derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
127 lines
17 KiB
Fortran
127 lines
17 KiB
Fortran
|
|
// RUN: fir-opt --split-input-file --cuf-device-global %s | FileCheck %s
|
|
// RUN: fir-opt --split-input-file --cuf-device-global="cuda-unified=true" %s | FileCheck %s --check-prefix=UNIFIED
|
|
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
fir.global @_QMmtestsEn(dense<[3, 4, 5, 6, 7]> : tensor<5xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<5xi32>
|
|
|
|
gpu.module @cuda_device_mod {
|
|
}
|
|
}
|
|
|
|
// CHECK: gpu.module @cuda_device_mo
|
|
// CHECK-NEXT: fir.global @_QMmtestsEn(dense<[3, 4, 5, 6, 7]> : tensor<5xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<5xi32>
|
|
|
|
// -----
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
fir.global @_QMm1ECb(dense<[90, 100, 110]> : tensor<3xi32>) constant : !fir.array<3xi32>
|
|
fir.global @_QMm2ECc(dense<[100, 200, 300]> : tensor<3xi32>) constant : !fir.array<3xi32>
|
|
}
|
|
|
|
// CHECK: fir.global @_QMm1ECb
|
|
// CHECK: fir.global @_QMm2ECc
|
|
// CHECK: gpu.module @cuda_device_mod
|
|
// CHECK-DAG: fir.global @_QMm2ECc
|
|
// CHECK-DAG: fir.global @_QMm1ECb
|
|
|
|
// -----
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
fir.global @_QMmEddarrays {data_attr = #cuf.cuda<managed>} : !fir.box<!fir.heap<!fir.array<?x!fir.type<_QMmTdevicearrays{phi_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_i:!fir.box<!fir.heap<!fir.array<?xf64>>>}>>>> {
|
|
%c0 = arith.constant 0 : index
|
|
%0 = fir.zero_bits !fir.heap<!fir.array<?x!fir.type<_QMmTdevicearrays{phi_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_i:!fir.box<!fir.heap<!fir.array<?xf64>>>}>>>
|
|
%1 = fir.shape %c0 : (index) -> !fir.shape<1>
|
|
%2 = fir.embox %0(%1) {allocator_idx = 3 : i32} : (!fir.heap<!fir.array<?x!fir.type<_QMmTdevicearrays{phi_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_i:!fir.box<!fir.heap<!fir.array<?xf64>>>}>>>, !fir.shape<1>) -> !fir.box<!fir.heap<!fir.array<?x!fir.type<_QMmTdevicearrays{phi_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_i:!fir.box<!fir.heap<!fir.array<?xf64>>>}>>>>
|
|
fir.has_value %2 : !fir.box<!fir.heap<!fir.array<?x!fir.type<_QMmTdevicearrays{phi_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,phi0_i:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_r:!fir.box<!fir.heap<!fir.array<?xf64>>>,buf_i:!fir.box<!fir.heap<!fir.array<?xf64>>>}>>>>
|
|
}
|
|
fir.global linkonce_odr @_QMmE.dt.devicearrays constant target : !fir.type<_QM__fortran_type_infoTderivedtype{binding:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTbinding{proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>}>>>>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>,sizeinbytes:i64,uninstantiated:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,kindparameter:!fir.box<!fir.ptr<!fir.array<?xi64>>>,lenparameterkind:!fir.box<!fir.ptr<!fir.array<?xi8>>>,component:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,genre:i8,category:i8,kind:i8,rank:i8,__padding0:!fir.array<4xi8>,offset:i64,characterlen:!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>,derived:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,lenvalue:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,bounds:!fir.box<!fir.ptr<!fir.array<?x?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>>>,procptr:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTprocptrcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,offset:i64,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,special:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTspecialbinding{which:i8,isargdescriptorset:i8,istypebound:i8,isargcontiguousset:i8,__padding0:!fir.array<4xi8>,proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,specialbitset:i32,hasparent:i8,noinitializationneeded:i8,nodestructionneeded:i8,nofinalizationneeded:i8,nodefinedassignment:i8,__padding0:!fir.array<3xi8>}> {
|
|
%0 = fir.undefined !fir.type<_QM__fortran_type_infoTderivedtype{binding:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTbinding{proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>}>>>>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>,sizeinbytes:i64,uninstantiated:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,kindparameter:!fir.box<!fir.ptr<!fir.array<?xi64>>>,lenparameterkind:!fir.box<!fir.ptr<!fir.array<?xi8>>>,component:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,genre:i8,category:i8,kind:i8,rank:i8,__padding0:!fir.array<4xi8>,offset:i64,characterlen:!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>,derived:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,lenvalue:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,bounds:!fir.box<!fir.ptr<!fir.array<?x?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>>>,procptr:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTprocptrcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,offset:i64,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,special:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTspecialbinding{which:i8,isargdescriptorset:i8,istypebound:i8,isargcontiguousset:i8,__padding0:!fir.array<4xi8>,proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,specialbitset:i32,hasparent:i8,noinitializationneeded:i8,nodestructionneeded:i8,nofinalizationneeded:i8,nodefinedassignment:i8,__padding0:!fir.array<3xi8>}>
