315 lines · plain
1// RUN: fir-opt --split-input-file --fir-to-llvm-ir="target=x86_64-unknown-linux-gnu" %s | FileCheck %s2 3module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<f80, dense<128> : vector<2xi64>>, #dlti.dl_entry<i128, dense<128> : vector<2xi64>>, #dlti.dl_entry<i64, dense<64> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr<272>, dense<64> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<271>, dense<32> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<270>, dense<32> : vector<4xi64>>, #dlti.dl_entry<f128, dense<128> : vector<2xi64>>, #dlti.dl_entry<f64, dense<64> : vector<2xi64>>, #dlti.dl_entry<f16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i32, dense<32> : vector<2xi64>>, #dlti.dl_entry<i16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i8, dense<8> : vector<2xi64>>, #dlti.dl_entry<i1, dense<8> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr, dense<64> : vector<4xi64>>, #dlti.dl_entry<"dlti.endianness", "little">, #dlti.dl_entry<"dlti.stack_alignment", 128 : i64>>} {4 func.func @_QQmain() attributes {fir.bindc_name = "cufkernel_global"} {5 %c0 = arith.constant 0 : index6 %0 = fir.address_of(@_QQclX3C737464696E3E00) : !fir.ref<!fir.char<1,8>>7 %c4_i32 = arith.constant 4 : i328 %c48 = arith.constant 48 : index9 %1 = fir.convert %c48 : (index) -> i6410 %2 = fir.convert %0 : (!fir.ref<!fir.char<1,8>>) -> !fir.ref<i8>11 %3 = fir.call @_FortranACUFAllocDescriptor(%1, %2, %c4_i32) : (i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>>12 %4 = fir.convert %3 : (!fir.ref<!fir.box<none>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>13 %5 = fir.zero_bits !fir.heap<!fir.array<?xi32>>14 %6 = fircg.ext_embox %5(%c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?xi32>>, index) -> !fir.box<!fir.heap<!fir.array<?xi32>>>15 fir.store %6 to %4 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>16 %8 = fir.load %3 : !fir.ref<!fir.box<none>>17 return18 }19 20 // CHECK-LABEL: llvm.func @_QQmain()21 // CHECK-COUNT-2: llvm.call @_FortranACUFAllocDescriptor 22 23 fir.global linkonce @_QQclX3C737464696E3E00 constant : !fir.char<1,8> {24 %0 = fir.string_lit "<stdin>\00"(8) : !fir.char<1,8>25 fir.has_value %0 : !fir.char<1,8>26 }27 func.func private @_FortranACUFAllocDescriptor(i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>> attributes {fir.runtime}28}29 30// -----31 32module attributes {dlti.dl_spec = #dlti.dl_spec<f80 = dense<128> : vector<2xi64>, i128 = dense<128> : vector<2xi64>, i64 = dense<64> : vector<2xi64>, !llvm.ptr<272> = dense<64> : vector<4xi64>, !llvm.ptr<271> = dense<32> : vector<4xi64>, !llvm.ptr<270> = dense<32> : vector<4xi64>, f128 = dense<128> : vector<2xi64>, f64 = dense<64> : vector<2xi64>, f16 = dense<16> : vector<2xi64>, i32 = dense<32> : vector<2xi64>, i16 = dense<16> : vector<2xi64>, i8 = dense<8> : vector<2xi64>, i1 = dense<8> : vector<2xi64>, !llvm.ptr = dense<64> : vector<4xi64>, "dlti.endianness" = "little", "dlti.stack_alignment" = 128 : i64>} {33 func.func @_QQmain() attributes {fir.bindc_name = "test"} {34 %c10 = arith.constant 10 : index35 %c20 = arith.constant 20 : index36 %0 = fir.address_of(@_QQclX64756D6D792E6D6C697200) : !fir.ref<!fir.char<1,11>>37 %c4 = arith.constant 4 : index38 %c200 = arith.constant 200 : index39 %1 = arith.muli %c200, %c4 : index40 %c6_i32 = arith.constant 6 : i3241 %c0_i32 = arith.constant 0 : i3242 %2 = fir.convert %1 : (index) -> i6443 %3 = fir.convert %0 : (!fir.ref<!fir.char<1,11>>) -> !fir.ref<i8>44 %4 = fir.call @_FortranACUFMemAlloc(%2, %c0_i32, %3, %c6_i32) : (i64, i32, !fir.ref<i8>, i32) -> !fir.llvm_ptr<i8>45 %5 = fir.convert %4 : (!fir.llvm_ptr<i8>) -> !fir.ref<!fir.array<10x20xi32>>46 %6 = fircg.ext_embox %5(%c10, %c20) : (!fir.ref<!fir.array<10x20xi32>>, index, index) -> !fir.box<!fir.array<10x20xi32>>47 return48 }49 fir.global linkonce @_QQclX64756D6D792E6D6C697200 constant : !fir.char<1,11> {50 %0 = fir.string_lit "dummy.mlir\00"(11) : !fir.char<1,11>51 fir.has_value %0 : !fir.char<1,11>52 }53 func.func private @_FortranACUFMemAlloc(i64, i32, !fir.ref<i8>, i32) -> !fir.llvm_ptr<i8> attributes {fir.runtime}54}55 56// CHECK-LABEL: llvm.func @_QQmain()57// CHECK: llvm.call @_FortranACUFMemAlloc58// CHECK: llvm.call @_FortranACUFAllocDescriptor59 60// -----61 62module attributes {dlti.dl_spec = #dlti.dl_spec<f80 = dense<128> : vector<2xi64>, i128 = dense<128> : vector<2xi64>, i64 = dense<64> : vector<2xi64>, !llvm.ptr<272> = dense<64> : vector<4xi64>, !llvm.ptr<271> = dense<32> : vector<4xi64>, !llvm.ptr<270> = dense<32> : vector<4xi64>, f128 = dense<128> : vector<2xi64>, f64 = dense<64> : vector<2xi64>, f16 = dense<16> : vector<2xi64>, i32 = dense<32> : vector<2xi64>, i16 = dense<16> : vector<2xi64>, i8 = dense<8> : vector<2xi64>, i1 = dense<8> : vector<2xi64>, !llvm.ptr = dense<64> : vector<4xi64>, "dlti.endianness" = "little", "dlti.stack_alignment" = 128 : i64>} {63 func.func @_QQmain() attributes {fir.bindc_name = "p1"} {64 %c1_i32 = arith.constant 1 : i3265 %c0_i32 = arith.constant 0 : i3266 %c16_i32 = arith.constant 16 : i3267 %c1 = arith.constant 1 : index68 %c0 = arith.constant 0 : index69 %0 = fir.alloca i32 {bindc_name = "iblk", uniq_name = "_QFEiblk"}70 %1 = fir.alloca i32 {bindc_name = "ithr", uniq_name = "_QFEithr"}71 %2 = fir.address_of(@_QQclX64756D6D792E6D6C697200) : !fir.ref<!fir.char<1,11>>72 %c14_i32 = arith.constant 14 : i3273 %c72 = arith.constant 72 : index74 %3 = fir.convert %c72 : (index) -> i6475 %4 = fir.convert %2 : (!fir.ref<!fir.char<1,11>>) -> !fir.ref<i8>76 %5 = fir.call @_FortranACUFAllocDescriptor(%3, %4, %c14_i32) : (i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>>77 %6 = fir.convert %5 : (!fir.ref<!fir.box<none>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>78 %7 = fir.zero_bits !fir.heap<!fir.array<?x?xf32>>79 %8 = fircg.ext_embox %7(%c0, %c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?x?xf32>>, index, index) -> !fir.box<!fir.heap<!fir.array<?x?xf32>>>80 fir.store %8 to %6 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>81 %9 = fir.address_of(@_QQclX64756D6D792E6D6C697200) : !fir.ref<!fir.char<1,11>>82 %c20_i32 = arith.constant 20 : i3283 %c48 = arith.constant 48 : index84 %10 = fir.convert %c48 : (index) -> i6485 %11 = fir.convert %9 : (!fir.ref<!fir.char<1,11>>) -> !fir.ref<i8>86 %12 = fir.call @_FortranACUFAllocDescriptor(%10, %11, %c20_i32) : (i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>>87 %13 = fir.convert %12 : (!fir.ref<!fir.box<none>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>88 %14 = fir.zero_bits !fir.heap<!fir.array<?xf32>>89 %15 = fircg.ext_embox %14(%c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?xf32>>, index) -> !fir.box<!fir.heap<!fir.array<?xf32>>>90 fir.store %15 to %13 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>91 %16 = fir.convert %6 