Update dependencies

.devcontainer/Dockerfile

@@ -43,7 +43,7 @@ ENTRYPOINT [ "/usr/local/share/docker-init.sh" ]
 CMD [ "sleep", "infinity" ]
 
 ARG MINIFORGE_NAME=Miniforge3
-ARG MINIFORGE_VERSION=24.3.0-0
+ARG MINIFORGE_VERSION=25.11.0-1
 ARG CONDA_DIR=/opt/conda
 
 # Based on https://github.com/conda-forge/miniforge-images/blob/master/ubuntu/Dockerfile
@@ -63,8 +63,8 @@ RUN wget --no-hsts --quiet https://github.com/conda-forge/miniforge/releases/dow
 # Create a conda environment from the environment file in the repo root.
 COPY --from=file-normalizer --chown=$USER_UID:conda /data/environment.yml /tmp/build/
 RUN umask 0002 \
-    && ${CONDA_DIR}/bin/mamba env create -f /tmp/build/environment.yml \
-    && ${CONDA_DIR}/bin/mamba clean -fy \
+    && ${CONDA_DIR}/bin/conda env create -f /tmp/build/environment.yml \
+    && ${CONDA_DIR}/bin/conda clean -fy \
     && sudo chown -R :conda ${CONDA_DIR}/envs
 
 # Add a file that is to be sourced from .bashrc and from the devops pipeline stages
@@ -90,7 +90,7 @@ RUN . /opt/conda/etc/profile.d/conda.sh \
 
 
 # Install the yardl tool
-ARG YARDL_VERSION=0.6.4
+ARG YARDL_VERSION=0.6.6
 RUN wget --quiet "https://github.com/microsoft/yardl/releases/download/v${YARDL_VERSION}/yardl_${YARDL_VERSION}_linux_x86_64.tar.gz" \
     && tar -xzf "yardl_${YARDL_VERSION}_linux_x86_64.tar.gz" \
     && mv yardl "/opt/conda/envs/${CONDA_ENVIRONMENT_NAME}/bin/" \

.github/actions/configure-mrd-build-environment/action.yml

@@ -17,13 +17,12 @@ runs:
         tar -xzf "yardl_${YARDL_VERSION}_linux_x86_64.tar.gz"
         pwd >> $GITHUB_PATH
 
-    - name: Setup Mambaforge
+    - name: Setup Miniforge
       uses: conda-incubator/setup-miniconda@v3
       with:
         miniforge-version: latest
         # Do not specify environment file - see Cache step below
         activate-environment: mrd
-        use-mamba: true
 
     - name: Get Date
       id: get-date
@@ -43,7 +42,7 @@ runs:
 
     - name: Update Environment
       shell: bash -el {0}
-      run: mamba env update -n mrd -f environment.yml
+      run: conda env update -n mrd -f environment.yml
       if: steps.cache-conda.outputs.cache-hit != 'true'
 
     - name: Set up MATLAB

.github/workflows/mrd_build.yml

@@ -11,7 +11,7 @@ on:
 env:
   # Increase this to manually reset the conda environment cache
   CONDA_CACHE_NUMBER: 0
-  YARDL_VERSION: 0.6.4
+  YARDL_VERSION: 0.6.6
 
 defaults:
   run:

VERSION

@@ -1 +1 @@
-2.1.2-preview
+2.2.0

cpp/CMakeLists.txt

@@ -28,7 +28,7 @@ endif ()
 message(STATUS "Build type: ${CMAKE_BUILD_TYPE}")
 
 if(NOT CMAKE_CXX_STANDARD)
-    set(CMAKE_CXX_STANDARD 17)
+    set(CMAKE_CXX_STANDARD 20)
 endif()
 message(STATUS "C++ Standard: ${CMAKE_CXX_STANDARD}")
 

cpp/conda/build.sh

@@ -9,7 +9,7 @@ echo 'Building mrd conda package using CMake...'
 
 cmake -GNinja \
     -DCMAKE_BUILD_TYPE=Release \
-    -DCMAKE_CXX_STANDARD=17 \
+    -DCMAKE_CXX_STANDARD=20 \
     -DCMAKE_INSTALL_PREFIX=${PREFIX} \
     -DCMAKE_OSX_SYSROOT=${CONDA_BUILD_SYSROOT} \
     ../

