From c1d2384e093f2792e6251041bc4c3f38782a59b9 Mon Sep 17 00:00:00 2001
From: paulfd <paulfd@outlook.fr>
Date: Sat, 7 Sep 2019 01:12:20 +0200
Subject: [PATCH] Added an AudioSpan and AudioBuffer, to better handle the
 memory for mono/stereo samples

---
 clients/jack_client.cpp |  16 +++---
 sfizz/AudioBuffer.h     |  72 +++++++++++++++------------
 sfizz/AudioSpan.h       |  67 ++++++++++++++++++-------
 sfizz/Config.h          |   2 +-
 sfizz/FilePool.cpp      |  30 ++++++-----
 sfizz/FilePool.h        |   7 ++-
 sfizz/Region.cpp        |   6 ++-
 sfizz/Region.h          |   6 +--
 sfizz/Synth.cpp         |   6 +--
 sfizz/Synth.h           |   6 +--
 sfizz/Voice.cpp         | 107 ++++++++++++++++++++++++++--------------
 sfizz/Voice.h           |  16 +++---
 tests/FilesT.cpp        |   2 -
 13 files changed, 204 insertions(+), 139 deletions(-)

diff --git a/clients/jack_client.cpp b/clients/jack_client.cpp
index 6e825005..c6bc369e 100644
--- a/clients/jack_client.cpp
+++ b/clients/jack_client.cpp
@@ -21,7 +21,7 @@
 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
-#include "StereoSpan.h"
+#include "AudioSpan.h"
 #include "Synth.h"
 #include <absl/flags/parse.h>
 #include <absl/types/span.h>
@@ -116,13 +116,9 @@ int process(jack_nframes_t numFrames, void* arg [[maybe_unused]])
         }
     }
 
-    StereoSpan<float> output {
-        jack_port_get_buffer(outputPort1, numFrames),
-        jack_port_get_buffer(outputPort2, numFrames), 
-        numFrames
-    };
-
-    synth->renderBlock(output);
+    auto leftOutput = reinterpret_cast<float*>(jack_port_get_buffer(outputPort1, numFrames));
+    auto rightOutput = reinterpret_cast<float*>(jack_port_get_buffer(outputPort2, numFrames));
+    synth->renderBlock({ { leftOutput, rightOutput }, numFrames });
 
     return 0;
 }
@@ -156,7 +152,7 @@ static void done(int sig [[maybe_unused]])
     std::cout << "Signal received" << '\n';
     shouldClose = true;
     // if (client != nullptr)
-        
+
     // exit(0);
 }
 
@@ -259,7 +255,7 @@ int main(int argc, char** argv)
 
     while (!shouldClose)
         sleep(1);
-    
+
     std::cout << "Closing..." << '\n';
     jack_client_close(client);
     return 0;
diff --git a/sfizz/AudioBuffer.h b/sfizz/AudioBuffer.h
index eef3b738..30735f07 100644
--- a/sfizz/AudioBuffer.h
+++ b/sfizz/AudioBuffer.h
@@ -50,6 +50,14 @@ public:
             buffers[i] = std::make_unique<buffer_type>(numFrames);
     }
 
+    bool resize(size_type newSize)
+    {
+        bool returnedOK = true;
+        for (auto i = 0; i < numChannels; ++i)
+            returnedOK &= buffers[i]->resize(newSize);
+        return returnedOK;
+    }
+
     iterator channelWriter(int channelIndex)
     {
         ASSERT(channelIndex < numChannels)
@@ -68,7 +76,7 @@ public:
         return {};
     }
 
-    const_iterator channelReader(int channelIndex)
+    const_iterator channelReader(int channelIndex) const
     {
         ASSERT(channelIndex < numChannels)
         if (channelIndex < numChannels)
@@ -77,7 +85,7 @@ public:
         return {};
     }
 
-    const_iterator channelReaderEnd(int channelIndex)
+    const_iterator channelReaderEnd(int channelIndex) const
     {
         ASSERT(channelIndex < numChannels)
         if (channelIndex < numChannels)
@@ -86,7 +94,7 @@ public:
         return {};
     }
 
-    absl::Span<value_type> getSpan(int channelIndex)
+    absl::Span<value_type> getSpan(int channelIndex) const
     {
         ASSERT(channelIndex < numChannels)
         if (channelIndex < numChannels)
@@ -95,45 +103,45 @@ public:
         return {};
     }
 
