2013-10-31 12:14:17 +00:00
|
|
|
; Test 64-bit atomic minimum and maximum. Here we match the z10 versions,
|
|
|
|
; which can't use LOCGR.
|
2013-05-06 16:17:29 +00:00
|
|
|
;
|
2013-10-31 12:14:17 +00:00
|
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
|
2013-05-06 16:17:29 +00:00
|
|
|
|
|
|
|
; Check signed minium.
|
|
|
|
define i64 @f1(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f1:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: [[LOOP:\.[^:]*]]:
|
|
|
|
; CHECK: lgr [[NEW:%r[0-9]+]], %r2
|
2013-05-28 10:41:11 +00:00
|
|
|
; CHECK: cgrjle %r2, %r4, [[KEEP:\..*]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lgr [[NEW]], %r4
|
|
|
|
; CHECK: csg %r2, [[NEW]], 0(%r3)
|
[SystemZ] Be more careful about inverting CC masks
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
|
|
|
; CHECK: jl [[LOOP]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: br %r14
|
|
|
|
%res = atomicrmw min i64 *%src, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check signed maximum.
|
|
|
|
define i64 @f2(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f2:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: [[LOOP:\.[^:]*]]:
|
|
|
|
; CHECK: lgr [[NEW:%r[0-9]+]], %r2
|
2013-05-28 10:41:11 +00:00
|
|
|
; CHECK: cgrjhe %r2, %r4, [[KEEP:\..*]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lgr [[NEW]], %r4
|
|
|
|
; CHECK: csg %r2, [[NEW]], 0(%r3)
|
[SystemZ] Be more careful about inverting CC masks
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
|
|
|
; CHECK: jl [[LOOP]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: br %r14
|
|
|
|
%res = atomicrmw max i64 *%src, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check unsigned minimum.
|
|
|
|
define i64 @f3(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f3:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: [[LOOP:\.[^:]*]]:
|
|
|
|
; CHECK: lgr [[NEW:%r[0-9]+]], %r2
|
2013-09-18 09:56:40 +00:00
|
|
|
; CHECK: clgrjle %r2, %r4, [[KEEP:\..*]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lgr [[NEW]], %r4
|
|
|
|
; CHECK: csg %r2, [[NEW]], 0(%r3)
|
[SystemZ] Be more careful about inverting CC masks
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
|
|
|
; CHECK: jl [[LOOP]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: br %r14
|
|
|
|
%res = atomicrmw umin i64 *%src, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check unsigned maximum.
|
|
|
|
define i64 @f4(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f4:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: [[LOOP:\.[^:]*]]:
|
|
|
|
; CHECK: lgr [[NEW:%r[0-9]+]], %r2
|
2013-09-18 09:56:40 +00:00
|
|
|
; CHECK: clgrjhe %r2, %r4, [[KEEP:\..*]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lgr [[NEW]], %r4
|
|
|
|
; CHECK: csg %r2, [[NEW]], 0(%r3)
|
[SystemZ] Be more careful about inverting CC masks
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
|
|
|
; CHECK: jl [[LOOP]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: br %r14
|
|
|
|
%res = atomicrmw umax i64 *%src, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the high end of the aligned CSG range.
|
|
|
|
define i64 @f5(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f5:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lg %r2, 524280(%r3)
|
|
|
|
; CHECK: csg %r2, {{%r[0-9]+}}, 524280(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i64 *%src, i64 65535
|
|
|
|
%res = atomicrmw min i64 *%ptr, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the next doubleword up, which requires separate address logic.
|
|
|
|
define i64 @f6(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f6:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: agfi %r3, 524288
|
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: csg %r2, {{%r[0-9]+}}, 0(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i64 *%src, i64 65536
|
|
|
|
%res = atomicrmw min i64 *%ptr, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the low end of the CSG range.
|
|
|
|
define i64 @f7(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f7:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lg %r2, -524288(%r3)
|
|
|
|
; CHECK: csg %r2, {{%r[0-9]+}}, -524288(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i64 *%src, i64 -65536
|
|
|
|
%res = atomicrmw min i64 *%ptr, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the next doubleword down, which requires separate address logic.
|
|
|
|
define i64 @f8(i64 %dummy, i64 *%src, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f8:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: agfi %r3, -524296
|
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: csg %r2, {{%r[0-9]+}}, 0(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr i64 *%src, i64 -65537
|
|
|
|
%res = atomicrmw min i64 *%ptr, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check that indexed addresses are not allowed.
|
|
|
|
define i64 @f9(i64 %dummy, i64 %base, i64 %index, i64 %b) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f9:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: agr %r3, %r4
|
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: csg %r2, {{%r[0-9]+}}, 0(%r3)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%add = add i64 %base, %index
|
|
|
|
%ptr = inttoptr i64 %add to i64 *
|
|
|
|
%res = atomicrmw min i64 *%ptr, i64 %b seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|
|
|
|
|
2013-06-27 09:38:48 +00:00
|
|
|
; Check that constants are handled.
|
2013-05-06 16:17:29 +00:00
|
|
|
define i64 @f10(i64 %dummy, i64 *%ptr) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f10:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: lghi [[LIMIT:%r[0-9]+]], 42
|
|
|
|
; CHECK: lg %r2, 0(%r3)
|
|
|
|
; CHECK: [[LOOP:\.[^:]*]]:
|
|
|
|
; CHECK: lgr [[NEW:%r[0-9]+]], %r2
|
2013-05-28 10:41:11 +00:00
|
|
|
; CHECK: cgrjle %r2, [[LIMIT]], [[KEEP:\..*]]
|
2013-06-27 09:38:48 +00:00
|
|
|
; CHECK: lghi [[NEW]], 42
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: csg %r2, [[NEW]], 0(%r3)
|
[SystemZ] Be more careful about inverting CC masks
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.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
|
|
|
; CHECK: jl [[LOOP]]
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: br %r14
|
|
|
|
%res = atomicrmw min i64 *%ptr, i64 42 seq_cst
|
|
|
|
ret i64 %res
|
|
|
|
}
|