WMA Voice now plays and seeks in the sim. The code is still in floating point, and is not added to the main build. There's still a bug with the decoder in the current state that it outputs a fewer number of samples than ffmpeg's.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27744 a1c6a512-1295-4272-9138-f99709370657

apps/codecs/libasf/asf.h

@@ -7,6 +7,7 @@
 #define ASF_CODEC_ID_WMAV1      0x160
 #define ASF_CODEC_ID_WMAV2      0x161
 #define ASF_CODEC_ID_WMAPRO     0x162
+#define ASF_CODEC_ID_WMAVOICE   0x00A
 
 enum asf_error_e {
     ASF_ERROR_INTERNAL       = -1,  /* incorrect input to API calls */
@@ -33,7 +34,7 @@ struct asf_waveformatex_s {
     uint16_t bitspersample;
     uint16_t datalen;
     uint16_t numpackets;
-    uint8_t data[18];
+    uint8_t data[46];
 };
 typedef struct asf_waveformatex_s asf_waveformatex_t;
 

apps/codecs/libwmavoice/SOURCES

@@ -0,0 +1,17 @@
+acelp_filters.c
+acelp_vectors.c
+avfft.c
+bitstream.c
+celp_filters.c
+celp_math.c
+dct.c
+fft.c
+lsp.c
+mdct.c
+rdft.c
+utils.c
+wmavoice.c
+libavutil/log.c
+libavutil/lzo.c
+libavutil/mem.c
+libavutil/mathematics.c

apps/codecs/libwmavoice/libwmavoice.make

@@ -0,0 +1,37 @@
+#             __________               __   ___.
+#   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
+#   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
+#   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
+#   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
+#                     \/            \/     \/    \/            \/
+# $Id: libwmavoice.make 27586 2010-07-27 06:48:15Z nls $
+#
+
+# libwmavoice
+WMAVOICELIB := $(CODECDIR)/libwmavoice.a
+WMAVOICELIB_SRC := $(call preprocess, $(APPSDIR)/codecs/libwmavoice/SOURCES)
+WMAVOICELIB_OBJ := $(call c2obj, $(WMAVOICELIB_SRC))
+OTHER_SRC += $(WMAVOICELIB_SRC)
+
+$(WMAVOICELIB): $(WMAVOICELIB_OBJ)
+	$(SILENT)$(shell rm -f $@)
+	$(call PRINTS,AR $(@F))$(AR) rcs $@ $^ >/dev/null
+
+WMAVOICEFLAGS = -I$(APPSDIR)/codecs/libwmavoice $(filter-out -O%,$(CODECFLAGS))
+
+ifeq ($(CPU),coldfire)
+	WMAVOICEFLAGS += -O2
+else
+	WMAVOICEFLAGS += -O1
+endif
+
+ifeq ($(APP_TYPE),sdl-sim)
+# wmavoice needs libm in the simulator
+$(CODECDIR)/wmavoice.codec: $(CODECDIR)/wmavoice.o
+	$(call PRINTS,LD $(@F))$(CC) $(CODECFLAGS) -o $(CODECDIR)/wmavoice.elf \
+	$(filter %.o, $^) \
+	$(filter %.a, $+) \
+	-lgcc -lm $(CODECLDFLAGS)
+	$(SILENT)cp $(CODECDIR)/wmavoice.elf $@
+endif
+

apps/codecs/libwmavoice/wmavoice.c

@@ -26,7 +26,7 @@
  */
 
 #include <math.h>
-#include "avcodec.h"
+#include "wmavoice.h"
 #include "get_bits.h"
 #include "put_bits.h"
 #include "wmavoice_data.h"
@@ -286,6 +286,10 @@ typedef struct {
      */
 } WMAVoiceContext;
 
