https://bugs.llvm.org/show_bug.cgi?id=45747
Bug ID: 45747
Summary: [VectorCombine] Generated code no longer uses
horizontal add/sub
Product: libraries
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: enhancement
Priority: P
Component: Scalar Optimizations
Assignee: unassignedb...@nondot.org
Reporter: wolfgang_p...@playstation.sony.com
CC: llvm-bugs@lists.llvm.org
After
https://github.com/llvm/llvm-project/commit/a69158c12acd635ee4bcb22b1060d92b29483042
the following test cases no longer generate horizontal add/sub for the
following test cases:
==========================================================================
typedef float __m128 __attribute__((__vector_size__(16), __aligned__(16)));
__m128 add_ps_002(__m128 a, __m128 b) {
__m128 r = (__m128){ a[0] + a[1], a[2] + a[3], b[0] + b[1], b[2] + b[3] };
return __builtin_shufflevector(r, a, -1, 1, -1, -1);
}
__m128 add_ps_013(__m128 a, __m128 b) {
__m128 r = (__m128){ a[0] + a[1], a[2] + a[3], b[0] + b[1], b[2] + b[3] };
return __builtin_shufflevector(r, a, 3, -1, -1, -1);
}
==========================================================================
The diffs (old code first):
< vhaddps %xmm0, %xmm0, %xmm0
< vmovshdup %xmm0, %xmm0 # xmm0 = xmm0[1,1,3,3]
---
> vmovshdup %xmm0, %xmm1 # xmm1 = xmm0[1,1,3,3]
> vaddps %xmm0, %xmm1, %xmm0
> vpermilps $232, %xmm0, %xmm0 # xmm0 = xmm0[0,2,2,3]
24,25c25,27
< vhaddps %xmm1, %xmm1, %xmm0
< vmovshdup %xmm0, %xmm0 # xmm0 = xmm0[1,1,3,3]
---
> vmovshdup %xmm1, %xmm0 # xmm0 = xmm1[1,1,3,3]
> vaddps %xmm1, %xmm0, %xmm0
> vpermilpd $1, %xmm0, %xmm0 # xmm0 = xmm0[1,0]
--
You are receiving this mail because:
You are on the CC list for the bug._______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
