diff --git a/cmake/SfizzSIMDSourceFiles.cmake b/cmake/SfizzSIMDSourceFiles.cmake
index 962d0fb0..edca20e6 100644
--- a/cmake/SfizzSIMDSourceFiles.cmake
+++ b/cmake/SfizzSIMDSourceFiles.cmake
@@ -2,10 +2,7 @@ macro(sfizz_add_simd_sources SOURCES_VAR PREFIX)
     # It needs a macro, otherwise the source properties cannot take effect.
 
     list (APPEND ${SOURCES_VAR}
-        ${PREFIX}/sfizz/SIMDSSE.cpp
-        ${PREFIX}/sfizz/SIMDHelpers.cpp
-        ${PREFIX}/sfizz/SIMDNEON.cpp
-        ${PREFIX}/sfizz/SIMDDummy.cpp)
+        ${PREFIX}/sfizz/SIMDHelpers.cpp)
 
     # For CPU-dispatched X86 sources
     # Always build them for all X86 targets.
diff --git a/dpf.mk b/dpf.mk
index a2734283..e9bfb051 100644
--- a/dpf.mk
+++ b/dpf.mk
@@ -96,9 +96,7 @@ SFIZZ_SOURCES = \
 	src/sfizz/sfizz_wrapper.cpp \
 	src/sfizz/SfzFilter.cpp \
 	src/sfizz/SfzHelpers.cpp \
-	src/sfizz/SIMDDummy.cpp \
-	src/sfizz/SIMDNEON.cpp \
-	src/sfizz/SIMDSSE.cpp \
+	src/sfizz/SIMDHelpers.cpp \
 	src/sfizz/Synth.cpp \
 	src/sfizz/Tuning.cpp \
 	src/sfizz/Voice.cpp \
diff --git a/scripts/run_clang_tidy.sh b/scripts/run_clang_tidy.sh
index 6c1f6f8a..370171df 100755
--- a/scripts/run_clang_tidy.sh
+++ b/scripts/run_clang_tidy.sh
@@ -17,7 +17,7 @@ clang-tidy \
   src/sfizz/sfizz.cpp \
   src/sfizz/Region.cpp \
   src/sfizz/SfzHelpers.cpp \
-  src/sfizz/SIMDSSE.cpp \
+  src/sfizz/SIMDHelpers.cpp \
   src/sfizz/Synth.cpp \
   src/sfizz/Voice.cpp \
   src/sfizz/effects/Eq.cpp \
diff --git a/src/sfizz/SIMDDummy.cpp b/src/sfizz/SIMDDummy.cpp
deleted file mode 100644
index 718d8a1c..00000000
--- a/src/sfizz/SIMDDummy.cpp
+++ /dev/null
@@ -1,177 +0,0 @@
-// SPDX-License-Identifier: BSD-2-Clause
-
-// This code is part of the sfizz library and is licensed under a BSD 2-clause
-// license. You should have receive a LICENSE.md file along with the code.
-// If not, contact the sfizz maintainers at https://github.com/sfztools/sfizz
-
-#include "SIMDConfig.h"
-
-#if !(SFIZZ_HAVE_SSE2 || SFIZZ_HAVE_NEON)
-
-#include "SIMDHelpers.h"
-
-template <>
-void sfz::readInterleaved<float, true>(absl::Span<const float> input, absl::Span<float> outputLeft, absl::Span<float> outputRight) noexcept
-{
-    readInterleaved<float, false>(input, outputLeft, outputRight);
-}
-
-template <>
-void sfz::writeInterleaved<float, true>(absl::Span<const float> inputLeft, absl::Span<const float> inputRight, absl::Span<float> output) noexcept
-{
-    writeInterleaved<float, false>(inputLeft, inputRight, output);
-}
-
-template <>
-void sfz::fill<float, true>(absl::Span<float> output, float value) noexcept
-{
-    fill<float, false>(output, value);
-}
-
-template <>
-void sfz::exp<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    exp<float, false>(input, output);
-}
-
-template <>
-void sfz::log<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    log<float, false>(input, output);
-}
-
-template <>
-void sfz::sin<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    sin<float, false>(input, output);
-}
-
-template <>
-void sfz::cos<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    cos<float, false>(input, output);
-}
-
-template <>
-void sfz::applyGain<float, true>(float gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    applyGain<float, false>(gain, input, output);
-}
-
-template <>
-void sfz::applyGain<float, true>(absl::Span<const float> gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    applyGain<float, false>(gain, input, output);
-}
-
-template <>
-void sfz::divide<float, true>(absl::Span<const float> input, absl::Span<const float> divisor, absl::Span<float> output) noexcept
-{
-    divide<float, false>(input, divisor, output);
-}
-
-template <>
-void sfz::multiplyAdd<float, true>(absl::Span<const float> gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    multiplyAdd<float, false>(gain, input, output);
-}
-
-template <>
-void sfz::multiplyAdd<float, true>(const float gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    multiplyAdd<float, false>(gain, input, output);
