xref: /llvm-project/llvm/test/CodeGen/AMDGPU/strict_fadd.f64.ll (revision f2c164c8150548d983565c4ddc0fde790f9e2a5b) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefix=GCN %s
3; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s
4; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX10 %s
5
6define double @v_constained_fadd_f64_fpexcept_strict(double %x, double %y) #0 {
7; GCN-LABEL: v_constained_fadd_f64_fpexcept_strict:
8; GCN:       ; %bb.0:
9; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
10; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
11; GCN-NEXT:    s_setpc_b64 s[30:31]
12;
13; GFX10-LABEL: v_constained_fadd_f64_fpexcept_strict:
14; GFX10:       ; %bb.0:
15; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
16; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
17; GFX10-NEXT:    s_setpc_b64 s[30:31]
18  %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
19  ret double %val
20}
21
22define double @v_constained_fadd_f64_fpexcept_ignore(double %x, double %y) #0 {
23; GCN-LABEL: v_constained_fadd_f64_fpexcept_ignore:
24; GCN:       ; %bb.0:
25; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
26; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
27; GCN-NEXT:    s_setpc_b64 s[30:31]
28;
29; GFX10-LABEL: v_constained_fadd_f64_fpexcept_ignore:
30; GFX10:       ; %bb.0:
31; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
32; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
33; GFX10-NEXT:    s_setpc_b64 s[30:31]
34  %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
35  ret double %val
36}
37
38define double @v_constained_fadd_f64_fpexcept_maytrap(double %x, double %y) #0 {
39; GCN-LABEL: v_constained_fadd_f64_fpexcept_maytrap:
40; GCN:       ; %bb.0:
41; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
42; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
43; GCN-NEXT:    s_setpc_b64 s[30:31]
44;
45; GFX10-LABEL: v_constained_fadd_f64_fpexcept_maytrap:
46; GFX10:       ; %bb.0:
47; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
48; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
49; GFX10-NEXT:    s_setpc_b64 s[30:31]
50  %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
51  ret double %val
52}
53
54define <2 x double> @v_constained_fadd_v2f64_fpexcept_strict(<2 x double> %x, <2 x double> %y) #0 {
55; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_strict:
56; GCN:       ; %bb.0:
57; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
58; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[4:5]
59; GCN-NEXT:    v_add_f64 v[2:3], v[2:3], v[6:7]
60; GCN-NEXT:    s_setpc_b64 s[30:31]
61;
62; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_strict:
63; GFX10:       ; %bb.0:
64; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
65; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[4:5]
66; GFX10-NEXT:    v_add_f64 v[2:3], v[2:3], v[6:7]
67; GFX10-NEXT:    s_setpc_b64 s[30:31]
68  %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
69  ret <2 x double> %val
70}
71
72define <2 x double> @v_constained_fadd_v2f64_fpexcept_ignore(<2 x double> %x, <2 x double> %y) #0 {
73; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_ignore:
74; GCN:       ; %bb.0:
75; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
76; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[4:5]
77; GCN-NEXT:    v_add_f64 v[2:3], v[2:3], v[6:7]
78; GCN-NEXT:    s_setpc_b64 s[30:31]
79;
80; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_ignore:
81; GFX10:       ; %bb.0:
82; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
83; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[4:5]
84; GFX10-NEXT:    v_add_f64 v[2:3], v[2:3], v[6:7]
85; GFX10-NEXT:    s_setpc_b64 s[30:31]
86  %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
87  ret <2 x double> %val
88}
89
90define <2 x double> @v_constained_fadd_v2f64_fpexcept_maytrap(<2 x double> %x, <2 x double> %y) #0 {
91; GCN-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap:
92; GCN:       ; %bb.0:
93; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
94; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[4:5]
95; GCN-NEXT:    v_add_f64 v[2:3], v[2:3], v[6:7]
96; GCN-NEXT:    s_setpc_b64 s[30:31]
97;
98; GFX10-LABEL: v_constained_fadd_v2f64_fpexcept_maytrap:
99; GFX10:       ; %bb.0:
100; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
101; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[4:5]
102; GFX10-NEXT:    v_add_f64 v[2:3], v[2:3], v[6:7]
103; GFX10-NEXT:    s_setpc_b64 s[30:31]
104  %val = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double> %x, <2 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
105  ret <2 x double> %val
106}
107
108define <3 x double> @v_constained_fadd_v3f64_fpexcept_strict(<3 x double> %x, <3 x double> %y) #0 {
109; GCN-LABEL: v_constained_fadd_v3f64_fpexcept_strict:
110; GCN:       ; %bb.0:
111; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
112; GCN-NEXT:    v_add_f64 v[0:1], v[0:1], v[6:7]
113; GCN-NEXT:    v_add_f64 v[2:3], v[2:3], v[8:9]
114; GCN-NEXT:    v_add_f64 v[4:5], v[4:5], v[10:11]
115; GCN-NEXT:    s_setpc_b64 s[30:31]
116;
117; GFX10-LABEL: v_constained_fadd_v3f64_fpexcept_strict:
118; GFX10:       ; %bb.0:
119; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
120; GFX10-NEXT:    v_add_f64 v[0:1], v[0:1], v[6:7]
121; GFX10-NEXT:    v_add_f64 v[2:3], v[2:3], v[8:9]
122; GFX10-NEXT:    v_add_f64 v[4:5], v[4:5], v[10:11]
123; GFX10-NEXT:    s_setpc_b64 s[30:31]
124  %val = call <3 x double> @llvm.experimental.constrained.fadd.v3f64(<3 x double> %x, <3 x double> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
125  ret <3 x double> %val
126}
127
128define amdgpu_ps <2 x float> @s_constained_fadd_f64_fpexcept_strict(double inreg %x, double inreg %y) #0 {
129; GCN-LABEL: s_constained_fadd_f64_fpexcept_strict:
130; GCN:       ; %bb.0:
131; GCN-NEXT:    v_mov_b32_e32 v0, s4
132; GCN-NEXT:    v_mov_b32_e32 v1, s5
133; GCN-NEXT:    v_add_f64 v[0:1], s[2:3], v[0:1]
134; GCN-NEXT:    ; return to shader part epilog
135;
136; GFX10-LABEL: s_constained_fadd_f64_fpexcept_strict:
137; GFX10:       ; %bb.0:
138; GFX10-NEXT:    v_add_f64 v[0:1], s[2:3], s[4:5]
139; GFX10-NEXT:    ; return to shader part epilog
140  %val = call double @llvm.experimental.constrained.fadd.f64(double %x, double %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
141  %cast = bitcast double %val to <2 x float>
142  ret <2 x float> %cast
143}
144
145declare double @llvm.experimental.constrained.fadd.f64(double, double, metadata, metadata) #1
146declare <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double>, <2 x double>, metadata, metadata) #1
147declare <3 x double> @llvm.experimental.constrained.fadd.v3f64(<3 x double>, <3 x double>, metadata, metadata) #1
148
149attributes #0 = { strictfp }
150attributes #1 = { inaccessiblememonly nounwind willreturn }
151