forked from OSchip/llvm-project
82 lines
3.6 KiB
LLVM
82 lines
3.6 KiB
LLVM
|
; RUN: llc -march=hexagon -O2 < %s | FileCheck %s
|
||
|
|
||
|
; Test that we convert 128B vcombine instructions to REG_SEQUENCE instructions.
|
||
|
|
||
|
; CHECK-LABEL: f0:
|
||
|
; CHECK-NOT: vcombine
|
||
|
define void @f0(i8* nocapture readonly %a0, i8* nocapture readonly %a1, i32 %a2, i8* nocapture %a3, i32 %a4, i32 %a5) #0 {
|
||
|
b0:
|
||
|
%v0 = bitcast i8* %a1 to i64*
|
||
|
%v1 = load i64, i64* %v0, align 8
|
||
|
%v2 = shl i64 %v1, 8
|
||
|
%v3 = trunc i64 %v2 to i32
|
||
|
%v4 = trunc i64 %v1 to i32
|
||
|
%v5 = and i32 %v4, 16777215
|
||
|
%v6 = bitcast i8* %a0 to <32 x i32>*
|
||
|
%v7 = load <32 x i32>, <32 x i32>* %v6, align 128
|
||
|
%v8 = getelementptr inbounds i8, i8* %a0, i32 32
|
||
|
%v9 = bitcast i8* %v8 to <32 x i32>*
|
||
|
%v10 = load <32 x i32>, <32 x i32>* %v9, align 128
|
||
|
%v11 = tail call <64 x i32> @llvm.hexagon.V6.vcombine.128B(<32 x i32> %v10, <32 x i32> %v7)
|
||
|
%v12 = tail call <64 x i32> @llvm.hexagon.V6.vrmpybusi.128B(<64 x i32> %v11, i32 %v5, i32 0)
|
||
|
%v13 = tail call <64 x i32> @llvm.hexagon.V6.vrmpybusi.128B(<64 x i32> %v11, i32 %v3, i32 0)
|
||
|
%v14 = tail call <32 x i32> @llvm.hexagon.V6.hi.128B(<64 x i32> %v12)
|
||
|
%v15 = tail call <32 x i32> @llvm.hexagon.V6.vasrwuhsat.128B(<32 x i32> %v14, <32 x i32> %v14, i32 %a2)
|
||
|
%v16 = tail call <32 x i32> @llvm.hexagon.V6.lo.128B(<64 x i32> %v13)
|
||
|
%v17 = tail call <32 x i32> @llvm.hexagon.V6.vasrwuhsat.128B(<32 x i32> %v16, <32 x i32> %v16, i32 %a2)
|
||
|
%v18 = getelementptr inbounds i8, i8* %a3, i32 32
|
||
|
%v19 = bitcast i8* %v18 to <32 x i32>*
|
||
|
store <32 x i32> %v15, <32 x i32>* %v19, align 128
|
||
|
%v20 = bitcast i8* %a3 to <32 x i32>*
|
||
|
store <32 x i32> %v17, <32 x i32>* %v20, align 128
|
||
|
ret void
|
||
|
}
|
||
|
|
||
|
; CHECK-LABEL: f1:
|
||
|
; CHECK-NOT: vcombine
|
||
|
define void @f1() #0 {
|
||
|
b0:
|
||
|
br i1 undef, label %b1, label %b3
|
||
|
|
||
|
b1: ; preds = %b1, %b0
|
||
|
%v0 = phi <64 x i32> [ %v6, %b1 ], [ undef, %b0 ]
|
||
|
%v1 = tail call <64 x i32> @llvm.hexagon.V6.vmpybus.acc.128B(<64 x i32> %v0, <32 x i32> undef, i32 16843009)
|
||
|
%v2 = tail call <64 x i32> @llvm.hexagon.V6.vmpabus.acc.128B(<64 x i32> %v1, <64 x i32> undef, i32 16843009)
|
||
|
%v3 = tail call <64 x i32> @llvm.hexagon.V6.vmpabus.acc.128B(<64 x i32> %v2, <64 x i32> undef, i32 16843009)
|
||
|
%v4 = tail call <64 x i32> @llvm.hexagon.V6.vmpabus.acc.128B(<64 x i32> %v3, <64 x i32> undef, i32 16843009)
|
||
|
%v5 = tail call <64 x i32> @llvm.hexagon.V6.vcombine.128B(<32 x i32> undef, <32 x i32> undef)
|
||
|
%v6 = tail call <64 x i32> @llvm.hexagon.V6.vmpabus.acc.128B(<64 x i32> %v4, <64 x i32> %v5, i32 16843009)
|
||
|
br i1 false, label %b2, label %b1
|
||
|
|
||
|
b2: ; preds = %b1
|
||
|
%v7 = tail call <32 x i32> @llvm.hexagon.V6.lo.128B(<64 x i32> %v6)
|
||
|
unreachable
|
||
|
|
||
|
b3: ; preds = %b0
|
||
|
ret void
|
||
|
}
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <64 x i32> @llvm.hexagon.V6.vcombine.128B(<32 x i32>, <32 x i32>) #1
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <64 x i32> @llvm.hexagon.V6.vrmpybusi.128B(<64 x i32>, i32, i32) #1
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <32 x i32> @llvm.hexagon.V6.vasrwuhsat.128B(<32 x i32>, <32 x i32>, i32) #1
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <32 x i32> @llvm.hexagon.V6.hi.128B(<64 x i32>) #1
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <32 x i32> @llvm.hexagon.V6.lo.128B(<64 x i32>) #1
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <64 x i32> @llvm.hexagon.V6.vmpybus.acc.128B(<64 x i32>, <32 x i32>, i32) #1
|
||
|
|
||
|
; Function Attrs: nounwind readnone
|
||
|
declare <64 x i32> @llvm.hexagon.V6.vmpabus.acc.128B(<64 x i32>, <64 x i32>, i32) #1
|
||
|
|
||
|
attributes #0 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length128b" }
|
||
|
attributes #1 = { nounwind readnone }
|