[llvm-bugs] [Bug 39454] New: [X86] f64->v2f64 broadcasts should be split into MOVDUP+SUBV_BROADCAST before isel instead of with an isel pattern

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=39454

            Bug ID: 39454
           Summary: [X86] f64->v2f64 broadcasts should be split into
                    MOVDUP+SUBV_BROADCAST before isel instead of with an
                    isel pattern
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: craig.topper at gmail.com
                CC: craig.topper at gmail.com, llvm-bugs at lists.llvm.org

This would allow the load to folded into MOVDDUP here properly. 

>From broadcast-elm-cross-splat-vec.ll
; AVX-64-LABEL: f64xi8_i64:
; AVX-64:       # %bb.0:
; AVX-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
; AVX-64-NEXT:    vmovsd {{.*#+}} xmm3 = mem[0],zero
; AVX-64-NEXT:    vmovddup {{.*#+}} xmm3 = xmm3[0,0]
; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
; AVX-64-NEXT:    vpaddb %xmm3, %xmm1, %xmm1
; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
; AVX-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
; AVX-64-NEXT:    vpaddb %xmm3, %xmm2, %xmm2
; AVX-64-NEXT:    vpaddb %xmm3, %xmm0, %xmm0
; AVX-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX-64-NEXT:    vinsertf128 $1, %xmm3, %ymm3, %ymm2
; AVX-64-NEXT:    vandps %ymm2, %ymm0, %ymm0
; AVX-64-NEXT:    vandps %ymm2, %ymm1, %ymm1
; AVX-64-NEXT:    retq

This issue was exposed by D53268 which removed the promotion of and/or/xor to
vXi64.

-- 
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/20181026/4ab40b7f/attachment.html>