-    absl::Span<const value_type> getConstSpan(int channelIndex)
+    absl::Span<const value_type> getConstSpan(int channelIndex) const
     {
         return getSpan(channelIndex);
     }
 
-   	void addChannel()
-   	{
-   		if (numChannels < MaxChannels)
-   			buffers[numChannels++] = std::make_unique<buffer_type>(numFrames);
-   	}
+    void addChannel()
+    {
+        if (numChannels < MaxChannels)
+            buffers[numChannels++] = std::make_unique<buffer_type>(numFrames);
+    }
 
-   	size_type getNumFrames()
-   	{
-   		return numFrames;
-   	}
+    size_type getNumFrames() const
+    {
+        return numFrames;
+    }
 
-	size_type getNumChannels()
-   	{
-   		return numChannels;
-   	}
+    int getNumChannels() const
+    {
+        return numChannels;
+    }
 
-   	bool empty()
-   	{
-   		return numFrames == 0;
-   	}
+    bool empty() const
+    {
+        return numFrames == 0;
+    }
 
-   	Type& getSample(int channelIndex, size_type frameIndex)
-   	{
-   		// Uhoh
-   		ASSERT(buffers[channelIndex] != nullptr);
-   		ASSERT(frameIndex < numFrames);
+    Type& getSample(int channelIndex, size_type frameIndex)
+    {
+        // Uhoh
+        ASSERT(buffers[channelIndex] != nullptr);
+        ASSERT(frameIndex < numFrames);
 
-   		return *(buffers[channelIndex]->data() + frameIndex);
-   	}
+        return *(buffers[channelIndex]->data() + frameIndex);
+    }
 
-   	Type& operator()(int channelIndex, size_type frameIndex)
-   	{
-   		return getSample(channelIndex, frameIndex);
-   	}
+    Type& operator()(int channelIndex, size_type frameIndex)
+    {
+        return getSample(channelIndex, frameIndex);
+    }
 
 private:
     using buffer_type = Buffer<Type, Alignment>;
diff --git a/sfizz/AudioSpan.h b/sfizz/AudioSpan.h
index 4cd6bb71..d796bf46 100644
--- a/sfizz/AudioSpan.h
+++ b/sfizz/AudioSpan.h
@@ -39,37 +39,48 @@ public:
     {
     }
 
+    AudioSpan(const std::array<Type*, MaxChannels>& spans, int numChannels, size_type offset, size_type size)
+        : numFrames(size)
+        , numChannels(numChannels)
+    {
+        ASSERT(static_cast<unsigned int>(numChannels) <= MaxChannels);
+        for (auto i = 0; i < numChannels; ++i)
+            this->spans[i] = spans[i] + offset;
+    }
+
     AudioSpan(std::initializer_list<Type*> spans, size_type numFrames)
         : numFrames(numFrames)
         , numChannels(spans.size())
     {
-    	static_assert(spans.size() <= MaxChannels);
-        for (auto i = 0; i < spans.size(); i++) {
+        ASSERT(spans.size() <= MaxChannels);
+        auto newSpan = spans.begin();
+        auto thisSpan = this->spans.begin();
+        for (; newSpan < spans.end() && thisSpan < this->spans.end(); thisSpan++, newSpan++) {
             // This will not end well...
-            ASSERT(spans[i] != nullptr);
-            this->spans[i] = spans[i];
+            ASSERT(*newSpan != nullptr);
+            *thisSpan = *newSpan;
         }
     }
 
     AudioSpan(std::initializer_list<absl::Span<Type>> spans)
         : numChannels(spans.size())
     {
-    	static_assert(spans.size() <= MaxChannels);
+        ASSERT(spans.size() <= MaxChannels);
         auto size = absl::Span<Type>::npos;
-        for (auto i = 0; i < spans.size(); i++) {
-            // This will not end well...
-            ASSERT(spans[i] != nullptr);
-            this->spans[i] = spans[i].data();
-            size = std::min(size, spans[i].size());
+        auto newSpan = spans.begin();
+        auto thisSpan = this->spans.begin();
+        for (; newSpan < spans.end() && thisSpan < this->spans.end(); thisSpan++, newSpan++) {
+            *thisSpan = newSpan->data();
+            size = std::min(size, newSpan->size());
         }
     }
 
