Change AudioTrack resampling buffers from 3 to 2

Move computation of minimum AudioTrack buffer size to server
for normal streaming PCM tracks.

Use server-side computation to exactly determine requirements
for the resampler to avoid triple buffering.

This reduces latency for normal audio tracks that require resampling,
and makes things consistent with the minimum buffer size.

Change-Id: I2f2ca0e599ee20e16559bc5c5dab61ed100da16c

1 files changed, 23 insertions, 59 deletions

diff --git a/media/libmedia/AudioTrack.cpp b/media/libmedia/AudioTrack.cpp
index d4bacc0..1d5fc95 100644
--- a/media/libmedia/AudioTrack.cpp
+++ b/media/libmedia/AudioTrack.cpp
@@ -66,12 +66,11 @@ status_t AudioTrack::getMinFrameCount(
         return BAD_VALUE;
     }
 
-    // FIXME merge with similar code in createTrack_l(), except we're missing
-    //       some information here that is available in createTrack_l():
+    // FIXME handle in server, like createTrack_l(), possible missing info:
     //          audio_io_handle_t output
     //          audio_format_t format
     //          audio_channel_mask_t channelMask
-    //          audio_output_flags_t flags
+    //          audio_output_flags_t flags (FAST)
     uint32_t afSampleRate;
     status_t status;
     status = AudioSystem::getOutputSamplingRate(&afSampleRate, streamType);
@@ -101,16 +100,16 @@ status_t AudioTrack::getMinFrameCount(
         minBufCount = 2;
     }
 
-    *frameCount = (sampleRate == 0) ? afFrameCount * minBufCount :
-            afFrameCount * minBufCount * uint64_t(sampleRate) / afSampleRate;
-    // The formula above should always produce a non-zero value, but return an error
-    // in the unlikely event that it does not, as that's part of the API contract.
+    *frameCount = minBufCount * sourceFramesNeeded(sampleRate, afFrameCount, afSampleRate);
+    // The formula above should always produce a non-zero value under normal circumstances:
+    // AudioTrack.SAMPLE_RATE_HZ_MIN <= sampleRate <= AudioTrack.SAMPLE_RATE_HZ_MAX.
+    // Return error in the unlikely event that it does not, as that's part of the API contract.
     if (*frameCount == 0) {
-        ALOGE("AudioTrack::getMinFrameCount failed for streamType %d, sampleRate %d",
+        ALOGE("AudioTrack::getMinFrameCount failed for streamType %d, sampleRate %u",
                 streamType, sampleRate);
         return BAD_VALUE;
     }
-    ALOGV("getMinFrameCount=%zu: afFrameCount=%zu, minBufCount=%d, afSampleRate=%d, afLatency=%d",
+    ALOGV("getMinFrameCount=%zu: afFrameCount=%zu, minBufCount=%u, afSampleRate=%u, afLatency=%u",
             *frameCount, afFrameCount, minBufCount, afSampleRate, afLatency);
     return NO_ERROR;
 }
@@ -1015,11 +1014,9 @@ status_t AudioTrack::createTrack_l()
     // The client's AudioTrack buffer is divided into n parts for purpose of wakeup by server, where
     //  n = 1   fast track with single buffering; nBuffering is ignored
     //  n = 2   fast track with double buffering
-    //  n = 2   normal track, no sample rate conversion
-    //  n = 3   normal track, with sample rate conversion
-    //          (pessimistic; some non-1:1 conversion ratios don't actually need triple-buffering)
-    //  n > 3   very high latency or very small notification interval; nBuffering is ignored
-    const uint32_t nBuffering = (mSampleRate == afSampleRate) ? 2 : 3;
+    //  n = 2   normal track, (including those with sample rate conversion)
+    //  n >= 3  very high latency or very small notification interval (unused).
+    const uint32_t nBuffering = 2;
 
     mNotificationFramesAct = mNotificationFramesReq;
 