+/* global decode context */
+static WMAVoiceContext globWMAVoiceCtx;
+
+
 /**
  * Set up the variable bit mode (VBM) tree from container extradata.
  * @param gb bit I/O context.
@@ -330,9 +334,10 @@ static av_cold int decode_vbmtree(GetBitContext *gb, int8_t vbm_tree[25])
 /**
  * Set up decoder with parameters from demuxer (extradata etc.).
  */
-static av_cold int wmavoice_decode_init(AVCodecContext *ctx)
+av_cold int wmavoice_decode_init(AVCodecContext *ctx)
 {
     int n, flags, pitch_range, lsp16_flag;
+    ctx->priv_data = &globWMAVoiceCtx;
     WMAVoiceContext *s = ctx->priv_data;
 
     /**
@@ -1743,7 +1748,7 @@ static int synth_superframe(AVCodecContext *ctx,
      * the wild yet. */
     if (!get_bits1(gb)) {
         av_log_missing_feature(ctx, "WMAPro-in-WMAVoice support", 1);
-        return -1;
+        return ERROR_WMAPRO_IN_WMAVOICE;
     }
 
     /* (optional) nr. of samples in superframe; always <= 480 and >= 0 */
@@ -1893,7 +1898,7 @@ static void copy_bits(PutBitContext *pb,
  *
  * For more information about frames, see #synth_superframe().
  */
-static int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
+int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
                                   int *data_size, AVPacket *avpkt)
 {
     WMAVoiceContext *s = ctx->priv_data;
@@ -1936,6 +1941,15 @@ static int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
                 s->sframe_cache_size += s->spillover_nbits;
                 if ((res = synth_superframe(ctx, data, data_size)) == 0 &&
                     *data_size > 0) {
+                    /* convert the float values to int32 for rockbox */
+                    int i;
+                    int32_t *iptr = data;
+                    float   *fptr = data;
+                    for(i = 0; i < *data_size/sizeof(float); i++)
+                    {
+                        fptr[i] *= (float)(INT32_MAX);
+                        iptr[i] = (int32_t)fptr[i];
+                    }
                     cnt += s->spillover_nbits;
                     s->skip_bits_next = cnt & 7;
                     return cnt >> 3;
@@ -1957,12 +1971,21 @@ static int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
     } else if (*data_size > 0) {
         int cnt = get_bits_count(gb);
         s->skip_bits_next = cnt & 7;
+        /* convert the float values to int32 for rockbox */
+        int i;
+        int32_t *iptr = data;
+        float   *fptr = data;
+        for(i = 0; i < *data_size/sizeof(float); i++)
+        {
+            fptr[i] *= (float)(INT32_MAX);
+            iptr[i] = (int32_t)fptr[i];
+        }
         return cnt >> 3;
     } else if ((s->sframe_cache_size = pos) > 0) {
         /* rewind bit reader to start of last (incomplete) superframe... */
         init_get_bits(gb, avpkt->data, size << 3);
         skip_bits_long(gb, (size << 3) - pos);
-        assert(get_bits_left(gb) == pos);
+        //assert(get_bits_left(gb) == pos);
 
         /* ...and cache it for spillover in next packet */
         init_put_bits(&s->pb, s->sframe_cache, SFRAME_CACHE_MAXSIZE);

apps/codecs/libwmavoice/wmavoice.h

@@ -0,0 +1,7 @@
+#include "avcodec.h"
+
+#define ERROR_WMAPRO_IN_WMAVOICE -0x162
+
+av_cold int wmavoice_decode_init(AVCodecContext *ctx);
+int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
+                                  int *data_size, AVPacket *avpkt);

