replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

LegalizeDAG: Try to use Overflow operations when expanding ADD/SUB 

v2: consider BooleanContents when processing overflow

Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewers: resistor, jholewinsky (nvidia parts)
Differential Revision: http://reviews.llvm.org/D6340

llvm-svn: 234755
This commit is contained in:
Jan Vesely 2015-04-13 15:32:01 +00:00
parent d9309774a6
commit a835555e40
5 changed files with 43 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp
View File

@ -1629,6 +1629,39 @@ void DAGTypeLegalizer::ExpandIntRes_ADDSUB(SDNode *N,
return;
}
bool hasOVF =
TLI.isOperationLegalOrCustom(N->getOpcode() == ISD::ADD ?
ISD::UADDO : ISD::USUBO,
TLI.getTypeToExpandTo(*DAG.getContext(), NVT));
if (hasOVF) {
SDVTList VTList = DAG.getVTList(NVT, NVT);
TargetLoweringBase::BooleanContent BoolType = TLI.getBooleanContents(NVT);
int RevOpc;
if (N->getOpcode() == ISD::ADD) {
RevOpc = ISD::SUB;
Lo = DAG.getNode(ISD::UADDO, dl, VTList, LoOps);
Hi = DAG.getNode(ISD::ADD, dl, NVT, makeArrayRef(HiOps, 2));
} else {
RevOpc = ISD::ADD;
Lo = DAG.getNode(ISD::USUBO, dl, VTList, LoOps);
Hi = DAG.getNode(ISD::SUB, dl, NVT, makeArrayRef(HiOps, 2));
}
SDValue OVF = Lo.getValue(1);
switch (BoolType) {
default:
case TargetLoweringBase::UndefinedBooleanContent:
OVF = DAG.getNode(ISD::AND, dl, NVT, DAG.getConstant(1, NVT), OVF);
// Fallthrough
case TargetLoweringBase::ZeroOrOneBooleanContent:
Hi = DAG.getNode(N->getOpcode(), dl, NVT, Hi, OVF);
break;
case TargetLoweringBase::ZeroOrNegativeOneBooleanContent:
Hi = DAG.getNode(RevOpc, dl, NVT, Hi, OVF);
}
return;
}
if (N->getOpcode() == ISD::ADD) {
Lo = DAG.getNode(ISD::ADD, dl, NVT, LoOps);
Hi = DAG.getNode(ISD::ADD, dl, NVT, makeArrayRef(HiOps, 2));

11
llvm/test/CodeGen/Hexagon/adde.ll
View File

@ -1,16 +1,9 @@
; RUN: llc -march=hexagon -hexagon-expand-condsets=0 < %s | FileCheck %s
; CHECK: r{{[0-9]+:[0-9]+}} = #0
; CHECK: r{{[0-9]+:[0-9]+}} = #1
; CHECK: r{{[0-9]+:[0-9]+}} = add(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = add(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: p{{[0-9]+}} = cmp.gtu(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: p{{[0-9]+}} = cmp.gtu(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = combine(r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = combine(r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, #1, #0)
; CHECK: r{{[0-9]+:[0-9]+}} = add(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})

9
llvm/test/CodeGen/Hexagon/sube.ll
View File

@ -1,13 +1,10 @@
; RUN: llc -march=hexagon -hexagon-expand-condsets=0 < %s | FileCheck %s
; CHECK: r{{[0-9]+:[0-9]+}} = #0
; CHECK: r{{[0-9]+:[0-9]+}} = #1
; CHECK: r{{[0-9]+:[0-9]+}} = sub(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = sub(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: p{{[0-9]+}} = cmp.gtu(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, r{{[0-9]+}}, r{{[0-9]+}})
; CHECK: r{{[0-9]+}} = mux(p{{[0-9]+}}, #1, #0)
; CHECK: r{{[0-9]+:[0-9]+}} = sub(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = sub(r{{[0-9]+:[0-9]+}}, r{{[0-9]+:[0-9]+}})
; CHECK: r{{[0-9]+:[0-9]+}} = combine(r{{[0-9]+}}, r{{[0-9]+}})
define void @check_sube_subc(i64 %AL, i64 %AH, i64 %BL, i64 %BH, i64* %RL, i64* %RH) {
entry:

4
llvm/test/CodeGen/NVPTX/add-128bit.ll
View File

@ -7,10 +7,8 @@ target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
define void @foo(i64 %a, i64 %add, i128* %retptr) {
; CHECK: add.s64
; CHECK: setp.lt.u64
; CHECK: setp.lt.u64
; CHECK: selp.b64
; CHECK: selp.b64
; CHECK: add.s64
; CHECK: sub.s64
%t1 = sext i64 %a to i128
%add2 = zext i64 %add to i128
%val = add i128 %t1, %add2

10
llvm/test/CodeGen/R600/sub.ll
View File

@ -58,11 +58,10 @@ define void @test_sub_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)
; SI: s_sub_u32
; SI: s_subb_u32
; EG-DAG: SETGE_UINT
; EG-DAG: CNDE_INT
; EG-DAG: SUB_INT
; EG-DAG: SUB_INT
; EG-DAG: SETGT_UINT
; EG-DAG: SUB_INT
; EG-DAG: ADD_INT
define void @s_sub_i64(i64 addrspace(1)* noalias %out, i64 %a, i64 %b) nounwind {
%result = sub i64 %a, %b
store i64 %result, i64 addrspace(1)* %out, align 8
@ -73,11 +72,10 @@ define void @s_sub_i64(i64 addrspace(1)* noalias %out, i64 %a, i64 %b) nounwind
; SI: v_sub_i32_e32
; SI: v_subb_u32_e32
; EG-DAG: SETGE_UINT
; EG-DAG: CNDE_INT
; EG-DAG: SUB_INT
; EG-DAG: SUB_INT
; EG-DAG: SETGT_UINT
; EG-DAG: SUB_INT
; EG-DAG: ADD_INT
define void @v_sub_i64(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %inA, i64 addrspace(1)* noalias %inB) nounwind {
%tid = call i32 @llvm.r600.read.tidig.x() readnone
%a_ptr = getelementptr i64, i64 addrspace(1)* %inA, i32 %tid