@@ -1060,39 +1057,9 @@ status_t AudioTrack::createTrack_l()
         // But when initializing a shared buffer AudioTrack via set(),
         // there _is_ a frameCount parameter.  We silently ignore it.
         frameCount = mSharedBuffer->size() / mFrameSize;
-
-    } else if (!(mFlags & AUDIO_OUTPUT_FLAG_FAST)) {
-
-        // FIXME move these calculations and associated checks to server
-
-        // Ensure that buffer depth covers at least audio hardware latency
-        uint32_t minBufCount = afLatency / ((1000 * afFrameCount)/afSampleRate);
-        ALOGV("afFrameCount=%zu, minBufCount=%d, afSampleRate=%u, afLatency=%d",
-                afFrameCount, minBufCount, afSampleRate, afLatency);
-        if (minBufCount <= nBuffering) {
-            minBufCount = nBuffering;
-        }
-
-        size_t minFrameCount = afFrameCount * minBufCount * uint64_t(mSampleRate) / afSampleRate;
-        ALOGV("minFrameCount: %zu, afFrameCount=%zu, minBufCount=%d, sampleRate=%u, afSampleRate=%u"
-                ", afLatency=%d",
-                minFrameCount, afFrameCount, minBufCount, mSampleRate, afSampleRate, afLatency);
-
-        if (frameCount == 0) {
-            frameCount = minFrameCount;
-        } else if (frameCount < minFrameCount) {
-            // not ALOGW because it happens all the time when playing key clicks over A2DP
-            ALOGV("Minimum buffer size corrected from %zu to %zu",
-                     frameCount, minFrameCount);
-            frameCount = minFrameCount;
-        }
-        // Make sure that application is notified with sufficient margin before underrun
-        if (mNotificationFramesAct == 0 || mNotificationFramesAct > frameCount/nBuffering) {
-            mNotificationFramesAct = frameCount/nBuffering;
-        }
-
     } else {
-        // For fast tracks, the frame count calculations and checks are done by server
+        // For fast and normal streaming tracks,
+        // the frame count calculations and checks are done by server
     }
 
     IAudioFlinger::track_flags_t trackFlags = IAudioFlinger::TRACK_DEFAULT;
@@ -1175,23 +1142,10 @@ status_t AudioTrack::createTrack_l()
         if (trackFlags & IAudioFlinger::TRACK_FAST) {
             ALOGV("AUDIO_OUTPUT_FLAG_FAST successful; frameCount %zu", frameCount);
             mAwaitBoost = true;
-            if (mSharedBuffer == 0) {
-                // Theoretically double-buffering is not required for fast tracks,
-                // due to tighter scheduling.  But in practice, to accommodate kernels with
-                // scheduling jitter, and apps with computation jitter, we use double-buffering.
-                if (mNotificationFramesAct == 0 || mNotificationFramesAct > frameCount/nBuffering) {
-                    mNotificationFramesAct = frameCount/nBuffering;
-                }
-            }
         } else {
             ALOGV("AUDIO_OUTPUT_FLAG_FAST denied by server; frameCount %zu", frameCount);
             // once denied, do not request again if IAudioTrack is re-created
             mFlags = (audio_output_flags_t) (mFlags & ~AUDIO_OUTPUT_FLAG_FAST);
-            if (mSharedBuffer == 0) {
-                if (mNotificationFramesAct == 0 || mNotificationFramesAct > frameCount/nBuffering) {
-                    mNotificationFramesAct = frameCount/nBuffering;
-                }
-            }
         }
     }
     if (mFlags & AUDIO_OUTPUT_FLAG_COMPRESS_OFFLOAD) {
@@ -1214,6 +1168,16 @@ status_t AudioTrack::createTrack_l()
             //return NO_INIT;
         }
     }
+    // Make sure that application is notified with sufficient margin before underrun
+    if (mSharedBuffer == 0 && audio_is_linear_pcm(mFormat)) {
+        // Theoretically double-buffering is not required for fast tracks,
+        // due to tighter scheduling.  But in practice, to accommodate kernels with
+        // scheduling jitter, and apps with computation jitter, we use double-buffering
+        // for fast tracks just like normal streaming tracks.
+        if (mNotificationFramesAct == 0 || mNotificationFramesAct > frameCount / nBuffering) {
+            mNotificationFramesAct = frameCount / nBuffering;
+        }
+    }
 
     // We retain a copy of the I/O handle, but don't own the reference
     mOutput = output;