feat(gpu): signed scalar div

Author: agnesLeroy

tfhe/benches/integer/signed_bench.rs

@@ -1813,6 +1813,12 @@ mod cuda {
         rng_func: default_signed_scalar
     );
 
+    define_cuda_server_key_bench_clean_input_scalar_signed_fn!(
+        method_name: unchecked_signed_scalar_div_rem,
+        display_name: div_rem,
+        rng_func: div_scalar
+    );
+
     //===========================================
     // Default
     //===========================================
@@ -2035,6 +2041,12 @@ mod cuda {
         rng_func: default_signed_scalar
     );
 
+    define_cuda_server_key_bench_clean_input_scalar_signed_fn!(
+        method_name: signed_scalar_div_rem,
+        display_name: div_rem,
+        rng_func: div_scalar
+    );
+
     criterion_group!(
         unchecked_cuda_ops,
         cuda_unchecked_add,
@@ -2081,6 +2093,7 @@ mod cuda {
         cuda_unchecked_scalar_le,
         cuda_unchecked_scalar_min,
         cuda_unchecked_scalar_max,
+        cuda_unchecked_signed_scalar_div_rem,
     );
 
     criterion_group!(
@@ -2146,6 +2159,7 @@ mod cuda {
         cuda_scalar_max,
         cuda_signed_overflowing_scalar_add,
         cuda_signed_overflowing_scalar_sub,
+        cuda_signed_scalar_div_rem,
     );
 
     fn cuda_bench_server_key_signed_cast_function<F>(

tfhe/src/core_crypto/gpu/entities/lwe_ciphertext_list.rs

@@ -41,6 +41,19 @@ impl<T: UnsignedInteger> CudaLweCiphertextList<T> {
     pub fn from_lwe_ciphertext_list<C: Container<Element = T>>(
         h_ct: &LweCiphertextList<C>,
         streams: &CudaStreams,
+    ) -> Self {
+        let res = unsafe { Self::from_lwe_ciphertext_list_async(h_ct, streams) };
+        streams.synchronize();
+        res
+    }
+
+    /// # Safety
+    ///
+    /// - `stream` __must__ be synchronized to guarantee computation has finished, and inputs must
+    ///   not be dropped until stream is synchronised
+    pub unsafe fn from_lwe_ciphertext_list_async<C: Container<Element = T>>(
+        h_ct: &LweCiphertextList<C>,
+        streams: &CudaStreams,
     ) -> Self {
         let lwe_dimension = h_ct.lwe_size().to_lwe_dimension();
         let lwe_ciphertext_count = h_ct.lwe_ciphertext_count();
@@ -56,7 +69,6 @@ impl<T: UnsignedInteger> CudaLweCiphertextList<T> {
         unsafe {
             d_vec.copy_from_cpu_async(h_input.as_ref(), streams, 0);
         }
-        streams.synchronize();
         let cuda_lwe_list = CudaLweList {
             d_vec,
             lwe_ciphertext_count,