i++)
-        opts[i] = filter_codec_opts(codec_opts, avcodec_find_decoder(s->streams[i]->codec->codec_id),
+        opts[i] = filter_codec_opts(codec_opts, s->streams[i]->codec->codec_id,
-                                    s, s->streams[i]);
+                                    s, s->streams[i], NULL);
     return opts;
 }
 

cmdutils.h

@@ -232,10 +232,12 @@ int check_stream_specifier(AVFormatContext *s, AVStream *st, const char *spec);
  *
  * @param s Corresponding format context.
  * @param st A stream from s for which the options should be filtered.
+ * @param codec The particular codec for which the options should be filtered.
+ *              If null, the default one is looked up according to the codec id.
  * @return a pointer to the created dictionary
  */
-AVDictionary *filter_codec_opts(AVDictionary *opts, AVCodec *codec,
+AVDictionary *filter_codec_opts(AVDictionary *opts, enum CodecID codec_id,
-                                AVFormatContext *s, AVStream *st);
+                                AVFormatContext *s, AVStream *st, AVCodec *codec);
 
 /**
  * Setup AVCodecContext options for avformat_find_stream_info().

configure

@@ -1265,6 +1265,7 @@ HAVE_LIST="
     struct_group_source_req
     struct_ip_mreq_source
     struct_ipv6_mreq
+    struct_pollfd
     struct_rusage_ru_maxrss
     struct_sockaddr_in6
     struct_sockaddr_sa_len
@@ -3125,6 +3126,7 @@ if enabled network; then
     check_type netinet/in.h "struct ip_mreq_source" -D_BSD_SOURCE
     check_type netinet/in.h "struct ipv6_mreq" -D_DARWIN_C_SOURCE
     check_type netinet/in.h "struct sockaddr_in6"
+    check_type poll.h "struct pollfd"
     check_type "sys/types.h sys/socket.h" "struct sockaddr_storage"
     check_struct "sys/types.h sys/socket.h" "struct sockaddr" sa_len
     check_header netinet/sctp.h
@@ -3141,6 +3143,7 @@ if enabled network; then
         check_type ws2tcpip.h "struct group_source_req"
         check_type ws2tcpip.h "struct ip_mreq_source"
         check_type ws2tcpip.h "struct ipv6_mreq"
+        check_type winsock2.h "struct pollfd"
         check_type ws2tcpip.h "struct sockaddr_in6"
         check_type ws2tcpip.h "struct sockaddr_storage"
         check_struct winsock2.h "struct sockaddr" sa_len

ffmpeg.c

@@ -4149,7 +4149,7 @@ static void add_input_streams(OptionsContext *o, AVFormatContext *ic)
         ist->file_index = nb_input_files;
         ist->discard = 1;
         st->discard  = AVDISCARD_ALL;
-        ist->opts = filter_codec_opts(codec_opts, choose_decoder(o, ic, st), ic, st);
+        ist->opts = filter_codec_opts(codec_opts, ist->st->codec->codec_id, ic, st, choose_decoder(o, ic, st));
 
         ist->ts_scale = 1.0;
         MATCH_PER_STREAM_OPT(ts_scale, dbl, ist->ts_scale, ic, st);
@@ -4505,7 +4505,7 @@ static OutputStream *new_output_stream(OptionsContext *o, AVFormatContext *oc, e
     st->codec->codec_type = type;
     choose_encoder(o, oc, ost);
     if (ost->enc) {
-        ost->opts  = filter_codec_opts(codec_opts, ost->enc, oc, st);
+        ost->opts  = filter_codec_opts(codec_opts, ost->enc->id, oc, st, ost->enc);
     }
 
     avcodec_get_context_defaults3(st->codec, ost->enc);

ffplay.c

@@ -2161,7 +2161,7 @@ static int stream_component_open(VideoState *is, int stream_index)
     avctx = ic->streams[stream_index]->codec;
 
     codec = avcodec_find_decoder(avctx->codec_id);
-    opts = filter_codec_opts(codec_opts, codec, ic, ic->streams[stream_index]);
+    opts = filter_codec_opts(codec_opts, avctx->codec_id, ic, ic->streams[stream_index], codec);
 
     switch(avctx->codec_type){
         case AVMEDIA_TYPE_AUDIO   : is->last_audio_stream    = stream_index; if(audio_codec_name   ) codec= avcodec_find_decoder_by_name(   audio_codec_name); break;

libavcodec/dnxhdenc.h

@@ -90,7 +90,7 @@ typedef struct DNXHDEncContext {
     RCCMPEntry *mb_cmp;
     RCEntry   (*mb_rc)[8160];
 
