mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-23 11:13:28 +01:00
c28d8cf19b
This commit removes the artificial types <512 x i1> and <1024 x i1>
from HVX intrinsics, and makes v512i1 and v1024i1 no longer legal on
Hexagon.
It may cause existing bitcode files to become invalid.
* Converting between vector predicates and vector registers must be
done explicitly via vandvrt/vandqrt instructions (their intrinsics),
i.e. (for 64-byte mode):
%Q = call <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32> %V, i32 -1)
%V = call <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1> %Q, i32 -1)
The conversion intrinsics are:
declare <64 x i1> @llvm.hexagon.V6.vandvrt(<16 x i32>, i32)
declare <128 x i1> @llvm.hexagon.V6.vandvrt.128B(<32 x i32>, i32)
declare <16 x i32> @llvm.hexagon.V6.vandqrt(<64 x i1>, i32)
declare <32 x i32> @llvm.hexagon.V6.vandqrt.128B(<128 x i1>, i32)
They are all pure.
* Vector predicate values cannot be loaded/stored directly. This directly
reflects the architecture restriction. Loading and storing or vector
predicates must be done indirectly via vector registers and explicit
conversions via vandvrt/vandqrt instructions.
138 lines
9.5 KiB
LLVM
138 lines
9.5 KiB
LLVM
; RUN: llc -march=hexagon -O3 < %s
|
|
; REQUIRES: asserts
|
|
|
|
; Test that we don't assert because of requiring too many scavenger spill
|
|
; slots. This happens because the kill flag wasn't added to the appropriate
|
|
; operands for the spill code.
|
|
|
|
define void @f0(i32 %a0, i8* noalias nocapture %a1) #0 {
|
|
b0:
|
|
%v0 = tail call <32 x i32> @llvm.hexagon.V6.vshuffh.128B(<32 x i32> undef)
|
|
%v1 = sdiv i32 %a0, 128
|
|
%v2 = icmp sgt i32 %a0, 127
|
|
br i1 %v2, label %b1, label %b3
|
|
|
|
b1: ; preds = %b0
|
|
%v3 = bitcast i8* %a1 to <32 x i32>*
|
|
br label %b2
|
|
|
|
b2: ; preds = %b2, %b1
|
|
%v4 = phi <32 x i32>* [ %v3, %b1 ], [ undef, %b2 ]
|
|
%v5 = tail call <32 x i32> @llvm.hexagon.V6.vlalignbi.128B(<32 x i32> undef, <32 x i32> zeroinitializer, i32 2)
|
|
%v6 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v5, <32 x i32> zeroinitializer)
|
|
%v7 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> zeroinitializer, <32 x i32> zeroinitializer, <32 x i32> zeroinitializer)
|
|
%v8 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v7, <32 x i32> zeroinitializer)
|
|
%v9 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> zeroinitializer, <32 x i32> %v8, <32 x i32> zeroinitializer)
|
|
%v10 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v9, <32 x i32> zeroinitializer)
|
|
%v11 = tail call <32 x i32> @llvm.hexagon.V6.vlalignbi.128B(<32 x i32> undef, <32 x i32> zeroinitializer, i32 4)
|
|
%v12 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v11, <32 x i32> zeroinitializer)
|
|
%v13 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> zeroinitializer, <32 x i32> zeroinitializer)
|
|
%v14 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> zeroinitializer, <32 x i32> undef)
|
|
%v15 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v12, <32 x i32> undef)
|
|
%v16 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v13, <32 x i32> undef)
|
|
%v17 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v10, <32 x i32> zeroinitializer)
|
|
%v18 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v14, <32 x i32> %v17, <32 x i32> zeroinitializer)
|
|
%v19 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v15, <32 x i32> %v18, <32 x i32> zeroinitializer)
|
|
%v20 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v16, <32 x i32> %v19, <32 x i32> zeroinitializer)
|
|
%v21 = getelementptr inbounds i8, i8* null, i32 undef
|
|
%v22 = bitcast i8* %v21 to <32 x i32>*
|
|
%v23 = load <32 x i32>, <32 x i32>* %v22, align 128, !tbaa !0
|
|
%v24 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v23, <32 x i32> zeroinitializer)
|
|
%v25 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v24, <32 x i32> undef)
|
|
%v26 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v25, <32 x i32> %v20, <32 x i32> zeroinitializer)
|
|
%v27 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v26, <32 x i32> zeroinitializer)
|
|
%v28 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v27, <32 x i32> zeroinitializer)
|
|
%v29 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v28, <32 x i32> zeroinitializer)
|
|
%v30 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v29, <32 x i32> zeroinitializer)
|
|
%v31 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v30, <32 x i32> zeroinitializer)
|
|
%v32 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v31, <32 x i32> zeroinitializer)
|
|
%v33 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v32, <32 x i32> zeroinitializer)
|
|
%v34 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v33, <32 x i32> zeroinitializer)
|
|
%v35 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v34, <32 x i32> zeroinitializer)
|
|
%v36 = tail call <32 x i32> @llvm.hexagon.V6.vlalignbi.128B(<32 x i32> undef, <32 x i32> undef, i32 1)
|
|
