c23b4fbdbb
Now that #149310 has restricted lifetime intrinsics to only work on allocas, we can also drop the explicit size argument. Instead, the size is implied by the alloca. This removes the ability to only mark a prefix of an alloca alive/dead. We never used that capability, so we should remove the need to handle that possibility everywhere (though many key places, including stack coloring, did not actually respect this).
192 lines
6.4 KiB
LLVM
192 lines
6.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature
|
|
; RUN: opt < %s -passes='require<globals-aa>,memcpyopt' -S -verify-memoryssa | FileCheck %s
|
|
|
|
target datalayout = "e"
|
|
|
|
declare void @foo(ptr)
|
|
declare void @llvm.memcpy.p0.p0.i32(ptr nocapture, ptr nocapture, i32, i1) nounwind
|
|
declare void @llvm.lifetime.start.p0(ptr nocapture)
|
|
declare void @llvm.lifetime.end.p0(ptr nocapture)
|
|
|
|
; Check that the transformation isn't applied if the called function can
|
|
; capture the pointer argument (i.e. the nocapture attribute isn't present)
|
|
define void @test() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR2]])
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr [[PTR1]], ptr [[PTR2]], i32 1, i1 false)
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca i8
|
|
%ptr2 = alloca i8
|
|
call void @foo(ptr %ptr2)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 1, i1 false)
|
|
call void @foo(ptr %ptr1)
|
|
ret void
|
|
}
|
|
|
|
; Same as previous test, but with a bitcasted argument.
|
|
define void @test_bitcast() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_bitcast() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca [2 x i8], align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca [2 x i8], align 1
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR2]])
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr [[PTR1]], ptr [[PTR2]], i32 2, i1 false)
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca [2 x i8]
|
|
%ptr2 = alloca [2 x i8]
|
|
call void @foo(ptr %ptr2)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 2, i1 false)
|
|
call void @foo(ptr %ptr1)
|
|
ret void
|
|
}
|
|
|
|
; Lifetime of %ptr2 ends before the potential use of the capture in the second
|
|
; call.
|
|
define void @test_lifetime_end() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_lifetime_end() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @llvm.lifetime.start.p0(ptr [[PTR2]])
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR1]])
|
|
; CHECK-NEXT: call void @llvm.lifetime.end.p0(ptr [[PTR2]])
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca i8
|
|
%ptr2 = alloca i8
|
|
call void @llvm.lifetime.start.p0(ptr %ptr2)
|
|
call void @foo(ptr %ptr2)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 1, i1 false)
|
|
call void @llvm.lifetime.end.p0(ptr %ptr2)
|
|
call void @foo(ptr %ptr1)
|
|
ret void
|
|
}
|
|
|
|
; Lifetime of %ptr2 ends before any potential use of the capture because we
|
|
; return from the function.
|
|
define void @test_function_end() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_function_end() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca i8
|
|
%ptr2 = alloca i8
|
|
call void @foo(ptr %ptr2)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 1, i1 false)
|
|
ret void
|
|
}
|
|
|
|
; A potential use of the capture occurs in a later block, can't be optimized.
|
|
define void @test_terminator() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_terminator() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR2]])
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr [[PTR1]], ptr [[PTR2]], i32 1, i1 false)
|
|
; CHECK-NEXT: br label [[NEXT:%.*]]
|
|
; CHECK: next:
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca i8
|
|
%ptr2 = alloca i8
|
|
call void @foo(ptr %ptr2)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 1, i1 false)
|
|
br label %next
|
|
|
|
next:
|
|
call void @foo(ptr %ptr1)
|
|
ret void
|
|
}
|
|
|
|
; This case can be optimized, but would require a scan across multiple blocks
|
|
; and is currently not performed.
|
|
define void @test_terminator2() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_terminator2() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR2]])
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr [[PTR1]], ptr [[PTR2]], i32 1, i1 false)
|
|
; CHECK-NEXT: br label [[NEXT:%.*]]
|
|
; CHECK: next:
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca i8
|
|
%ptr2 = alloca i8
|
|
call void @foo(ptr %ptr2)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 1, i1 false)
|
|
br label %next
|
|
|
|
next:
|
|
ret void
|
|
}
|
|
|
|
declare void @capture(ptr)
|
|
|
|
; This case should not be optimized, because dest is captured before the call.
|
|
define void @test_dest_captured_before_alloca() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_dest_captured_before_alloca() {
|
|
; CHECK-NEXT: [[PTR1:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: [[PTR2:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @capture(ptr [[PTR1]])
|
|
; CHECK-NEXT: call void @foo(ptr [[PTR2]]) #[[ATTR2:[0-9]+]]
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr [[PTR1]], ptr [[PTR2]], i32 1, i1 false)
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr1 = alloca i8
|
|
%ptr2 = alloca i8
|
|
call void @capture(ptr %ptr1)
|
|
call void @foo(ptr %ptr2) argmemonly
|
|
call void @llvm.memcpy.p0.p0.i32(ptr %ptr1, ptr %ptr2, i32 1, i1 false)
|
|
ret void
|
|
}
|
|
|
|
|
|
@g = internal global i8 0
|
|
|
|
; This case should not be optimized, because @g is captured before the call
|
|
; (being a global) and @icmp_g might depend on its identity.
|
|
define void @test_dest_captured_before_global() {
|
|
; CHECK-LABEL: define {{[^@]+}}@test_dest_captured_before_global() {
|
|
; CHECK-NEXT: [[PTR:%.*]] = alloca i8, align 1
|
|
; CHECK-NEXT: call void @icmp_g(ptr [[PTR]])
|
|
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr @g, ptr [[PTR]], i32 1, i1 false)
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%ptr = alloca i8
|
|
call void @icmp_g(ptr %ptr)
|
|
call void @llvm.memcpy.p0.p0.i32(ptr @g, ptr %ptr, i32 1, i1 false)
|
|
ret void
|
|
}
|
|
|
|
define void @icmp_g(ptr %p) {
|
|
; CHECK-LABEL: define {{[^@]+}}@icmp_g
|
|
; CHECK-SAME: (ptr [[P:%.*]]) {
|
|
; CHECK-NEXT: [[C:%.*]] = icmp eq ptr [[P]], @g
|
|
; CHECK-NEXT: br i1 [[C]], label [[IF:%.*]], label [[ELSE:%.*]]
|
|
; CHECK: if:
|
|
; CHECK-NEXT: store i8 1, ptr [[P]], align 1
|
|
; CHECK-NEXT: ret void
|
|
; CHECK: else:
|
|
; CHECK-NEXT: store i8 2, ptr [[P]], align 1
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
%c = icmp eq ptr %p, @g
|
|
br i1 %c, label %if, label %else
|
|
|
|
if:
|
|
store i8 1, ptr %p
|
|
ret void
|
|
|
|
else:
|
|
store i8 2, ptr %p
|
|
ret void
|
|
}
|