Change audio SHM buffer offsets to be in bytes

CLAP allows mixed precision ports. For instance, the main IO ports can
be 32-bit only, while some CV ports allow 64-bit audio. The plugin only
knows the sample format it receives when the process call happens, so
we'll always need to allocate enough space for that to work.

src/common/audio-shm.h

@@ -75,17 +75,22 @@ class AudioShmBuffer {
         uint32_t size;
 
         /**
-         * Offsets **in samples** within the shared memory object for an input
+         * Offsets **in bytes** within the shared memory object for an input
          * audio channel, indexed by `[bus][channel]`. For VST2 plugins the bus
          * will always be 0. This can be used later to retrieve a pointer to the
          * audio channel.
+         *
+         * The offsets are in bytes rather than in samples to accommodate mixed
+         * 32-bit and 64-bit IO. This is allowed in CLAP. If a port can do
+         * either 32-bit or 64-bit audio, we'll reserve enough space for 64-bit
+         * samples and then just not use the latter half if the host ends up
+         * using 32-bit samples.
          */
         std::vector<std::vector<uint32_t>> input_offsets;
         /**
-         * Offsets **in samples** within the shared memory object for an output
-         * audio channel, indexed by `[bus][channel]`. For VST2 plugins the bus
-         * will always be 0. This can be used later to retrieve a pointer to the
-         * audio channel.
+         * Offsets **in bytes** within the shared memory object for an output
+         * audio channel, indexed by `[bus][channel]`. See `input_offsets` for
+         * more details.
          */
         std::vector<std::vector<uint32_t>> output_offsets;
 
@@ -151,16 +156,16 @@ class AudioShmBuffer {
     template <typename T>
     // NOLINTNEXTLINE(bugprone-easily-swappable-parameters)
     T* input_channel_ptr(const uint32_t bus, const uint32_t channel) noexcept {
-        return reinterpret_cast<T*>(shm_bytes_) +
-               config_.input_offsets[bus][channel];
+        return reinterpret_cast<T*>(shm_bytes_ +
+                                    config_.input_offsets[bus][channel]);
     }
 
     template <typename T>
     // NOLINTNEXTLINE(bugprone-easily-swappable-parameters)
     const T* input_channel_ptr(const uint32_t bus,
                                const uint32_t channel) const noexcept {
-        return reinterpret_cast<const T*>(shm_bytes_) +
-               config_.input_offsets[bus][channel];
+        return reinterpret_cast<const T*>(shm_bytes_ +
+                                          config_.input_offsets[bus][channel]);
     }
 
     /**
@@ -171,16 +176,16 @@ class AudioShmBuffer {
     template <typename T>
     // NOLINTNEXTLINE(bugprone-easily-swappable-parameters)
     T* output_channel_ptr(const uint32_t bus, const uint32_t channel) noexcept {
-        return reinterpret_cast<T*>(shm_bytes_) +
-               config_.output_offsets[bus][channel];
+        return reinterpret_cast<T*>(shm_bytes_ +
+                                    config_.output_offsets[bus][channel]);
     }
 
     template <typename T>
     // NOLINTNEXTLINE(bugprone-easily-swappable-parameters)
     const T* output_channel_ptr(const uint32_t bus,
                                 const uint32_t channel) const noexcept {
-        return reinterpret_cast<const T*>(shm_bytes_) +
-               config_.output_offsets[bus][channel];
+        return reinterpret_cast<const T*>(shm_bytes_ +
+                                          config_.output_offsets[bus][channel]);
     }
 