apps/codecs/wmavoice.c

@@ -0,0 +1,195 @@
+/***************************************************************************
+ *             __________               __   ___.
+ *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
+ *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
+ *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
+ *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
+ *                     \/            \/     \/    \/            \/
+ * $Id$
+ *
+ * Copyright (C) 2010 Mohamed Tarek
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
+ * KIND, either express or implied.
+ *
+ ****************************************************************************/
+
+#include "codeclib.h"
+#include "libasf/asf.h"
+#include "libwmavoice/wmavoice.h"
+
+CODEC_HEADER
+
+static AVCodecContext avctx;
+static AVPacket avpkt;
+
+#define MAX_FRAMES           3   /*maximum number of frames per superframe*/
+#define MAX_FRAMESIZE        160 /* maximum number of samples per frame */
+#define BUFSIZE              MAX_FRAMES*MAX_FRAMESIZE
+static int32_t decoded[BUFSIZE] IBSS_ATTR;
+
+
+/* This function initialises AVCodecContext with the data needed for the wmapro
+ * decoder to work. The required data is taken from asf_waveformatex_t because that's
+ * what the rockbox asf metadata parser fill/work with. In the future, when the 
+ * codec is being optimised for on-target playback this function should not be needed. */
+static void init_codec_ctx(AVCodecContext *avctx, asf_waveformatex_t *wfx)
+{
+    /* Copy the extra-data */
+    avctx->extradata_size = wfx->datalen;
+    avctx->extradata = (uint8_t *)malloc(wfx->datalen*sizeof(uint8_t));
+    memcpy(avctx->extradata, wfx->data, wfx->datalen*sizeof(uint8_t));
+    
+    avctx->block_align = wfx->blockalign;
+    avctx->sample_rate = wfx->rate;
+    avctx->channels    = wfx->channels;
+    
+}
+
+/* this is the codec entry point */
+enum codec_status codec_main(void)
+{
+    uint32_t elapsedtime;
+    int retval;
+    asf_waveformatex_t wfx;     /* Holds the stream properties */
+    size_t resume_offset;
+    int res;                    /* Return values from asf_read_packet() and decode_packet() */
+    uint8_t* audiobuf;          /* Pointer to the payload of one wma pro packet */
+    int audiobufsize;           /* Payload size */
+    int packetlength = 0;       /* Logical packet size (minus the header size) */          
+    int outlen = 0;             /* Number of bytes written to the output buffer */
+    int pktcnt = 0;             /* Count of the packets played */
+
+    /* Generic codec initialisation */
+    ci->configure(DSP_SET_SAMPLE_DEPTH, 31);
+    
+
+next_track:
+
+    /* Wait for the metadata to be read */
+    while (!*ci->taginfo_ready && !ci->stop_codec)
+        ci->sleep(1);
+
+    retval = CODEC_OK;
+
+    /* Remember the resume position */
+    resume_offset = ci->id3->offset;
+ restart_track:   
+    if (codec_init()) {
+        LOGF("(WMA Voice) Error: Error initialising codec\n");
+        retval = CODEC_ERROR;
+        goto done;
+    }
+
+    /* Copy the format metadata we've stored in the id3 TOC field.  This
+       saves us from parsing it again here. */
+    memcpy(&wfx, ci->id3->toc, sizeof(wfx));
+    memset(&avctx, 0, sizeof(AVCodecContext));
+    memset(&avpkt, 0, sizeof(AVPacket));
+    
+    ci->configure(DSP_SWITCH_FREQUENCY, wfx.rate);
+    ci->configure(DSP_SET_STEREO_MODE, wfx.channels == 1 ?
+                  STEREO_MONO : STEREO_INTERLEAVED);
+    codec_set_replaygain(ci->id3);
+    
+    /* Initialise the AVCodecContext */
+    init_codec_ctx(&avctx, &wfx);
+
+    if (wmavoice_decode_init(&avctx) < 0) {
+        LOGF("(WMA Voice) Error: Unsupported or corrupt file\n");
+        retval = CODEC_ERROR;
+        goto done;
+    }
+
+    /* Now advance the file position to the first frame */
+    ci->seek_buffer(ci->id3->first_frame_offset);
+    
+    elapsedtime = 0;
+    resume_offset = 0;
+    
+    /* The main decoding loop */
+
+    while (pktcnt < wfx.numpackets)
+    {
+        ci->yield();
+        if (ci->stop_codec || ci->new_track) {
+            goto done;
+        }
+        
+        /* Deal with any pending seek requests */
+        if (ci->seek_time){
+
+            if (ci->seek_time == 1) {
+                ci->seek_complete();
+                goto restart_track; /* Pretend you never saw this... */
+            }
+
+            elapsedtime = asf_seek(ci->seek_time, &wfx);
+            if (elapsedtime < 1){
+                ci->seek_complete();
+                goto next_track;
+            }
+
+            ci->set_elapsed(elapsedtime);
+            ci->seek_complete();
+        }
+        
+new_packet:
+        res = asf_read_packet(&audiobuf, &audiobufsize, &packetlength, &wfx);
+
+        if (res < 0) {
+            LOGF("(WMA Voice) read_packet error %d\n",res);
+            goto done;
+        } else {
+            avpkt.data = audiobuf;
+            avpkt.size = audiobufsize;
+            pktcnt++;
+            
+            while(avpkt.size > 0)
+            {
+                /* wmavoice_decode_packet checks for the output buffer size to 
+                   avoid overflows */
+                outlen = BUFSIZE*sizeof(int32_t);
+
+                res = wmavoice_decode_packet(&avctx, decoded, &outlen, &avpkt);
+                if(res < 0) {
+                    LOGF("(WMA Voice) Error: decode_packet returned %d", res);
+                    if(res == ERROR_WMAPRO_IN_WMAVOICE){
+                    /* Just skip this packet */
+                        ci->advance_buffer(packetlength);
+                        goto new_packet;    
+                    }
+                    else
+                        goto done;
+                }
+                avpkt.data += res;
+                avpkt.size -= res;
+                if(outlen) {
+                    ci->yield ();
+                    outlen /= sizeof(int32_t);
+                    ci->pcmbuf_insert(decoded, NULL, outlen);
+                    elapsedtime += outlen*10/(wfx.rate/100);
+                    ci->set_elapsed(elapsedtime);
+                    ci->yield ();
+                }
+            }
+
+        }
+
+        /* Advance to the next logical packet */
+        ci->advance_buffer(packetlength);
+    }
+    retval = CODEC_OK;
+
+done:
+    if (ci->request_next_track())
+        goto next_track;
+    
+    return retval;
+}
+