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>) -> !fir.ref<!fir.box<none>>92 %17 = fir.convert %c1 : (index) -> i6493 %18 = fir.convert %c16_i32 : (i32) -> i6494 fir.call @_FortranAAllocatableSetBounds(%16, %c0_i32, %17, %18) fastmath<contract> : (!fir.ref<!fir.box<none>>, i32, i64, i64) -> ()95 fir.call @_FortranAAllocatableSetBounds(%16, %c1_i32, %17, %18) fastmath<contract> : (!fir.ref<!fir.box<none>>, i32, i64, i64) -> ()96 %21 = fir.address_of(@_QQclX64756D6D792E6D6C697200) : !fir.ref<!fir.char<1,11>>97 %c31_i32 = arith.constant 31 : i3298 %false = arith.constant false99 %22 = fir.absent !fir.box<none>100 %c-1_i64 = arith.constant -1 : i64101 %23 = fir.convert %6 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>) -> !fir.ref<!fir.box<none>>102 %24 = fir.convert %21 : (!fir.ref<!fir.char<1,11>>) -> !fir.ref<i8>103 %25 = fir.call @_FortranACUFAllocatableAllocate(%23, %c-1_i64, %false, %22, %24, %c31_i32) : (!fir.ref<!fir.box<none>>, i64, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32104 %26 = fir.convert %13 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>105 fir.call @_FortranAAllocatableSetBounds(%26, %c0_i32, %17, %18) fastmath<contract> : (!fir.ref<!fir.box<none>>, i32, i64, i64) -> ()106 %28 = fir.address_of(@_QQclX64756D6D792E6D6C697200) : !fir.ref<!fir.char<1,11>>107 %c34_i32 = arith.constant 34 : i32108 %false_0 = arith.constant false109 %29 = fir.absent !fir.box<none>110 %c-1_i64_1 = arith.constant -1 : i64111 %30 = fir.convert %13 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>112 %31 = fir.convert %28 : (!fir.ref<!fir.char<1,11>>) -> !fir.ref<i8>113 %32 = fir.call @_FortranACUFAllocatableAllocate(%30, %c-1_i64_1, %false_0, %29, %31, %c34_i32) : (!fir.ref<!fir.box<none>>, i64, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32114 %33 = fir.load %6 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>115 %34 = fircg.ext_rebox %33 : (!fir.box<!fir.heap<!fir.array<?x?xf32>>>) -> !fir.box<!fir.array<?x?xf32>>116 return117 }118 func.func private @_FortranAAllocatableSetBounds(!fir.ref<!fir.box<none>>, i32, i64, i64) -> () attributes {fir.runtime}119 fir.global linkonce @_QQclX64756D6D792E6D6C697200 constant : !fir.char<1,11> {120 %0 = fir.string_lit "dummy.mlir\00"(11) : !fir.char<1,11>121 fir.has_value %0 : !fir.char<1,11>122 }123 func.func private @_FortranACUFAllocDescriptor(i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>> attributes {fir.runtime}124 func.func private @_FortranACUFAllocatableAllocate(!fir.ref<!fir.box<none>>, i64, i1, !fir.box<none>, !fir.ref<i8>, i32) -> i32 attributes {fir.runtime}125}126 127// CHECK-LABEL: llvm.func @_QQmain()128// CHECK-COUNT-4: llvm.call @_FortranACUFAllocDescriptor129 130// -----131 132module attributes {dlti.dl_spec = #dlti.dl_spec<!llvm.ptr<270> = dense<32> : vector<4xi64>, f128 = dense<128> : vector<2xi64>, f64 = dense<64> : vector<2xi64>, f16 = dense<16> : vector<2xi64>, i32 = dense<32> : vector<2xi64>, i64 = dense<64> : vector<2xi64>, !llvm.ptr<272> = dense<64> : vector<4xi64>, !llvm.ptr<271> = dense<32> : vector<4xi64>, f80 = dense<128> : vector<2xi64>, i128 = dense<128> : vector<2xi64>, i16 = dense<16> : vector<2xi64>, i8 = dense<8> : vector<2xi64>, !llvm.ptr = dense<64> : vector<4xi64>, i1 = dense<8> : vector<2xi64>, "dlti.endianness" = "little", "dlti.stack_alignment" = 128 : i64>, fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module, llvm.data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128", llvm.ident = "flang version 20.0.0 (git@github.com:clementval/llvm-project.git efc2415bcce8e8a9e73e77aa122c8aba1c1fbbd2)", llvm.target_triple = "x86_64-unknown-linux-gnu"} {133 func.func @_QQmain() {134 %c1_i32 = arith.constant 1 : i32135 %c2 = arith.constant 2 : index136 %c10 = arith.constant 10 : index137 %c1 = arith.constant 1 : index138 %c0 = arith.constant 0 : index139 %0 = fir.address_of(@_QQclX3C737464696E3E00) : !fir.ref<!fir.char<1,8>>140 %c11_i32 = arith.constant 11 : i32141 %c72 = arith.constant 72 : index142 %1 = fir.convert %c72 : (index) -> i64143 %2 = fir.convert %0 : (!fir.ref<!fir.char<1,8>>) -> !fir.ref<i8>144 %3 = fir.call @_FortranACUFAllocDescriptor(%1, %2, %c11_i32) : (i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>>145 %4 = fir.convert %3 : (!fir.ref<!fir.box<none>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>146 %5 = fir.zero_bits !fir.heap<!fir.array<?x?xi32>>147 %6 = fircg.ext_embox %5(%c0, %c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?x?xi32>>, index, index) -> !fir.box<!fir.heap<!fir.array<?x?xi32>>>148 fir.store %6 to %4 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>149 %7 = fir.load %4 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>150 %8 = fir.box_addr %7 : (!fir.box<!fir.heap<!fir.array<?x?xi32>>>) -> !fir.heap<!fir.array<?x?xi32>>151 %c0_0 = arith.constant 0 : index152 %9:3 = fir.box_dims %7, %c0_0 : (!fir.box<!fir.heap<!fir.array<?x?xi32>>>, index) -> (index, index, index)153 %c1_1 = arith.constant 1 : index154 %10:3 = fir.box_dims %7, %c1_1 : (!fir.box<!fir.heap<!fir.array<?x?xi32>>>, index) -> (index, index, index)155 %11 = fircg.ext_embox %8(%9#1, %10#1) origin %9#0, %10#0[%c1, %c10, %c1, %c1, %c2, %c1] : (!fir.heap<!fir.array<?x?xi32>>, index, index, index, index, index, index, index, index, index, index) -> !fir.box<!fir.array<10x2xi32>>156 return157 }158 gpu.module @cuda_device_mod {159 gpu.func @_QMassumedPglob(%arg0: !fir.box<!fir.array<?x?xi32>>) kernel {160 gpu.return161 }162 }163 fir.global linkonce @_QQclX3C737464696E3E00 constant : !fir.char<1,8> {164 %0 = fir.string_lit "<stdin>\00"(8) : !fir.char<1,8>165 fir.has_value %0 : !fir.char<1,8>166 }167 func.func private @_FortranACUFAllocDescriptor(i64, !fir.ref<i8>, i32) -> !fir.ref<!fir.box<none>> attributes {fir.runtime}168 func.func private @_FortranACUFFreeDescriptor(!fir.ref<!fir.box<none>>, !fir.ref<i8>, i32) -> () attributes {fir.runtime}169}170 171// CHECK-LABEL: llvm.func @_QQmain()172// CHECK-COUNT-3: llvm.call @_FortranACUFAllocDescriptor173 174// -----175 176module attributes {dlti.dl_spec = #dlti.dl_spec<!llvm.ptr<270> = dense<32> : vector<4xi64>, f128 = dense<128> : vector<2xi64>, f64 = dense<64> : vector<2xi64>, f16 = dense<16> : vector<2xi64>, i32 = dense<32> : vector<2xi64>, i64 = dense<64> : vector<2xi64>, !llvm.ptr<272> = dense<64> : vector<4xi64>, !llvm.ptr<271> = dense<32> : vector<4xi64>, f80 = dense<128> : vector<2xi64>, i128 = dense<128> : vector<2xi64>, i16 = dense<16> : vector<2xi64>, i8 = dense<8> : vector<2xi64>, !llvm.ptr = dense<64> : vector<4xi64>, i1 = dense<8> : vector<2xi64>, "dlti.endianness" = "little", "dlti.stack_alignment" = 128 : i64>, fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", gpu.container_module, llvm.data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128", llvm.ident = "flang version 20.0.0 (git@github.com:clementval/llvm-project.git