|
|
fir.has_value %0 : !fir.type<_QM__fortran_type_infoTderivedtype{binding:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTbinding{proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>}>>>>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>,sizeinbytes:i64,uninstantiated:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,kindparameter:!fir.box<!fir.ptr<!fir.array<?xi64>>>,lenparameterkind:!fir.box<!fir.ptr<!fir.array<?xi8>>>,component:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,genre:i8,category:i8,kind:i8,rank:i8,__padding0:!fir.array<4xi8>,offset:i64,characterlen:!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>,derived:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,lenvalue:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,bounds:!fir.box<!fir.ptr<!fir.array<?x?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>>>,procptr:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTprocptrcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,offset:i64,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,special:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTspecialbinding{which:i8,isargdescriptorset:i8,istypebound:i8,isargcontiguousset:i8,__padding0:!fir.array<4xi8>,proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,specialbitset:i32,hasparent:i8,noinitializationneeded:i8,nodestructionneeded:i8,nofinalizationneeded:i8,nodefinedassignment:i8,__padding0:!fir.array<3xi8>}>
|
|
}
|
|
}
|
|
|
|
// CHECK-NAG: fir.global @_QMmEddarrays
|
|
// CHECK-NAG: fir.global linkonce_odr @_QMmE.dt.devicearrays
|
|
// CHECK: gpu.module @cuda_device_mod
|
|
// CHECK-NAG: fir.global @_QMmEddarrays
|
|
// CHECK-NAG: fir.global linkonce_odr @_QMmE.dt.devicearrays
|
|
|
|
// -----
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
func.func @_QMmod1Ptest1() attributes {cuf.proc_attr = #cuf.cuda_proc<global>} {
|
|
%34 = fir.alloca !fir.type<_QMvector_typesTv2real2{x:f16,y:f16}> {bindc_name = "cv", uniq_name = "_QMmod1Ftest1Ecv"}
|
|
return
|
|
}
|
|
fir.global linkonce_odr @_QMvector_typesE.dt.v2real2 constant : !fir.type<_QM__fortran_type_infoTderivedtype{binding:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTbinding{proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>}>>>>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>,sizeinbytes:i64,uninstantiated:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,kindparameter:!fir.box<!fir.ptr<!fir.array<?xi64>>>,lenparameterkind:!fir.box<!fir.ptr<!fir.array<?xi8>>>,component:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,genre:i8,category:i8,kind:i8,rank:i8,memoryspace:i8,__padding0:!fir.array<3xi8>,offset:i64,characterlen:!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>,derived:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,lenvalue:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,bounds:!fir.box<!fir.ptr<!fir.array<?x?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>>>,procptr:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTprocptrcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,offset:i64,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,special:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTspecialbinding{which:i8,isargdescriptorset:i8,istypebound:i8,specialcaseflag:i8,__padding0:!fir.array<4xi8>,proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,specialbitset:i32,hasparent:i8,noinitializationneeded:i8,nodestructionneeded:i8,nofinalizationneeded:i8,nodefinedassignment:i8,__padding0:!fir.array<3xi8>}> {
|
|
%0 = fir.undefined !fir.type<_QM__fortran_type_infoTderivedtype{binding:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTbinding{proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>}>>>>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>,sizeinbytes:i64,uninstantiated:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,kindparameter:!fir.box<!fir.ptr<!fir.array<?xi64>>>,lenparameterkind:!fir.box<!fir.ptr<!fir.array<?xi8>>>,component:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,genre:i8,category:i8,kind:i8,rank:i8,memoryspace:i8,__padding0:!fir.array<3xi8>,offset:i64,characterlen:!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>,derived:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,lenvalue:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,bounds:!fir.box<!fir.ptr<!fir.array<?x?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>>>,procptr:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTprocptrcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,offset:i64,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,special:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTspecialbinding{which:i8,isargdescriptorset:i8,istypebound:i8,specialcaseflag:i8,__padding0:!fir.array<4xi8>,proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,specialbitset:i32,hasparent:i8,noinitializationneeded:i8,nodestructionneeded:i8,nofinalizationneeded:i8,nodefinedassignment:i8,__padding0:!fir.array<3xi8>}>
