aboutsummaryrefslogtreecommitdiffstats
path: root/alc
diff options
context:
space:
mode:
authorChris Robinson <[email protected]>2022-05-28 06:17:53 -0700
committerChris Robinson <[email protected]>2022-05-28 06:56:41 -0700
commitd7367aeee04764ad7595ef3f5de8e01b7a6d93ad (patch)
tree65c0bb2be6e40d16d7f83e3ec1412d5e08f15db3 /alc
parentecb444e1e02bd966626445ab201024b942981186 (diff)
Fix requeueing logic for OpenSLCapture::captureSamples
The write vector represents the chunks queued for OpenSL, and the write pointer only increments when OpenSL fills in more samples. So requeueing from the start of the write vector is requeueing chunks that are already queued, which is obviously wrong. It instead needs to queue chunks that become available as they're read. Because the ring buffer holds more elements than can be written to at a givem time, the read vector does not represent the next writable chunks to queue. Advancing the read pointer increases the number of writable elements, so instead of queueing from the read vector or the start of the write vector, queue from the end of the write vector given the number of chunks read.
Diffstat (limited to 'alc')
-rw-r--r--alc/backends/opensl.cpp46
1 files changed, 29 insertions, 17 deletions
diff --git a/alc/backends/opensl.cpp b/alc/backends/opensl.cpp
index 3a2100b5..49e5c268 100644
--- a/alc/backends/opensl.cpp
+++ b/alc/backends/opensl.cpp
@@ -881,7 +881,6 @@ void OpenSLCapture::captureSamples(al::byte *buffer, uint samples)
{
const uint update_size{mDevice->UpdateSize};
const uint chunk_size{update_size * mFrameSize};
- const auto silence = (mDevice->FmtType == DevFmtUByte) ? al::byte{0x80} : al::byte{0};
/* Read the desired samples from the ring buffer then advance its read
* pointer.
@@ -910,39 +909,52 @@ void OpenSLCapture::captureSamples(al::byte *buffer, uint samples)
i += rem;
}
- mRing->readAdvance(adv_count);
SLAndroidSimpleBufferQueueItf bufferQueue{};
- if LIKELY(mDevice->Connected.load(std::memory_order_acquire))
+ if(likely(mDevice->Connected.load(std::memory_order_acquire)))
{
const SLresult result{VCALL(mRecordObj,GetInterface)(SL_IID_ANDROIDSIMPLEBUFFERQUEUE,
&bufferQueue)};
PRINTERR(result, "recordObj->GetInterface");
- if UNLIKELY(SL_RESULT_SUCCESS != result)
+ if(unlikely(SL_RESULT_SUCCESS != result))
{
mDevice->handleDisconnect("Failed to get capture buffer queue: 0x%08x", result);
bufferQueue = nullptr;
}
}
+ if(unlikely(!bufferQueue) || adv_count == 0)
+ return;
+
+ /* For each buffer chunk that was fully read, queue another writable buffer
+ * chunk to keep the OpenSL queue full. This is rather convulated, as a
+ * result of the ring buffer holding more elements than are writable at a
+ * given time. The end of the write vector increments when the read pointer
+ * advances, which will "expose" a previously unwritable element. So for
+ * every element that we've finished reading, we queue that many elements
+ * from the end of the write vector.
+ */
+ mRing->readAdvance(adv_count);
- if LIKELY(bufferQueue)
+ SLresult result{SL_RESULT_SUCCESS};
+ auto wdata = mRing->getWriteVector();
+ if(likely(adv_count > wdata.second.len))
{
- SLresult result{SL_RESULT_SUCCESS};
- auto wdata = mRing->getWriteVector();
- std::fill_n(wdata.first.buf, wdata.first.len*chunk_size, silence);
- for(size_t i{0u};i < wdata.first.len && SL_RESULT_SUCCESS == result;i++)
+ auto len1 = std::min(wdata.first.len, adv_count-wdata.second.len);
+ auto buf1 = wdata.first.buf + chunk_size*(wdata.first.len-len1);
+ for(size_t i{0u};i < len1 && SL_RESULT_SUCCESS == result;i++)
{
- result = VCALL(bufferQueue,Enqueue)(wdata.first.buf + chunk_size*i, chunk_size);
+ result = VCALL(bufferQueue,Enqueue)(buf1 + chunk_size*i, chunk_size);
PRINTERR(result, "bufferQueue->Enqueue");
}
- if(wdata.second.len > 0)
+ }
+ if(wdata.second.len > 0)
+ {
+ auto len2 = std::min(wdata.second.len, adv_count);
+ auto buf2 = wdata.second.buf + chunk_size*(wdata.second.len-len2);
+ for(size_t i{0u};i < len2 && SL_RESULT_SUCCESS == result;i++)
{
- std::fill_n(wdata.second.buf, wdata.second.len*chunk_size, silence);
- for(size_t i{0u};i < wdata.second.len && SL_RESULT_SUCCESS == result;i++)
- {
- result = VCALL(bufferQueue,Enqueue)(wdata.second.buf + chunk_size*i, chunk_size);
- PRINTERR(result, "bufferQueue->Enqueue");
- }
+ result = VCALL(bufferQueue,Enqueue)(buf2 + chunk_size*i, chunk_size);
+ PRINTERR(result, "bufferQueue->Enqueue");
}
}
}