1ddf18057a
Introduce a FIR operation to do memcopy/memmove of compile time constant size types. This is to avoid requiring derived type copies to done with load/store which is badly supported in LLVM when the aggregate type is "big" (no threshold can easily be defined here, better to always avoid them for fir.type). This was the root cause of the regressions caused by #114002 which introduced a load/store of fir.type<> which caused hand/asserts to fire in LLVM on several benchmarks. See https://llvm.org/docs/Frontend/PerformanceTips.html#avoid-creating-values-of-aggregate-type
36 lines
1.9 KiB
Plaintext
36 lines
1.9 KiB
Plaintext
// Test fir.copy codegen.
|
|
// RUN: fir-opt --fir-to-llvm-ir %s -o - | FileCheck %s
|
|
|
|
!t=!fir.type<sometype{i:!fir.array<9xi32>}>
|
|
|
|
module attributes {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"} {
|
|
|
|
func.func @test_copy_1(%arg0: !fir.ref<!t>, %arg1: !fir.ref<!t>) {
|
|
fir.copy %arg0 to %arg1 no_overlap : !fir.ref<!t>, !fir.ref<!t>
|
|
return
|
|
}
|
|
// CHECK-LABEL: llvm.func @test_copy_1(
|
|
// CHECK-SAME: %[[VAL_0:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: !llvm.ptr,
|
|
// CHECK-SAME: %[[VAL_1:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: !llvm.ptr) {
|
|
// CHECK: %[[VAL_2:.*]] = llvm.mlir.zero : !llvm.ptr
|
|
// CHECK: %[[VAL_3:.*]] = llvm.getelementptr %[[VAL_2]][1] : (!llvm.ptr) -> !llvm.ptr, !llvm.struct<"sometype", (array<9 x i32>)>
|
|
// CHECK: %[[VAL_4:.*]] = llvm.ptrtoint %[[VAL_3]] : !llvm.ptr to i64
|
|
// CHECK: "llvm.intr.memcpy"(%[[VAL_1]], %[[VAL_0]], %[[VAL_4]]) <{isVolatile = false}> : (!llvm.ptr, !llvm.ptr, i64) -> ()
|
|
// CHECK: llvm.return
|
|
// CHECK: }
|
|
|
|
func.func @test_copy_2(%arg0: !fir.ref<!t>, %arg1: !fir.ref<!t>) {
|
|
fir.copy %arg0 to %arg1 : !fir.ref<!t>, !fir.ref<!t>
|
|
return
|
|
}
|
|
// CHECK-LABEL: llvm.func @test_copy_2(
|
|
// CHECK-SAME: %[[VAL_0:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: !llvm.ptr,
|
|
// CHECK-SAME: %[[VAL_1:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: !llvm.ptr) {
|
|
// CHECK: %[[VAL_2:.*]] = llvm.mlir.zero : !llvm.ptr
|
|
// CHECK: %[[VAL_3:.*]] = llvm.getelementptr %[[VAL_2]][1] : (!llvm.ptr) -> !llvm.ptr, !llvm.struct<"sometype", (array<9 x i32>)>
|
|
// CHECK: %[[VAL_4:.*]] = llvm.ptrtoint %[[VAL_3]] : !llvm.ptr to i64
|
|
// CHECK: "llvm.intr.memmove"(%[[VAL_1]], %[[VAL_0]], %[[VAL_4]]) <{isVolatile = false}> : (!llvm.ptr, !llvm.ptr, i64) -> ()
|
|
// CHECK: llvm.return
|
|
// CHECK: }
|
|
}
|