llvm-6502/test/Transforms/TailCallElim/basic.ll

191 lines
4.9 KiB
LLVM
Raw Normal View History

; RUN: opt < %s -tailcallelim -S | FileCheck %s
declare void @noarg()
declare void @use(i32*)
Teach TailRecursionElimination to handle certain cases of nocapture escaping allocas. Without the changes introduced into this patch, if TRE saw any allocas at all, TRE would not perform TRE *or* mark callsites with the tail marker. Because TRE runs after mem2reg, this inadequacy is not a death sentence. But given a callsite A without escaping alloca argument, A may not be able to have the tail marker placed on it due to a separate callsite B having a write-back parameter passed in via an argument with the nocapture attribute. Assume that B is the only other callsite besides A and B only has nocapture escaping alloca arguments (*NOTE* B may have other arguments that are not passed allocas). In this case not marking A with the tail marker is unnecessarily conservative since: 1. By assumption A has no escaping alloca arguments itself so it can not access the caller's stack via its arguments. 2. Since all of B's escaping alloca arguments are passed as parameters with the nocapture attribute, we know that B does not stash said escaping allocas in a manner that outlives B itself and thus could be accessed indirectly by A. With the changes introduced by this patch: 1. If we see any escaping allocas passed as a capturing argument, we do nothing and bail early. 2. If we do not see any escaping allocas passed as captured arguments but we do see escaping allocas passed as nocapture arguments: i. We do not perform TRE to avoid PR962 since the code generator produces significantly worse code for the dynamic allocas that would be created by the TRE algorithm. ii. If we do not return twice, mark call sites without escaping allocas with the tail marker. *NOTE* This excludes functions with escaping nocapture allocas. 3. If we do not see any escaping allocas at all (whether captured or not): i. If we do not have usage of setjmp, mark all callsites with the tail marker. ii. If there are no dynamic/variable sized allocas in the function, attempt to perform TRE on all callsites in the function. Based off of a patch by Nick Lewycky. rdar://14324281. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186057 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-11 04:40:01 +00:00
declare void @use_nocapture(i32* nocapture)
declare void @use2_nocapture(i32* nocapture, i32* nocapture)
; Trivial case. Mark @noarg with tail call.
define void @test0() {
; CHECK: tail call void @noarg()
call void @noarg()
ret void
}
; PR615. Make sure that we do not move the alloca so that it interferes with the tail call.
define i32 @test1() {
; CHECK: i32 @test1()
; CHECK-NEXT: alloca
%A = alloca i32 ; <i32*> [#uses=2]
store i32 5, i32* %A
call void @use(i32* %A)
; CHECK: tail call i32 @test1
%X = tail call i32 @test1() ; <i32> [#uses=1]
ret i32 %X
}
; This function contains intervening instructions which should be moved out of the way
define i32 @test2(i32 %X) {
; CHECK: i32 @test2
; CHECK-NOT: call
; CHECK: ret i32
entry:
%tmp.1 = icmp eq i32 %X, 0 ; <i1> [#uses=1]
br i1 %tmp.1, label %then.0, label %endif.0
then.0: ; preds = %entry
%tmp.4 = add i32 %X, 1 ; <i32> [#uses=1]
ret i32 %tmp.4
endif.0: ; preds = %entry
%tmp.10 = add i32 %X, -1 ; <i32> [#uses=1]
%tmp.8 = call i32 @test2(i32 %tmp.10) ; <i32> [#uses=1]
%DUMMY = add i32 %X, 1 ; <i32> [#uses=0]
ret i32 %tmp.8
}
; Though this case seems to be fairly unlikely to occur in the wild, someone
; plunked it into the demo script, so maybe they care about it.