-    void (*get_pixels_8x4_sym)(DCTELEM */*align 16*/, const uint8_t *, int);
+    void (*get_pixels_8x4_sym)(DCTELEM * /*align 16*/, const uint8_t *, int);
 } DNXHDEncContext;
 
 void ff_dnxhd_init_mmx(DNXHDEncContext *ctx);

libavcodec/x86/dsputilenc_mmx.c

@@ -1128,7 +1128,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
 #endif
 
         c->pix_norm1 = pix_norm1_mmx;
-        c->sse[0] = (HAVE_YASM && mm_flags & AV_CPU_FLAG_SSE2) ? ff_sse16_sse2 : sse16_mmx;
+        c->sse[0] = sse16_mmx;
         c->sse[1] = sse8_mmx;
         c->vsad[4]= vsad_intra16_mmx;
 
@@ -1165,9 +1165,12 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
             if (bit_depth <= 8)
                 c->get_pixels = get_pixels_sse2;
             c->sum_abs_dctelem= sum_abs_dctelem_sse2;
-#if HAVE_YASM && HAVE_ALIGNED_STACK
+#if HAVE_YASM
+            c->sse[0] = ff_sse16_sse2;
+#if HAVE_ALIGNED_STACK
             c->hadamard8_diff[0]= ff_hadamard8_diff16_sse2;
             c->hadamard8_diff[1]= ff_hadamard8_diff_sse2;
+#endif
 #endif
         }
 

libavcodec/x86/fmtconvert.asm

@@ -115,6 +115,84 @@ FLOAT_TO_INT16 sse, 0
 FLOAT_TO_INT16 3dnow, 0
 %undef cvtps2pi
 
+;------------------------------------------------------------------------------
+; void ff_float_to_int16_step(int16_t *dst, const float *src, long len, long step);
+;------------------------------------------------------------------------------
+%macro FLOAT_TO_INT16_STEP 2
+cglobal float_to_int16_step_%1, 4,7,%2, dst, src, len, step, step3, v1, v2
+    add       lenq, lenq
+    lea       srcq, [srcq+2*lenq]
+    lea     step3q, [stepq*3]
+    neg       lenq
+.loop:
+%ifidn %1, sse2
+    cvtps2dq    m0, [srcq+2*lenq   ]
+    cvtps2dq    m1, [srcq+2*lenq+16]
+    packssdw    m0, m1
+    movd       v1d, m0
+    psrldq      m0, 4
+    movd       v2d, m0
+    psrldq      m0, 4
+    mov     [dstq], v1w
+    mov  [dstq+stepq*4], v2w
+    shr        v1d, 16
+    shr        v2d, 16
+    mov  [dstq+stepq*2], v1w
+    mov  [dstq+step3q*2], v2w
+    lea       dstq, [dstq+stepq*8]
+    movd       v1d, m0
+    psrldq      m0, 4
+    movd       v2d, m0
+    mov     [dstq], v1w
+    mov  [dstq+stepq*4], v2w
+    shr        v1d, 16
+    shr        v2d, 16
+    mov  [dstq+stepq*2], v1w
+    mov  [dstq+step3q*2], v2w
+    lea       dstq, [dstq+stepq*8]
+%else
+    cvtps2pi    m0, [srcq+2*lenq   ]
+    cvtps2pi    m1, [srcq+2*lenq+ 8]
+    cvtps2pi    m2, [srcq+2*lenq+16]
+    cvtps2pi    m3, [srcq+2*lenq+24]
+    packssdw    m0, m1
+    packssdw    m2, m3
+    movd       v1d, m0
+    psrlq       m0, 32
+    movd       v2d, m0
+    mov     [dstq], v1w
+    mov  [dstq+stepq*4], v2w
+    shr        v1d, 16
+    shr        v2d, 16
+    mov  [dstq+stepq*2], v1w
+    mov  [dstq+step3q*2], v2w
+    lea       dstq, [dstq+stepq*8]
+    movd       v1d, m2
+    psrlq       m2, 32
+    movd       v2d, m2
+    mov     [dstq], v1w
+    mov  [dstq+stepq*4], v2w
+    shr        v1d, 16
+    shr        v2d, 16
+    mov  [dstq+stepq*2], v1w
+    mov  [dstq+step3q*2], v2w
+    lea       dstq, [dstq+stepq*8]
+%endif
+    add       lenq, 16
+    js .loop
+%ifnidn %1, sse2
+    emms
+%endif
+    REP_RET
+%endmacro
+
+INIT_XMM
+FLOAT_TO_INT16_STEP sse2, 2
+INIT_MMX
+FLOAT_TO_INT16_STEP sse, 0
+%define cvtps2pi pf2id
+FLOAT_TO_INT16_STEP 3dnow, 0
+%undef cvtps2pi
 
