• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mcpu=pwr9 -mtriple=powerpc64le-unknown-unknown \
3; RUN:   -ppc-asm-full-reg-names -verify-machineinstrs -O2 < %s | FileCheck %s \
4; RUN:   --check-prefix=CHECK-P9
5
6define i32 @t(i32 %n, i32 %delta, i32 %a) {
7; CHECK-P9-LABEL: t:
8; CHECK-P9:       # %bb.0: # %entry
9; CHECK-P9-NEXT:    lis r7, 0
10; CHECK-P9-NEXT:    li r6, 0
11; CHECK-P9-NEXT:    li r8, 0
12; CHECK-P9-NEXT:    li r9, 0
13; CHECK-P9-NEXT:    ori r7, r7, 65535
14; CHECK-P9-NEXT:    .p2align 5
15; CHECK-P9-NEXT:  .LBB0_1: # %header
16; CHECK-P9-NEXT:    #
17; CHECK-P9-NEXT:    addi r9, r9, 1
18; CHECK-P9-NEXT:    cmpw r9, r3
19; CHECK-P9-NEXT:    blt cr0, .LBB0_4
20; CHECK-P9-NEXT:  # %bb.2: # %cont
21; CHECK-P9-NEXT:    #
22; CHECK-P9-NEXT:    add r8, r8, r4
23; CHECK-P9-NEXT:    cmpw r8, r7
24; CHECK-P9-NEXT:    bgt cr0, .LBB0_1
25; CHECK-P9-NEXT:  # %bb.3: # %cont.1
26; CHECK-P9-NEXT:    addi r6, r5, 1024
27; CHECK-P9-NEXT:  .LBB0_4: # %return
28; CHECK-P9-NEXT:    addi r3, r5, 1024
29; CHECK-P9-NEXT:    mullw r3, r6, r3
30; CHECK-P9-NEXT:    blr
31entry:
32  br label %header
33
34header:
35  %sum = phi i32 [ 0, %entry ], [ %sum.1, %cont ]
36  %i = phi i32 [ 0, %entry ], [ %i.1, %cont ]
37  %i.1 = add nsw i32 %i, 1
38  %lt = icmp slt i32 %i.1, %n
39  br i1 %lt, label %return, label %cont
40
41cont:
42  %sum.1 = add nsw i32 %sum, %delta
43  %lt.1 = icmp slt i32 %sum.1, 65536
44  br i1 %lt.1, label %cont.1, label %header
45
46cont.1:
47  %delta.1 = add nsw i32 %a, 1024
48  br label %return
49
50return:
51  %delta.2 = phi i32 [ %delta.1, %cont.1 ], [ 0, %header ]
52  %delta.3 = add nsw i32 %a, 1024
53  %ret = mul i32 %delta.2, %delta.3
54  ret i32 %ret
55}
56
57define dso_local signext i32 @foo(i32 signext %x, i32 signext %y) nounwind {
58; CHECK-P9-LABEL: foo:
59; CHECK-P9:       # %bb.0: # %entry
60; CHECK-P9-NEXT:    mflr r0
61; CHECK-P9-NEXT:    std r27, -40(r1) # 8-byte Folded Spill
62; CHECK-P9-NEXT:    std r28, -32(r1) # 8-byte Folded Spill
63; CHECK-P9-NEXT:    std r29, -24(r1) # 8-byte Folded Spill
64; CHECK-P9-NEXT:    std r30, -16(r1) # 8-byte Folded Spill
65; CHECK-P9-NEXT:    std r0, 16(r1)
66; CHECK-P9-NEXT:    stdu r1, -80(r1)
67; CHECK-P9-NEXT:    mr r30, r4
68; CHECK-P9-NEXT:    mr r29, r3
69; CHECK-P9-NEXT:    lis r3, 21845
70; CHECK-P9-NEXT:    add r28, r30, r29
71; CHECK-P9-NEXT:    ori r27, r3, 21846
72; CHECK-P9-NEXT:    b .LBB1_4
73; CHECK-P9-NEXT:    .p2align 4
74; CHECK-P9-NEXT:  .LBB1_1: # %sw.bb3
75; CHECK-P9-NEXT:    #
76; CHECK-P9-NEXT:    mulli r3, r30, 23
77; CHECK-P9-NEXT:  .LBB1_2: # %sw.epilog
78; CHECK-P9-NEXT:    #
79; CHECK-P9-NEXT:    add r28, r3, r28
80; CHECK-P9-NEXT:  .LBB1_3: # %sw.epilog
81; CHECK-P9-NEXT:    #
82; CHECK-P9-NEXT:    cmpwi r28, 1025
83; CHECK-P9-NEXT:    bge cr0, .LBB1_7
84; CHECK-P9-NEXT:  .LBB1_4: # %while.cond
85; CHECK-P9-NEXT:    #
86; CHECK-P9-NEXT:    extsw r3, r29
87; CHECK-P9-NEXT:    bl bar
88; CHECK-P9-NEXT:    nop
89; CHECK-P9-NEXT:    mr r29, r3
90; CHECK-P9-NEXT:    extsw r3, r30
91; CHECK-P9-NEXT:    bl bar
92; CHECK-P9-NEXT:    nop
93; CHECK-P9-NEXT:    mr r30, r3
94; CHECK-P9-NEXT:    mulhw r3, r28, r27
95; CHECK-P9-NEXT:    srwi r4, r3, 31
96; CHECK-P9-NEXT:    add r3, r3, r4
97; CHECK-P9-NEXT:    slwi r4, r3, 1
98; CHECK-P9-NEXT:    add r3, r3, r4
99; CHECK-P9-NEXT:    sub r3, r28, r3
100; CHECK-P9-NEXT:    cmplwi r3, 1
101; CHECK-P9-NEXT:    beq cr0, .LBB1_1
102; CHECK-P9-NEXT:  # %bb.5: # %while.cond
