GeezerVison

ROCm release v7.2.4

ROCm release v7.2.3

ROCm release v7.2.2

ROCm release v7.2.1

These are the release notes for AOMP 23.0-0. AOMP uses AMD developer modifications to the upstream LLVM development trunk. These differences are managed in a branch called "amd-staging". This branch is found in a mirror of upstream LLVM found at https://github.com/ROCm/llvm-project. The amd-staging branch is constantly changing as it merges the upstream development trunk with its downstream development updates. The AMD modifications are experimental while under review for the upstream trunk. AOMP uses a snapshot of amd-staging at the commit ids and dates listed below. AOMP also includes builds of related ROCm components. We call AOMP a "standalone" build as it does not use or require ROCm with the exception of the kernel module (amdgpu-dkms) and libdrm which are often part of the Linux distribution. AOMP is isolated from any ROCm installations by installing into /usr/lib/aomp and the use of RPATH for runtime libraries.

For AOMP 23.0-0, the last LLVM trunk commit is a 82447b92c37842b33e708146ba4ec794bb9fae98 on February 12, 2026. The last amd-only commit is ae56654bd26b8475fc000b1a2a8a80557fe7c83f on February 12, 2026. These commits form a frozen branch now called "aomp-23.0-0". See https://github.com/ROCm/llvm-project/tree/aomp-23.0-0.
The integrated ROCm components for this AOMP release were built with ROCM 7.2.0 sources.
This is the 1st AOMP release based on upstream LLVM 23 development.

These are the changes since 22.0-2:

• We now distribute release binaries in a tar file that was built on ManyLinux (AlmaLinux 8)
• Deprecated docs for installing debians/rpms on Ubuntu, SLES, RHEL, and CentOS
• Switched to ROCm 7.2 sources
• This release does not include the hip math libraries.

Flang updates:

• Fix type error when calling EXIT intrinsic
• Fix recursive mapper emission.
• Fix mapping of constant arrays
  • Fixes error: 'hlfir.designate' op using value defined outside the region
• Allow assumed-size arrays on USE_DEVICE_ADDR clause
• Allow ALLOC/RELEASE in place of STORAGE in 6.0
• Add FIR and LLVM lowering support for prefetch directive
• Hoist static parallel region allocas to the entry block on the CPU
• Hoist static allocas emitted by private init regions to the allocation IP
• Add support for implicit linearization in omp.simd

ROCm release v7.2.0

These are the release notes for AOMP 22.0-2. AOMP uses AMD developer modifications to the upstream LLVM development trunk. These differences are managed in a branch called "amd-staging". This branch is found in a mirror of upstream LLVM found at https://github.com/ROCm/llvm-project. The amd-staging branch is constantly changing as it merges the upstream development trunk with its downstream development updates. The AMD modifications are experimental while under review for the upstream trunk. AOMP uses a snapshot of amd-staging at the commit ids and dates listed below. AOMP also includes builds of related ROCm components. We call AOMP a "standalone" build as it does not use or require ROCm with the exception of the kernel module (amdgpu-dkms) and libdrm which are often part of the Linux distribution. AOMP is isolated from any ROCm installations by installing into /usr/lib/aomp and the use of RPATH for runtime libraries.

For AOMP 22.0-2, the last LLVM trunk commit is a0e0775d91859d6ca32efc5fb8644b6281e98593 on January 12, 2026. The last amd-only commit is 7652844035d910d0d9b4b3070ba515903ad3ca19 on January 18, 2026. These commits form a frozen branch now called "aomp-22.0-2". See https://github.com/ROCm/llvm-project/tree/aomp-22.0-2.
The integrated ROCm components for this AOMP release were built with ROCM 7.1.1 sources.
This is the 2nd AOMP release based on upstream LLVM 22 development.

These are the changes since 22.0-1:

• Switch to ROCm 7.1.1 sources
• Added a small patch to Tensile to fix issue with v_dot4 op_sel modifiers
• We recommend using ROCm 7.1 amdgpu-dkms driver with this release

Updates to flang:

• Fix declare target to implicit mapping
• Fix assert SIInstrInfo.cpp:907 AMDGPU::VGPR_32RegClass.contains(SrcReg)
• Defer descriptor mapping for assumed dummy argument types
• Fix negative array indexing with allocatable derived type array maps
• Add standalone tile support
• Implicitly map nested allocatable components in derived types
• Add support for -fno-fast-real-mod (now enabled as part of -ffast-math)
• Enable no-loop kernels
• User defined reductions are now supported
• Improved support for split distribute and parallel
• Add debug type support for procedure pointers
• Swap to attach semantics for descriptor mapping
• Emit module summary for Full LTO to improve link times
• Set REQUIRES flags on program unit symbol
• Fix USM close semantics and use_device_ptr
• Fixed flang issue (error: failure in HLFIR intrinsic simplification) with GenASiS

Errata:

• Flang - UMT runtime failure

ROCm release v7.1.1

ROCm release v7.1.0

ROCm release v7.0.2