1 ; RUN: opt -S -loop-vectorize -instcombine -force-vector-width=4 -force-vector-interleave=1 -enable-interleaved-mem-accesses=true < %s | FileCheck %s
3 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
5 ; Check that the interleaved-mem-access analysis identifies the access
6 ; to array 'in' as interleaved, despite the possibly wrapping unsigned
9 ; In this test the interleave-groups are full (have no gaps), so no wrapping
10 ; checks are necessary. We can call getPtrStride with Assume=false and
11 ; ShouldCheckWrap=false to safely figure out that the stride is 2.
20 ; Complex() : real_(0), imaginary_(0) { }
21 ; Complex(float real, float imaginary) : real_(real), imaginary_(imaginary) { }
22 ; Complex(const Complex &rhs) : real_(rhs.real()), imaginary_(rhs.imaginary()) { }
24 ; inline float real() const { return real_; }
25 ; inline float imaginary() const { return imaginary_; }
28 ;void test(Complex * __restrict__ out, Complex * __restrict__ in, size_t out_start, size_t size)
30 ; for (size_t out_offset = 0; out_offset < size; ++out_offset)
32 ; size_t out_ix = out_start + out_offset;
33 ; Complex t0 = in[out_ix];
39 ; CHECK: %wide.vec = load <8 x i32>, <8 x i32>* {{.*}}, align 4
40 ; CHECK: shufflevector <8 x i32> %wide.vec, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
41 ; CHECK: shufflevector <8 x i32> %wide.vec, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
43 %class.Complex = type { float, float }
45 define void @_Z4testP7ComplexS0_mm(%class.Complex* noalias nocapture %out, %class.Complex* noalias nocapture readonly %in, i64 %out_start, i64 %size) local_unnamed_addr {
47 %cmp9 = icmp eq i64 %size, 0
48 br i1 %cmp9, label %for.cond.cleanup, label %for.body.preheader
53 for.cond.cleanup.loopexit:
54 br label %for.cond.cleanup
60 %out_offset.010 = phi i64 [ %inc, %for.body ], [ 0, %for.body.preheader ]
61 %add = add i64 %out_offset.010, %out_start
62 %arrayidx = getelementptr inbounds %class.Complex, %class.Complex* %in, i64 %add
63 %0 = bitcast %class.Complex* %arrayidx to i32*
64 %1 = load i32, i32* %0, align 4
65 %imaginary_.i.i = getelementptr inbounds %class.Complex, %class.Complex* %in, i64 %add, i32 1
66 %2 = bitcast float* %imaginary_.i.i to i32*
67 %3 = load i32, i32* %2, align 4
68 %arrayidx1 = getelementptr inbounds %class.Complex, %class.Complex* %out, i64 %add
69 %4 = bitcast %class.Complex* %arrayidx1 to i64*
70 %t0.sroa.4.0.insert.ext = zext i32 %3 to i64
71 %t0.sroa.4.0.insert.shift = shl nuw i64 %t0.sroa.4.0.insert.ext, 32
72 %t0.sroa.0.0.insert.ext = zext i32 %1 to i64
73 %t0.sroa.0.0.insert.insert = or i64 %t0.sroa.4.0.insert.shift, %t0.sroa.0.0.insert.ext
74 store i64 %t0.sroa.0.0.insert.insert, i64* %4, align 4
75 %inc = add nuw i64 %out_offset.010, 1
76 %exitcond = icmp eq i64 %inc, %size
77 br i1 %exitcond, label %for.cond.cleanup.loopexit, label %for.body