chore(gpu): fix warnings detection

Author: andrei-stoian-zama

Makefile

@@ -697,8 +697,9 @@ test_high_level_api_gpu_debug: install_rs_build_toolchain install_cargo_nextest
 test_integer_hl_test_gpu_check_warnings: install_rs_build_toolchain
 	RUSTFLAGS="$(RUSTFLAGS)" cargo $(CARGO_RS_BUILD_TOOLCHAIN) build \
 		--features=integer,internal-keycache,gpu-debug,zk-pok -vv -p $(TFHE_SPEC) &> /tmp/gpu_compile_output
-	WARNINGS=$$(cat /tmp/gpu_compile_output | grep ": warning:" | grep "\[tfhe-cuda-backend" | grep -v "inline function" || true) && \
+	WARNINGS=$$(cat /tmp/gpu_compile_output | grep ": warning #" | grep "\[tfhe-cuda-backend" | grep -v "inline qualifier" || true) && \
 	if [[ "$${WARNINGS}" != "" ]]; then \
+	    echo "FAILING BECAUSE CUDA COMPILATION WARNINGS WERE DETECTED: " && \
 		echo "$${WARNINGS}" && exit 1; \
 	fi
 

backends/tfhe-cuda-backend/cuda/include/integer/integer_utilities.h

@@ -3769,8 +3769,6 @@ template <typename Torus> struct int_tree_sign_reduction_buffer {
     gpu_memory_allocated = allocate_gpu_memory;
     this->params = params;
 
-    uint64_t big_size = (params.big_lwe_dimension + 1) * sizeof(Torus);
-
     block_selector_f = [](Torus msb, Torus lsb) -> Torus {
       if (msb == IS_EQUAL) // EQUAL
         return lsb;
@@ -3866,8 +3864,6 @@ template <typename Torus> struct int_comparison_diff_buffer {
       }
     };
 
-    uint64_t big_size = (params.big_lwe_dimension + 1) * sizeof(Torus);
-
     tmp_packed = new CudaRadixCiphertextFFI;
     create_zero_radix_ciphertext_async<Torus>(
         streams[0], gpu_indexes[0], tmp_packed, num_radix_blocks,

backends/tfhe-cuda-backend/cuda/src/polynomial/dot_product.cuh

@@ -100,7 +100,6 @@ __host__ void host_glwe_wrapping_polynomial_mul_one_to_many(
     cudaStream_t stream, uint32_t gpu_index, Torus *result,
     const Torus *glwe_lhs, int8_t *circulant, const Torus *poly_rhs,
     uint32_t polynomial_size, uint32_t glwe_dimension, uint32_t n_rhs) {
-  uint64_t const *glwe_lhs_t = static_cast<uint64_t const *>(glwe_lhs);
 
   for (unsigned i = 0; i < glwe_dimension + 1; ++i) {
     host_wrapping_polynomial_mul_one_to_many<uint64_t, ulonglong4>(