bounds

Signed-off-by: Robert Kruszewski <[email protected]>

Author: robert3005

vortex-gpu/src/rle_decompress.rs

@@ -19,8 +19,6 @@ use vortex_dtype::{
 use vortex_error::{VortexExpect, VortexResult, vortex_err, vortex_panic};
 use vortex_fastlanes::RLEArray;
 
-use crate::take::cuda_take_kernel;
-
 pub fn cuda_rle_decompress(rle: &RLEArray, ctx: Arc<CudaContext>) -> VortexResult<ArrayRef> {
     match_each_native_ptype!(rle.values().dtype().as_ptype(), |V| {
         match_each_unsigned_integer_ptype!(rle.values().dtype().as_ptype(), |O| {
@@ -56,10 +54,11 @@ fn cuda_rle_decompress_typed<Values, Indices, Offsets>(
 where
     Values: NativePType + DeviceRepr + ValidAsZeroBits,
     Indices: UnsignedPType + DeviceRepr,
+    Offsets: UnsignedPType + DeviceRepr,
 {
     assert_eq!(indices.len() % 1024, 0);
 
-    let kernel_func = cuda_take_kernel::<Indices, Values>(false, ctx.clone())?;
+    let kernel_func = cuda_rle_kernel::<Indices, Values, Offsets>(ctx.clone())?;
     let num_chunks =
         u32::try_from(indices.len().div_ceil(1024)).vortex_expect("num chunks overflow");
     let stream = ctx.default_stream();
@@ -107,10 +106,7 @@ where
     Ok(PrimitiveArray::new(buffer, Validity::NonNullable).into_array())
 }
 
-fn cuda_rle_kernel<Indices, Values, Offsets>(
-    mask: bool,
-    ctx: Arc<CudaContext>,
-) -> VortexResult<CudaFunction>
+fn cuda_rle_kernel<Indices, Values, Offsets>(ctx: Arc<CudaContext>) -> VortexResult<CudaFunction>
 where
     Indices: UnsignedPType,
     Values: NativePType,
@@ -122,9 +118,9 @@ where
 
     let kernel_name = format!(
         "rle_decompress_i{}_v{}_o{}",
-        if mask { "_masked" } else { "" },
         &Indices::PTYPE,
-        &Values::PTYPE & Offsets::PTYPE,
+        &Values::PTYPE,
+        &Offsets::PTYPE,
     );
 
     module