     template <class U, unsigned int N, unsigned int Alignment, typename = std::enable_if<N <= MaxChannels>>
-    AudioSpan(const AudioBuffer<U, N, Alignment>& audioBuffer)
+    AudioSpan(AudioBuffer<U, N, Alignment>& audioBuffer)
         : numFrames(audioBuffer.getNumFrames())
         , numChannels(audioBuffer.getNumChannels())
     {
-        for (auto i = 0; i < N; i++) {
+        for (int i = 0; i < numChannels; i++) {
             if constexpr (std::is_const<Type>::value)
                 this->spans[i] = audioBuffer.channelReader(i);
             else
@@ -82,7 +93,7 @@ public:
         : numFrames(other.getNumFrames())
         , numChannels(other.getNumChannels())
     {
-        for (auto i = 0; i < N; i++) {
+        for (int i = 0; i < numChannels; i++) {
             this->spans[i] = other.getChannel(i);
         }
     }
@@ -105,7 +116,7 @@ public:
         return {};
     }
 
-    absl::Span<const std::remove_cv<Type>> getConstSpan(int channelIndex)
+    absl::Span<const Type> getConstSpan(int channelIndex)
     {
         ASSERT(channelIndex < numChannels);
         if (channelIndex < numChannels)
@@ -117,23 +128,23 @@ public:
     void fill(Type value) noexcept
     {
         for (int i = 0; i < numChannels; ++i)
-            ::fill<Type>({ getChannel(i), numFrames }, value);
+            ::fill<Type>(getSpan(i), value);
     }
 
     void applyGain(absl::Span<const Type> gain) noexcept
     {
         for (int i = 0; i < numChannels; ++i)
-            ::applyGain<Type>({ getChannel(i), numFrames }, gain);
+            ::applyGain<Type>(gain, getSpan(i));
     }
 
     void applyGain(Type gain) noexcept
     {
         for (int i = 0; i < numChannels; ++i)
-            ::applyGain<Type>({ getChannel(i), numFrames }, gain);
+            ::applyGain<Type>(gain, getSpan(i));
     }
 
     template <class U, unsigned int N, typename = std::enable_if<N <= MaxChannels>>
-    void add(const AudioSpan<U, N>& other)
+    void add(AudioSpan<U, N>& other)
     {
         ASSERT(other.getNumChannels() == numChannels);
         if (other.getNumChannels() == numChannels) {
@@ -143,7 +154,7 @@ public:
     }
 
     template <class U, unsigned int N, typename = std::enable_if<N <= MaxChannels>>
-    void copy(const AudioSpan<U, N>& other)
+    void copy(AudioSpan<U, N>& other)
     {
         ASSERT(other.getNumChannels() == numChannels);
         if (other.getNumChannels() == numChannels) {
@@ -162,6 +173,24 @@ public:
         return numChannels;
     }
 
+    AudioSpan<Type> first(size_type length)
+    {
+        ASSERT(length <= numFrames);
+        return { spans, numChannels, 0, length };
+    }
+
+    AudioSpan<Type> last(size_type length)
+    {
+        ASSERT(length <= numFrames);
+        return { spans, numChannels, numFrames - length, length };
+    }
+
+    AudioSpan<Type> subspan(size_type offset, size_type length)
+    {
+        ASSERT(length + offset <= numFrames);
+        return { spans, numChannels, offset, length };
+    }
+
 private:
     std::array<Type*, MaxChannels> spans;
     size_type numFrames { 0 };
diff --git a/sfizz/Config.h b/sfizz/Config.h
index 1bb304d7..b5854847 100644
--- a/sfizz/Config.h
+++ b/sfizz/Config.h
@@ -28,7 +28,7 @@ namespace sfz {
 namespace config {
     constexpr float defaultSampleRate { 48000 };
     constexpr int defaultSamplesPerBlock { 1024 };
-    constexpr int preloadSize { 8192 * 2 };
+    constexpr int preloadSize { 8192 };
     constexpr int numChannels { 2 };
     constexpr int numVoices { 64 };
     constexpr int numLoadingThreads { 4 };
diff --git a/sfizz/FilePool.cpp b/sfizz/FilePool.cpp
index edbab518..6dd709b4 100644
--- a/sfizz/FilePool.cpp
+++ b/sfizz/FilePool.cpp
@@ -22,26 +22,27 @@
 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include "FilePool.h"
+#include "AudioBuffer.h"
 #include "Config.h"
 #include "Debug.h"
 #include "absl/types/span.h"
 #include <chrono>
+#include <memory>
 #include <sndfile.hh>
 using namespace std::chrono_literals;
 