cpp/conda/meta.yaml

@@ -17,21 +17,18 @@ requirements:
     - cmake
     - ninja
   host:
-    - fftw=3.3.9
-    - fmt=12.0.0
+    - fftw=3.3.*
     - howardhinnant_date=3.0.3
-    - hdf5=1.10.6
-    - imagemagick=7.0.11_13
-    - ismrmrd>=1.14.2
+    - hdf5=1.14.*
+    - imagemagick>=7.1.2_8
+    - ismrmrd>=1.15.0
     - nlohmann_json=3.12.0
-    - xtensor=0.25.0
-    - xtensor-fftw=0.2.5
+    - xtensor=0.27.1
   run:
-    - fmt=12.0.0
-    - hdf5=1.10.6
-    - imagemagick=7.0.11_13
-    - ismrmrd>=1.14.2
-    - fftw=3.3.9
+    - fftw=3.3.*
+    - hdf5=1.14.*
+    - imagemagick=7.1.2_8
+    - ismrmrd>=1.15.0
 
 about:
   home: https://ismrmrd.github.io/mrd

cpp/mrd-tools/CMakeLists.txt

@@ -10,7 +10,9 @@ add_executable(mrd_stream_to_hdf5 mrd_stream_to_hdf5.cc)
 target_link_libraries(mrd_stream_to_hdf5 mrd_generated)
 
 add_executable(mrd_stream_recon mrd_stream_recon.cc)
-target_link_libraries(mrd_stream_recon fftw3f mrd_generated)
+target_link_libraries(mrd_stream_recon mrd_generated fftw3f)
+# Suppress false positive warnings from GCC 15+ with xtensor complex operations
+target_compile_options(mrd_stream_recon PRIVATE "-Wno-array-bounds" "-Wno-stringop-overflow")
 
 install(TARGETS
     mrd_hdf5_to_stream
@@ -38,17 +40,16 @@ endif()
 
 
 find_package(ImageMagick COMPONENTS Magick++)
-find_package(fmt)
 
-if (ImageMagick_FOUND AND fmt_FOUND)
-  message(STATUS "ImageMagick and fmt found.")
+if (ImageMagick_FOUND)
+  message(STATUS "ImageMagick found.")
   include_directories(${ImageMagick_INCLUDE_DIRS})
 
   add_executable(mrd_image_stream_to_png mrd_image_stream_to_png.cc)
   target_compile_options(mrd_image_stream_to_png PRIVATE "-DMAGICKCORE_QUANTUM_DEPTH=8" "-DMAGICKCORE_HDRI_ENABLE=0")
-  target_link_libraries(mrd_image_stream_to_png mrd_generated ${ImageMagick_LIBRARIES} fmt::fmt)
+  target_link_libraries(mrd_image_stream_to_png mrd_generated ${ImageMagick_LIBRARIES})
 
   install(TARGETS mrd_image_stream_to_png DESTINATION bin)
 else()
-  message(STATUS "ImageMagick or fmt not found. Skipping mrd_image_stream_to_png.")
+  message(STATUS "ImageMagick not found. Skipping mrd_image_stream_to_png.")
 endif()

cpp/mrd-tools/converters.cc

@@ -4,7 +4,7 @@
 #include <ismrmrd/version.h>
 
 #include <date/date.h>
-#include <xtensor/xview.hpp>
+#include <xtensor/views/xview.hpp>
 
