Repository navigation
hipfort 0.11.0 for ROCm 10.2.0
Added
- Regenerated the bindings against the ROCm 10.2 API, adding 77 routines in
HIP, hipSOLVER, rocSOLVER and rocSPARSE. - The hipFFT Xt API, in the new
hipfort_hipfftxtmodule (hipfort::hipfftxt
CMake component).use hipfort_hipfftxtgives the whole hipFFT API.
hipfort_hipfftkeeps thehipfft.hAPI only. The multi-process routines
exist only in a hipFFT built with MPI, which the ROCm packages are not. - Array overloads for every routine with a data buffer, including the ILP64
(_64) routines of rocBLAS, hipBLAS and rocSOLVER, with_devptrtwins that
take atype(c_ptr)wherever an argument may be a device pointer. X_strided_rankarray overloads for the strided-batched routines of rocBLAS,
hipBLAS, rocSOLVER and hipSOLVER, which take the batch as one more dimension,
as inA(lda, n, batch_count).- The optional arguments of
hipStreamGetCaptureInfo_sptand
hipStreamGetCaptureInfo_v2_spt.
Changed
-
Breaking. Arguments are now classified against the C documentation.
- The single values that rocSOLVER writes on the GPU (
info,nev,
n_sweeps,residual,rcond, ...) and theinfoof
hipsolverDnXpotrfandhipsolverDnXpotrsare taken as one value on the
device, not as an array. - Host values (the
infoofhipblasXgeqrfandhipblasXgetrs, and a few
rocSPARSE and hipSPARSE arguments) lose their_devptrforms. Pass a
Fortran variable. - The array overloads of the hipSOLVER batched routines take
devInfoas an
array ofbatch_countentries.
Calls through the generic names with a
type(c_ptr)compile as before, but a call
to a specific procedure by name may need its_devptrform. - The single values that rocSOLVER writes on the GPU (
-
Breaking. The array overloads of
hipMalloc,hipFree,hipMemcpyand
related routines are now part of the Fortran 2008 interfaces, so a build with
-DHIPFORT_USE_FPOINTER_INTERFACES=OFFno longer has them. -
Breaking.
hipfftXtSetWorkAreamoves fromhipfort_hipfftto the new
hipfort_hipfftxtmodule, with the rest of the hipFFT Xt API. Code that calls
it writesuse hipfort_hipfftxt. -
Breaking for calls by keyword. The
hipblasSetVector/GetVector/
SetMatrix/GetMatrixfamilies name their arguments as in the hipBLAS
headers (elemSize,AP,BP). -
hipblasXrotBatchedandhipblasXrotStridedBatchedtakecandsas
scalars read per the pointer mode, as rocBLAS does. -
The archives are compiled with one section per procedure. Link with
-Wl,--gc-sectionsto keep only the wrappers you call and to link against an
older ROCm. A shared library also needs
-Wl,--exclude-libs,libhipfort-amdgcn.a. -
The assumed-rank overloads (
-DHIPFORT_ASSUMED_RANK=ON) no longer declare
their arrayscontiguous, like the per-rank ones, so a section is passed as
it is, without a copy. A submatrix works when the leading dimension is that of
the whole array, and a strided vector when the increment matches the stride. -
The modules compile 2 to 3 times faster with gfortran and amdflang, and much
faster with the Cray compiler.
Fixed
- On the CUDA backend, the GPUDirect RDMA device attributes had their HIP
values instead of their CUDA values. - A failed
hipMalloc,hipMallocManagedorhipHostMallocarray overload
now leaves its pointer unassociated. -DHIPFORT_ASSUMED_RANK=ONwith nvfortran now falls back to the per-rank
interfaces, since itsc_locof an assumed-rank argument is wrong.- Warnings with the Cray compiler.
- The
find_packagetest on platforms that install tolib64.