|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 |
| 2 | +; Test long double atomic stores on z14. |
| 3 | +; |
| 4 | +; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z14 | FileCheck %s |
| 5 | + |
| 6 | +define void @f1(ptr %dst, ptr %src) { |
| 7 | +; CHECK-LABEL: f1: |
| 8 | +; CHECK: # %bb.0: |
| 9 | +; CHECK-NEXT: lg %r1, 8(%r3) |
| 10 | +; CHECK-NEXT: lg %r0, 0(%r3) |
| 11 | +; CHECK-NEXT: stpq %r0, 0(%r2) |
| 12 | +; CHECK-NEXT: bcr 14, %r0 |
| 13 | +; CHECK-NEXT: br %r14 |
| 14 | + %val = load fp128, ptr %src, align 8 |
| 15 | + store atomic fp128 %val, ptr %dst seq_cst, align 16 |
| 16 | + ret void |
| 17 | +} |
| 18 | + |
| 19 | +define void @f1_fpsrc(ptr %dst, ptr %src) { |
| 20 | +; CHECK-LABEL: f1_fpsrc: |
| 21 | +; CHECK: # %bb.0: |
| 22 | +; CHECK-NEXT: vl %v0, 0(%r3), 3 |
| 23 | +; CHECK-NEXT: wfaxb %v0, %v0, %v0 |
| 24 | +; CHECK-NEXT: vlgvg %r1, %v0, 1 |
| 25 | +; CHECK-NEXT: vlgvg %r0, %v0, 0 |
| 26 | +; CHECK-NEXT: stpq %r0, 0(%r2) |
| 27 | +; CHECK-NEXT: bcr 14, %r0 |
| 28 | +; CHECK-NEXT: br %r14 |
| 29 | + %val = load fp128, ptr %src, align 8 |
| 30 | + %add = fadd fp128 %val, %val |
| 31 | + store atomic fp128 %add, ptr %dst seq_cst, align 16 |
| 32 | + ret void |
| 33 | +} |
| 34 | + |
| 35 | +define void @f2(ptr %dst, ptr %src) { |
| 36 | +; CHECK-LABEL: f2: |
| 37 | +; CHECK: # %bb.0: |
| 38 | +; CHECK-NEXT: stmg %r14, %r15, 112(%r15) |
| 39 | +; CHECK-NEXT: .cfi_offset %r14, -48 |
| 40 | +; CHECK-NEXT: .cfi_offset %r15, -40 |
| 41 | +; CHECK-NEXT: aghi %r15, -176 |
| 42 | +; CHECK-NEXT: .cfi_def_cfa_offset 336 |
| 43 | +; CHECK-NEXT: vl %v0, 0(%r3), 3 |
| 44 | +; CHECK-NEXT: lgr %r0, %r2 |
| 45 | +; CHECK-NEXT: la %r4, 160(%r15) |
| 46 | +; CHECK-NEXT: lghi %r2, 16 |
| 47 | +; CHECK-NEXT: lgr %r3, %r0 |
| 48 | +; CHECK-NEXT: lhi %r5, 5 |
| 49 | +; CHECK-NEXT: vst %v0, 160(%r15), 3 |
| 50 | +; CHECK-NEXT: brasl %r14, __atomic_store@PLT |
| 51 | +; CHECK-NEXT: lmg %r14, %r15, 288(%r15) |
| 52 | +; CHECK-NEXT: br %r14 |
| 53 | + %val = load fp128, ptr %src, align 8 |
| 54 | + store atomic fp128 %val, ptr %dst seq_cst, align 8 |
| 55 | + ret void |
| 56 | +} |
| 57 | + |
| 58 | +define void @f2_fpuse(ptr %dst, ptr %src) { |
| 59 | +; CHECK-LABEL: f2_fpuse: |
| 60 | +; CHECK: # %bb.0: |
| 61 | +; CHECK-NEXT: stmg %r14, %r15, 112(%r15) |
| 62 | +; CHECK-NEXT: .cfi_offset %r14, -48 |
| 63 | +; CHECK-NEXT: .cfi_offset %r15, -40 |
| 64 | +; CHECK-NEXT: aghi %r15, -176 |
| 65 | +; CHECK-NEXT: .cfi_def_cfa_offset 336 |
| 66 | +; CHECK-NEXT: vl %v0, 0(%r3), 3 |
| 67 | +; CHECK-NEXT: wfaxb %v0, %v0, %v0 |
| 68 | +; CHECK-NEXT: lgr %r0, %r2 |
| 69 | +; CHECK-NEXT: la %r4, 160(%r15) |
| 70 | +; CHECK-NEXT: lghi %r2, 16 |
| 71 | +; CHECK-NEXT: lgr %r3, %r0 |
| 72 | +; CHECK-NEXT: lhi %r5, 5 |
| 73 | +; CHECK-NEXT: vst %v0, 160(%r15), 3 |
| 74 | +; CHECK-NEXT: brasl %r14, __atomic_store@PLT |
| 75 | +; CHECK-NEXT: lmg %r14, %r15, 288(%r15) |
| 76 | +; CHECK-NEXT: br %r14 |
| 77 | + %val = load fp128, ptr %src, align 8 |
| 78 | + %add = fadd fp128 %val, %val |
| 79 | + store atomic fp128 %add, ptr %dst seq_cst, align 8 |
| 80 | + ret void |
| 81 | +} |
0 commit comments