define i32 @test3(i32 %c) {
; CHECK: i32 @test3
; CHECK-NOT: call
; CHECK: ret i32 0
entry:
%tmp.1 = icmp eq i32 %c, 0 ; <i1> [#uses=1]
br i1 %tmp.1, label %return, label %else
else: ; preds = %entry
%tmp.5 = add i32 %c, -1 ; <i32> [#uses=1]
%tmp.3 = call i32 @test3(i32 %tmp.5) ; <i32> [#uses=0]
ret i32 0
return: ; preds = %entry
ret i32 0
}
Teach TailRecursionElimination to handle certain cases of nocapture escaping allocas. Without the changes introduced into this patch, if TRE saw any allocas at all, TRE would not perform TRE *or* mark callsites with the tail marker. Because TRE runs after mem2reg, this inadequacy is not a death sentence. But given a callsite A without escaping alloca argument, A may not be able to have the tail marker placed on it due to a separate callsite B having a write-back parameter passed in via an argument with the nocapture attribute. Assume that B is the only other callsite besides A and B only has nocapture escaping alloca arguments (*NOTE* B may have other arguments that are not passed allocas). In this case not marking A with the tail marker is unnecessarily conservative since: 1. By assumption A has no escaping alloca arguments itself so it can not access the caller's stack via its arguments. 2. Since all of B's escaping alloca arguments are passed as parameters with the nocapture attribute, we know that B does not stash said escaping allocas in a manner that outlives B itself and thus could be accessed indirectly by A. With the changes introduced by this patch: 1. If we see any escaping allocas passed as a capturing argument, we do nothing and bail early. 2. If we do not see any escaping allocas passed as captured arguments but we do see escaping allocas passed as nocapture arguments: i. We do not perform TRE to avoid PR962 since the code generator produces significantly worse code for the dynamic allocas that would be created by the TRE algorithm. ii. If we do not return twice, mark call sites without escaping allocas with the tail marker. *NOTE* This excludes functions with escaping nocapture allocas. 3. If we do not see any escaping allocas at all (whether captured or not): i. If we do not have usage of setjmp, mark all callsites with the tail marker. ii. If there are no dynamic/variable sized allocas in the function, attempt to perform TRE on all callsites in the function. Based off of a patch by Nick Lewycky. rdar://14324281. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186057 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-11 04:40:01 +00:00
; Make sure that a nocapture pointer does not stop adding a tail call marker to
; an unrelated call and additionally that we do not mark the nocapture call with
; a tail call.
;
; rdar://14324281
define void @test4() {
; CHECK: void @test4
; CHECK-NOT: tail call void @use_nocapture
; CHECK: tail call void @noarg()
; CHECK: ret void
%a = alloca i32
call void @use_nocapture(i32* %a)
call void @noarg()
ret void
}
; Make sure that we do not perform TRE even with a nocapture use. This is due to
; bad codegen caused by PR962.
;
; rdar://14324281.
define i32* @test5(i32* nocapture %A, i1 %cond) {
; CHECK: i32* @test5
; CHECK-NOT: tailrecurse:
; CHECK: ret i32* null
%B = alloca i32
br i1 %cond, label %cond_true, label %cond_false
cond_true:
call i32* @test5(i32* %B, i1 false)
ret i32* null
cond_false:
call void @use2_nocapture(i32* %A, i32* %B)
call void @noarg()
ret i32* null
}
; PR14143: Make sure that we do not mark functions with nocapture allocas with tail.
;
; rdar://14324281.
define void @test6(i32* %a, i32* %b) {
; CHECK-LABEL: @test6(
Teach TailRecursionElimination to handle certain cases of nocapture escaping allocas. Without the changes introduced into this patch, if TRE saw any allocas at all, TRE would not perform TRE *or* mark callsites with the tail marker. Because TRE runs after mem2reg, this inadequacy is not a death sentence. But given a callsite A without escaping alloca argument, A may not be able to have the tail marker placed on it due to a separate callsite B having a write-back parameter passed in via an argument with the nocapture attribute. Assume that B is the only other callsite besides A and B only has nocapture escaping alloca arguments (*NOTE* B may have other arguments that are not passed allocas). In this case not marking A with the tail marker is unnecessarily conservative since: 1. By assumption A has no escaping alloca arguments itself so it can not access the caller's stack via its arguments. 2. Since all of B's escaping alloca arguments are passed as parameters with the nocapture attribute, we know that B does not stash said escaping allocas in a manner that outlives B itself and thus could be accessed indirectly by A. With the changes introduced by this patch: 1. If we see any escaping allocas passed as a capturing argument, we do nothing and bail early. 2. If we do not see any escaping allocas passed as captured arguments but we do see escaping allocas passed as nocapture arguments: i. We do not perform TRE to avoid PR962 since the code generator produces significantly worse code for the dynamic allocas that would be created by the TRE algorithm. ii. If we do not return twice, mark call sites without escaping allocas with the tail marker. *NOTE* This excludes functions with escaping nocapture allocas. 3. If we do not see any escaping allocas at all (whether captured or not): i. If we do not have usage of setjmp, mark all callsites with the tail marker. ii. If there are no dynamic/variable sized allocas in the function, attempt to perform TRE on all callsites in the function. Based off of a patch by Nick Lewycky. rdar://14324281. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186057 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-11 04:40:01 +00:00
; CHECK-NOT: tail call
; CHECK: ret void
%c = alloca [100 x i8], align 16
%tmp = bitcast [100 x i8]* %c to i32*
call void @use2_nocapture(i32* %b, i32* %tmp)
ret void
}
; PR14143: Make sure that we do not mark functions with nocapture allocas with tail.
