2013-06-27 09:27:40 +00:00
|
|
|
; Test f32 conditional stores that are presented as selects.
|
|
|
|
;
|
|
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
|
|
|
|
|
|
|
|
declare void @foo(float *)
|
|
|
|
|
|
|
|
; Test with the loaded value first.
|
|
|
|
define void @f1(float *%ptr, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f1:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: ste %f0, 0(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; ...and with the loaded value second
|
|
|
|
define void @f2(float *%ptr, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f2:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
[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: jhe [[LABEL:[^ ]*]]
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: ste %f0, 0(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %alt, float %orig
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the high end of the aligned STE range.
|
|
|
|
define void @f3(float *%base, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f3:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: ste %f0, 4092(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr float *%base, i64 1023
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the next word up, which should use STEY instead of STE.
|
|
|
|
define void @f4(float *%base, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f4:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: stey %f0, 4096(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr float *%base, i64 1024
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the high end of the aligned STEY range.
|
|
|
|
define void @f5(float *%base, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f5:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: stey %f0, 524284(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr float *%base, i64 131071
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the next word up, which needs separate address logic.
|
|
|
|
; Other sequences besides this one would be OK.
|
|
|
|
define void @f6(float *%base, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f6:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: agfi %r2, 524288
|
|
|
|
; CHECK: ste %f0, 0(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr float *%base, i64 131072
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the low end of the STEY range.
|
|
|
|
define void @f7(float *%base, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f7:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: stey %f0, -524288(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr float *%base, i64 -131072
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check the next word down, which needs separate address logic.
|
|
|
|
; Other sequences besides this one would be OK.
|
|
|
|
define void @f8(float *%base, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f8:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: agfi %r2, -524292
|
|
|
|
; CHECK: ste %f0, 0(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = getelementptr float *%base, i64 -131073
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check that STEY allows an index.
|
|
|
|
define void @f9(i64 %base, i64 %index, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f9:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r2
|
|
|
|
; CHECK: stey %f0, 4096(%r3,%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: br %r14
|
|
|
|
%add1 = add i64 %base, %index
|
|
|
|
%add2 = add i64 %add1, 4096
|
|
|
|
%ptr = inttoptr i64 %add2 to float *
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Check that volatile loads are not matched.
|
|
|
|
define void @f10(float *%ptr, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f10:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK: le {{%f[0-5]}}, 0(%r2)
|
|
|
|
; CHECK: {{jl|jnl}} [[LABEL:[^ ]*]]
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: ste {{%f[0-5]}}, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load volatile float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; ...likewise stores. In this case we should have a conditional load into %f0.
|
|
|
|
define void @f11(float *%ptr, float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f11:
|
[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: jhe [[LABEL:[^ ]*]]
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK: le %f0, 0(%r2)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: ste %f0, 0(%r2)
|
|
|
|
; CHECK: br %r14
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store volatile float %res, float *%ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
; Try a frame index base.
|
|
|
|
define void @f12(float %alt, i32 %limit) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f12:
|
2013-06-27 09:27:40 +00:00
|
|
|
; CHECK: brasl %r14, foo@PLT
|
|
|
|
; CHECK-NOT: %r15
|
|
|
|
; CHECK: jl [[LABEL:[^ ]*]]
|
|
|
|
; CHECK-NOT: %r15
|
|
|
|
; CHECK: ste {{%f[0-9]+}}, {{[0-9]+}}(%r15)
|
|
|
|
; CHECK: [[LABEL]]:
|
|
|
|
; CHECK: brasl %r14, foo@PLT
|
|
|
|
; CHECK: br %r14
|
|
|
|
%ptr = alloca float
|
|
|
|
call void @foo(float *%ptr)
|
|
|
|
%cond = icmp ult i32 %limit, 42
|
|
|
|
%orig = load float *%ptr
|
|
|
|
%res = select i1 %cond, float %orig, float %alt
|
|
|
|
store float %res, float *%ptr
|
|
|
|
call void @foo(float *%ptr)
|
|
|
|
ret void
|
|
|
|
}
|