refactor!: Add invariant to GpuIndex

This commit does 2 things:

* It adds to GpuIndex the invariant that the index corresponds
  to a valid GPU, to do so, the inner u32 is made private
  and new/try_new method are now used to construct a GpuIndex
  these methods checks that the index is valid
* It makes GpuIndex transparent, allowing to safely cast a *const
  GpuIndex to *const u32, this is to save same copies made to transform
  Vec<GpuIndex> to Vec<u32> that was used to get a *const u32

BREAKING CHANGES: GpuIndex(some_value) is no longer valid and
GpuIndex::new(some_value) / GpuIndex::try_new(some_value) has to be
used

tfhe/benches/core_crypto/ks_bench.rs

@@ -330,7 +330,7 @@ mod cuda {
             SecretRandomGenerator::<DefaultRandomGenerator>::new(seeder.seed());
 
         let gpu_index = 0;
-        let streams = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+        let streams = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
         for (name, params) in parameters.iter() {
             let lwe_dimension = params.lwe_dimension.unwrap();
@@ -433,7 +433,7 @@ mod cuda {
             SecretRandomGenerator::<DefaultRandomGenerator>::new(seeder.seed());
 
         let gpu_index = 0;
-        let streams = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+        let streams = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
         for (name, params) in parameters.iter() {
             let lwe_dimension = params.lwe_dimension.unwrap();

tfhe/benches/core_crypto/pbs_bench.rs

@@ -901,7 +901,7 @@ mod cuda {
             SecretRandomGenerator::<DefaultRandomGenerator>::new(seeder.seed());
 
         let gpu_index = 0;
-        let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+        let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
         for (name, params) in parameters.iter() {
             // Create the LweSecretKey
@@ -1022,7 +1022,7 @@ mod cuda {
             SecretRandomGenerator::<DefaultRandomGenerator>::new(seeder.seed());
 
         let gpu_index = 0;
-        let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+        let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
         for (name, params, grouping_factor) in parameters.iter() {
             // Create the LweSecretKey
@@ -1144,7 +1144,7 @@ mod cuda {
             SecretRandomGenerator::<DefaultRandomGenerator>::new(seeder.seed());
 
         let gpu_index = 0;
-        let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+        let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
         for (name, params) in parameters.iter() {
             let input_lwe_secret_key = allocate_and_generate_new_binary_lwe_secret_key(
@@ -1284,7 +1284,7 @@ mod cuda {
             SecretRandomGenerator::<DefaultRandomGenerator>::new(seeder.seed());
 
         let gpu_index = 0;
-        let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+        let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
         for (name, params, grouping_factor) in parameters.iter() {
             let input_lwe_secret_key = allocate_and_generate_new_binary_lwe_secret_key(

tfhe/src/core_crypto/gpu/algorithms/glwe_sample_extraction.rs

@@ -46,7 +46,7 @@ pub unsafe fn cuda_extract_lwe_samples_from_glwe_ciphertext_list_async<Scalar>(
     let nth_array: Vec<u32> = vec_nth.iter().map(|x| x.0 as u32).collect_vec();
     let gpu_indexes = &streams.gpu_indexes;
     unsafe {
-        let d_nth_array = CudaVec::from_cpu_async(&nth_array, streams, gpu_indexes[0].0);
+        let d_nth_array = CudaVec::from_cpu_async(&nth_array, streams, gpu_indexes[0].get());
         extract_lwe_samples_from_glwe_ciphertext_list_async(
             streams,
             &mut output_lwe_list.0.d_vec,

tfhe/src/core_crypto/gpu/algorithms/test/glwe_sample_extraction.rs

@@ -27,7 +27,7 @@ fn glwe_encrypt_sample_extract_decrypt_custom_mod<Scalar: UnsignedTorus + Send +
     let delta: Scalar = encoding_with_padding / msg_modulus;
 
     let gpu_index = 0;
-    let streams = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+    let streams = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
     let mut msgs = vec![];
 

tfhe/src/core_crypto/gpu/algorithms/test/lwe_keyswitch.rs

@@ -18,7 +18,7 @@ fn lwe_encrypt_ks_decrypt_custom_mod<Scalar: UnsignedTorus + CastFrom<usize>>(
     let ks_decomp_base_log = params.ks_base_log;
     let ks_decomp_level_count = params.ks_level;
 