-}
-
-template <>
-float sfz::loopingSFZIndex<float, true>(absl::Span<const float> jumps, absl::Span<float> leftCoeff, absl::Span<float> rightCoeff, absl::Span<int> indices, float floatIndex, float loopEnd, float loopStart) noexcept
-{
-    return loopingSFZIndex<float, false>(jumps, leftCoeff, rightCoeff, indices, floatIndex, loopEnd, loopStart);
-}
-
-template <>
-float sfz::saturatingSFZIndex<float, true>(absl::Span<const float> jumps, absl::Span<float> leftCoeff, absl::Span<float> rightCoeff, absl::Span<int> indices, float floatIndex, float loopEnd) noexcept
-{
-    return saturatingSFZIndex<float, false>(jumps, leftCoeff, rightCoeff, indices, floatIndex, loopEnd);
-}
-
-
-template <>
-float sfz::linearRamp<float, true>(absl::Span<float> output, float start, float step) noexcept
-{
-    return linearRamp<float, false>(output, start, step);
-}
-
-template <>
-float sfz::multiplicativeRamp<float, true>(absl::Span<float> output, float start, float step) noexcept
-{
-    return multiplicativeRamp<float, false>(output, start, step);
-}
-
-template <>
-void sfz::add<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    add<float, false>(input, output);
-}
-
-template <>
-void sfz::add<float, true>(float value, absl::Span<float> output) noexcept
-{
-    add<float, false>(value, output);
-}
-
-template <>
-void sfz::subtract<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    subtract<float, false>(input, output);
-}
-
-template <>
-void sfz::subtract<float, true>(const float value, absl::Span<float> output) noexcept
-{
-    subtract<float, false>(value, output);
-}
-
-
-template <>
-void sfz::copy<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    copy<float, false>(input, output);
-}
-
-template <>
-void sfz::pan<float, true>(absl::Span<const float> panEnvelope, absl::Span<float> leftBuffer, absl::Span<float> rightBuffer) noexcept
-{
-    pan<float, false>(panEnvelope, leftBuffer, rightBuffer);
-}
-
-template <>
-float sfz::mean<float, true>(absl::Span<const float> vector) noexcept
-{
-    return mean<float, false>(vector);
-}
-
-template <>
-float sfz::meanSquared<float, true>(absl::Span<const float> vector) noexcept
-{
-    return meanSquared<float, false>(vector);
-}
-
-template <>
-void sfz::cumsum<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    cumsum<float, false>(input, output);
-}
-
-template<>
-void sfz::sfzInterpolationCast<float, true>(absl::Span<const float> floatJumps, absl::Span<int> jumps, absl::Span<float> coeffs) noexcept
-{
-    sfzInterpolationCast<float, false>(floatJumps, jumps, coeffs);
-}
-
-template <>
-void sfz::diff<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    diff<float, false>(input, output);
-}
-
-#endif // !(SFIZZ_HAVE_SSE2 || SFIZZ_HAVE_NEON)
diff --git a/src/sfizz/SIMDHelpers.cpp b/src/sfizz/SIMDHelpers.cpp
index da87bc8e..5fd4d226 100644
--- a/src/sfizz/SIMDHelpers.cpp
+++ b/src/sfizz/SIMDHelpers.cpp
@@ -9,6 +9,10 @@
 #include <emmintrin.h>
 #endif
 
+#if SFIZZ_HAVE_NEON
+#include <arm_neon.h>
+#endif
+
 namespace sfz {
 
 static std::array<bool, static_cast<unsigned>(SIMDOps::_sentinel)> simdStatus;
@@ -100,6 +104,12 @@ void readInterleaved(const float* input, float* outputLeft, float* outputRight,
             // Fallthrough from lastAligned to sentinel
         }
 #endif
+
+#if 0 // NEON wip
+        auto reg = vld2q_f32(in);
+        vst1q_f32(lOut, reg.val[0]);
+        vst1q_f32(rOut, reg.val[1]);
+#endif
     }
 
     while (input < sentinel)
diff --git a/src/sfizz/SIMDHelpers.h b/src/sfizz/SIMDHelpers.h
index 5d3fea02..acc2c526 100644
--- a/src/sfizz/SIMDHelpers.h
+++ b/src/sfizz/SIMDHelpers.h
@@ -12,16 +12,8 @@
  * These functions are templated to apply on
  * various underlying buffer types, and this file contains the generic version of the
  * function. Some templates specializations exists for different architecture that try
- * to make use of SIMD intrinsics; you can find such a file in SIMDSSE.cpp and possibly
- * someday SIMDNEON.cpp for ARM platforms.
