1 ; RUN: opt < %s -reassociate -S | FileCheck %s
3 ; Check that a*c+b*c is turned into (a+b)*c
4 define <4 x float> @test1(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
6 ; CHECK-NEXT: %tmp = fadd fast <4 x float> %b, %a
7 ; CHECK-NEXT: %tmp1 = fmul fast <4 x float> %tmp, %c
8 ; CHECK-NEXT: ret <4 x float> %tmp1
10 %mul = fmul fast <4 x float> %a, %c
11 %mul1 = fmul fast <4 x float> %b, %c
12 %add = fadd fast <4 x float> %mul, %mul1
16 ; Check that a*a*b+a*a*c is turned into a*(a*(b+c)).
17 define <2 x float> @test2(<2 x float> %a, <2 x float> %b, <2 x float> %c) {
19 ; CHECK-NEXT: fadd fast <2 x float> %c, %b
20 ; CHECK-NEXT: fmul fast <2 x float> %a, %tmp2
21 ; CHECK-NEXT: fmul fast <2 x float> %tmp3, %a
22 ; CHECK-NEXT: ret <2 x float>
24 %t0 = fmul fast <2 x float> %a, %b
25 %t1 = fmul fast <2 x float> %a, %t0
26 %t2 = fmul fast <2 x float> %a, %c
27 %t3 = fmul fast <2 x float> %a, %t2
28 %t4 = fadd fast <2 x float> %t1, %t3
32 ; Check that a*b+a*c+d is turned into a*(b+c)+d.
33 define <2 x double> @test3(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x double> %d) {
35 ; CHECK-NEXT: fadd fast <2 x double> %c, %b
36 ; CHECK-NEXT: fmul fast <2 x double> %tmp, %a
37 ; CHECK-NEXT: fadd fast <2 x double> %tmp1, %d
38 ; CHECK-NEXT: ret <2 x double>
40 %t0 = fmul fast <2 x double> %a, %b
41 %t1 = fmul fast <2 x double> %a, %c
42 %t2 = fadd fast <2 x double> %t1, %d
43 %t3 = fadd fast <2 x double> %t0, %t2
48 define <2 x float> @test4(<2 x float> %A) {
50 ; CHECK-NEXT: %X = fadd <2 x float> %A, <float 1.000000e+00, float 1.000000e+00>
51 ; CHECK-NEXT: %Y = fadd <2 x float> %A, <float 1.000000e+00, float 1.000000e+00>
52 ; CHECK-NEXT: %R = fsub <2 x float> %X, %Y
53 ; CHECK-NEXT: ret <2 x float> %R
55 %X = fadd <2 x float> %A, < float 1.000000e+00, float 1.000000e+00 >
56 %Y = fadd <2 x float> %A, < float 1.000000e+00, float 1.000000e+00 >
57 %R = fsub <2 x float> %X, %Y
61 ; Check 47*X + 47*X -> 94*X.
62 define <2 x float> @test5(<2 x float> %X) {
64 ; CHECK-NEXT: fmul fast <2 x float> %X, <float 9.400000e+01, float 9.400000e+01>
65 ; CHECK-NEXT: ret <2 x float>
67 %Y = fmul fast <2 x float> %X, <float 4.700000e+01, float 4.700000e+01>
68 %Z = fadd fast <2 x float> %Y, %Y
73 define <2 x float> @test6(<2 x float> %X) {
75 ; CHECK-NEXT: fmul fast <2 x float> %X, <float 3.000000e+00, float 3.000000e+00>
76 ; CHECK-NEXT: ret <2 x float>
78 %Y = fadd fast <2 x float> %X ,%X
79 %Z = fadd fast <2 x float> %Y, %X
83 ; Check 127*W+50*W -> 177*W.
84 define <2 x double> @test7(<2 x double> %W) {
86 ; CHECK-NEXT: fmul fast <2 x double> %W, <double 1.770000e+02, double 1.770000e+02>
87 ; CHECK-NEXT: ret <2 x double>
89 %X = fmul fast <2 x double> %W, <double 127.0, double 127.0>
90 %Y = fmul fast <2 x double> %W, <double 50.0, double 50.0>
91 %Z = fadd fast <2 x double> %Y, %X
95 ; Check X*12*12 -> X*144.
96 define <2 x float> @test8(<2 x float> %arg) {
98 ; CHECK: fmul fast <2 x float> %arg, <float 1.440000e+02, float 1.440000e+02>
99 ; CHECK-NEXT: ret <2 x float> %tmp2
101 %tmp1 = fmul fast <2 x float> <float 1.200000e+01, float 1.200000e+01>, %arg
102 %tmp2 = fmul fast <2 x float> %tmp1, <float 1.200000e+01, float 1.200000e+01>
103 ret <2 x float> %tmp2
106 ; Check (b+(a+1234))+-a -> b+1234.
107 define <2 x double> @test9(<2 x double> %b, <2 x double> %a) {
108 ; CHECK-LABEL: @test9
109 ; CHECK: fadd fast <2 x double> %b, <double 1.234000e+03, double 1.234000e+03>
110 ; CHECK-NEXT: ret <2 x double>
112 %1 = fadd fast <2 x double> %a, <double 1.234000e+03, double 1.234000e+03>
113 %2 = fadd fast <2 x double> %b, %1
114 %3 = fsub fast <2 x double> <double 0.000000e+00, double 0.000000e+00>, %a
115 %4 = fadd fast <2 x double> %2, %3
119 ; Check -(-(z*40)*a) -> a*40*z.