-    let stream = CudaStreams::new_single_gpu(GpuIndex(0));
+    let stream = CudaStreams::new_single_gpu(GpuIndex::new(0));
 
     let mut rsc = TestResources::new();
 

tfhe/src/core_crypto/gpu/algorithms/test/lwe_multi_bit_programmable_bootstrapping.rs

@@ -28,7 +28,7 @@ fn lwe_encrypt_multi_bit_pbs_decrypt_custom_mod<
     let grouping_factor = params.grouping_factor;
 
     let gpu_index = 0;
-    let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+    let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
     let mut rsc = TestResources::new();
 

tfhe/src/core_crypto/gpu/algorithms/test/lwe_packing_keyswitch.rs

@@ -68,7 +68,7 @@ where
     let delta: Scalar = encoding_with_padding / msg_modulus;
 
     let gpu_index = 0;
-    let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+    let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
     while msg != Scalar::ZERO {
         msg = msg.wrapping_sub(Scalar::ONE);
@@ -152,7 +152,7 @@ where
     let delta: Scalar = encoding_with_padding / msg_modulus;
 
     let gpu_index = 0;
-    let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+    let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
     while msg != Scalar::ZERO {
         msg = msg.wrapping_sub(Scalar::ONE);

tfhe/src/core_crypto/gpu/algorithms/test/lwe_programmable_bootstrapping.rs

@@ -26,7 +26,7 @@ fn lwe_encrypt_pbs_decrypt<
     let decomp_level_count = params.pbs_level;
 
     let gpu_index = 0;
-    let stream = CudaStreams::new_single_gpu(GpuIndex(gpu_index));
+    let stream = CudaStreams::new_single_gpu(GpuIndex::new(gpu_index));
 
     let mut rsc = TestResources::new();
 

tfhe/src/core_crypto/gpu/algorithms/test/noise_distribution/lwe_multi_bit_programmable_bootstrapping_noise.rs

@@ -40,7 +40,7 @@ where
     let grouping_factor = params.grouping_factor;
     let number_of_messages = 1;
 
-    let gpu_index = GpuIndex(0);
+    let gpu_index = GpuIndex::new(0);
     let stream = CudaStreams::new_single_gpu(gpu_index);
 
     let modulus_as_f64 = if ciphertext_modulus.is_native_modulus() {

tfhe/src/core_crypto/gpu/algorithms/test/noise_distribution/lwe_programmable_bootstrapping_noise.rs

@@ -33,7 +33,7 @@ where
     let pbs_decomposition_level_count = params.pbs_level;
     let number_of_messages = 1;
 
-    let gpu_index = GpuIndex(0);
+    let gpu_index = GpuIndex::new(0);
     let stream = CudaStreams::new_single_gpu(gpu_index);
 
     let modulus_as_f64 = if ciphertext_modulus.is_native_modulus() {

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

@@ -128,7 +128,7 @@ impl<T: UnsignedInteger> CudaLweCiphertextList<T> {
                 first_item.0.d_vec.as_c_ptr(0),
                 size as u64,
                 streams.ptr[0],
-                streams.gpu_indexes[0].0,
+                streams.gpu_indexes[0].get(),
             );
             ptr = ptr.wrapping_byte_add(size);
             for list in cuda_ciphertexts_list_vec {
@@ -137,7 +137,7 @@ impl<T: UnsignedInteger> CudaLweCiphertextList<T> {
                     list.0.d_vec.as_c_ptr(0),
                     size as u64,
                     streams.ptr[0],
-                    streams.gpu_indexes[0].0,
+                    streams.gpu_indexes[0].get(),
                 );
                 ptr = ptr.wrapping_byte_add(size);
             }