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
242 lines
8.7 KiB
Plaintext
242 lines
8.7 KiB
Plaintext
// RUN: fir-opt %s --acc-optimize-firstprivate-map -split-input-file | FileCheck %s
|
|
|
|
// Test: Integer variable - should optimize
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_trivial_scalar_hoist
|
|
func.func @test_trivial_scalar_hoist() {
|
|
%scalar = fir.alloca i32 {bindc_name = "scalar_var"}
|
|
%decl = fir.declare %scalar {uniq_name = "_QFtest_trivial_scalarEscalar_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%fpmap = acc.firstprivate_map varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[DECL]] : !fir.ref<i32>
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: fir.call @use_i32(%[[LOAD]])
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Inside offload region - should optimize by getting rid of firstprivate_map
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_inside_offload_region
|
|
func.func @test_inside_offload_region() {
|
|
%scalar = fir.alloca i32 {bindc_name = "scalar_var"}
|
|
%decl = fir.declare %scalar {uniq_name = "_QFtest_inside_offloadEscalar_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%copyin = acc.copyin varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK: %[[COPYIN:.*]] = acc.copyin varPtr(%[[DECL]] : !fir.ref<i32>) -> !fir.ref<i32>
|
|
acc.parallel dataOperands(%copyin : !fir.ref<i32>) {
|
|
%fpmap = acc.firstprivate_map varPtr(%copyin : !fir.ref<i32>) -> !fir.ref<i32>
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: acc.parallel dataOperands(%[[COPYIN]] : !fir.ref<i32>)
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[COPYIN]] : !fir.ref<i32>
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Local alloca inside offload region - should optimize (erase)
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_local_alloca_inside_offload
|
|
func.func @test_local_alloca_inside_offload() {
|
|
acc.parallel {
|
|
%local = fir.alloca i32 {bindc_name = "local_var"}
|
|
%decl = fir.declare %local {uniq_name = "_QFtest_local_allocaElocal_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%fpmap = acc.firstprivate_map varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: acc.parallel
|
|
// CHECK: %[[LOCAL:.*]] = fir.alloca i32
|
|
// CHECK: %[[DECL:.*]] = fir.declare %[[LOCAL]]
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[DECL]] : !fir.ref<i32>
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: firstprivate_map with acc.private input - should optimize (erase)
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_private_input
|
|
func.func @test_private_input() {
|
|
%scalar = fir.alloca i32 {bindc_name = "scalar_var"}
|
|
%decl = fir.declare %scalar {uniq_name = "_QFtest_private_inputEscalar_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%private = acc.private varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK: %[[PRIVATE:.*]] = acc.private varPtr(%[[DECL]] : !fir.ref<i32>) -> !fir.ref<i32>
|
|
acc.parallel private(%private : !fir.ref<i32>) {
|
|
%fpmap = acc.firstprivate_map varPtr(%private : !fir.ref<i32>) -> !fir.ref<i32>
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: acc.parallel private(%[[PRIVATE]] : !fir.ref<i32>)
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[PRIVATE]] : !fir.ref<i32>
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Variable defined outside offload region but firstprivate_map inside -
|
|
// should hoist out and then optimize
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_hoist_from_offload_region
|
|
func.func @test_hoist_from_offload_region() {
|
|
%scalar = fir.alloca i32 {bindc_name = "scalar_var"}
|
|
%decl = fir.declare %scalar {uniq_name = "_QFtest_hoistEscalar_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[DECL]] : !fir.ref<i32>
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%fpmap = acc.firstprivate_map varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: fir.call @use_i32(%[[LOAD]])
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Optional variable - should NOT optimize
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_optional_no_hoist
|
|
