89bbcbe285
If v2i64 scalar_to_vector is made custom, llc can crash in certain legalization cases where v2i64 vectors are injected, even if they weren't otherwise present. The code generated would be fine, but that operation is not handled in ReplaceNodeResults. Add handling.
56 lines
1.9 KiB
LLVM
56 lines
1.9 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc -mcpu=ppc64 -mtriple=powerpc64-unknown-linux-gnu \
|
|
; RUN: -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
|
|
; RUN: < %s | FileCheck %s
|
|
|
|
define void @_blah() {
|
|
; CHECK-LABEL: _blah:
|
|
; CHECK: # %bb.0: # %entry
|
|
; CHECK-NEXT: li r3, 0
|
|
; CHECK-NEXT: li r4, 15
|
|
; CHECK-NEXT: lvx v3, 0, r4
|
|
; CHECK-NEXT: addi r5, r1, -64
|
|
; CHECK-NEXT: lvx v4, 0, r3
|
|
; CHECK-NEXT: lvsl v2, 0, r3
|
|
; CHECK-NEXT: vperm v2, v4, v3, v2
|
|
; CHECK-NEXT: lwz r4, 16(0)
|
|
; CHECK-NEXT: stvx v2, 0, r5
|
|
; CHECK-NEXT: lhz r5, -64(r1)
|
|
; CHECK-NEXT: lhz r6, -58(r1)
|
|
; CHECK-NEXT: lhz r7, -52(r1)
|
|
; CHECK-NEXT: sth r4, -34(r1)
|
|
; CHECK-NEXT: sth r3, -36(r1)
|
|
; CHECK-NEXT: sth r3, -40(r1)
|
|
; CHECK-NEXT: sth r3, -44(r1)
|
|
; CHECK-NEXT: sth r3, -48(r1)
|
|
; CHECK-NEXT: addi r3, r1, -48
|
|
; CHECK-NEXT: sth r7, -38(r1)
|
|
; CHECK-NEXT: sth r6, -42(r1)
|
|
; CHECK-NEXT: sth r5, -46(r1)
|
|
; CHECK-NEXT: lvx v2, 0, r3
|
|
; CHECK-NEXT: addi r3, r1, -32
|
|
; CHECK-NEXT: vsldoi v3, v2, v2, 8
|
|
; CHECK-NEXT: vmaxuw v2, v2, v3
|
|
; CHECK-NEXT: vspltw v3, v2, 1
|
|
; CHECK-NEXT: vmaxuw v2, v2, v3
|
|
; CHECK-NEXT: stvx v2, 0, r3
|
|
; CHECK-NEXT: lwz r3, -32(r1)
|
|
; CHECK-NEXT: cmplwi r3, 0
|
|
; CHECK-NEXT: blr
|
|
entry:
|
|
%wide.vec904 = load <12 x i16>, ptr null, align 2
|
|
%strided.vec905 = shufflevector <12 x i16> %wide.vec904, <12 x i16> zeroinitializer, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
|
|
%0 = zext <4 x i16> %strided.vec905 to <4 x i32>
|
|
%1 = tail call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> %0)
|
|
%cmp55.not823 = icmp ugt i32 1, %1
|
|
br i1 %cmp55.not823, label %for.cond.cleanup56, label %for.body57.lr.ph
|
|
|
|
for.body57.lr.ph: ; preds = %entry
|
|
ret void
|
|
|
|
for.cond.cleanup56: ; preds = %entry
|
|
ret void
|
|
}
|
|
|
|
declare i32 @llvm.vector.reduce.umax.v4i32(<4 x i32>)
|