 namespace mrd::converters {
 

cpp/mrd-tools/fftw_wrappers.h

@@ -0,0 +1,114 @@
+#pragma once
+
+#include <complex>
+#include <cstring>
+#include <fftw3.h>
+#include <xtensor.hpp>
+#include <xtensor/misc/xmanipulation.hpp>
+
+namespace fftw_wrappers {
+
+namespace detail {
+
+// Helper function for 1D FFTW operations
+template <class E>
+inline auto fftw_1d_impl(E&& e, int direction, bool normalize) -> xt::xarray<std::complex<float>> {
+  auto size = e.size();
+
+  std::vector<size_t> shape = {size};
+  xt::xarray<std::complex<float>> result = xt::zeros<std::complex<float>>(shape);
+  xt::xarray<std::complex<float>> input = e;
+
+  fftwf_complex* in = reinterpret_cast<fftwf_complex*>(input.data());
+  fftwf_complex* out = reinterpret_cast<fftwf_complex*>(result.data());
+
+  fftwf_plan plan = fftwf_plan_dft_1d(size, in, out, direction, FFTW_ESTIMATE);
+  fftwf_execute(plan);
+  fftwf_destroy_plan(plan);
+
+  if (normalize) {
+    result /= static_cast<float>(size);
+  }
+
+  return result;
+}
+
+// Helper function for 2D FFTW operations
+template <class E>
+inline auto fftw_2d_impl(E&& e, int direction, bool normalize) -> xt::xarray<std::complex<float>> {
+  auto shape = e.shape();
+
+  if (shape.size() != 2) {
+    throw std::runtime_error(direction == FFTW_FORWARD ? "fft2 requires 2D input" : "ifft2 requires 2D input");
+  }
+
+  auto n0 = shape[0];
+  auto n1 = shape[1];
+
+  std::vector<size_t> result_shape = {n0, n1};
+  xt::xarray<std::complex<float>> result = xt::zeros<std::complex<float>>(result_shape);
+  xt::xarray<std::complex<float>> input = e;
+
+  fftwf_complex* in = reinterpret_cast<fftwf_complex*>(input.data());
+  fftwf_complex* out = reinterpret_cast<fftwf_complex*>(result.data());
+
+  fftwf_plan plan = fftwf_plan_dft_2d(n0, n1, in, out, direction, FFTW_ESTIMATE);
+  fftwf_execute(plan);
+  fftwf_destroy_plan(plan);
+
+  if (normalize) {
+    result /= static_cast<float>(n0 * n1);
+  }
+
+  return result;
+}
+} // namespace detail
+
+// FFT shift functions
+template <class E>
+inline auto fftshift(E&& e) -> xt::xarray<typename std::decay_t<E>::value_type> {
+  xt::xarray<typename std::decay_t<E>::value_type> result = e;
+
+  for (std::size_t ax = 0; ax < result.dimension(); ++ax) {
+    auto size = result.shape()[ax];
+    auto shift = size / 2;
+    result = xt::roll(result, shift, ax);
+  }
+  return result;
+}
+
+template <class E>
+inline auto ifftshift(E&& e) -> xt::xarray<typename std::decay_t<E>::value_type> {
+  xt::xarray<typename std::decay_t<E>::value_type> result = e;
+
+  for (std::size_t ax = 0; ax < result.dimension(); ++ax) {
+    auto size = result.shape()[ax];
+    auto shift = (size + 1) / 2; // Different from fftshift for odd sizes
+    result = xt::roll(result, shift, ax);
+  }
+  return result;
+}
+
+// 1D FFT/IFFT wrappers - now using shared implementation
+template <class E>
+inline auto fft(E&& e) -> xt::xarray<std::complex<float>> {
+  return detail::fftw_1d_impl(std::forward<E>(e), FFTW_FORWARD, false);
+}
+
+template <class E>
+inline auto ifft(E&& e) -> xt::xarray<std::complex<float>> {
+  return detail::fftw_1d_impl(std::forward<E>(e), FFTW_BACKWARD, true);
+}
+
+// 2D FFT/IFFT wrappers - now using shared implementation
+template <class E>
+inline auto fft2(E&& e) -> xt::xarray<std::complex<float>> {
+  return detail::fftw_2d_impl(std::forward<E>(e), FFTW_FORWARD, false);
+}
+
+template <class E>
+inline auto ifft2(E&& e) -> xt::xarray<std::complex<float>> {
+  return detail::fftw_2d_impl(std::forward<E>(e), FFTW_BACKWARD, true);
+}
+
+} // namespace fftw_wrappers

cpp/mrd-tools/ismrmrd_to_mrd.cc

@@ -6,11 +6,9 @@
 #include <ismrmrd/meta.h>
 #include <ismrmrd/serialization_iostream.h>
 #include <ismrmrd/version.h>
-#include <xtensor/xview.hpp>
 
 #include <date/date.h>
 
-
 void print_usage(std::string program_name) {
   std::cerr << "Usage: " << program_name << std::endl;
   std::cerr << "  -i|--input   <input ISMRMRD stream> (default: stdin)" << std::endl;

cpp/mrd-tools/mrd_image_stream_to_png.cc

@@ -1,14 +1,7 @@
-// NOTE:
-// GCC 13+ appears to emit false positive warnings for array-bounds
-// when using xt::views (specifically the calls to `xt::amax` below).
-// The statements are correct and there's no overflow, so we can suppress the warnings.
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Warray-bounds"
 #include "mrd/binary/protocols.h"
-#pragma GCC diagnostic pop
 
 #include <Magick++.h>
-#include <fmt/format.h>
+#include <format>
 #include <iostream>
 
 void print_usage(std::string program_name) {
@@ -97,8 +90,8 @@ int main(int argc, char** argv) {
     image.depth(8);
     image.type(Magick::GrayscaleType);
 
-    // Use fmt to generate filename from prefix and increment with 6 digits (e.g. prefix_000001.png)
-    std::string filename = fmt::format("{}{:05d}.png", prefix, image_count++);
+    // Generate filename from prefix and increment with 5 digits (e.g. prefix_00001.png)
+    std::string filename = std::format("{}{:05d}.png", prefix, image_count++);
     image.write(filename);
     if (verbose) {
       std::cerr << "Generated image " << filename << std::endl;

cpp/mrd-tools/mrd_phantom.cc

@@ -1,13 +1,12 @@
 #include "mrd/binary/protocols.h"
 #include "shepp_logan_phantom.h"
+#include "fftw_wrappers.h"
 