%v37 = tail call <32 x i32> @llvm.hexagon.V6.valignbi.128B(<32 x i32> undef, <32 x i32> undef, i32 1)
|
|
%v38 = tail call <32 x i32> @llvm.hexagon.V6.vlalignbi.128B(<32 x i32> undef, <32 x i32> undef, i32 2)
|
|
%v39 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v36, <32 x i32> zeroinitializer)
|
|
%v40 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v37, <32 x i32> zeroinitializer)
|
|
%v41 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v38, <32 x i32> zeroinitializer)
|
|
%v42 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v39, <32 x i32> undef)
|
|
%v43 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v40, <32 x i32> undef)
|
|
%v44 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v41, <32 x i32> undef)
|
|
%v45 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> undef, <32 x i32> undef)
|
|
%v46 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v42, <32 x i32> %v35, <32 x i32> zeroinitializer)
|
|
%v47 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v43, <32 x i32> %v46, <32 x i32> zeroinitializer)
|
|
%v48 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v44, <32 x i32> %v47, <32 x i32> zeroinitializer)
|
|
%v49 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v45, <32 x i32> %v48, <32 x i32> zeroinitializer)
|
|
%v50 = tail call <32 x i32> @llvm.hexagon.V6.vlalignbi.128B(<32 x i32> undef, <32 x i32> undef, i32 4)
|
|
%v51 = tail call <32 x i32> @llvm.hexagon.V6.valignbi.128B(<32 x i32> undef, <32 x i32> undef, i32 4)
|
|
%v52 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> undef, <32 x i32> zeroinitializer)
|
|
%v53 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v50, <32 x i32> zeroinitializer)
|
|
%v54 = tail call <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32> %v51, <32 x i32> zeroinitializer)
|
|
%v55 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v52, <32 x i32> undef)
|
|
%v56 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v53, <32 x i32> undef)
|
|
%v57 = tail call <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32> %v54, <32 x i32> undef)
|
|
%v58 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> undef, <32 x i32> %v49, <32 x i32> zeroinitializer)
|
|
%v59 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v55, <32 x i32> %v58, <32 x i32> zeroinitializer)
|
|
%v60 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v56, <32 x i32> %v59, <32 x i32> zeroinitializer)
|
|
%v61 = tail call <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1> %v57, <32 x i32> %v60, <32 x i32> zeroinitializer)
|
|
%v62 = tail call <64 x i32> @llvm.hexagon.V6.vlutvwh.oracc.128B(<64 x i32> zeroinitializer, <32 x i32> %v61, <32 x i32> undef, i32 5)
|
|
%v63 = tail call <64 x i32> @llvm.hexagon.V6.vmpyuhv.128B(<32 x i32> undef, <32 x i32> undef)
|
|
%v64 = tail call <32 x i32> @llvm.hexagon.V6.hi.128B(<64 x i32> %v62)
|
|
%v65 = tail call <64 x i32> @llvm.hexagon.V6.vmpyuhv.128B(<32 x i32> zeroinitializer, <32 x i32> %v64)
|
|
%v66 = tail call <32 x i32> @llvm.hexagon.V6.hi.128B(<64 x i32> %v63)
|
|
%v67 = tail call <32 x i32> @llvm.hexagon.V6.lo.128B(<64 x i32> %v63)
|
|
%v68 = tail call <32 x i32> @llvm.hexagon.V6.vasrwh.128B(<32 x i32> %v66, <32 x i32> %v67, i32 14)
|
|
%v69 = tail call <32 x i32> @llvm.hexagon.V6.hi.128B(<64 x i32> %v65)
|
|
%v70 = tail call <32 x i32> @llvm.hexagon.V6.vasrwh.128B(<32 x i32> %v69, <32 x i32> undef, i32 14)
|
|
%v71 = tail call <32 x i32> @llvm.hexagon.V6.vshuffeb.128B(<32 x i32> %v70, <32 x i32> %v68)
|
|
store <32 x i32> %v71, <32 x i32>* %v4, align 128, !tbaa !0
|
|
%v72 = icmp slt i32 0, %v1
|
|
br i1 %v72, label %b2, label %b3
|
|
|
|
b3: ; preds = %b2, %b0
|
|
ret void
|
|
}
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.vshuffh.128B(<32 x i32>) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.vabsdiffub.128B(<32 x i32>, <32 x i32>) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <128 x i1> @llvm.hexagon.V6.vgtub.128B(<32 x i32>, <32 x i32>) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.vaddbnq.128B(<128 x i1>, <32 x i32>, <32 x i32>) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.vlalignbi.128B(<32 x i32>, <32 x i32>, i32) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.valignbi.128B(<32 x i32>, <32 x i32>, i32) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <64 x i32> @llvm.hexagon.V6.vlutvwh.oracc.128B(<64 x i32>, <32 x i32>, <32 x i32>, i32) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <64 x i32> @llvm.hexagon.V6.vmpyuhv.128B(<32 x i32>, <32 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 <32 x i32> @llvm.hexagon.V6.hi.128B(<64 x i32>) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.vasrwh.128B(<32 x i32>, <32 x i32>, i32) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare <32 x i32> @llvm.hexagon.V6.vshuffeb.128B(<32 x i32>, <32 x i32>) #1
|
|
|
|
attributes #0 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length128b" }
|
|
attributes #1 = { nounwind readnone }
|
|
|
|
!0 = !{!1, !1, i64 0}
|
|
!1 = !{!"omnipotent char", !2, i64 0}
|
|
!2 = !{!"Simple C/C++ TBAA"}
|