- *
- * If you want to write specializations for float buffers the idea is to start from the SIMDDummy
- * file that just calls back the generic implementation, and implement the specializations you
- * wish from this list. You can then either activate or deactivate a SIMD version by default
- * using the variables in Config.h, or call e.g. writeInterleaved<float, true>(...) to use the
- * SIMD version of writeInterleaved. To implement e.g. double template specializations you
- * will need to amend this file to pre-declare the specializations, and create a file similar to
- * SIMDxxx.cpp.
+ * to make use of SIMD intrinsics in SIMDHelpers.cpp. A runtime dispatch can help if you
+ * write implementation for larger/newer SIMD operations.
  *
  * All the SIMD functions are benchmarked. If you run the benchmark for a given function you can check
  * if it is interesting to run the SIMD version by default. The interest is that you can activate
diff --git a/src/sfizz/SIMDNEON.cpp b/src/sfizz/SIMDNEON.cpp
deleted file mode 100644
index e229d954..00000000
--- a/src/sfizz/SIMDNEON.cpp
+++ /dev/null
@@ -1,241 +0,0 @@
-// Copyright (c) 2019, Paul Ferrand
-// All rights reserved.
-
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are met:
-
-// 1. Redistributions of source code must retain the above copyright notice, this
-//    list of conditions and the following disclaimer.
-// 2. Redistributions in binary form must reproduce the above copyright notice,
-//    this list of conditions and the following disclaimer in the documentation
-//    and/or other materials provided with the distribution.
-
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
-// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
-// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
-// ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-// (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-// LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
-// ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-#include "SIMDConfig.h"
-
-#if SFIZZ_HAVE_NEON
-
-#include "SIMDHelpers.h"
-#include <arm_neon.h>
-
-using Type = float;
-constexpr uintptr_t TypeAlignment { 4 };
-constexpr uintptr_t ByteAlignment { TypeAlignment * sizeof(Type) };
-constexpr uintptr_t ByteAlignmentMask { ByteAlignment - 1 };
-
-float* nextAligned(const float* ptr)
-{
-    return reinterpret_cast<float*>((reinterpret_cast<uintptr_t>(ptr) + ByteAlignmentMask) & (~ByteAlignmentMask));
-}
-
-float* prevAligned(const float* ptr)
-{
-    return reinterpret_cast<float*>(reinterpret_cast<uintptr_t>(ptr) & (~ByteAlignmentMask));
-}
-
-bool unaligned(const float* ptr)
-{
-    return (reinterpret_cast<uintptr_t>(ptr) & ByteAlignmentMask) != 0;
-}
-
-template<class... Args>
-bool unaligned(const float* ptr1, Args... rest)
-{
-    return unaligned(ptr1) || unaligned(rest...);
-}
-
-template <>
-void sfz::readInterleaved<float, true>(absl::Span<const float> input, absl::Span<float> outputLeft, absl::Span<float> outputRight) noexcept
-{
-    // The size of the outputs is not big enough for the input...
-    ASSERT(outputLeft.size() >= input.size() / 2);
-    ASSERT(outputRight.size() >= input.size() / 2);
-    // Input is too small
-    ASSERT(input.size() > 1);
-
-    auto* in = input.begin();
-    auto* lOut = outputLeft.begin();
-    auto* rOut = outputRight.begin();
-
-    const auto size = std::min(input.size(), std::min(outputLeft.size() * 2, outputRight.size() * 2));
-    const auto* lastAligned = prevAligned(input.begin() + size - TypeAlignment);
-
-    while (unaligned(in, lOut, rOut) && in < lastAligned)
-        _internals::snippetRead<float>(in, lOut, rOut);
-
-    while (in < lastAligned) {
-        auto reg = vld2q_f32(in);
-        vst1q_f32(lOut, reg.val[0]);
-        vst1q_f32(rOut, reg.val[1]);
-        // *lOut = reg.val[0];
-        // *rOut = reg.val[1];
-        incrementAll<TypeAlignment>(in, in, lOut, rOut);
-    }
-
-    while (in < input.end() - 1)
-        _internals::snippetRead<float>(in, lOut, rOut);
-}
-
-template <>
-void sfz::writeInterleaved<float, true>(absl::Span<const float> inputLeft, absl::Span<const float> inputRight, absl::Span<float> output) noexcept
-{
-    writeInterleaved<float, false>(inputLeft, inputRight, output);
-}
-
-template <>
-void sfz::fill<float, true>(absl::Span<float> output, float value) noexcept
-{
-    fill<float, false>(output, value);
-}
-
-template <>
-void sfz::exp<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    exp<float, false>(input, output);
-}
-
-template <>