 #include <iostream>
 #include <random>
-#include <xtensor-fftw/basic.hpp>
-#include <xtensor-fftw/helper.hpp>
-#include <xtensor/xio.hpp>
-#include <xtensor/xrandom.hpp>
-#include <xtensor/xview.hpp>
+#include <xtensor/io/xio.hpp>
+#include <xtensor/generators/xrandom.hpp>
+#include <xtensor/views/xview.hpp>
 
 using namespace mrd;
 
@@ -329,7 +328,7 @@ int main(int argc, char** argv) {
   // Apply FFT
   coil_images = fftshift(coil_images);
   for (unsigned int c = 0; c < ncoils; c++) {
-    auto tmp1 = xt::fftw::fft2(xt::xarray<std::complex<float>>(xt::view(coil_images, c, 0, xt::all(), xt::all())));
+    auto tmp1 = fftw_wrappers::fft2(xt::xarray<std::complex<float>>(xt::view(coil_images, c, 0, xt::all(), xt::all())));
     tmp1 /= std::sqrt(1.0f * tmp1.size());
     xt::view(coil_images, c, 0, xt::all(), xt::all()) = tmp1;
   }

cpp/mrd-tools/mrd_stream_recon.cc

@@ -1,24 +1,9 @@
 
+#include "fftw_wrappers.h"
 #include "mrd/binary/protocols.h"
 #include "mrd/protocols.h"
 #include "mrd/types.h"
-
-#include <xtensor-fftw/basic.hpp>
-#include <xtensor-fftw/helper.hpp>
-
-// NOTE:
-// GCC 13+ appears to emit false positive warnings for array-bounds and stringop-overflow
-// when using xt::view with xt::range (specifically in `remove_oversampling` below).
-// The assignment is correct and there's no overflow, so we can suppress the warnings.
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Warray-bounds"
-#pragma GCC diagnostic ignored "-Wstringop-overflow"
-#include <xtensor/xview.hpp>
-#pragma GCC diagnostic pop
-
-xt::xtensor<std::complex<float>, 4> fftshift(xt::xtensor<std::complex<float>, 4> x) {
-  return xt::roll(xt::roll(x, x.shape(3) / 2, 3), x.shape(2) / 2, 2);
-}
+#include <xtensor/views/xview.hpp>
 
 void remove_oversampling(mrd::Acquisition& acq, const mrd::EncodingType& enc) {
   auto new_shape = acq.data.shape();
@@ -30,16 +15,20 @@ void remove_oversampling(mrd::Acquisition& acq, const mrd::EncodingType& enc) {
   xt::xtensor<std::complex<float>, 2> new_data = xt::zeros<std::complex<float>>(new_shape);
   for (size_t c = 0; c < acq.Coils(); c++) {
     auto ft_line = xt::xarray<std::complex<float>>(xt::view(acq.data, c, xt::all()));
-    ft_line = xt::fftw::fftshift(xt::fftw::ifft(xt::fftw::ifftshift(ft_line)));
+    ft_line = fftw_wrappers::fftshift(fftw_wrappers::ifft(fftw_wrappers::ifftshift(ft_line)));
     ft_line *= std::sqrt(1.0f * ft_line.size());
     ft_line = xt::view(ft_line, xt::range(x_pad, rNx + x_pad));
-    ft_line = xt::fftw::fftshift(xt::fftw::fft(xt::fftw::ifftshift(ft_line)));
+    ft_line = fftw_wrappers::fftshift(fftw_wrappers::fft(fftw_wrappers::ifftshift(ft_line)));
     ft_line /= std::sqrt(1.0f * ft_line.size());
     xt::view(new_data, c, xt::all()) = ft_line;
   }
   acq.data = new_data;
 }
 
+xt::xtensor<std::complex<float>, 4> fftshift(xt::xtensor<std::complex<float>, 4> x) {
+  return xt::roll(xt::roll(x, x.shape(3) / 2, 3), x.shape(2) / 2, 2);
+}
+
 void print_usage(std::string program_name) {
   std::cerr << "Usage: " << program_name << std::endl;
   std::cerr << "  -i|--input   <input MRD stream> (default: stdin)" << std::endl;
@@ -190,7 +179,7 @@ int main(int argc, char** argv) {
           auto slice = xt::view(buffer, c, s, xt::all(), xt::all(), xt::all(), xt::all());
           slice = fftshift(slice);
           for (unsigned int coil = 0; coil < slice.shape(0); coil++) {
-            auto tmp1 = xt::fftw::ifft2(xt::xarray<std::complex<float>>(xt::view(slice, coil, 0, xt::all(), xt::all())));
+            auto tmp1 = fftw_wrappers::ifft2(xt::xarray<std::complex<float>>(xt::view(slice, coil, 0, xt::all(), xt::all())));
             tmp1 *= std::sqrt(1.0f * tmp1.size());
             xt::view(slice, coil, 0, xt::all(), xt::all()) = tmp1;
           }