     Config config_;

src/wine-host/bridges/vst2.cpp

@@ -795,30 +795,34 @@ intptr_t Vst2Bridge::dispatch_wrapper(AEffect* plugin,
 }
 
 AudioShmBuffer::Config Vst2Bridge::setup_shared_audio_buffers() {
+    assert(max_samples_per_block_);
+
     // We'll first compute the size and channel offsets for our buffer based on
     // the information already passed to us by the host. The offsets for each
-    // audio channel are in samples (since they'll be used with pointer
-    // arithmetic in `AudioShmBuffer`), and we'll only use the first bus (since
-    // VST2 plugins don't have multiple audio busses).
-    assert(max_samples_per_block_);
+    // audio channel are in bytes because CLAP allows some ports to be 32-bit
+    // only while other others are mixed 32-bit and 64-bit if the plugin opts in
+    // to it, and the plugin only knows what format it receives during the
+    // process call.
+    const size_t sample_size =
+        (double_precision_ ? sizeof(double) : sizeof(float));
+
     uint32_t current_offset = 0;
 
     std::vector<uint32_t> input_channel_offsets(plugin_->numInputs);
     for (int channel = 0; channel < plugin_->numInputs; channel++) {
         input_channel_offsets[channel] = current_offset;
-        current_offset += *max_samples_per_block_;
+        current_offset += *max_samples_per_block_ * sample_size;
     }
 
     std::vector<uint32_t> output_channel_offsets(plugin_->numOutputs);
     for (int channel = 0; channel < plugin_->numOutputs; channel++) {
         output_channel_offsets[channel] = current_offset;
-        current_offset += *max_samples_per_block_;
+        current_offset += *max_samples_per_block_ * sample_size;
     }
 
     // The size of the buffer is in bytes, and it will depend on whether the
     // host is going to pass 32-bit or 64-bit audio to the plugin
-    const uint32_t buffer_size =
-        current_offset * (double_precision_ ? sizeof(double) : sizeof(float));
+    const uint32_t buffer_size = current_offset;
 
     // We'll set up these shared memory buffers on the Wine side first, and then
     // when this request returns we'll do the same thing on the native plugin

src/wine-host/bridges/vst3.cpp

@@ -1376,8 +1376,15 @@ std::optional<AudioShmBuffer::Config> Vst3Bridge::setup_shared_audio_buffers(
 
     // We'll query the plugin for its audio bus layouts, and then create
     // calculate the offsets in a large memory buffer for the different audio
-    // channels. The offsets for each audio channel are in samples (since
-    // they'll be used with pointer arithmetic in `AudioShmBuffer`).
+    // channels. The offsets for each audio channel are in bytes because CLAP
+    // allows some ports to be 32-bit only while other others are mixed 32-bit
+    // and 64-bit if the plugin opts in to it, and the plugin only knows what
+    // format it receives during the process call.
+    const bool double_precision =
+        instance.process_setup->symbolicSampleSize == Steinberg::Vst::kSample64;
+    const size_t sample_size =
+        (double_precision ? sizeof(double) : sizeof(float));
+
     uint32_t current_offset = 0;
 
     auto create_bus_offsets = [&, &setup = instance.process_setup](
@@ -1404,7 +1411,7 @@ std::optional<AudioShmBuffer::Config> Vst3Bridge::setup_shared_audio_buffers(
 
             for (size_t channel = 0; channel < num_channels; channel++) {
                 bus_offsets[bus][channel] = current_offset;
-                current_offset += setup->maxSamplesPerBlock;
+                current_offset += setup->maxSamplesPerBlock * sample_size;
             }
         }
 
@@ -1419,10 +1426,7 @@ std::optional<AudioShmBuffer::Config> Vst3Bridge::setup_shared_audio_buffers(
 
     // The size of the buffer is in bytes, and it will depend on whether the
     // host is going to pass 32-bit or 64-bit audio to the plugin
-    const bool double_precision =
-        instance.process_setup->symbolicSampleSize == Steinberg::Vst::kSample64;
-    const uint32_t buffer_size =
-        current_offset * (double_precision ? sizeof(double) : sizeof(float));
+    const uint32_t buffer_size = current_offset;
 
     // If this function has been called previously and the size did not change,
     // then we should not do any work