 template <class T>
-void readFromFile(SndfileHandle& sndFile, int numFrames, StereoBuffer<T>& output)
+std::unique_ptr<AudioBuffer<T>> readFromFile(SndfileHandle& sndFile, int numFrames)
 {
+    auto returnedBuffer = std::make_unique<AudioBuffer<T>>(sndFile.channels(), numFrames);
     if (sndFile.channels() == 1) {
-        auto tempReadBuffer = std::make_unique<Buffer<float>>(numFrames);
-        sndFile.readf(tempReadBuffer->data(), numFrames);
-        std::copy(tempReadBuffer->begin(), tempReadBuffer->end(), output.begin(Channel::left));
-        std::copy(tempReadBuffer->begin(), tempReadBuffer->end(), output.begin(Channel::right));
+        sndFile.readf(returnedBuffer->channelWriter(0), numFrames);
     } else if (sndFile.channels() == 2) {
-        auto tempReadBuffer = std::make_unique<Buffer<float>>(2 * numFrames);
-        sndFile.readf(tempReadBuffer->data(), numFrames);
-        output.readInterleaved(*tempReadBuffer);
+        auto tempReadBuffer = std::make_unique<AudioBuffer<float>>(1, 2 * numFrames);
+        sndFile.readf(tempReadBuffer->channelWriter(0), numFrames);
+        ::readInterleaved<float>(tempReadBuffer->getSpan(0), returnedBuffer->getSpan(0), returnedBuffer->getSpan(1));
     }
+    return returnedBuffer;
 }
 
 std::optional<sfz::FilePool::FileInformation> sfz::FilePool::getFileInformation(std::string_view filename) noexcept
@@ -52,12 +53,11 @@ std::optional<sfz::FilePool::FileInformation> sfz::FilePool::getFileInformation(
 
     SndfileHandle sndFile(reinterpret_cast<const char*>(file.c_str()));
     if (sndFile.channels() != 1 && sndFile.channels() != 2) {
-        DBG("Missing logic for " << sndFile.channels() << ", discarding sample " << filename);
+        DBG("Missing logic for " << sndFile.channels() << " channels, discarding sample " << filename);
         return {};
     }
 
     FileInformation returnedValue;
-    returnedValue.numChannels = sndFile.channels();
     returnedValue.end = static_cast<uint32_t>(sndFile.frames());
     returnedValue.sampleRate = static_cast<double>(sndFile.samplerate());
 
@@ -78,8 +78,7 @@ std::optional<sfz::FilePool::FileInformation> sfz::FilePool::getFileInformation(
     if (preloadedData.contains(filename)) {
         returnedValue.preloadedData = preloadedData[filename];
     } else {
-        returnedValue.preloadedData = std::make_shared<StereoBuffer<float>>(preloadedSize);
-        readFromFile(sndFile, preloadedSize, *returnedValue.preloadedData);
+        returnedValue.preloadedData = std::shared_ptr<AudioBuffer<float>>(readFromFile<float>(sndFile, preloadedSize));
         preloadedData[filename] = returnedValue.preloadedData;
     }
 
@@ -97,7 +96,7 @@ void sfz::FilePool::loadingThread() noexcept
 {
     FileLoadingInformation fileToLoad {};
     while (!quitThread) {
-        if (!loadingQueue.wait_dequeue_timed(fileToLoad, 1ms))
+        if (!loadingQueue.wait_dequeue_timed(fileToLoad, 100ms))
             continue;
 
         if (fileToLoad.voice == nullptr) {
@@ -113,8 +112,7 @@ void sfz::FilePool::loadingThread() noexcept
         }
 
         SndfileHandle sndFile(reinterpret_cast<const char*>(file.c_str()));
-        auto fileLoaded = std::make_unique<StereoBuffer<float>>(fileToLoad.numFrames);
-        readFromFile(sndFile, fileToLoad.numFrames, *fileLoaded);
-        fileToLoad.voice->setFileData(std::move(fileLoaded));
+        auto fileLoaded = std::make_unique<AudioBuffer<float>>(sndFile.channels(), fileToLoad.numFrames);
+        fileToLoad.voice->setFileData(readFromFile<float>(sndFile, fileToLoad.numFrames));
     }
 }
\ No newline at end of file
diff --git a/sfizz/FilePool.h b/sfizz/FilePool.h
index bb4912b6..f27bb870 100644
--- a/sfizz/FilePool.h
+++ b/sfizz/FilePool.h
@@ -24,7 +24,7 @@
 #pragma once
 #include "Defaults.h"
 #include "LeakDetector.h"
-#include "StereoBuffer.h"
+#include "AudioBuffer.h"
 #include "Voice.h"
 #include "readerwriterqueue.h"
 #include <absl/container/flat_hash_map.h>
@@ -50,12 +50,11 @@ public:
     size_t getNumPreloadedSamples() const noexcept { return preloadedData.size(); }
 
