diff options
| author | Arnold Schwaighofer <aschwaighofer@apple.com> | 2015-05-08 23:52:00 +0000 |
|---|---|---|
| committer | Arnold Schwaighofer <aschwaighofer@apple.com> | 2015-05-08 23:52:00 +0000 |
| commit | f54b73d681d6d65a75cbd887939a75d652dbc24e (patch) | |
| tree | a7bcb3fe3cd39f0f19d4669de937fef7c42031e9 /llvm/test/CodeGen/AArch64/tailcall_misched_graph.ll | |
| parent | 2ddfa0c5b2b8026cd4d6981da2cfd4ab03026d7c (diff) | |
| download | bcm5719-llvm-f54b73d681d6d65a75cbd887939a75d652dbc24e.tar.gz bcm5719-llvm-f54b73d681d6d65a75cbd887939a75d652dbc24e.zip | |
ScheduleDAGInstrs: In functions with tail calls PseudoSourceValues are not non-aliasing distinct objects
The code that builds the dependence graph assumes that two PseudoSourceValues
don't alias. In a tail calling function two FixedStackObjects might refer to the
same location. Worse 'immutable' fixed stack objects like function arguments are
not immutable and will be clobbered.
Change this so that a load from a FixedStackObject is not invariant in a tail
calling function and don't return a PseudoSourceValue for an instruction in tail
calling functions when building the dependence graph so that we handle function
arguments conservatively.
Fix for PR23459.
rdar://20740035
llvm-svn: 236916
Diffstat (limited to 'llvm/test/CodeGen/AArch64/tailcall_misched_graph.ll')
| -rw-r--r-- | llvm/test/CodeGen/AArch64/tailcall_misched_graph.ll | 42 |
1 files changed, 42 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AArch64/tailcall_misched_graph.ll b/llvm/test/CodeGen/AArch64/tailcall_misched_graph.ll new file mode 100644 index 00000000000..57f96874c0d --- /dev/null +++ b/llvm/test/CodeGen/AArch64/tailcall_misched_graph.ll @@ -0,0 +1,42 @@ +; RUN: llc -mcpu=cyclone -debug-only=misched < %s 2>&1 | FileCheck %s + +; REQUIRE: asserts + +target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128" +target triple = "arm64-apple-ios7.0.0" + +define void @caller2(i8* %a0, i8* %a1, i8* %a2, i8* %a3, i8* %a4, i8* %a5, i8* %a6, i8* %a7, i8* %a8, i8* %a9) { +entry: + tail call void @callee2(i8* %a1, i8* %a2, i8* %a3, i8* %a4, i8* %a5, i8* %a6, i8* %a7, i8* %a8, i8* %a9, i8* %a0) + ret void +} + +declare void @callee2(i8*, i8*, i8*, i8*, i8*, + i8*, i8*, i8*, i8*, i8*) + +; Make sure there is a dependence between the load and store to the same stack +; location during a tail call. Tail calls clobber the incoming argument area and +; therefore it is not safe to assume argument locations are invariant. +; PR23459 has a test case that we where miscompiling because of this at the +; time. + +; CHECK: Frame Objects +; CHECK: fi#-4: {{.*}} fixed, at location [SP+8] +; CHECK: fi#-3: {{.*}} fixed, at location [SP] +; CHECK: fi#-2: {{.*}} fixed, at location [SP+8] +; CHECK: fi#-1: {{.*}} fixed, at location [SP] + +; CHECK: [[VRA:%vreg.*]]<def> = LDRXui <fi#-1> +; CHECK: [[VRB:%vreg.*]]<def> = LDRXui <fi#-2> +; CHECK: STRXui %vreg{{.*}}, <fi#-4> +; CHECK: STRXui [[VRB]], <fi#-3> + +; Make sure that there is an dependence edge between fi#-2 and fi#-4. +; Without this edge the scheduler would be free to move the store accross the load. + +; CHECK: SU({{.*}}): [[VRB]]<def> = LDRXui <fi#-2> +; CHECK-NOT: SU +; CHECK: Successors: +; CHECK: ch SU([[DEPSTORE:.*]]): Latency=0 + +; CHECK: SU([[DEPSTORE]]): STRXui %vreg0, <fi#-4> |

