[llvm-bugs] [Bug 47924] New: Better codegen for vector elem set at variable index

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=47924

            Bug ID: 47924
           Summary: Better codegen for vector elem set at variable index
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: david.bolvansky at gmail.com
                CC: craig.topper at gmail.com, llvm-bugs at lists.llvm.org,
                    llvm-dev at redking.me.uk, pengfei.wang at intel.com,
                    spatel+llvm at rotateright.com

#define N 32
typedef int T;
typedef T V __attribute__((vector_size(N)));
V vec_set (V v, int idx, T val)
{
  v[idx] = val;
  return v;
}


=>

V vec_set_opt (V v, int idx, T val)
{
  V valv = (V){idx, idx, idx, idx, idx, idx, idx, idx};
  V mask = ((V){0, 1, 2, 3, 4, 5, 6, 7} == valv);
  v = (v & ~mask) | (valv & mask);
  return v;
}

Clang -O3 -mavx2

vec_set(int __vector(8), int, int):                      # @vec_set(int
__vector(8), int, int)
        push    rbp
        mov     rbp, rsp
        and     rsp, -32
        sub     rsp, 64
        vmovaps ymmword ptr [rsp], ymm0
        and     edi, 7
        mov     dword ptr [rsp + 4*rdi], esi
        vmovaps ymm0, ymmword ptr [rsp]
        mov     rsp, rbp
        pop     rbp
        ret
.LCPI1_0:
        .long   0                               # 0x0
        .long   1                               # 0x1
        .long   2                               # 0x2
        .long   3                               # 0x3
        .long   4                               # 0x4
        .long   5                               # 0x5
        .long   6                               # 0x6
        .long   7                               # 0x7
vec_set_opt(int __vector(8), int, int):                 # @vec_set_opt(int
__vector(8), int, int)
        vmovd   xmm1, edi
        vpbroadcastd    ymm1, xmm1
        vmovdqa ymm2, ymmword ptr [rip + .LCPI1_0] # ymm2 = [0,1,2,3,4,5,6,7]
        vpcmpeqd        ymm1, ymm1, ymm2
        vblendvps       ymm0, ymm0, ymm2, ymm1
        ret


https://godbolt.org/z/zvYMzT

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20201020/282aeaa3/attachment.html>