     struct FileInformation {
-        int numChannels { 1 };
         uint32_t end { Default::sampleEndRange.getEnd() };
         uint32_t loopBegin { Default::loopRange.getStart() };
         uint32_t loopEnd { Default::loopRange.getEnd() };
         double sampleRate { config::defaultSampleRate };
-        std::shared_ptr<StereoBuffer<float>> preloadedData;
+        std::shared_ptr<AudioBuffer<float>> preloadedData;
     };
     std::optional<FileInformation> getFileInformation(std::string_view filename) noexcept;
     void enqueueLoading(Voice* voice, std::string_view sample, int numFrames) noexcept;
@@ -71,7 +70,7 @@ private:
     void loadingThread() noexcept;
     std::thread fileLoadingThread;
     bool quitThread { false };
-    absl::flat_hash_map<std::string_view, std::shared_ptr<StereoBuffer<float>>> preloadedData;
+    absl::flat_hash_map<std::string_view, std::shared_ptr<AudioBuffer<float>>> preloadedData;
     LEAK_DETECTOR(FilePool);
 };
 }
\ No newline at end of file
diff --git a/sfizz/Region.cpp b/sfizz/Region.cpp
index 67e11e2b..02df3818 100644
--- a/sfizz/Region.cpp
+++ b/sfizz/Region.cpp
@@ -22,6 +22,7 @@
 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include "Region.h"
+#include "MathHelpers.h"
 #include "Debug.h"
 #include "StringViewHelpers.h"
 #include "absl/strings/str_replace.h"
@@ -651,7 +652,10 @@ bool sfz::Region::canUsePreloadedData() const noexcept
 
 bool sfz::Region::isStereo() const noexcept
 {
-    return this->numChannels == 2;
+    if (isGenerator())
+        return 1;
+
+    return (this->preloadedData->getNumChannels() == 2);
 }
 
 template<class T, class U>
diff --git a/sfizz/Region.h b/sfizz/Region.h
index 567364b7..b4bec2dd 100644
--- a/sfizz/Region.h
+++ b/sfizz/Region.h
@@ -27,7 +27,7 @@
 #include "Defaults.h"
 #include "EGDescription.h"
 #include "Opcode.h"
-#include "StereoBuffer.h"
+#include "AudioBuffer.h"
 #include <bitset>
 #include <optional>
 #include <random>
@@ -148,9 +148,7 @@ struct Region {
     EGDescription filterEG;
 
     double sampleRate { config::defaultSampleRate };
-    int numChannels { 1 };
-    std::shared_ptr<StereoBuffer<float>> preloadedData { nullptr };
-
+    std::shared_ptr<AudioBuffer<float>> preloadedData { nullptr };
 private:
     bool keySwitched { true };
     bool previousKeySwitched { true };
diff --git a/sfizz/Synth.cpp b/sfizz/Synth.cpp
index 79f9c751..976aa4e8 100644
--- a/sfizz/Synth.cpp
+++ b/sfizz/Synth.cpp
@@ -197,8 +197,6 @@ bool sfz::Synth::loadSfzFile(const std::filesystem::path& filename)
                 lastRegion--;
                 continue;
             }
-
-            region->numChannels = fileInformation->numChannels;
             region->sampleEnd = std::min(region->sampleEnd, fileInformation->end);
             region->loopRange.shrinkIfSmaller(fileInformation->loopBegin, fileInformation->loopEnd);
             region->preloadedData = fileInformation->preloadedData;
@@ -274,11 +272,11 @@ void sfz::Synth::setSampleRate(float sampleRate) noexcept
         voice->setSampleRate(sampleRate);
 }
 