efc2415bcce8e8a9e73e77aa122c8aba1c1fbbd2)", llvm.target_triple = "x86_64-unknown-linux-gnu"} {177 func.func @_QPouter(%arg0: !fir.ref<!fir.array<100x100xf64>> {cuf.data_attr = #cuf.cuda<device>, fir.bindc_name = "a"}) {178 %c0_i32 = arith.constant 0 : i32179 %c100 = arith.constant 100 : index180 %0 = fir.alloca tuple<!fir.box<!fir.array<100x100xf64>>>181 %1 = fir.coordinate_of %0, %c0_i32 : (!fir.ref<tuple<!fir.box<!fir.array<100x100xf64>>>>, i32) -> !fir.ref<!fir.box<!fir.array<100x100xf64>>>182 %2 = fircg.ext_embox %arg0(%c100, %c100) : (!fir.ref<!fir.array<100x100xf64>>, index, index) -> !fir.box<!fir.array<100x100xf64>>183 fir.store %2 to %1 : !fir.ref<!fir.box<!fir.array<100x100xf64>>>184 return185 }186}187 188// CHECK-LABEL: llvm.func @_QPouter189// CHECK: _FortranACUFAllocDescriptor190 191// -----192 193func.func @_QMm1Psub1(%arg0: !fir.box<!fir.array<?xi32>> {cuf.data_attr = #cuf.cuda<device>, fir.bindc_name = "da"}, %arg1: !fir.box<!fir.array<?xi32>> {cuf.data_attr = #cuf.cuda<device>, fir.bindc_name = "db"}, %arg2: !fir.ref<i32> {fir.bindc_name = "n"}) {194 %0 = fircg.ext_rebox %arg0 : (!fir.box<!fir.array<?xi32>>) -> !fir.box<!fir.array<?xi32>>195 %1 = fircg.ext_rebox %arg1 : (!fir.box<!fir.array<?xi32>>) -> !fir.box<!fir.array<?xi32>>196 return197}198 199// CHECK-LABEL: llvm.func @_QMm1Psub1200// CHECK-COUNT-2: _FortranACUFAllocDescriptor201 202// -----203 204fir.global common @_QPshared_static__shared_mem(dense<0> : vector<28xi8>) {alignment = 8 : i64, data_attr = #cuf.cuda<shared>} : !fir.array<28xi8>205 206// CHECK: llvm.mlir.global common @_QPshared_static__shared_mem(dense<0> : vector<28xi8>) {addr_space = 3 : i32, alignment = 8 : i64} : !llvm.array<28 x i8>207 208// -----209 210module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<f80, dense<128> : vector<2xi64>>, #dlti.dl_entry<i128, dense<128> : vector<2xi64>>, #dlti.dl_entry<i64, dense<64> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr<272>, dense<64> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<271>, dense<32> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<270>, dense<32> : vector<4xi64>>, #dlti.dl_entry<f128, dense<128> : vector<2xi64>>, #dlti.dl_entry<f64, dense<64> : vector<2xi64>>, #dlti.dl_entry<f16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i32, dense<32> : vector<2xi64>>, #dlti.dl_entry<i16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i8, dense<8> : vector<2xi64>>, #dlti.dl_entry<i1, dense<8> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr, dense<64> : vector<4xi64>>, #dlti.dl_entry<"dlti.endianness", "little">, #dlti.dl_entry<"dlti.stack_alignment", 128 : i64>>} {211 func.func @_QQmain() attributes {fir.bindc_name = "cufkernel_global"} {212 %c0 = arith.constant 0 : index213 %3 = fir.call @__tgt_acc_get_deviceptr() : () -> !fir.ref<!fir.box<none>>214 %4 = fir.convert %3 : (!fir.ref<!fir.box<none>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>215 %5 = fir.load %4 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>216 return217 }218 219 // CHECK-LABEL: llvm.func @_QQmain()220 // CHECK: llvm.call @_FortranACUFAllocDescriptor221 222 func.func private @__tgt_acc_get_deviceptr() -> !fir.ref<!fir.box<none>>223}224 225// -----226 227module attributes {gpu.container_module, dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<f80, dense<128> : vector<2xi64>>, #dlti.dl_entry<i128, dense<128> : vector<2xi64>>, #dlti.dl_entry<i64, dense<64> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr<272>, dense<64> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<271>, dense<32> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<270>, dense<32> : vector<4xi64>>, #dlti.dl_entry<f128, dense<128> : vector<2xi64>>, #dlti.dl_entry<f64, dense<64> : vector<2xi64>>, #dlti.dl_entry<f16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i32, dense<32> : vector<2xi64>>, #dlti.dl_entry<i16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i8, dense<8> : vector<2xi64>>, #dlti.dl_entry<i1, dense<8> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr, dense<64> : vector<4xi64>>, #dlti.dl_entry<"dlti.endianness", "little">, #dlti.dl_entry<"dlti.stack_alignment", 128 : i64>>} {228 fir.global @_QMm1Eda {data_attr = #cuf.cuda<device>} : !fir.box<!fir.heap<!fir.array<?x?xf32>>> {229 %c0 = arith.constant 0 : index230 %0 = fir.zero_bits !fir.heap<!fir.array<?x?xf32>>231 %1 = fircg.ext_embox %0(%c0, %c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?x?xf32>>, index, index) -> !fir.box<!fir.heap<!fir.array<?x?xf32>>>232 fir.has_value %1 : !fir.box<!fir.heap<!fir.array<?x?xf32>>>233 }234 func.func @_QQmain() attributes {fir.bindc_name = "P", target_cpu = "x86-64", target_features = #llvm.target_features<["+cmov", "+mmx", "+sse", "+sse2", "+cx8", "+x87", "+fxsr"]>} {235 %c64 = arith.constant 64 : index236 %c1 = arith.constant 1 : index237 %c0_i32 = arith.constant 0 : i32238 %0 = fir.address_of(@_QMm1Eda) : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>239 %8 = fir.load %0 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>240 %9 = fircg.ext_rebox %8 : (!fir.box<!fir.heap<!fir.array<?x?xf32>>>) -> !fir.box<!fir.array<?x?xf32>>241 gpu.launch_func @cuda_device_mod::@_QMm1Psub2 blocks in (%c1, %c1, %c1) threads in (%c64, %c1, %c1) dynamic_shared_memory_size %c0_i32 args(%9 : !fir.box<!fir.array<?x?xf32>>) {cuf.proc_attr = #cuf.cuda_proc<global>}242 return243 }244 gpu.module @cuda_device_mod {245 fir.global @_QMm1Eda {data_attr = #cuf.cuda<device>} : !fir.box<!fir.heap<!fir.array<?x?xf32>>> {246 %c0 = arith.constant 0 : index247 %0 = fir.zero_bits !fir.heap<!fir.array<?x?xf32>>248 %1 = fircg.ext_embox %0(%c0, %c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?x?xf32>>, index, index) -> !fir.box<!fir.heap<!fir.array<?x?xf32>>>249 fir.has_value %1 : !fir.box<!fir.heap<!fir.array<?x?xf32>>>250 }251 gpu.func @_QMm1Psub2(%arg0: !fir.box<!fir.array<?x?xf32>>) kernel {252 gpu.return253 }254 }255}256 257// CHECK-LABEL: llvm.func @_QQmain()258// CHECK: llvm.call @_FortranACUFAllocDescriptor259 260// -----261 262module attributes {gpu.container_module, dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<f80, dense<128> : vector<2xi64>>, #dlti.dl_entry<i128, dense<128> : vector<2xi64>>, #dlti.dl_entry<i64, dense<64> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr<272>, dense<64> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<271>, dense<32> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<270>, dense<32> : vector<4xi64>>, #dlti.dl_entry<f128, dense<128> : vector<2xi64>>, #dlti.dl_entry<f64, dense<64> : vector<2xi64>>, #dlti.dl_entry<f16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i32, dense<32> : vector<2xi64>>, #dlti.dl_entry<i16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i8, dense<8> : vector<2xi64>>, #dlti.dl_entry<i1, dense<8> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr, dense<64> : vector<4xi64>>, #dlti.dl_entry<"dlti.endianness", "little">, #dlti.dl_entry<"dlti.stack_alignment", 