AMDGPU.jl

AMDGPU v0.2.2

Diff since v0.2.1

Closed issues:

• Implement RNGs (#14)
• Allow 0-argument kernels (#10)
• Add tests to match CUDAnative (#22)
• Build fails on OSX (#19)
• Add options to at-roc to allow initializing globals (#36)
• Some errors during test. Are they cause for concern? (#70)

Merged pull requests:

• Implement OCKL wavefront ops (#49) (@jpsamaroo)
• Accept hooks for user-defined globals (#55) (@jpsamaroo)
• Add at-rocprintf support (#61) (@jpsamaroo)
• UB GPUArrays to 5.0/5.1 (#71) (@jpsamaroo)
• rocRAND (#72) (@antholzer)
• Update to Julia 1.6 (#75) (@jpsamaroo)
• Some fixes for Julia 1.6 (#78) (@jpsamaroo)
• Bump GPUArrays to 6 (#79) (@jpsamaroo)
• Add Buildkite CI (#80) (@jpsamaroo)
• Allow 0-argument kernels (#81) (@jpsamaroo)
• Skip build if in AutoMerge (#82) (@jpsamaroo)

AMDGPU v0.2.1

Diff since v0.2.0

Closed issues:

• Throw error in wait() call on queue error (#7)
• Test code doesn't work (#43)
• Broken build script? (#46)
• HSASignal and HSAKernelInstance references can be accidentally GC'd (#63)

Merged pull requests:

• Made build and initialization quieter (#47) (@jpsamaroo)
• Refcount HSA objects for safe shutdown (#52) (@jpsamaroo)
• Adapt to GPUCompiler disk cache changes (#53) (@jpsamaroo)
• Disable broken GPUArrays tests (#54) (@jpsamaroo)
• Fixed malloc return type (#59) (@jpsamaroo)
• Remove HSA C files from repo (#60) (@jpsamaroo)
• Don't do set_used on GVs (#62) (@jpsamaroo)
• Add barrier packet support (#65) (@jpsamaroo)
• Bump to 0.2.1, use GPUCompiler 0.8 (#66) (@jpsamaroo)

AMDGPU v0.2.0

Diff since v0.1.2

Merged pull requests:

• Reimplement ROCArray using GPUArrays (#32) (@jpsamaroo)

AMDGPU v0.1.2

Diff since v0.1.1

Merged pull requests:

• CompatHelper: bump compat for "Setfield" to "0.7" (#28) (@github-actions[bot])
• fix for finding HIP binaries (#35) (@wsphillips)
• CompatHelper: bump compat for "GPUCompiler" to "0.6" (#37) (@github-actions[bot])
• CompatHelper: bump compat for "GPUArrays" to "5.1" (#38) (@github-actions[bot])
• Test on Julia 1.5 (#39) (@jpsamaroo)
• Switch to GPUCompiler master (#40) (@jpsamaroo)
• Move to stateless LLVM contexts (#41) (@jpsamaroo)

AMDGPU v0.1.1

Diff since v0.1.0

Closed issues:

• What is needed to build? (#29)

Merged pull requests:

• CompatHelper: bump compat for "GPUArrays" to "4.0" (#3) (@github-actions[bot])
• Small cleanup in deps/ (#30) (@jpsamaroo)
• Improve kernel-thrown exceptions (#31) (@jpsamaroo)

AMDGPU v0.1.0

Merged pull requests:

• Changes for the package merge (#1) (@jpsamaroo)