CSDK-2183 daily-python: align memory to i16

Author: aconchillo

src/media/native_vad.rs

@@ -1,3 +1,5 @@
+use crate::util::memory::AlignedI16Data;
+
 use webrtc_daily::sys::vad::NativeWebrtcVad;
 
 use daily_core::prelude::daily_core_context_vad_analyze;
@@ -76,26 +78,26 @@ impl PyNativeVad {
     /// :return: The probability (from 0 to 1.0) that speech was detected
     /// :rtype: float
     fn analyze_frames(&self, frames: &PyBytes) -> PyResult<f32> {
-        // TODO(aleix): Should this be i16 aligned?
-        let bytes = frames.as_bytes();
-
-        let bytes_length = frames.len()?;
+        let num_bytes = frames.len()?;
         let bytes_per_sample = 2;
 
         // libwebrtc needs 16-bit linear PCM samples
-        if bytes_length % bytes_per_sample != 0 {
+        if num_bytes % bytes_per_sample != 0 {
             return Err(exceptions::PyValueError::new_err(
                 "frames bytestring should contain 16-bit samples",
             ));
         }
 
-        let num_frames = (bytes_length / bytes_per_sample) / self.channels as usize;
+        let num_frames = (num_bytes / bytes_per_sample) / self.channels as usize;
+
+        let bytes = frames.as_bytes();
+        let aligned = AlignedI16Data::new(bytes);
 
         let confidence = Python::with_gil(|py| {
             py.allow_threads(move || unsafe {
                 daily_core_context_vad_analyze(
                     self.webrtc_vad.as_ref().unwrap().as_ptr() as *mut _,
-                    bytes.as_ptr() as *const _,
+                    aligned.as_ptr(),
                     num_frames,
                 )
             })

src/media/virtual_microphone_device.rs

@@ -1,6 +1,7 @@
 use std::sync::atomic::{AtomicU64, Ordering};
 use std::{collections::HashMap, sync::Mutex};
 
+use crate::util::memory::AlignedI16Data;
 use crate::GIL_MUTEX_HACK;
 
 use webrtc_daily::sys::virtual_microphone_device::NativeVirtualMicrophoneDevice;
@@ -114,28 +115,28 @@ impl PyVirtualMicrophoneDevice {
             ));
         }
 
-        let bytes_length = frames.len()?;
+        let num_bytes = frames.len()?;
         let bytes_per_sample = 2;
 
         // libwebrtc needs 16-bit linear PCM samples
-        if bytes_length % bytes_per_sample != 0 {
+        if num_bytes % bytes_per_sample != 0 {
             return Err(exceptions::PyValueError::new_err(
                 "frames bytestring should contain 16-bit samples",
             ));
         }
 
-        let num_frames = (bytes_length / bytes_per_sample) / self.channels as usize;
+        let num_frames = (num_bytes / bytes_per_sample) / self.channels as usize;
 
-        // TODO(aleix): Should this be i16 aligned?
         let bytes = frames.as_bytes();
+        let aligned = AlignedI16Data::new(bytes);
 
         let request_id = self.maybe_register_completion(completion);
 
         Python::with_gil(|py| {
             let frames_written = py.allow_threads(move || unsafe {
                 daily_core_context_virtual_microphone_device_write_frames(
                     self.audio_device.as_ref().unwrap().as_ptr() as *mut _,
-                    bytes.as_ptr() as *const _,
+                    aligned.as_ptr(),
                     num_frames,
                     request_id,
                     on_write_frames,

src/util.rs

@@ -1 +1,2 @@
 pub(crate) mod dict;
+pub(crate) mod memory;

src/util/memory.rs

@@ -0,0 +1,37 @@
+pub(crate) enum AlignedI16Data<'a> {
+    AlreadyAligned(&'a [u8]),
+    Copied(Vec<i16>),
+}
+
+impl<'a> AlignedI16Data<'a> {
+    pub fn new(src: &'a [u8]) -> Self {
+        let bytes_ptr = src.as_ptr();
+
+        // If `src`'s memory is not 16-bit aligned, create a new 16-bit aligned
+        // memory area and copy the contents of `src` to it. Otherwise, simply
+        // keep the original slice.
+        if bytes_ptr as usize % 2 == 0 {
+            AlignedI16Data::AlreadyAligned(src)
+        } else {
+            let num_bytes = src.len();
+            let num_words = num_bytes / 2;
+
+            let mut words = Vec::<i16>::with_capacity(num_words);
+            let words_ptr = words.as_mut_ptr() as *mut u8;
+
+            unsafe {
+                std::ptr::copy_nonoverlapping(bytes_ptr, words_ptr, num_bytes);
+                words.set_len(num_words);
+            }
+
+            AlignedI16Data::Copied(words)
+        }
+    }
+
+    pub fn as_ptr(&self) -> *const i16 {
+        match self {
+            AlignedI16Data::AlreadyAligned(d) => d.as_ptr() as *const i16,
+            AlignedI16Data::Copied(d) => d.as_ptr(),
+        }
+    }
+}