-void sfz::Synth::renderBlock(StereoSpan<float> buffer) noexcept
+void sfz::Synth::renderBlock(AudioSpan<float> buffer) noexcept
 {
     ScopedFTZ ftz;
     buffer.fill(0.0f);
-    StereoSpan<float> tempSpan { tempBuffer, buffer.size() };
+    auto tempSpan = AudioSpan<float>(tempBuffer).first(buffer.getNumFrames());
     for (auto& voice : voices) {
         voice->renderBlock(tempSpan);
         buffer.add(tempSpan);
diff --git a/sfizz/Synth.h b/sfizz/Synth.h
index f0e82832..7c7f58d9 100644
--- a/sfizz/Synth.h
+++ b/sfizz/Synth.h
@@ -26,7 +26,7 @@
 #include "Parser.h"
 #include "Region.h"
 #include "LeakDetector.h"
-#include "StereoSpan.h"
+#include "AudioSpan.h"
 #include "absl/types/span.h"
 #include <optional>
 #include <random>
@@ -51,7 +51,7 @@ public:
 
     void setSamplesPerBlock(int samplesPerBlock) noexcept;
     void setSampleRate(float sampleRate) noexcept;
-    void renderBlock(StereoSpan<float> buffer) noexcept;
+    void renderBlock(AudioSpan<float> buffer) noexcept;
     void noteOn(int delay, int channel, int noteNumber, uint8_t velocity) noexcept;
     void noteOff(int delay, int channel, int noteNumber, uint8_t velocity) noexcept;
     void cc(int delay, int channel, int ccNumber, uint8_t ccValue) noexcept;
@@ -91,7 +91,7 @@ private:
     std::array<RegionPtrVector, 128> noteActivationLists;
     std::array<RegionPtrVector, 128> ccActivationLists;
 
-    StereoBuffer<float> tempBuffer { config::defaultSamplesPerBlock };
+    AudioBuffer<float> tempBuffer { 2, config::defaultSamplesPerBlock };
     int samplesPerBlock { config::defaultSamplesPerBlock };
     float sampleRate { config::defaultSampleRate };
 
diff --git a/sfizz/Voice.cpp b/sfizz/Voice.cpp
index d235150c..77480901 100644
--- a/sfizz/Voice.cpp
+++ b/sfizz/Voice.cpp
@@ -57,7 +57,7 @@ void sfz::Voice::startVoice(Region* region, int delay, int channel, int number,
     if (region->amplitudeCC)
         baseGain *= normalizeCC(ccState[region->amplitudeCC->first]) * normalizePercents(region->amplitudeCC->second);
     amplitudeEnvelope.reset(baseGain);
-    
+
     sourcePosition = region->getOffset();
     initialDelay = delay + region->getDelay();
     baseFrequency = midiNoteFrequency(number) * pitchRatio;
@@ -80,8 +80,9 @@ void sfz::Voice::prepareEGEnvelope(int delay, uint8_t velocity) noexcept
         normalizePercents(region->amplitudeEG.getStart(ccState, velocity)));
 }
 
-void sfz::Voice::setFileData(std::unique_ptr<StereoBuffer<float>> file) noexcept
+void sfz::Voice::setFileData(std::unique_ptr<AudioBuffer<float>> file) noexcept
 {
+    // DBG("File data set for sample " << region->sample);
     fileData = std::move(file);
     dataReady.store(true);
 }
@@ -161,10 +162,9 @@ void sfz::Voice::setSamplesPerBlock(int samplesPerBlock) noexcept
     indexSpan = absl::MakeSpan(indexBuffer);
 }
 
-void sfz::Voice::renderBlock(StereoSpan<float> buffer) noexcept
+void sfz::Voice::renderBlock(AudioSpan<float> buffer) noexcept
 {
-    const auto numSamples = buffer.size();
-    ASSERT(static_cast<int>(numSamples) <= samplesPerBlock);
+    ASSERT(static_cast<int>(buffer.getNumFrames()) <= samplesPerBlock);
     buffer.fill(0.0f);
 
     if (state == State::idle || region == nullptr)
@@ -175,34 +175,59 @@ void sfz::Voice::renderBlock(StereoSpan<float> buffer) noexcept
     else
         fillWithData(buffer);
 
-    auto envelopeSpan = tempSpan1.first(numSamples);
-    amplitudeEnvelope.getBlock(envelopeSpan);
-    buffer.applyGain(envelopeSpan);
-
-    egEnvelope.getBlock(envelopeSpan);
-    buffer.applyGain(envelopeSpan);
-
+    if (region->isStereo())
+        processStereo(buffer);
+    else
+        processMono(buffer);
+    
     if (!egEnvelope.isSmoothing())
         reset();
 }
 