apps/metadata.c

@@ -182,6 +182,9 @@ const struct afmt_entry audio_formats[AFMT_NUM_CODECS] =
     /* True Audio */
     [AFMT_TTA] =
         AFMT_ENTRY("TTA",  "tta",   NULL,          "tta\0"      ),
+    /* WMA Voice in ASF */
+    [AFMT_WMAVOICE] =
+        AFMT_ENTRY("WMAVoice",  "wmavoice",     NULL,    "wma\0wmv\0asf\0"   ),
 #endif
 };
 
@@ -297,7 +300,7 @@ bool get_metadata(struct mp3entry* id3, int fd, const char* trackname)
         }
 
         break;
-
+    
     case AFMT_WMA:
         if (!get_asf_metadata(fd, id3))
         {

apps/metadata.h

@@ -84,6 +84,7 @@ enum
     AFMT_VOX,          /* VOX */
     AFMT_WAVE64,       /* Wave64 */
     AFMT_TTA,          /* True Audio */
+    AFMT_WMAVOICE,     /* WMA Voice in ASF */
 #endif
 
     /* add new formats at any index above this line to have a sensible order -

apps/metadata/asf.c

@@ -357,6 +357,11 @@ static int asf_parse_header(int fd, struct mp3entry* id3,
                             wfx->audiostream = flags&0x7f;
                             /* Correct codectype to redirect playback to the proper .codec */
                             id3->codectype = AFMT_WMAPRO;
+                        } else if (wfx->codec_id == ASF_CODEC_ID_WMAVOICE) {
+                            read(fd, wfx->data, wfx->datalen);
+                            lseek(fd,current.size - 24 - 72 - wfx->datalen,SEEK_CUR);
+                            wfx->audiostream = flags&0x7f;
+                            id3->codectype = AFMT_WMAVOICE;
                         } else {
                             DEBUGF("Unsupported WMA codec (Lossless, Voice, etc)\n");
                             lseek(fd,current.size - 24 - 72,SEEK_CUR);