103; CHECK-P9-NEXT:    #
104; CHECK-P9-NEXT:    cmplwi r3, 0
105; CHECK-P9-NEXT:    bne cr0, .LBB1_3
106; CHECK-P9-NEXT:  # %bb.6: # %sw.bb
107; CHECK-P9-NEXT:    #
108; CHECK-P9-NEXT:    mulli r3, r29, 13
109; CHECK-P9-NEXT:    b .LBB1_2
110; CHECK-P9-NEXT:  .LBB1_7: # %while.end
111; CHECK-P9-NEXT:    lis r3, -13108
112; CHECK-P9-NEXT:    lis r4, 13107
113; CHECK-P9-NEXT:    ori r3, r3, 52429
114; CHECK-P9-NEXT:    ori r4, r4, 13108
115; CHECK-P9-NEXT:    mullw r3, r28, r3
116; CHECK-P9-NEXT:    cmplw r3, r4
117; CHECK-P9-NEXT:    blt cr0, .LBB1_9
118; CHECK-P9-NEXT:  # %bb.8: # %if.then8
119; CHECK-P9-NEXT:    mulli r3, r29, 13
120; CHECK-P9-NEXT:    mulli r5, r30, 23
121; CHECK-P9-NEXT:    extsw r4, r28
122; CHECK-P9-NEXT:    extsw r3, r3
123; CHECK-P9-NEXT:    extsw r5, r5
124; CHECK-P9-NEXT:    sub r3, r4, r3
125; CHECK-P9-NEXT:    sub r4, r5, r4
126; CHECK-P9-NEXT:    rldicl r3, r3, 1, 63
127; CHECK-P9-NEXT:    rldicl r4, r4, 1, 63
128; CHECK-P9-NEXT:    or r3, r3, r4
129; CHECK-P9-NEXT:    b .LBB1_10
130; CHECK-P9-NEXT:  .LBB1_9: # %cleanup20
131; CHECK-P9-NEXT:    li r3, 0
132; CHECK-P9-NEXT:  .LBB1_10: # %cleanup20
133; CHECK-P9-NEXT:    addi r1, r1, 80
134; CHECK-P9-NEXT:    ld r0, 16(r1)
135; CHECK-P9-NEXT:    ld r30, -16(r1) # 8-byte Folded Reload
136; CHECK-P9-NEXT:    ld r29, -24(r1) # 8-byte Folded Reload
137; CHECK-P9-NEXT:    ld r28, -32(r1) # 8-byte Folded Reload
138; CHECK-P9-NEXT:    mtlr r0
139; CHECK-P9-NEXT:    ld r27, -40(r1) # 8-byte Folded Reload
140; CHECK-P9-NEXT:    blr
141entry:
142  %add = add nsw i32 %y, %x
143  br label %while.cond
144
145while.cond:                                       ; preds = %sw.epilog, %entry
146  %sum.0 = phi i32 [ %add, %entry ], [ %sum.1, %sw.epilog ]
147  %y.addr.0 = phi i32 [ %y, %entry ], [ %call1, %sw.epilog ]
148  %x.addr.0 = phi i32 [ %x, %entry ], [ %call, %sw.epilog ]
149  %call = tail call signext i32 @bar(i32 signext %x.addr.0) #2
150  %call1 = tail call signext i32 @bar(i32 signext %y.addr.0) #2
151  %rem = srem i32 %sum.0, 3
152  switch i32 %rem, label %sw.epilog [
153    i32 0, label %sw.bb
154    i32 1, label %sw.bb3
155  ]
156
157sw.bb:                                            ; preds = %while.cond
158  %mul = mul nsw i32 %call, 13
159  %add2 = add nsw i32 %mul, %sum.0
160  br label %sw.epilog
161
162sw.bb3:                                           ; preds = %while.cond
163  %mul4 = mul nsw i32 %call1, 23
164  %add5 = add nsw i32 %mul4, %sum.0
165  br label %sw.epilog
166
167sw.epilog:                                        ; preds = %while.cond, %sw.bb3, %sw.bb
168  %sum.1 = phi i32 [ %sum.0, %while.cond ], [ %add5, %sw.bb3 ], [ %add2, %sw.bb ]
169  %cmp = icmp slt i32 %sum.1, 1025
170  br i1 %cmp, label %while.cond, label %while.end
171
172while.end:                                        ; preds = %sw.epilog
173  %rem739 = urem i32 %sum.1, 5
174  %tobool = icmp eq i32 %rem739, 0
175  br i1 %tobool, label %cleanup20, label %if.then8
176
177if.then8:                                         ; preds = %while.end
178  %mul9 = mul nsw i32 %call, 13
179  %cmp11 = icmp slt i32 %sum.1, %mul9
180  %mul10 = mul nsw i32 %call1, 23
181  %cmp12 = icmp sgt i32 %sum.1, %mul10
182  %or.cond = or i1 %cmp11, %cmp12
183  %spec.select = zext i1 %or.cond to i32
184  ret i32 %spec.select
185
186cleanup20:                                        ; preds = %while.end
187  ret i32 0
188}
189
190declare signext i32 @bar(i32 signext) local_unnamed_addr #1
191