-void sfz::log<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    log<float, false>(input, output);
-}
-
-template <>
-void sfz::sin<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    sin<float, false>(input, output);
-}
-
-template <>
-void sfz::cos<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    cos<float, false>(input, output);
-}
-
-template <>
-void sfz::applyGain<float, true>(float gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    applyGain<float, false>(gain, input, output);
-}
-
-template <>
-void sfz::applyGain<float, true>(absl::Span<const float> gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    applyGain<float, false>(gain, input, output);
-}
-
-template <>
-void sfz::divide<float, true>(absl::Span<const float> input, absl::Span<const float> divisor, absl::Span<float> output) noexcept
-{
-    divide<float, false>(input, divisor, output);
-}
-
-template <>
-void sfz::multiplyAdd<float, true>(absl::Span<const float> gain, absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    multiplyAdd<float, false>(gain, input, output);
-}
-
-template <>
-float sfz::loopingSFZIndex<float, true>(absl::Span<const float> jumps, absl::Span<float> leftCoeff, absl::Span<float> rightCoeff, absl::Span<int> indices, float floatIndex, float loopEnd, float loopStart) noexcept
-{
-    return loopingSFZIndex<float, false>(jumps, leftCoeff, rightCoeff, indices, floatIndex, loopEnd, loopStart);
-}
-
-template <>
-float sfz::saturatingSFZIndex<float, true>(absl::Span<const float> jumps, absl::Span<float> leftCoeff, absl::Span<float> rightCoeff, absl::Span<int> indices, float floatIndex, float loopEnd) noexcept
-{
-    return saturatingSFZIndex<float, false>(jumps, leftCoeff, rightCoeff, indices, floatIndex, loopEnd);
-}
-
-
-template <>
-float sfz::linearRamp<float, true>(absl::Span<float> output, float start, float step) noexcept
-{
-    return linearRamp<float, false>(output, start, step);
-}
-
-template <>
-float sfz::multiplicativeRamp<float, true>(absl::Span<float> output, float start, float step) noexcept
-{
-    return multiplicativeRamp<float, false>(output, start, step);
-}
-
-template <>
-void sfz::add<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    add<float, false>(input, output);
-}
-
-template <>
-void sfz::add<float, true>(float value, absl::Span<float> output) noexcept
-{
-    add<float, false>(value, output);
-}
-
-template <>
-void sfz::subtract<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    subtract<float, false>(input, output);
-}
-
-template <>
-void sfz::subtract<float, true>(const float value, absl::Span<float> output) noexcept
-{
-    subtract<float, false>(value, output);
-}
-
-
-template <>
-void sfz::copy<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    copy<float, false>(input, output);
-}
-
-template <>
-void sfz::pan<float, true>(absl::Span<const float> panEnvelope, absl::Span<float> leftBuffer, absl::Span<float> rightBuffer) noexcept
-{
-    pan<float, false>(panEnvelope, leftBuffer, rightBuffer);
-}
-
-template <>
-float sfz::mean<float, true>(absl::Span<const float> vector) noexcept
-{
-    return mean<float, false>(vector);
-}
-
-template <>
-float sfz::meanSquared<float, true>(absl::Span<const float> vector) noexcept
-{
-    return meanSquared<float, false>(vector);
-}
-
-template <>
-void sfz::cumsum<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    cumsum<float, false>(input, output);
-}
-
-template<>
-void sfz::sfzInterpolationCast<float, true>(absl::Span<const float> floatJumps, absl::Span<int> jumps, absl::Span<float> coeffs) noexcept
-{
-    sfzInterpolationCast<float, false>(floatJumps, jumps, coeffs);
-}
-
-template <>
-void sfz::diff<float, true>(absl::Span<const float> input, absl::Span<float> output) noexcept
-{
-    diff<float, false>(input, output);
-}
-
-#endif // SFIZZ_HAVE_NEON
diff --git a/src/sfizz/SIMDSSE.cpp b/src/sfizz/SIMDSSE.cpp
deleted file mode 100644
index d3c023db..00000000
--- a/src/sfizz/SIMDSSE.cpp
+++ /dev/null
@@ -1,19 +0,0 @@
-// SPDX-License-Identifier: BSD-2-Clause
-
-// This code is part of the sfizz library and is licensed under a BSD 2-clause
-// license. You should have receive a LICENSE.md file along with the code.
-// If not, contact the sfizz maintainers at https://github.com/sfztools/sfizz
-
-#include "SIMDConfig.h"
-
-#if SFIZZ_HAVE_SSE2
-
-#include "SIMDHelpers.h"
-#include <array>
-#include <xmmintrin.h>
-#include <emmintrin.h>
-#include "mathfuns/sse_mathfun.h"
-
-constexpr uintptr_t TypeAlignment = 4;
-
-#endif // SFIZZ_HAVE_SSE2