音声合成の処理を丸ごとスレッド分離して実行する (#692)

VOICEVOX · Nov 25, 2023 · 402c994 · 402c994
1 parent 280cc79
commit 402c994
Show file tree

Hide file tree

Showing 10 changed files with 380 additions and 173 deletions.
diff --git a/crates/voicevox_core/src/__internal.rs b/crates/voicevox_core/src/__internal.rs
@@ -1,4 +1,5 @@
 pub mod doctest_fixtures;
+pub mod interop;
 
 // VOICEVOX CORE内のラッパー向けの実装
 // FIXME: 要議論: https://github.com/VOICEVOX/voicevox_core/issues/595

diff --git a/crates/voicevox_core/src/__internal/interop.rs b/crates/voicevox_core/src/__internal/interop.rs
@@ -0,0 +1 @@
+pub use crate::synthesizer::PerformInference;
diff --git a/crates/voicevox_core/src/engine/open_jtalk.rs b/crates/voicevox_core/src/engine/open_jtalk.rs
@@ -52,7 +52,7 @@ impl OpenJtalk {
     pub async fn new(open_jtalk_dict_dir: impl AsRef<Path>) -> crate::result::Result<Self> {
         let open_jtalk_dict_dir = open_jtalk_dict_dir.as_ref().to_owned();
 
-        tokio::task::spawn_blocking(move || {
+        crate::task::asyncify(move || {
             let mut s = Self::new_without_dic();
             s.load(open_jtalk_dict_dir).map_err(|()| {
                 // FIXME: 「システム辞書を読もうとしたけど読めなかった」というエラーをちゃんと用意する
@@ -61,7 +61,6 @@ impl OpenJtalk {
             Ok(s)
         })
         .await
-        .unwrap()
     }
 
     // 先に`load`を呼ぶ必要がある。
@@ -80,7 +79,7 @@ impl OpenJtalk {
 
         let words = user_dict.to_mecab_format();
 
-        let result = tokio::task::spawn_blocking(move || -> crate::Result<_> {
+        let result = crate::task::asyncify(move || -> crate::Result<_> {
             // ユーザー辞書用のcsvを作成
             let mut temp_csv =
                 NamedTempFile::new().map_err(|e| ErrorRepr::UseUserDict(e.into()))?;
@@ -111,8 +110,7 @@ impl OpenJtalk {
 
             Ok(mecab.load_with_userdic(dict_dir.as_ref(), Some(Path::new(&temp_dict_path))))
         })
-        .await
-        .unwrap()?;
+        .await?;
 
         if !result {
             return Err(ErrorRepr::UseUserDict(anyhow!("辞書のコンパイルに失敗しました")).into());

diff --git a/crates/voicevox_core/src/infer/status.rs b/crates/voicevox_core/src/infer/status.rs
@@ -90,10 +90,16 @@ impl<R: InferenceRuntime, D: InferenceDomain> Status<R, D> {
         self.is_loaded_model_by_style_id(style_id)
     }
 
+    /// 推論を実行する。
+    ///
+    /// # Performance
+    ///
+    /// CPU/GPU-boundな操作であるため、非同期ランタイム上では直接実行されるべきではない。
+    ///
     /// # Panics
     ///
     /// `self`が`model_id`を含んでいないとき、パニックする。
-    pub(crate) async fn run_session<I>(
+    pub(crate) fn run_session<I>(
         &self,
         model_id: &VoiceModelId,
         input: I,
@@ -103,10 +109,7 @@ impl<R: InferenceRuntime, D: InferenceDomain> Status<R, D> {
         I::Signature: InferenceSignature<Domain = D>,
     {
         let sess = self.loaded_models.lock().unwrap().get(model_id);
-
-        tokio::task::spawn_blocking(move || sess.run(input))
-            .await
-            .unwrap()
+        sess.run(input)
     }
 }
 

diff --git a/crates/voicevox_core/src/lib.rs b/crates/voicevox_core/src/lib.rs
@@ -11,11 +11,11 @@ mod metas;
 mod numerics;
 mod result;
 mod synthesizer;
+mod task;
 mod user_dict;
 mod version;
 mod voice_model;
 
-#[doc(hidden)]
 pub mod __internal;
 
 #[cfg(test)]
@@ -31,7 +31,9 @@ pub use self::result::*;
 pub use self::voice_model::*;
 pub use devices::*;
 pub use manifest::*;
-pub use synthesizer::*;
+pub use synthesizer::{
+    AccelerationMode, InitializeOptions, SynthesisOptions, Synthesizer, TtsOptions,
+};
 pub use user_dict::*;
 pub use version::*;