120 define <2 x float> @test10(<2 x float> %a, <2 x float> %b, <2 x float> %z) {
121 ; CHECK-LABEL: @test10
122 ; CHECK: fmul fast <2 x float> %a, <float 4.000000e+01, float 4.000000e+01>
123 ; CHECK-NEXT: fmul fast <2 x float> %e, %z
124 ; CHECK-NEXT: ret <2 x float>
126 %d = fmul fast <2 x float> %z, <float 4.000000e+01, float 4.000000e+01>
127 %c = fsub fast <2 x float> <float 0.000000e+00, float 0.000000e+00>, %d
128 %e = fmul fast <2 x float> %a, %c
129 %f = fsub fast <2 x float> <float 0.000000e+00, float 0.000000e+00>, %e
133 ; Check x*y+y*x -> x*y*2.
134 define <2 x double> @test11(<2 x double> %x, <2 x double> %y) {
135 ; CHECK-LABEL: @test11
136 ; CHECK-NEXT: %factor = fmul fast <2 x double> %y, <double 2.000000e+00, double 2.000000e+00>
137 ; CHECK-NEXT: %tmp1 = fmul fast <2 x double> %factor, %x
138 ; CHECK-NEXT: ret <2 x double> %tmp1
140 %1 = fmul fast <2 x double> %x, %y
141 %2 = fmul fast <2 x double> %y, %x
142 %3 = fadd fast <2 x double> %1, %2
146 ; FIXME: shifts should be converted to mul to assist further reassociation.
147 define <2 x i64> @test12(<2 x i64> %b, <2 x i64> %c) {
148 ; CHECK-LABEL: @test12
149 ; CHECK-NEXT: %mul = mul <2 x i64> %c, %b
150 ; CHECK-NEXT: %shl = shl <2 x i64> %mul, <i64 5, i64 5>
151 ; CHECK-NEXT: ret <2 x i64> %shl
153 %mul = mul <2 x i64> %c, %b
154 %shl = shl <2 x i64> %mul, <i64 5, i64 5>
158 ; FIXME: expressions with a negative const should be canonicalized to assist
159 ; further reassociation.
160 ; We would expect (-5*b)+a -> a-(5*b) but only the constant operand is commuted.
161 define <4 x float> @test13(<4 x float> %a, <4 x float> %b) {
162 ; CHECK-LABEL: @test13
163 ; CHECK-NEXT: %mul = fmul fast <4 x float> %b, <float -5.000000e+00, float -5.000000e+00, float -5.000000e+00, float -5.000000e+00>
164 ; CHECK-NEXT: %add = fadd fast <4 x float> %mul, %a
165 ; CHECK-NEXT: ret <4 x float> %add
167 %mul = fmul fast <4 x float> <float -5.000000e+00, float -5.000000e+00, float -5.000000e+00, float -5.000000e+00>, %b
168 %add = fadd fast <4 x float> %mul, %a
172 ; Break up subtract to assist further reassociation.
173 ; Check a+b-c -> a+b+-c.
174 define <2 x i64> @test14(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) {
175 ; CHECK-LABEL: @test14
176 ; CHECK-NEXT: %add = add <2 x i64> %b, %a
177 ; CHECK-NEXT: %c.neg = sub <2 x i64> zeroinitializer, %c
178 ; CHECK-NEXT: %sub = add <2 x i64> %add, %c.neg
179 ; CHECK-NEXT: ret <2 x i64> %sub
181 %add = add <2 x i64> %b, %a
182 %sub = sub <2 x i64> %add, %c
186 define <2 x i32> @test15(<2 x i32> %x, <2 x i32> %y) {
187 ; CHECK-LABEL: test15
188 ; CHECK-NEXT: %tmp3 = and <2 x i32> %y, %x
189 ; CHECK-NEXT: ret <2 x i32> %tmp3
191 %tmp1 = and <2 x i32> %x, %y
192 %tmp2 = and <2 x i32> %y, %x
193 %tmp3 = and <2 x i32> %tmp1, %tmp2
197 define <2 x i32> @test16(<2 x i32> %x, <2 x i32> %y) {
198 ; CHECK-LABEL: test16
199 ; CHECK-NEXT: %tmp3 = or <2 x i32> %y, %x
200 ; CHECK-NEXT: ret <2 x i32> %tmp3
202 %tmp1 = or <2 x i32> %x, %y
203 %tmp2 = or <2 x i32> %y, %x
204 %tmp3 = or <2 x i32> %tmp1, %tmp2
208 ; FIXME: Optimize vector xor. Currently only commute operands.
209 define <2 x i32> @test17(<2 x i32> %x, <2 x i32> %y) {
210 ; CHECK-LABEL: test17
211 ; CHECK-NEXT: %tmp1 = xor <2 x i32> %x, %y
212 ; CHECK-NEXT: %tmp2 = xor <2 x i32> %x, %y
213 ; CHECK-NEXT: %tmp3 = xor <2 x i32> %tmp1, %tmp2
215 %tmp1 = xor <2 x i32> %x, %y
216 %tmp2 = xor <2 x i32> %y, %x
217 %tmp3 = xor <2 x i32> %tmp1, %tmp2