Revert "Attempt at making decoding faster with SIMD"

This reverts commit e85658b42b5373ef7e54295b100d1f083d55dd8d.

Author: sfc-gh-pgaur

cpp/src/arrow/util/alp/Alp.cc

@@ -24,7 +24,6 @@
 #include <map>
 
 #include "arrow/util/alp/AlpConstants.h"
-#include "arrow/util/alp/alp_simd_internal.h"
 #include "arrow/util/bit_stream_utils_internal.h"
 #include "arrow/util/bpacking_internal.h"
 #include "arrow/util/logging.h"
@@ -681,38 +680,17 @@ void AlpCompression<T>::DecodeVector(TargetType* output_vector,
   const ExactType* data = input_vector.data();
   const ExactType frame_of_ref = vector_info.frame_of_reference;
 
-  // Use SIMD-optimized path when T == TargetType (no type conversion needed)
-  if constexpr (std::is_same_v<T, TargetType>) {
-    // Get the two factors separately to preserve exact floating-point precision
-    // Original decode: value * GetFactor(factor) * Constants::GetFactor(exponent)
-    const int64_t int_factor =
-        AlpConstants::GetFactor(vector_info.exponent_and_factor.factor);
-    const T float_factor =
-        Constants::GetFactor(vector_info.exponent_and_factor.exponent);
-
-    if constexpr (std::is_same_v<T, double>) {
-      internal::DecodeVectorSimdDouble(data, output_vector, num_elements, frame_of_ref,
-                                       int_factor, float_factor);
-    } else {
-      static_assert(std::is_same_v<T, float>);
-      internal::DecodeVectorSimdFloat(data, output_vector, num_elements, frame_of_ref,
-                                      int_factor, float_factor);
-    }
-  } else {
-    // Fall back to scalar loop for type conversions (e.g., float -> double)
-    // to preserve exact floating-point behavior
 #pragma GCC unroll AlpConstants::kLoopUnrolls
 #pragma GCC ivdep
-    for (size_t i = 0; i < num_elements; ++i) {
-      // 1. Apply frame of reference (unFOR) - unsigned arithmetic
-      const ExactType unfored_value = data[i] + frame_of_ref;
-      // 2. Reinterpret as signed integer for decode
-      SignedExactType signed_value;
-      std::memcpy(&signed_value, &unfored_value, sizeof(SignedExactType));
-      // 3. Decode using original function to preserve exact floating-point behavior
-      output_vector[i] =
-          AlpInlines<T>::DecodeValue(signed_value, vector_info.exponent_and_factor);
-    }
+  for (size_t i = 0; i < num_elements; ++i) {
+    // 1. Apply frame of reference (unFOR) - unsigned arithmetic
+    const ExactType unfored_value = data[i] + frame_of_ref;
+    // 2. Reinterpret as signed integer for decode
+    SignedExactType signed_value;
+    std::memcpy(&signed_value, &unfored_value, sizeof(SignedExactType));
+    // 3. Decode using original function to preserve exact floating-point behavior
+    output_vector[i] =
+        AlpInlines<T>::DecodeValue(signed_value, vector_info.exponent_and_factor);
   }
 }