This repository provides pre-built MLIR binaries.
Standalone zstd executables are also provided as separate assets for each supported platform to facilitate decompression.
For installation instructions, please refer to the setup-mlir repository.
The repository provides
- an action for setting up MLIR in GitHub Actions and
- installation scripts for setting up MLIR locally.
If desired, you can run the staged build scripts directly. Refer to
scripts/toolchain/linux/build-zstd.sh,scripts/toolchain/linux/build-mold.sh, andscripts/toolchain/linux/build-mlir.shfor Linux,scripts/toolchain/macos/build-zstd.shandscripts/toolchain/macos/build-mlir.shfor macOS, andscripts/toolchain/windows/build-zstd.ps1,scripts/toolchain/windows/build-lld.ps1, andscripts/toolchain/windows/build-mlir.ps1for Windows.
The usage is documented in each script. Linux builds run in a manylinux container and therefore require Docker on the host system.
Windows Debug MLIR archives may be published as split assets named
llvm-mlir_<version>_<target>_debug.tar.zst.part-000, .part-001, ... to stay
below GitHub's per-asset size limit. Reassemble the parts (in order) before
extracting the archive.
PowerShell:
$parts = Get-ChildItem -File "llvm-mlir_*_debug.tar.zst.part-*" | Sort-Object Name
$out = "llvm-mlir_debug.tar.zst"
if (Test-Path $out) { Remove-Item $out -Force }
$target = [System.IO.File]::Open($out, [System.IO.FileMode]::CreateNew)
try {
foreach ($p in $parts) {
$in = [System.IO.File]::OpenRead($p.FullName)
try { $in.CopyTo($target) } finally { $in.Dispose() }
}
} finally {
$target.Dispose()
}Bash:
cat llvm-mlir_*_debug.tar.zst.part-* > llvm-mlir_debug.tar.zst