chore: clean up #[inline] annotations (#7000)

Remove #[inline] from private, pub(crate), and inner functions where the compiler already has full visibility, and from large (>20 line) functions where code bloat outweighs benefit.

Add #[inline] to BitBuffer operator trait impls (BitOr, BitAnd, Not, BitXor) — non-generic impls on a concrete type that are otherwise opaque cross-crate with LTO disabled.

Signed-off-by: Alexander Droste <alexander.droste@protonmail.com>

Author: 0ax1

encodings/fastlanes/src/bit_transpose/aarch64.rs

@@ -114,7 +114,6 @@ static SCATTER_8X8_NEON: [[u8; 16]; 4] = [
 
 /// Perform 8x8 bit transpose on two u64s packed in a `uint64x2_t`.
 #[allow(unsafe_op_in_unsafe_fn)]
-#[inline]
 unsafe fn bit_transpose_8x8_neon(mut v: uint64x2_t) -> uint64x2_t {
     let mask1 = vdupq_n_u64(TRANSPOSE_2X2);
     let t = vandq_u64(veorq_u64(v, vshrq_n_u64::<7>(v)), mask1);

encodings/fastlanes/src/bit_transpose/scalar.rs

@@ -15,7 +15,6 @@ use crate::bit_transpose::TRANSPOSE_8X8;
 #[allow(dead_code)]
 pub fn transpose_bits_scalar(input: &[u8; 128], output: &mut [u8; 128]) {
     // Helper to perform 8x8 bit transpose on a u64 (each byte becomes a row)
-    #[inline]
     fn transpose_8x8(mut x: u64) -> u64 {
         // Step 1: Transpose 2x2 bit blocks
         let t = (x ^ (x >> 7)) & TRANSPOSE_2X2;
@@ -29,7 +28,6 @@ pub fn transpose_bits_scalar(input: &[u8; 128], output: &mut [u8; 128]) {
     }
 
     // Helper to gather 8 bytes at stride 16 into a u64
-    #[inline]
     fn gather(input: &[u8; 128], base: usize) -> u64 {
         u64::from(input[base])
             | (u64::from(input[base + 16]) << 8)
@@ -89,7 +87,6 @@ pub fn transpose_bits_scalar(input: &[u8; 128], output: &mut [u8; 128]) {
 #[inline(never)]
 #[allow(dead_code)]
 pub fn untranspose_bits_scalar(input: &[u8; 128], output: &mut [u8; 128]) {
-    #[inline]
     fn transpose_8x8(mut x: u64) -> u64 {
         let t = (x ^ (x >> 7)) & TRANSPOSE_2X2;
         x = x ^ t ^ (t << 7);
@@ -99,7 +96,6 @@ pub fn untranspose_bits_scalar(input: &[u8; 128], output: &mut [u8; 128]) {
         x ^ t ^ (t << 28)
     }
 
-    #[inline]
     fn gather_transposed(input: &[u8; 128], base_group: usize, offset: usize) -> u64 {
         let mut result: u64 = 0;
         for bit_pos in 0..8 {
@@ -108,7 +104,6 @@ pub fn untranspose_bits_scalar(input: &[u8; 128], output: &mut [u8; 128]) {
         result
     }
 
-    #[inline]
     fn scatter(output: &mut [u8; 128], base: usize, val: u64) {
         output[base] = val as u8;
         output[base + 16] = (val >> 8) as u8;

encodings/fastlanes/src/bit_transpose/x86.rs

@@ -49,7 +49,6 @@ pub fn has_vbmi() -> bool {
 #[allow(unsafe_op_in_unsafe_fn)]
 pub unsafe fn transpose_bits_bmi2(input: &[u8; 128], output: &mut [u8; 128]) {
     // Helper to gather 8 bytes at stride 16 into a u64
-    #[inline]
     fn gather(input: &[u8; 128], base: usize) -> u64 {
         (input[base] as u64)
             | ((input[base + 16] as u64) << 8)
@@ -238,7 +237,6 @@ pub unsafe fn transpose_bits_bmi2(input: &[u8; 128], output: &mut [u8; 128]) {
 #[allow(unsafe_op_in_unsafe_fn)]
 pub unsafe fn untranspose_bits_bmi2(input: &[u8; 128], output: &mut [u8; 128]) {
     // Helper: scatter a u64 to 8 output bytes at stride 16
-    #[inline]
     fn scatter(output: &mut [u8; 128], base: usize, val: u64) {
         output[base] = val as u8;
         output[base + 16] = (val >> 8) as u8;

encodings/fastlanes/src/bitpacking/array/unpack_iter.rs

@@ -256,12 +256,10 @@ impl<T: PhysicalPType, S: UnpackStrategy<T>> UnpackedChunks<T, S> {
         })
     }
 
-    #[inline]
     fn last_chunk_is_sliced(&self) -> bool {
         self.last_chunk_length != 0
     }
 
-    #[inline]
     fn first_chunk_is_sliced(&self) -> bool {
         self.offset != 0
     }

encodings/fastlanes/src/delta/array/mod.rs

@@ -161,7 +161,6 @@ impl DeltaArray {
         &self.deltas
     }
 
-    #[inline]
     pub(crate) fn lanes(&self) -> usize {
         let ptype =
             PType::try_from(self.dtype()).vortex_expect("DeltaArray DType must be primitive");
@@ -189,17 +188,14 @@ impl DeltaArray {
         self.offset
     }
 
-    #[inline]
     pub(crate) fn bases_len(&self) -> usize {
         self.bases.len()
     }
 
-    #[inline]
     pub(crate) fn deltas_len(&self) -> usize {
         self.deltas.len()
     }
 
-    #[inline]
     pub(crate) fn stats_set(&self) -> &ArrayStats {
         &self.stats_set
     }

encodings/fastlanes/src/for/array/mod.rs

@@ -63,7 +63,6 @@ impl FoRArray {
         &self.reference
     }
 
-    #[inline]
     pub(crate) fn stats_set(&self) -> &ArrayStats {
         &self.stats_set
     }

encodings/fastlanes/src/rle/array/mod.rs

@@ -207,7 +207,6 @@ impl RLEArray {
         self.offset
     }
 
-    #[inline]
     pub(crate) fn stats_set(&self) -> &ArrayStats {
         &self.stats_set
     }

encodings/runend/src/iter.rs

@@ -6,7 +6,6 @@ use std::cmp::min;
 use vortex_array::dtype::IntegerPType;
 use vortex_error::vortex_panic;
 
-#[inline]
 pub fn trimmed_ends_iter<E: IntegerPType>(
     run_ends: &[E],
     offset: usize,

vortex-array/src/aggregate_fn/fns/nan_count/primitive.rs

@@ -15,7 +15,6 @@ pub(super) fn accumulate_primitive(count: &mut u64, p: &PrimitiveArray) -> Vorte
     Ok(())
 }
 
-#[inline]
 fn compute_nan_count_with_validity<T: NativePType>(values: &[T], validity: Mask) -> usize {
     match validity {
         Mask::AllTrue(_) => values.iter().filter(|v| v.is_nan()).count(),

vortex-array/src/arrays/decimal/compute/min_max.rs

@@ -28,7 +28,6 @@ impl MinMaxKernel for Decimal {
 
 register_kernel!(MinMaxKernelAdapter(Decimal).lift());
 
-#[inline]
 fn compute_min_max_with_validity<D>(array: &DecimalArray) -> VortexResult<Option<MinMaxResult>>
 where
     D: Into<DecimalValue> + NativeDecimalType,