|
|
fir.has_value %0 : !fir.type<_QM__fortran_type_infoTderivedtype{binding:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTbinding{proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>}>>>>,name:!fir.box<!fir.ptr<!fir.char<1,?>>>,sizeinbytes:i64,uninstantiated:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,kindparameter:!fir.box<!fir.ptr<!fir.array<?xi64>>>,lenparameterkind:!fir.box<!fir.ptr<!fir.array<?xi8>>>,component:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,genre:i8,category:i8,kind:i8,rank:i8,memoryspace:i8,__padding0:!fir.array<3xi8>,offset:i64,characterlen:!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>,derived:!fir.box<!fir.ptr<!fir.type<_QM__fortran_type_infoTderivedtype>>>,lenvalue:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,bounds:!fir.box<!fir.ptr<!fir.array<?x?x!fir.type<_QM__fortran_type_infoTvalue{genre:i8,__padding0:!fir.array<7xi8>,value:i64}>>>>,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>>>,procptr:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTprocptrcomponent{name:!fir.box<!fir.ptr<!fir.char<1,?>>>,offset:i64,initialization:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,special:!fir.box<!fir.ptr<!fir.array<?x!fir.type<_QM__fortran_type_infoTspecialbinding{which:i8,isargdescriptorset:i8,istypebound:i8,specialcaseflag:i8,__padding0:!fir.array<4xi8>,proc:!fir.type<_QM__fortran_builtinsT__builtin_c_funptr{__address:i64}>}>>>>,specialbitset:i32,hasparent:i8,noinitializationneeded:i8,nodestructionneeded:i8,nofinalizationneeded:i8,nodefinedassignment:i8,__padding0:!fir.array<3xi8>}>
|
|
}
|
|
}
|
|
|
|
// CHECK-LABEL: func.func @_QMmod1Ptest1()
|
|
// CHECK: fir.global linkonce_odr @_QMvector_typesE.dt.v2real2
|
|
// CHECK-LABEL: gpu.module @cuda_device_mod
|
|
// CHECK: fir.global linkonce_odr @_QMvector_typesE.dt.v2real2
|
|
|
|
// -----
|
|
|
|
// Test that when a global already exists in the gpu.module, the pass
|
|
// continues cloning the remaining candidates instead of stopping.
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
fir.global @_QMmEa(dense<[1, 2, 3]> : tensor<3xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<3xi32>
|
|
fir.global @_QMmEb(dense<[10, 20, 30]> : tensor<3xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<3xi32>
|
|
fir.global @_QMmEc(dense<[100, 200, 300]> : tensor<3xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<3xi32>
|
|
|
|
gpu.module @cuda_device_mod {
|
|
fir.global @_QMmEa(dense<[1, 2, 3]> : tensor<3xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<3xi32>
|
|
}
|
|
}
|
|
|
|
// CHECK: gpu.module @cuda_device_mod
|
|
// CHECK-DAG: fir.global @_QMmEa
|
|
// CHECK-DAG: fir.global @_QMmEb
|
|
// CHECK-DAG: fir.global @_QMmEc
|
|
|
|
// -----
|
|
|
|
// Under -gpu=mem:unified (cuda-unified=true), plain host module-scope
|
|
// variables referenced from device code are mirrored as no-body external
|
|
// declarations in the GPU module. PTX lowers them as `.extern .global ...`.
|
|
// CUFAddConstructor + the runtime then map the device-side extern to the
|
|
// host pointer via __cudaRegisterHostVar.
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
fir.global @_QMmtestsEm(dense<[1, 2, 3, 4, 5]> : tensor<5xi32>) : !fir.array<5xi32>
|
|
func.func @_QMmtestsPg1() attributes {cuf.proc_attr = #cuf.cuda_proc<global>} {
|
|
%0 = fir.address_of(@_QMmtestsEm) : !fir.ref<!fir.array<5xi32>>
|
|
return
|
|
}
|
|
}
|
|
|
|
// Host-side definition is preserved.
|
|
// UNIFIED: fir.global @_QMmtestsEm(dense<[1, 2, 3, 4, 5]> : tensor<5xi32>) : !fir.array<5xi32>
|
|
// GPU-module clone is an external declaration (no init body, no `dense<...>`).
|
|
// UNIFIED: gpu.module @cuda_device_mod
|
|
// UNIFIED: fir.global @_QMmtestsEm : !fir.array<5xi32>
|
|
// UNIFIED-NOT: fir.global @_QMmtestsEm{{.*}}dense
|
|
|
|
// -----
|
|
|
|
// Globals with an explicit CUF data attribute (device, managed, constant)
|
|
// keep their existing definition-clone path even with cuda-unified=true.
|
|
|
|
module attributes {fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module} {
|
|
fir.global @_QMmtestsEdev(dense<[1, 2, 3]> : tensor<3xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<3xi32>
|
|
func.func @_QMmtestsPg1() attributes {cuf.proc_attr = #cuf.cuda_proc<global>} {
|
|
%0 = fir.address_of(@_QMmtestsEdev) : !fir.ref<!fir.array<3xi32>>
|
|
return
|
|
}
|
|
}
|
|
|
|
// UNIFIED: gpu.module @cuda_device_mod
|
|
// UNIFIED: fir.global @_QMmtestsEdev(dense<[1, 2, 3]> : tensor<3xi32>) {data_attr = #cuf.cuda<device>} : !fir.array<3xi32>
|