mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-24 03:33:20 +01:00
b3ecd3b03e
System z branches have a mask to select which of the 4 CC values should cause the branch to be taken. We can invert a branch by inverting the mask. However, not all instructions can produce all 4 CC values, so inverting the branch like this can lead to some oddities. For example, integer comparisons only produce a CC of 0 (equal), 1 (less) or 2 (greater). If an integer EQ is reversed to NE before instruction selection, the branch will test for 1 or 2. If instead the branch is reversed after instruction selection (by inverting the mask), it will test for 1, 2 or 3. Both are correct, but the second isn't really canonical. This patch therefore keeps track of which CC values are possible and uses this when inverting a mask. Although this is mostly cosmestic, it fixes undefined behavior for the CIJNLH in branch-08.ll. Another fix would have been to mask out bit 0 when generating the fused compare and branch, but the point of this patch is that we shouldn't need to do that in the first place. The patch also makes it easier to reuse CC results from other instructions. llvm-svn: 187495
125 lines
3.1 KiB
LLVM
125 lines
3.1 KiB
LLVM
; Test STOCGs that are presented as selects.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z196 | FileCheck %s
|
|
|
|
declare void @foo(i64 *)
|
|
|
|
; Test with the loaded value first.
|
|
define void @f1(i64 *%ptr, i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: clfi %r4, 42
|
|
; CHECK: stocghe %r3, 0(%r2)
|
|
; CHECK: br %r14
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %orig, i64 %alt
|
|
store i64 %res, i64 *%ptr
|
|
ret void
|
|
}
|
|
|
|
; ...and with the loaded value second
|
|
define void @f2(i64 *%ptr, i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: clfi %r4, 42
|
|
; CHECK: stocgl %r3, 0(%r2)
|
|
; CHECK: br %r14
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %alt, i64 %orig
|
|
store i64 %res, i64 *%ptr
|
|
ret void
|
|
}
|
|
|
|
; Check the high end of the aligned STOCG range.
|
|
define void @f3(i64 *%base, i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: clfi %r4, 42
|
|
; CHECK: stocghe %r3, 524280(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr i64 *%base, i64 65535
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %orig, i64 %alt
|
|
store i64 %res, i64 *%ptr
|
|
ret void
|
|
}
|
|
|
|
; Check the next doubleword up. Other sequences besides this one would be OK.
|
|
define void @f4(i64 *%base, i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: agfi %r2, 524288
|
|
; CHECK: clfi %r4, 42
|
|
; CHECK: stocghe %r3, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr i64 *%base, i64 65536
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %orig, i64 %alt
|
|
store i64 %res, i64 *%ptr
|
|
ret void
|
|
}
|
|
|
|
; Check the low end of the STOCG range.
|
|
define void @f5(i64 *%base, i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: clfi %r4, 42
|
|
; CHECK: stocghe %r3, -524288(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr i64 *%base, i64 -65536
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %orig, i64 %alt
|
|
store i64 %res, i64 *%ptr
|
|
ret void
|
|
}
|
|
|
|
; Check the next doubleword down, with the same comments as f4.
|
|
define void @f6(i64 *%base, i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: agfi %r2, -524296
|
|
; CHECK: clfi %r4, 42
|
|
; CHECK: stocghe %r3, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr i64 *%base, i64 -65537
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %orig, i64 %alt
|
|
store i64 %res, i64 *%ptr
|
|
ret void
|
|
}
|
|
|
|
; Try a frame index base.
|
|
define void @f7(i64 %alt, i32 %limit) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: brasl %r14, foo@PLT
|
|
; CHECK: stocghe {{%r[0-9]+}}, {{[0-9]+}}(%r15)
|
|
; CHECK: brasl %r14, foo@PLT
|
|
; CHECK: br %r14
|
|
%ptr = alloca i64
|
|
call void @foo(i64 *%ptr)
|
|
%cond = icmp ult i32 %limit, 42
|
|
%orig = load i64 *%ptr
|
|
%res = select i1 %cond, i64 %orig, i64 %alt
|
|
store i64 %res, i64 *%ptr
|
|
call void @foo(i64 *%ptr)
|
|
ret void
|
|
}
|
|
|
|
; Test that conditionally-executed stores do not use STOC, since STOC
|
|
; is allowed to trap even when the condition is false.
|
|
define void @f8(i64 %a, i64 %b, i64 *%dest) {
|
|
; CHECK-LABEL: f8:
|
|
; CHECK-NOT: stocg %r3, 0(%r4)
|
|
; CHECK: br %r14
|
|
entry:
|
|
%cmp = icmp ule i64 %a, %b
|
|
br i1 %cmp, label %store, label %exit
|
|
|
|
store:
|
|
store i64 %b, i64 *%dest
|
|
br label %exit
|
|
|
|
exit:
|
|
ret void
|
|
}
|