rename

Signed-off-by: Andrew Duffy <andrew@a10y.dev>

Author: a10y

vortex-cuda/kernels/src/alp.cu

@@ -13,13 +13,13 @@
 // only encoded as i32, and f64 is only encoded as i64. The i64 → double cast
 // is lossless for all values ALP can produce.
 template <typename EncT, typename FloatT>
-__device__ void _alp_device(const EncT *__restrict in,
-                            FloatT *__restrict out,
-                            FloatT f,
-                            FloatT e,
-                            uint64_t array_len,
-                            int thread_idx,
-                            GPUPatches &patches) {
+__device__ void alp_device(const EncT *__restrict in,
+                           FloatT *__restrict out,
+                           FloatT f,
+                           FloatT e,
+                           uint64_t array_len,
+                           int thread_idx,
+                           GPUPatches &patches) {
     constexpr int ThreadCount = 32;
     // ThreadCount == 32 (one warp) is baked into this kernel:
     //   - __syncwarp() below is only sufficient because all threads live in one warp.
@@ -83,7 +83,7 @@ __device__ void _alp_device(const EncT *__restrict in,
         int thread_idx = threadIdx.x;                                                                        \
         auto in = full_in + (blockIdx.x * 1024);                                                             \
         auto out = full_out + (blockIdx.x * 1024);                                                           \
-        _alp_device<EncT, FloatT>(in, out, f, e, array_len, thread_idx, patches);                            \
+        alp_device<EncT, FloatT>(in, out, f, e, array_len, thread_idx, patches);                             \
     }
 
 // The only ALPInt bindings produced by the encoder are (f32, i32) and (f64, i64).