 ;-------------------------------------------------------------------------------
 ; void ff_float_to_int16_interleave2(int16_t *dst, const float **src, long len);

libavcodec/x86/fmtconvert_mmx.c

@@ -25,6 +25,7 @@
 #include "libavutil/cpu.h"
 #include "libavutil/x86_cpu.h"
 #include "libavcodec/fmtconvert.h"
+#include "libavcodec/dsputil.h"
 
 #if HAVE_YASM
 
@@ -35,6 +36,10 @@ void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
 void ff_float_to_int16_sse  (int16_t *dst, const float *src, long len);
 void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
 
+void ff_float_to_int16_step_3dnow(int16_t *dst, const float *src, long len, long step);
+void ff_float_to_int16_step_sse  (int16_t *dst, const float *src, long len, long step);
+void ff_float_to_int16_step_sse2 (int16_t *dst, const float *src, long len, long step);
+
 void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
 void ff_float_to_int16_interleave2_sse  (int16_t *dst, const float **src, long len);
 void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
@@ -48,12 +53,9 @@ void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len
 #define FLOAT_TO_INT16_INTERLEAVE(cpu) \
 /* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
 static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
-    DECLARE_ALIGNED(16, int16_t, tmp)[len];\
+    int c;\
-    int i,j,c;\
     for(c=0; c<channels; c++){\
-        ff_float_to_int16_##cpu(tmp, src[c], len);\
+        ff_float_to_int16_step_##cpu(dst+c, src[c], len, channels);\
-        for(i=0, j=c; i<len; i++, j+=channels)\
-            dst[j] = tmp[i];\
     }\
 }\
 \

libavcodec/x86/vc1dsp_mmx.c

@@ -809,7 +809,7 @@ void ff_vc1dsp_init_mmx(VC1DSPContext *dsp)
 #if HAVE_YASM
     if (mm_flags & AV_CPU_FLAG_MMX) {
     }
-    return;
+
     if (mm_flags & AV_CPU_FLAG_MMX2) {
         ASSIGN_LF(mmx2);
     }

libavcodec/x86/vc1dsp_yasm.asm

@@ -119,7 +119,9 @@ section .text
     pand    m2, m6
     pand    m3, m2  ; d final
 
-    PSIGNW  m3, m7
+    psraw   m7, 15
+    pxor    m3, m7
+    psubw   m3, m7
     psubw   m0, m3
     paddw   m1, m3
     packuswb m0, m0
@@ -284,7 +286,6 @@ cglobal vc1_h_loop_filter8_sse2, 3,6,8
     RET
 
 %define PABSW PABSW_SSSE3
-%define PSIGNW PSIGNW_SSSE3
 
 INIT_MMX
 ; void ff_vc1_v_loop_filter4_ssse3(uint8_t *src, int stride, int pq)

libavformat/os_support.c

@@ -28,10 +28,12 @@
 #include "os_support.h"
 