-void sfz::Voice::fillWithData(StereoSpan<float> buffer) noexcept
+void sfz::Voice::processMono(AudioSpan<float> buffer) noexcept
+{
+    const auto numSamples = buffer.getNumFrames();
+    auto leftBuffer = buffer.getSpan(0);
+    auto rightBuffer = buffer.getSpan(1);
+    
+    auto envelopeSpan = tempSpan1.first(numSamples);
+    amplitudeEnvelope.getBlock(envelopeSpan);
+    ::applyGain<float>(envelopeSpan, leftBuffer);
+
+    egEnvelope.getBlock(envelopeSpan);
+    ::applyGain<float>(envelopeSpan, leftBuffer);
+
+    ::copy<float>(leftBuffer, rightBuffer);
+}
+void sfz::Voice::processStereo(AudioSpan<float> buffer) noexcept
+{
+    const auto numSamples = buffer.getNumFrames();
+    auto envelopeSpan = tempSpan1.first(numSamples);
+   
+    amplitudeEnvelope.getBlock(envelopeSpan);
+    buffer.applyGain(envelopeSpan);
+    
+    egEnvelope.getBlock(envelopeSpan);
+    buffer.applyGain(envelopeSpan);
+}
+
+void sfz::Voice::fillWithData(AudioSpan<float> buffer) noexcept
 {
     auto source { [&]() {
         if (region->canUsePreloadedData() || !dataReady)
-            return StereoSpan<const float>(*region->preloadedData);
+            return AudioSpan<const float>(*region->preloadedData);
         else
-            return StereoSpan<const float>(*fileData);
+            return AudioSpan<const float>(*fileData);
     }() };
 
-    auto indices = indexSpan.first(buffer.size());
-    auto jumps = tempSpan1.first(buffer.size());
-    auto leftCoeffs = tempSpan1.first(buffer.size());
-    auto rightCoeffs = tempSpan2.first(buffer.size());
+    auto indices = indexSpan.first(buffer.getNumFrames());
+    auto jumps = tempSpan1.first(buffer.getNumFrames());
+    auto leftCoeffs = tempSpan1.first(buffer.getNumFrames());
+    auto rightCoeffs = tempSpan2.first(buffer.getNumFrames());
 
     ::fill<float>(jumps, pitchRatio * speedRatio);
 
-    if (region->shouldLoop() && region->trueSampleEnd() <= source.size()) {
+    if (region->shouldLoop() && region->trueSampleEnd() <= source.getNumFrames()) {
         floatPosition = ::loopingSFZIndex<float, false>(
             jumps,
             leftCoeffs,
@@ -218,32 +243,42 @@ void sfz::Voice::fillWithData(StereoSpan<float> buffer) noexcept
             rightCoeffs,
             indices,
             floatPosition,
-            source.size() - 1);
+            source.getNumFrames() - 1);
     }
 
     auto ind = indices.data();
     auto leftCoeff = leftCoeffs.data();
     auto rightCoeff = rightCoeffs.data();
-    auto left = buffer.left().data();
-    auto right = buffer.right().data();
-    while (ind < indices.end()) {
-        *left = source.left()[*ind] * (*leftCoeff) + source.left()[*ind + 1] * (*rightCoeff);
-        *right = source.right()[*ind] * (*leftCoeff) + source.right()[*ind + 1] * (*rightCoeff);
-        left++;
-        right++;
-        ind++;
-        leftCoeff++;
-        rightCoeff++;
+    auto left = buffer.getChannel(0);
+    if (source.getNumChannels() == 1) {
+        while (ind < indices.end()) {
+            *left = source.getChannel(0)[*ind] * (*leftCoeff) + source.getChannel(0)[*ind + 1] * (*rightCoeff);
+            left++;
+            ind++;
+            leftCoeff++;
+            rightCoeff++;
+        }
+    } else {
+        auto right = buffer.getChannel(1);
+        while (ind < indices.end()) {
+            *left = source.getChannel(0)[*ind] * (*leftCoeff) + source.getChannel(0)[*ind + 1] * (*rightCoeff);
+            *right = source.getChannel(1)[*ind] * (*leftCoeff) + source.getChannel(1)[*ind + 1] * (*rightCoeff);
+            left++;
+            right++;
+            ind++;
+            leftCoeff++;
+            rightCoeff++;
+        }
     }
 
-    if (!region->shouldLoop() && (floatPosition + 1.01) > source.size()) {
+    if (!region->shouldLoop() && (floatPosition + 1.01) > source.getNumFrames()) {
         DBG("Releasing " << region->sample);
         auto last = std::distance(indices.begin(), absl::c_find(indices, region->trueSampleEnd() - 1));
         release(last);
     }
 }
 