func.func @test_optional_no_hoist(%arg0: !fir.ref<i32>) {
|
|
%decl = fir.declare %arg0 {fortran_attrs = #fir.var_attrs<optional>, uniq_name = "_QFtest_optionalEopt_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%fpmap = acc.firstprivate_map varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare {{.*}} {fortran_attrs = #fir.var_attrs<optional>
|
|
// CHECK: %[[FPMAP:.*]] = acc.firstprivate_map varPtr(%[[DECL]] : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[FPMAP]] : !fir.ref<i32>
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Array type (non-trivial) - should NOT optimize
|
|
|
|
func.func private @use_array(!fir.array<10xi32>)
|
|
|
|
// CHECK-LABEL: func.func @test_array_no_hoist
|
|
func.func @test_array_no_hoist() {
|
|
%c10 = arith.constant 10 : index
|
|
%array = fir.alloca !fir.array<10xi32> {bindc_name = "array_var"}
|
|
%shape = fir.shape %c10 : (index) -> !fir.shape<1>
|
|
%decl = fir.declare %array(%shape) {uniq_name = "_QFtest_arrayEarray_var"} : (!fir.ref<!fir.array<10xi32>>, !fir.shape<1>) -> !fir.ref<!fir.array<10xi32>>
|
|
%fpmap = acc.firstprivate_map varPtr(%decl : !fir.ref<!fir.array<10xi32>>) -> !fir.ref<!fir.array<10xi32>>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK: %[[FPMAP:.*]] = acc.firstprivate_map varPtr(%[[DECL]] : !fir.ref<!fir.array<10xi32>>) -> !fir.ref<!fir.array<10xi32>>
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%load = fir.load %fpmap : !fir.ref<!fir.array<10xi32>>
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[FPMAP]] : !fir.ref<!fir.array<10xi32>>
|
|
fir.call @use_array(%load) : (!fir.array<10xi32>) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Multiple loads from same firstprivate_map - should optimize and hoist all
|
|
|
|
func.func private @use_i32_i32(i32, i32)
|
|
|
|
// CHECK-LABEL: func.func @test_multiple_loads_hoist
|
|
func.func @test_multiple_loads_hoist() {
|
|
%scalar = fir.alloca i32 {bindc_name = "scalar_var"}
|
|
%decl = fir.declare %scalar {uniq_name = "_QFtest_multiple_loadsEscalar_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%fpmap = acc.firstprivate_map varPtr(%decl : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK-DAG: %[[LOAD1:.*]] = fir.load %[[DECL]] : !fir.ref<i32>
|
|
// CHECK-DAG: %[[LOAD2:.*]] = fir.load %[[DECL]] : !fir.ref<i32>
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%load1 = fir.load %fpmap : !fir.ref<i32>
|
|
%load2 = fir.load %fpmap : !fir.ref<i32>
|
|
fir.call @use_i32_i32(%load1, %load2) : (i32, i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Variable through fir.convert - should optimize
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_through_convert
|
|
func.func @test_through_convert() {
|
|
%scalar = fir.alloca i32 {bindc_name = "scalar_var"}
|
|
%decl = fir.declare %scalar {uniq_name = "_QFtest_convertEscalar_var"} : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%convert = fir.convert %decl : (!fir.ref<i32>) -> !fir.ref<i32>
|
|
%fpmap = acc.firstprivate_map varPtr(%convert : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[DECL:.*]] = fir.declare
|
|
// CHECK: %[[CONVERT:.*]] = fir.convert %[[DECL]]
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[CONVERT]] : !fir.ref<i32>
|
|
// CHECK-NOT: acc.firstprivate_map
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: fir.call @use_i32(%[[LOAD]])
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|
|
|
|
// -----
|
|
|
|
// Test: Block argument (unknown origin) - should NOT optimize
|
|
|
|
func.func private @use_i32(i32)
|
|
|
|
// CHECK-LABEL: func.func @test_block_arg_no_hoist
|
|
func.func @test_block_arg_no_hoist(%arg0: !fir.ref<i32>) {
|
|
// No declare op, so we can't determine if it's optional - conservative no-op
|
|
%fpmap = acc.firstprivate_map varPtr(%arg0 : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: %[[FPMAP:.*]] = acc.firstprivate_map varPtr(%arg0 : !fir.ref<i32>) -> !fir.ref<i32>
|
|
// CHECK: acc.parallel
|
|
acc.parallel {
|
|
%load = fir.load %fpmap : !fir.ref<i32>
|
|
// CHECK: %[[LOAD:.*]] = fir.load %[[FPMAP]] : !fir.ref<i32>
|
|
fir.call @use_i32(%load) : (i32) -> ()
|
|
acc.yield
|
|
}
|
|
return
|
|
}
|