Fix session kernel benchmark setup

Initialize direct encoding benchmark sessions with their crate-level kernel registrations so they exercise the session execute-parent path instead of fallback materialization.

Cache the ArrayKernels handle in ExecutionCtx to avoid repeated session lookups while trying execute-parent kernels.

Signed-off-by: "Nicholas Gates" <nick@nickgates.com>

Author: gatesn

encodings/alp/benches/alp_compress.rs

@@ -50,7 +50,11 @@ const BENCH_ARGS: &[(usize, f64, f64)] = &[
     (10_000, 0.1, 1.0),
 ];
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_alp::initialize(&session);
+    session
+});
 
 #[divan::bench(types = [f32, f64], args = BENCH_ARGS)]
 fn compress_alp<T: ALPFloat + NativePType>(bencher: Bencher, args: (usize, f64, f64)) {

encodings/experimental/onpair/benches/decode.rs

@@ -79,7 +79,11 @@ impl DecodeInputs {
 use vortex_onpair::onpair_compress;
 use vortex_session::VortexSession;
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_onpair::initialize(&session);
+    session
+});
 
 #[derive(Copy, Clone, Debug)]
 enum Shape {

encodings/fastlanes/benches/bitpack_compare.rs

@@ -35,7 +35,11 @@ fn main() {
     divan::main();
 }
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fastlanes::initialize(&session);
+    session
+});
 
 const LENS: &[usize] = &[1024, 64 * 1024];
 const BIT_WIDTHS: &[u8] = &[4, 16];

encodings/fastlanes/benches/bitpack_compare_sweep.rs

@@ -39,7 +39,11 @@ fn main() {
     divan::main();
 }
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fastlanes::initialize(&session);
+    session
+});
 
 /// Number of elements per benchmarked array (64 full FastLanes blocks).
 const LEN: usize = 64 * 1024;

encodings/fastlanes/benches/bitpacking_take.rs

@@ -26,7 +26,11 @@ fn main() {
     divan::main();
 }
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fastlanes::initialize(&session);
+    session
+});
 
 #[divan::bench]
 fn take_10_stratified(bencher: Bencher) {

encodings/fastlanes/benches/canonicalize_bench.rs

@@ -34,7 +34,11 @@ const BENCH_ARGS: &[(usize, usize, f64)] = &[
     (10000, 1000, 0.00),
 ];
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fastlanes::initialize(&session);
+    session
+});
 
 #[cfg(not(codspeed))]
 #[divan::bench(args = BENCH_ARGS)]

encodings/fastlanes/benches/cast_bitpacked.rs

@@ -38,7 +38,11 @@ fn main() {
     divan::main();
 }
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fastlanes::initialize(&session);
+    session
+});
 
 const U32: DType = DType::Primitive(PType::U32, Nullability::NonNullable);
 

encodings/fastlanes/benches/compute_between.rs

@@ -24,7 +24,11 @@ fn main() {
     divan::main();
 }
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fastlanes::initialize(&session);
+    session
+});
 
 fn generate_primitive_array<T: NativePType + NumCast>(
     rng: &mut StdRng,

encodings/fsst/benches/chunked_dict_fsst_builder.rs

@@ -28,7 +28,11 @@ const BENCH_ARGS: &[(usize, usize, usize)] = &[
     (1000, 1000, 100),
 ];
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fsst::initialize(&session);
+    session
+});
 
 fn make_dict_fsst_chunks<T: NativePType>(
     len: usize,

encodings/fsst/benches/fsst_compress.rs

@@ -31,7 +31,11 @@ fn main() {
     divan::main();
 }
 
-static SESSION: LazyLock<VortexSession> = LazyLock::new(vortex_array::array_session);
+static SESSION: LazyLock<VortexSession> = LazyLock::new(|| {
+    let session = vortex_array::array_session();
+    vortex_fsst::initialize(&session);
+    session
+});
 
 // [(string_count, avg_len, unique_chars)]
 const BENCH_ARGS: &[(usize, usize, u8)] = &[