blob: 9352118a32f8a6dea18ebec047cf273b9c7894b7 [file] [log] [blame]
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; Test 64-bit atomic minimum and maximum. Here we match the z10 versions,
; which can't use LOCGR.
;
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
; Todo: If-converter no longer producing CondReturns (with AtomicExpand pass).
; Check signed minimum.
define i64 @f1(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f1:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB0_2
; CHECK-NEXT: .LBB0_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB0_4
; CHECK-NEXT: .LBB0_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjle %r2, %r4, .LBB0_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB0_1
; CHECK-NEXT: .LBB0_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%res = atomicrmw min ptr %src, i64 %b seq_cst
ret i64 %res
}
; Check signed maximum.
define i64 @f2(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f2:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB1_2
; CHECK-NEXT: .LBB1_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB1_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB1_4
; CHECK-NEXT: .LBB1_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjh %r2, %r4, .LBB1_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB1_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB1_1
; CHECK-NEXT: .LBB1_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%res = atomicrmw max ptr %src, i64 %b seq_cst
ret i64 %res
}
; Check unsigned minimum.
define i64 @f3(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f3:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB2_2
; CHECK-NEXT: .LBB2_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB2_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB2_4
; CHECK-NEXT: .LBB2_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: clgrjle %r2, %r4, .LBB2_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB2_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB2_1
; CHECK-NEXT: .LBB2_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%res = atomicrmw umin ptr %src, i64 %b seq_cst
ret i64 %res
}
; Check unsigned maximum.
define i64 @f4(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f4:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB3_2
; CHECK-NEXT: .LBB3_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB3_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB3_4
; CHECK-NEXT: .LBB3_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: clgrjh %r2, %r4, .LBB3_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB3_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB3_1
; CHECK-NEXT: .LBB3_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%res = atomicrmw umax ptr %src, i64 %b seq_cst
ret i64 %res
}
; Check the high end of the aligned CSG range.
define i64 @f5(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f5:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 524280(%r3)
; CHECK-NEXT: j .LBB4_2
; CHECK-NEXT: .LBB4_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB4_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 524280(%r3)
; CHECK-NEXT: je .LBB4_4
; CHECK-NEXT: .LBB4_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjle %r2, %r4, .LBB4_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB4_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB4_1
; CHECK-NEXT: .LBB4_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%ptr = getelementptr i64, ptr %src, i64 65535
%res = atomicrmw min ptr %ptr, i64 %b seq_cst
ret i64 %res
}
; Check the next doubleword up, which requires separate address logic.
define i64 @f6(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f6:
; CHECK: # %bb.0:
; CHECK-NEXT: agfi %r3, 524288
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB5_2
; CHECK-NEXT: .LBB5_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB5_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB5_4
; CHECK-NEXT: .LBB5_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjle %r2, %r4, .LBB5_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB5_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB5_1
; CHECK-NEXT: .LBB5_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%ptr = getelementptr i64, ptr %src, i64 65536
%res = atomicrmw min ptr %ptr, i64 %b seq_cst
ret i64 %res
}
; Check the low end of the CSG range.
define i64 @f7(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f7:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, -524288(%r3)
; CHECK-NEXT: j .LBB6_2
; CHECK-NEXT: .LBB6_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB6_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, -524288(%r3)
; CHECK-NEXT: je .LBB6_4
; CHECK-NEXT: .LBB6_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjle %r2, %r4, .LBB6_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB6_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB6_1
; CHECK-NEXT: .LBB6_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%ptr = getelementptr i64, ptr %src, i64 -65536
%res = atomicrmw min ptr %ptr, i64 %b seq_cst
ret i64 %res
}
; Check the next doubleword down, which requires separate address logic.
define i64 @f8(i64 %dummy, ptr %src, i64 %b) {
; CHECK-LABEL: f8:
; CHECK: # %bb.0:
; CHECK-NEXT: agfi %r3, -524296
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB7_2
; CHECK-NEXT: .LBB7_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB7_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB7_4
; CHECK-NEXT: .LBB7_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjle %r2, %r4, .LBB7_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB7_2 Depth=1
; CHECK-NEXT: lgr %r0, %r4
; CHECK-NEXT: j .LBB7_1
; CHECK-NEXT: .LBB7_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%ptr = getelementptr i64, ptr %src, i64 -65537
%res = atomicrmw min ptr %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) {
; CHECK-LABEL: f9:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 0(%r4,%r3)
; CHECK-NEXT: agr %r3, %r4
; CHECK-NEXT: j .LBB8_2
; CHECK-NEXT: .LBB8_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB8_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB8_4
; CHECK-NEXT: .LBB8_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgrjle %r2, %r5, .LBB8_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB8_2 Depth=1
; CHECK-NEXT: lgr %r0, %r5
; CHECK-NEXT: j .LBB8_1
; CHECK-NEXT: .LBB8_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%add = add i64 %base, %index
%ptr = inttoptr i64 %add to ptr
%res = atomicrmw min ptr %ptr, i64 %b seq_cst
ret i64 %res
}
; Check that constants are handled.
define i64 @f10(i64 %dummy, ptr %ptr) {
; CHECK-LABEL: f10:
; CHECK: # %bb.0:
; CHECK-NEXT: lg %r2, 0(%r3)
; CHECK-NEXT: j .LBB9_2
; CHECK-NEXT: .LBB9_1: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB9_2 Depth=1
; CHECK-NEXT: csg %r2, %r0, 0(%r3)
; CHECK-NEXT: je .LBB9_4
; CHECK-NEXT: .LBB9_2: # %atomicrmw.start
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: lgr %r0, %r2
; CHECK-NEXT: cgijl %r2, 43, .LBB9_1
; CHECK-NEXT: # %bb.3: # %atomicrmw.start
; CHECK-NEXT: # in Loop: Header=BB9_2 Depth=1
; CHECK-NEXT: lghi %r0, 42
; CHECK-NEXT: j .LBB9_1
; CHECK-NEXT: .LBB9_4: # %atomicrmw.end
; CHECK-NEXT: br %r14
%res = atomicrmw min ptr %ptr, i64 42 seq_cst
ret i64 %res
}