 #if defined(_WIN32) && !defined(__MINGW32CE__)
+#undef open
+#include <fcntl.h>
+#include <io.h>
 #include <windows.h>
 #include <share.h>
 
-#undef open
 int ff_win32_open(const char *filename_utf8, int oflag, int pmode)
 {
     int fd;
@@ -265,7 +267,7 @@ int ff_socket_nonblock(int socket, int enable)
 }
 
 #if !HAVE_POLL_H
-int poll(struct pollfd *fds, nfds_t numfds, int timeout)
+int ff_poll(struct pollfd *fds, nfds_t numfds, int timeout)
 {
     fd_set read_set;
     fd_set write_set;
@@ -285,7 +287,7 @@ int poll(struct pollfd *fds, nfds_t numfds, int timeout)
     FD_ZERO(&write_set);
     FD_ZERO(&exception_set);
 
-    n = -1;
+    n = 0;
     for(i = 0; i < numfds; i++) {
         if (fds[i].fd < 0)
             continue;
@@ -300,22 +302,22 @@ int poll(struct pollfd *fds, nfds_t numfds, int timeout)
         if (fds[i].events & POLLOUT) FD_SET(fds[i].fd, &write_set);
         if (fds[i].events & POLLERR) FD_SET(fds[i].fd, &exception_set);
 
-        if (fds[i].fd > n)
+        if (fds[i].fd >= n)
-            n = fds[i].fd;
+            n = fds[i].fd + 1;
     };
 
-    if (n == -1)
+    if (n == 0)
         /* Hey!? Nothing to poll, in fact!!! */
         return 0;
 
     if (timeout < 0)
-        rc = select(n+1, &read_set, &write_set, &exception_set, NULL);
+        rc = select(n, &read_set, &write_set, &exception_set, NULL);
     else {
         struct timeval    tv;
 
         tv.tv_sec = timeout / 1000;
         tv.tv_usec = 1000 * (timeout % 1000);
-        rc = select(n+1, &read_set, &write_set, &exception_set, &tv);
+        rc = select(n, &read_set, &write_set, &exception_set, &tv);
     };
 
     if (rc < 0)

libavformat/os_support.h

@@ -29,6 +29,8 @@
 
 #include "config.h"
 
+#include <sys/stat.h>
+
 #if defined(__MINGW32__) && !defined(__MINGW32CE__)
 #  include <fcntl.h>
 #  ifdef lseek
@@ -58,6 +60,13 @@ static inline int is_dos_path(const char *path)
 #define SHUT_RD SD_RECEIVE
 #define SHUT_WR SD_SEND
 #define SHUT_RDWR SD_BOTH
+
+#ifndef S_IRUSR
+#define S_IRUSR S_IREAD
+#endif
+#ifndef S_IWUSR
+#define S_IWUSR S_IWRITE
+#endif
 #endif
 
 #if defined(_WIN32) && !defined(__MINGW32CE__)
@@ -78,6 +87,10 @@ typedef int socklen_t;
 #if !HAVE_POLL_H
 typedef unsigned long nfds_t;
 
+#if HAVE_WINSOCK2_H
+#include <winsock2.h>
+#endif
+#if !HAVE_STRUCT_POLLFD
 struct pollfd {
     int fd;
     short events;  /* events to look for */
@@ -97,9 +110,11 @@ struct pollfd {
 #define POLLERR    0x0004  /* errors pending */
 #define POLLHUP    0x0080  /* disconnected */
 #define POLLNVAL   0x1000  /* invalid file descriptor */
+#endif
 
 
-int poll(struct pollfd *fds, nfds_t numfds, int timeout);
+int ff_poll(struct pollfd *fds, nfds_t numfds, int timeout);
+#define poll ff_poll
 #endif /* HAVE_POLL_H */
 #endif /* CONFIG_NETWORK */
 

libavutil/attributes.h

@@ -35,6 +35,8 @@
 #ifndef av_always_inline
 #if AV_GCC_VERSION_AT_LEAST(3,1)
 #    define av_always_inline __attribute__((always_inline)) inline
+#elif defined(_MSC_VER)
+#    define av_always_inline __forceinline
 #else
 #    define av_always_inline inline
 #endif