;
; rdar://14324281
define void @test7(i32* %a, i32* %b) nounwind uwtable {
entry:
; CHECK-LABEL: @test7(
Teach TailRecursionElimination to handle certain cases of nocapture escaping allocas. Without the changes introduced into this patch, if TRE saw any allocas at all, TRE would not perform TRE *or* mark callsites with the tail marker. Because TRE runs after mem2reg, this inadequacy is not a death sentence. But given a callsite A without escaping alloca argument, A may not be able to have the tail marker placed on it due to a separate callsite B having a write-back parameter passed in via an argument with the nocapture attribute. Assume that B is the only other callsite besides A and B only has nocapture escaping alloca arguments (*NOTE* B may have other arguments that are not passed allocas). In this case not marking A with the tail marker is unnecessarily conservative since: 1. By assumption A has no escaping alloca arguments itself so it can not access the caller's stack via its arguments. 2. Since all of B's escaping alloca arguments are passed as parameters with the nocapture attribute, we know that B does not stash said escaping allocas in a manner that outlives B itself and thus could be accessed indirectly by A. With the changes introduced by this patch: 1. If we see any escaping allocas passed as a capturing argument, we do nothing and bail early. 2. If we do not see any escaping allocas passed as captured arguments but we do see escaping allocas passed as nocapture arguments: i. We do not perform TRE to avoid PR962 since the code generator produces significantly worse code for the dynamic allocas that would be created by the TRE algorithm. ii. If we do not return twice, mark call sites without escaping allocas with the tail marker. *NOTE* This excludes functions with escaping nocapture allocas. 3. If we do not see any escaping allocas at all (whether captured or not): i. If we do not have usage of setjmp, mark all callsites with the tail marker. ii. If there are no dynamic/variable sized allocas in the function, attempt to perform TRE on all callsites in the function. Based off of a patch by Nick Lewycky. rdar://14324281. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186057 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-11 04:40:01 +00:00
; CHECK-NOT: tail call
; CHECK: ret void
%c = alloca [100 x i8], align 16
%0 = bitcast [100 x i8]* %c to i32*
call void @use2_nocapture(i32* %0, i32* %a)
call void @use2_nocapture(i32* %b, i32* %0)
ret void
}
; If we have a mix of escaping captured/non-captured allocas, ensure that we do
; not do anything including marking callsites with the tail call marker.
;
; rdar://14324281.
define i32* @test8(i32* nocapture %A, i1 %cond) {
; CHECK: i32* @test8
; CHECK-NOT: tailrecurse:
; CHECK-NOT: tail call
; CHECK: ret i32* null
%B = alloca i32
%B2 = alloca i32
br i1 %cond, label %cond_true, label %cond_false
cond_true:
call void @use(i32* %B2)
call i32* @test8(i32* %B, i1 false)
ret i32* null
cond_false:
call void @use2_nocapture(i32* %A, i32* %B)
call void @noarg()
ret i32* null
}
; Don't tail call if a byval arg is captured.
define void @test9(i32* byval %a) {
; CHECK-LABEL: define void @test9(
; CHECK: {{^ *}}call void @use(
call void @use(i32* %a)
ret void
}
%struct.X = type { i8* }
declare void @ctor(%struct.X*)
define void @test10(%struct.X* noalias sret %agg.result, i1 zeroext %b) {
; CHECK-LABEL @test10
entry:
%x = alloca %struct.X, align 8
br i1 %b, label %if.then, label %if.end
if.then: ; preds = %entry
call void @ctor(%struct.X* %agg.result)
; CHECK: tail call void @ctor
br label %return
if.end:
call void @ctor(%struct.X* %x)
; CHECK: call void @ctor
br label %return
return:
ret void
}
declare void @test11_helper1(i8** nocapture, i8*)
declare void @test11_helper2(i8*)
define void @test11() {
; CHECK-LABEL: @test11
; CHECK-NOT: tail
%a = alloca i8*
%b = alloca i8
call void @test11_helper1(i8** %a, i8* %b) ; a = &b
%c = load i8** %a
call void @test11_helper2(i8* %c)
; CHECK: call void @test11_helper2
ret void
}