diff --git a/image_analysis/image_preprocessing/ImagePreprocessorBuffer.cpp b/image_analysis/image_preprocessing/ImagePreprocessorBuffer.cpp
index abdb2c53..0a2728a3 100644
--- a/image_analysis/image_preprocessing/ImagePreprocessorBuffer.cpp
+++ b/image_analysis/image_preprocessing/ImagePreprocessorBuffer.cpp
@@ -5,35 +5,6 @@
 
 ImagePreprocessorBuffer::ImagePreprocessorBuffer(size_t npixels) : buffer(npixels) {}
 
-// Standard CPU operation
-std::vector<int32_t> &ImagePreprocessorBuffer::getBuffer() {
-    return buffer;
-}
-
-const std::vector<int32_t> &ImagePreprocessorBuffer::getBuffer() const {
-    return buffer;
-}
-
-int32_t &ImagePreprocessorBuffer::operator[](size_t i) {
-    return buffer[i];
-}
-
-const int32_t &ImagePreprocessorBuffer::operator[](size_t i) const {
-    return buffer[i];
-}
-
-size_t ImagePreprocessorBuffer::size() const {
-    return buffer.size();
-}
-
-int32_t *ImagePreprocessorBuffer::data() {
-    return buffer.data();
-}
-
-const int32_t *ImagePreprocessorBuffer::data() const {
-    return buffer.data();
-}
-
 void ImagePreprocessorBuffer::Gather(const std::vector<uint32_t> &npixel, std::vector<int32_t> &values) const {
     values.resize(npixel.size());
     for (size_t i = 0; i < npixel.size(); i++)
diff --git a/image_analysis/image_preprocessing/ImagePreprocessorBuffer.h b/image_analysis/image_preprocessing/ImagePreprocessorBuffer.h
index 3644ff03..a0ef88c5 100644
--- a/image_analysis/image_preprocessing/ImagePreprocessorBuffer.h
+++ b/image_analysis/image_preprocessing/ImagePreprocessorBuffer.h
@@ -14,15 +14,19 @@ public:
     explicit ImagePreprocessorBuffer(size_t npixels);
     virtual ~ImagePreprocessorBuffer() = default;
 
-    // Standard CPU operation
-    std::vector<int32_t> &getBuffer();
-    const std::vector<int32_t> &getBuffer() const;
-    int32_t &operator[](size_t i);
-    const int32_t &operator[](size_t i) const;
-    size_t size() const;
+    // Standard CPU operation. Defined here rather than in the .cpp: these are called once per pixel by
+    // the CPU preprocessor, the CPU azimuthal integrator and the CPU spot finder, and the build has no
+    // link-time optimisation - so out of line they are a real call per pixel, and they stop the callers
+    // vectorising at all. Inlining them costs about a third of the per-image CPU work on every path
+    // that does not have a GPU, and leaves the results bit-identical.
+    std::vector<int32_t> &getBuffer() { return buffer; }
+    const std::vector<int32_t> &getBuffer() const { return buffer; }
+    int32_t &operator[](size_t i) { return buffer[i]; }
+    const int32_t &operator[](size_t i) const { return buffer[i]; }
+    size_t size() const { return buffer.size(); }
 
-    int32_t *data();
-    const int32_t *data() const;
+    int32_t *data() { return buffer.data(); }
+    const int32_t *data() const { return buffer.data(); }
 
     // GPU operations (overriden in ImagePreprocessorBufferGPU
     virtual int32_t *getGPUBuffer();