1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 11:13:28 +01:00
llvm-mirror/test/CodeGen/Hexagon/swp-resmii-1.ll
James Molloy 25cc728d30 [ModuloSchedule] Peel out prologs and epilogs, generate actual code
Summary:
This extends the PeelingModuloScheduleExpander to generate prolog and epilog code,
and correctly stitch uses through the prolog, kernel, epilog DAG.

The key concept in this patch is to ensure that all transforms are *local*; only a
function of a block and its immediate predecessor and successor. By defining the problem in this way
we can inductively rewrite the entire DAG using only local knowledge that is easy to
reason about.

For example, we assume that all prologs and epilogs are near-perfect clones of the
steady-state kernel. This means that if a block has an instruction that is predicated out,
we can redirect all users of that instruction to that equivalent instruction in our
immediate predecessor. As all blocks are clones, every instruction must have an equivalent in
every other block.

Similarly we can make the assumption by construction that if a value defined in a block is used
outside that block, the only possible user is its immediate successors. We maintain this
even for values that are used outside the loop by creating a limited form of LCSSA.

This code isn't small, but it isn't complex.

Enabled a bunch of testing from Hexagon. There are a couple of tests not enabled yet;
I'm about 80% sure there isn't buggy codegen but the tests are checking for patterns
that we don't produce. Those still need a bit more investigation. In the meantime we
(Google) are happy with the code produced by this on our downstream SMS implementation,
and believe it generates correct code.

Subscribers: mgorny, hiraditya, jsji, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68205

llvm-svn: 373462
2019-10-02 12:46:44 +00:00

115 lines
4.8 KiB
LLVM

; RUN: llc -march=hexagon -enable-pipeliner -debug-only=pipeliner < %s -o - 2>&1 > /dev/null -pipeliner-experimental-cg=true | FileCheck %s
; REQUIRES: asserts
; Test that checks that we compute the correct ResMII for haar.
; CHECK: MII = 4 MAX_II = 14 (rec=1, res=4)
; Function Attrs: nounwind
define void @f0(i16* noalias nocapture readonly %a0, i32 %a1, i32 %a2, i32 %a3, i8* noalias nocapture %a4, i32 %a5) #0 {
b0:
%v0 = ashr i32 %a3, 2
%v1 = ashr i32 %a3, 1
%v2 = add i32 %v1, %v0
%v3 = icmp sgt i32 %a2, 0
br i1 %v3, label %b1, label %b8
b1: ; preds = %b0
%v4 = sdiv i32 %a1, 64
%v5 = icmp sgt i32 %a1, 63
br label %b2
b2: ; preds = %b6, %b1
%v6 = phi i32 [ 0, %b1 ], [ %v56, %b6 ]
%v7 = ashr exact i32 %v6, 1
%v8 = mul nsw i32 %v7, %a3
br i1 %v5, label %b3, label %b6
b3: ; preds = %b2
%v9 = add nsw i32 %v6, 1
%v10 = mul nsw i32 %v9, %a5
%v11 = mul nsw i32 %v6, %a5
%v12 = add i32 %v2, %v8
%v13 = add i32 %v8, %v0
%v14 = add i32 %v8, %v1
%v15 = getelementptr inbounds i8, i8* %a4, i32 %v10
%v16 = getelementptr inbounds i8, i8* %a4, i32 %v11
%v17 = getelementptr inbounds i16, i16* %a0, i32 %v12
%v18 = getelementptr inbounds i16, i16* %a0, i32 %v13
%v19 = getelementptr inbounds i16, i16* %a0, i32 %v14
%v20 = getelementptr inbounds i16, i16* %a0, i32 %v8
%v21 = bitcast i8* %v15 to <16 x i32>*
%v22 = bitcast i8* %v16 to <16 x i32>*
%v23 = bitcast i16* %v17 to <16 x i32>*
%v24 = bitcast i16* %v18 to <16 x i32>*
%v25 = bitcast i16* %v19 to <16 x i32>*
%v26 = bitcast i16* %v20 to <16 x i32>*
br label %b4
b4: ; preds = %b4, %b3
%v27 = phi i32 [ 0, %b3 ], [ %v54, %b4 ]
%v28 = phi <16 x i32>* [ %v26, %b3 ], [ %v34, %b4 ]
%v29 = phi <16 x i32>* [ %v25, %b3 ], [ %v36, %b4 ]
%v30 = phi <16 x i32>* [ %v24, %b3 ], [ %v38, %b4 ]
%v31 = phi <16 x i32>* [ %v23, %b3 ], [ %v40, %b4 ]
%v32 = phi <16 x i32>* [ %v21, %b3 ], [ %v53, %b4 ]
%v33 = phi <16 x i32>* [ %v22, %b3 ], [ %v52, %b4 ]
%v34 = getelementptr inbounds <16 x i32>, <16 x i32>* %v28, i32 1
%v35 = load <16 x i32>, <16 x i32>* %v28, align 64
%v36 = getelementptr inbounds <16 x i32>, <16 x i32>* %v29, i32 1
%v37 = load <16 x i32>, <16 x i32>* %v29, align 64
%v38 = getelementptr inbounds <16 x i32>, <16 x i32>* %v30, i32 1
%v39 = load <16 x i32>, <16 x i32>* %v30, align 64
%v40 = getelementptr inbounds <16 x i32>, <16 x i32>* %v31, i32 1
%v41 = load <16 x i32>, <16 x i32>* %v31, align 64
%v42 = tail call <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32> %v35, <16 x i32> %v37)
%v43 = tail call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %v35, <16 x i32> %v37)
%v44 = tail call <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32> %v39, <16 x i32> %v41)
%v45 = tail call <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32> %v39, <16 x i32> %v41)
%v46 = tail call <16 x i32> @llvm.hexagon.V6.vavgh(<16 x i32> %v42, <16 x i32> %v44)
%v47 = tail call <16 x i32> @llvm.hexagon.V6.vnavgh(<16 x i32> %v42, <16 x i32> %v44)
%v48 = tail call <16 x i32> @llvm.hexagon.V6.vavgh(<16 x i32> %v43, <16 x i32> %v45)
%v49 = tail call <16 x i32> @llvm.hexagon.V6.vnavgh(<16 x i32> %v43, <16 x i32> %v45)
%v50 = tail call <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32> %v47, <16 x i32> %v46)
%v51 = tail call <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32> %v49, <16 x i32> %v48)
%v52 = getelementptr inbounds <16 x i32>, <16 x i32>* %v33, i32 1
store <16 x i32> %v50, <16 x i32>* %v33, align 64
%v53 = getelementptr inbounds <16 x i32>, <16 x i32>* %v32, i32 1
store <16 x i32> %v51, <16 x i32>* %v32, align 64
%v54 = add nsw i32 %v27, 1
%v55 = icmp slt i32 %v54, %v4
br i1 %v55, label %b4, label %b5
b5: ; preds = %b4
br label %b6
b6: ; preds = %b5, %b2
%v56 = add nsw i32 %v6, 2
%v57 = icmp slt i32 %v56, %a2
br i1 %v57, label %b2, label %b7
b7: ; preds = %b6
br label %b8
b8: ; preds = %b7, %b0
ret void
}
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vaddh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsubh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vavgh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vnavgh(<16 x i32>, <16 x i32>) #1
; Function Attrs: nounwind readnone
declare <16 x i32> @llvm.hexagon.V6.vsathub(<16 x i32>, <16 x i32>) #1
attributes #0 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length64b" }
attributes #1 = { nounwind readnone }