-void sfz::Voice::fillWithGenerator(StereoSpan<float> buffer) noexcept
+void sfz::Voice::fillWithGenerator(AudioSpan<float> buffer) noexcept
 {
     if (region->sample != "*sine")
         return;
@@ -251,10 +286,10 @@ void sfz::Voice::fillWithGenerator(StereoSpan<float> buffer) noexcept
     float step = baseFrequency * twoPi<float> / sampleRate;
     phase = ::linearRamp<float>(tempSpan1, phase, step);
 
-    ::sin<float>(tempSpan1.first(buffer.size()), buffer.left());
-    absl::c_copy(buffer.left(), buffer.right().begin());
+    ::sin<float>(tempSpan1.first(buffer.getNumFrames()), buffer.getSpan(0));
+    ::copy<float>(buffer.getSpan(0), buffer.getSpan(1));
 
-    sourcePosition += buffer.size();
+    sourcePosition += buffer.getNumFrames();
 }
 
 bool sfz::Voice::checkOffGroup(int delay, uint32_t group) noexcept
diff --git a/sfizz/Voice.h b/sfizz/Voice.h
index 9f940bc2..da4e1ab7 100644
--- a/sfizz/Voice.h
+++ b/sfizz/Voice.h
@@ -26,8 +26,8 @@
 #include "Config.h"
 #include "LinearEnvelope.h"
 #include "Region.h"
-#include "StereoBuffer.h"
-#include "StereoSpan.h"
+#include "AudioBuffer.h"
+#include "AudioSpan.h"
 #include "LeakDetector.h"
 #include <absl/types/span.h>
 #include <atomic>
@@ -47,7 +47,7 @@ public:
     
     void startVoice(Region* region, int delay, int channel, int number, uint8_t value, TriggerType triggerType) noexcept;
 
-    void setFileData(std::unique_ptr<StereoBuffer<float>> file) noexcept;
+    void setFileData(std::unique_ptr<AudioBuffer<float>> file) noexcept;
     void registerNoteOff(int delay, int channel, int noteNumber, uint8_t velocity) noexcept;
     void registerCC(int delay, int channel, int ccNumber, uint8_t ccValue) noexcept;
     void registerPitchWheel(int delay, int channel, int pitch) noexcept;
@@ -55,7 +55,7 @@ public:
     void registerTempo(int delay, float secondsPerQuarter) noexcept;
     bool checkOffGroup(int delay, uint32_t group) noexcept;
 
-    void renderBlock(StereoSpan<float> buffer) noexcept;
+    void renderBlock(AudioSpan<float, 2> buffer) noexcept;
 
     bool isFree() const noexcept;
     int getTriggerNumber() const noexcept;
@@ -66,9 +66,11 @@ public:
     void reset() noexcept;
     void garbageCollect() noexcept;
 private:
-    void fillWithData(StereoSpan<float> buffer) noexcept;
-    void fillWithGenerator(StereoSpan<float> buffer) noexcept;
+    void fillWithData(AudioSpan<float> buffer) noexcept;
+    void fillWithGenerator(AudioSpan<float> buffer) noexcept;
     void prepareEGEnvelope(int delay, uint8_t velocity) noexcept;
+    void processMono(AudioSpan<float> buffer) noexcept;
+    void processStereo(AudioSpan<float> buffer) noexcept;
     void release(int delay) noexcept;
     Region* region { nullptr };
 
@@ -96,7 +98,7 @@ private:
     uint32_t initialDelay { 0 };
 
     std::atomic<bool> dataReady { false };
-    std::unique_ptr<StereoBuffer<float>> fileData { nullptr };
+    std::unique_ptr<AudioBuffer<float>> fileData { nullptr };
 
     Buffer<float> tempBuffer1;
     Buffer<float> tempBuffer2;
diff --git a/tests/FilesT.cpp b/tests/FilesT.cpp
index c43de0fb..a68f3b26 100644
--- a/tests/FilesT.cpp
+++ b/tests/FilesT.cpp
@@ -257,10 +257,8 @@ TEST_CASE("[Files] Channels (channels.sfz)")
     synth.loadSfzFile(std::filesystem::current_path() / "tests/TestFiles/channels.sfz");
     REQUIRE(synth.getNumRegions() == 2);
     REQUIRE(synth.getRegionView(0)->sample == "mono_sample.wav");
-    REQUIRE(synth.getRegionView(0)->numChannels == 1);
     REQUIRE(!synth.getRegionView(0)->isStereo());
     REQUIRE(synth.getRegionView(1)->sample == "stereo_sample.wav");
-    REQUIRE(synth.getRegionView(1)->numChannels == 2);
     REQUIRE(synth.getRegionView(1)->isStereo());
 }