128 : i64>>} {263 fir.global @_QMm1Eda {data_attr = #cuf.cuda<device>} : !fir.box<!fir.heap<!fir.array<?x?xf32>>> {264 %c0 = arith.constant 0 : index265 %0 = fir.zero_bits !fir.heap<!fir.array<?x?xf32>>266 %1 = fircg.ext_embox %0(%c0, %c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?x?xf32>>, index, index) -> !fir.box<!fir.heap<!fir.array<?x?xf32>>>267 fir.has_value %1 : !fir.box<!fir.heap<!fir.array<?x?xf32>>>268 }269 gpu.module @cuda_device_mod {270 fir.global @_QMm1Eda {data_attr = #cuf.cuda<device>} : !fir.box<!fir.heap<!fir.array<?x?xf32>>> {271 %c0 = arith.constant 0 : index272 %0 = fir.zero_bits !fir.heap<!fir.array<?x?xf32>>273 %1 = fircg.ext_embox %0(%c0, %c0) {allocator_idx = 2 : i32} : (!fir.heap<!fir.array<?x?xf32>>, index, index) -> !fir.box<!fir.heap<!fir.array<?x?xf32>>>274 fir.has_value %1 : !fir.box<!fir.heap<!fir.array<?x?xf32>>>275 }276 func.func @_QQxxx() {277 %0 = fir.address_of(@_QMm1Eda) : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>278 %8 = fir.load %0 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf32>>>>279 return280 }281 }282}283 284// CHECK-LABEL: llvm.func @_QQxxx()285// CHECK: llvm.alloca %{{.*}} x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8, array<2 x array<3 x i64>>)> {alignment = 8 : i64} : (i32) -> !llvm.ptr286// CHECK-NOT: llvm.call @_FortranACUFAllocDescriptor287 288// -----289 290module attributes {gpu.container_module, dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<f80, dense<128> : vector<2xi64>>, #dlti.dl_entry<i128, dense<128> : vector<2xi64>>, #dlti.dl_entry<i64, dense<64> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr<272>, dense<64> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<271>, dense<32> : vector<4xi64>>, #dlti.dl_entry<!llvm.ptr<270>, dense<32> : vector<4xi64>>, #dlti.dl_entry<f128, dense<128> : vector<2xi64>>, #dlti.dl_entry<f64, dense<64> : vector<2xi64>>, #dlti.dl_entry<f16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i32, dense<32> : vector<2xi64>>, #dlti.dl_entry<i16, dense<16> : vector<2xi64>>, #dlti.dl_entry<i8, dense<8> : vector<2xi64>>, #dlti.dl_entry<i1, dense<8> : vector<2xi64>>, #dlti.dl_entry<!llvm.ptr, dense<64> : vector<4xi64>>, #dlti.dl_entry<"dlti.endianness", "little">, #dlti.dl_entry<"dlti.stack_alignment", 128 : i64>>} {291 gpu.module @cuda_device_mod {292 fir.global @_QMkernelsEinitial_val {data_attr = #cuf.cuda<constant>} : i32 {293 %0 = fir.zero_bits i32294 fir.has_value %0 : i32295 }296 gpu.func @_QMkernelsPassign(%arg0: !fir.ref<!fir.array<?xi32>>) kernel {297 %c-1 = arith.constant -1 : index298 %c1_i32 = arith.constant 1 : i32299 %0 = arith.constant 1 : i32300 %1 = arith.addi %0, %c1_i32 : i32301 %2 = fir.address_of(@_QMkernelsEinitial_val) : !fir.ref<i32>302 %4 = fir.load %2 : !fir.ref<i32>303 %5 = fir.convert %1 : (i32) -> i64304 %6 = fircg.ext_array_coor %arg0(%c-1)<%5> : (!fir.ref<!fir.array<?xi32>>, index, i64) -> !fir.ref<i32>305 fir.store %4 to %6 : !fir.ref<i32>306 gpu.return307 }308 }309}310 311// CHECK: llvm.mlir.global external @_QMkernelsEinitial_val() {addr_space = 4 : i32} : i32312// CHECK-LABEL: gpu.func @_QMkernelsPassign313// CHECK: %[[ADDROF:.*]] = llvm.mlir.addressof @_QMkernelsEinitial_val : !llvm.ptr<4>314// CHECK: %{{.*}} = llvm.addrspacecast %[[ADDROF]] : !llvm.ptr<4> to !llvm.ptr315