diff --git a/Cargo.toml b/Cargo.toml
index cda9752..0b126cb 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -28,4 +28,6 @@ native-cpu = ["snmalloc-sys/native-cpu"]
 local_dynamic_tls = ["snmalloc-sys/local_dynamic_tls"]
 win8compat = ["snmalloc-sys/win8compat"]
 usecxx17 = ["snmalloc-sys/usecxx17"]
-check = ["snmalloc-sys/check"]
\ No newline at end of file
+check = ["snmalloc-sys/check"]
+lto = ["snmalloc-sys/lto"]
+notls = ["snmalloc-sys/notls"]
diff --git a/snmalloc-sys/Cargo.toml b/snmalloc-sys/Cargo.toml
index dcedae7..85452c8 100644
--- a/snmalloc-sys/Cargo.toml
+++ b/snmalloc-sys/Cargo.toml
@@ -28,3 +28,5 @@ local_dynamic_tls = []
 win8compat = []
 usecxx17 = []
 check = []
+lto = []
+notls = []
diff --git a/snmalloc-sys/build.rs b/snmalloc-sys/build.rs
index a431248..a1d02d8 100644
--- a/snmalloc-sys/build.rs
+++ b/snmalloc-sys/build.rs
@@ -108,12 +108,22 @@ fn main() {
     }
 
     if cfg!(feature = "cache-friendly") {
-        eprintln!("cache-friendly feature flag is deprecated and no longer has any effect. \
-            it may be removed in a future release");
+        eprintln!(
+            "cache-friendly feature flag is deprecated and no longer has any effect. \
+            it may be removed in a future release"
+        );
         // The following code no longer works
         // build.define("CACHE_FRIENDLY_OFFSET", "64");
     }
 
+    if cfg!(feature = "lto") {
+        build.define("SNMALLOC_IPO", "ON");
+    }
+
+    if cfg!(feature = "notls") {
+        build.define("SNMALLOC_ENABLE_DYNAMIC_LOADING", "ON");
+    }
+
     build.compile(target);
 
     if target_env == "msvc" {
@@ -140,14 +150,16 @@ fn main() {
         // no need
     } else {
         // link c++ runtime
-        println!("cargo:rustc-link-lib={}",
-                 std::env::var("CXXSTDLIB").unwrap_or_else(|_| {
-                     if cfg!(target_os = "macos") || cfg!(target_os = "openbsd") {
-                         "c++".to_string()
-                     } else {
-                         "stdc++".to_string()
-                     }
-                 }))
+        println!(
+            "cargo:rustc-link-lib={}",
+            std::env::var("CXXSTDLIB").unwrap_or_else(|_| {
+                if cfg!(target_os = "macos") || cfg!(target_os = "openbsd") {
+                    "c++".to_string()
+                } else {
+                    "stdc++".to_string()
+                }
+            })
+        )
     }
 }
 
@@ -305,13 +317,15 @@ fn main() {
         // not need for explicit c++ runtime
     } else {
         // link c++ runtime
-        println!("cargo:rustc-link-lib={}",
-                 std::env::var("CXXSTDLIB").unwrap_or_else(|_| {
-                     if cfg!(target_os = "macos") || cfg!(target_os = "openbsd") {
-                         "c++".to_string()
-                     } else {
-                         "stdc++".to_string()
-                     }
-                 }))
+        println!(
+            "cargo:rustc-link-lib={}",
+            std::env::var("CXXSTDLIB").unwrap_or_else(|_| {
+                if cfg!(target_os = "macos") || cfg!(target_os = "openbsd") {
+                    "c++".to_string()
+                } else {
+                    "stdc++".to_string()
+                }
+            })
+        )
     }
 }
diff --git a/snmalloc-sys/snmalloc b/snmalloc-sys/snmalloc
index dc12688..b8e9e99 160000
--- a/snmalloc-sys/snmalloc
+++ b/snmalloc-sys/snmalloc
@@ -1 +1 @@
-Subproject commit dc1268886a5d49d38a54e5d1402b5924a71fee0b
+Subproject commit b8e9e99cf096357d37c67d423ccb657652879ba1
diff --git a/src/lib.rs b/src/lib.rs
index cc31c31..457fcbf 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -44,7 +44,6 @@ unsafe impl GlobalAlloc for SnMalloc {
         ffi::sn_rust_alloc(layout.align(), layout.size()) as _
     }
 
-
     /// De-allocate the memory at the given address with the given alignment and size.
     /// The client must assure the following things:
     /// - the memory is acquired using the same allocator and the pointer points to the start position.
@@ -113,4 +112,15 @@ mod tests {
             alloc.dealloc(ptr, layout);
         }
     }
+
+    #[test]
+    fn it_frees_large_alloc() {
+        unsafe {
+            let layout = Layout::from_size_align(1 << 20, 32).unwrap();
+            let alloc = SnMalloc;
+
+            let ptr = alloc.alloc(layout);
+            alloc.dealloc(ptr, layout);
+        }
+    }
 }