[SVE] Lower fixed length vXi32/vXi64 SDIV

[cameron.mcinally]

Summary

Here's a patch to lower SDIV for vectors of i32/i64 to scalable vectors.

And a couple follow-up questions:

1. v2i32 and v4i64 do not have NEON support, so I mapped those to scalable vectors. Is that the correct thing to do?

2. How do we want to handle i8/i16 vectors? Shall we sign extend to i32 vectors, then truncate the result?

[efriedma]

v2i32 and v4i64 do not have NEON support, so I mapped those to scalable vectors. Is that the correct thing to do?

Makes sense to me.

How do we want to handle i8/i16 vectors? Shall we sign extend to i32 vectors, then truncate the result?

We have code to do this for scalable types in AArch64TargetLowering::LowerDIV; I think it makes sense to extend it to fixed types. We can leave this for a followup, though.

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

1140

Did you mean to add udiv testcases?

[cameron.mcinally]

Remove UDIV setOperationAction.

[cameron.mcinally]

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

1140

Whoops. Not yet. I figure it makes sense to finalize the SDIV flowerings before starting on UDIV.

[cameron.mcinally]

Remove unneeded braces.

[efriedma]

LGTM

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

if (useSVEForFixedLengthVectors()) {
       for (auto VT : {MVT::v16i8, MVT::v8i16, MVT::v4i32})
         setOperationAction(ISD::TRUNCATE, VT, Custom);
       for (auto VT : {MVT::v8f16, MVT::v4f32})
         setOperationAction(ISD::FP_ROUND, VT, Expand);
 
       // These operations are not supported on NEON but SVE can do them.
       setOperationAction(ISD::MUL, MVT::v1i64, Custom);
       setOperationAction(ISD::MUL, MVT::v2i64, Custom);
+      setOperationAction(ISD::SDIV, MVT::v2i32, Custom);
+      setOperationAction(ISD::SDIV, MVT::v4i32, Custom);
+      setOperationAction(ISD::SDIV, MVT::v1i64, Custom);
+      setOperationAction(ISD::SDIV, MVT::v2i64, Custom);
       setOperationAction(ISD::SMAX, MVT::v1i64, Custom);
       setOperationAction(ISD::SMAX, MVT::v2i64, Custom);
       setOperationAction(ISD::SMIN, MVT::v1i64, Custom);
       setOperationAction(ISD::SMIN, MVT::v2i64, Custom);
       setOperationAction(ISD::UMAX, MVT::v1i64, Custom);
       setOperationAction(ISD::UMAX, MVT::v2i64, Custom);
       setOperationAction(ISD::UMIN, MVT::v1i64, Custom);
       setOperationAction(ISD::UMIN, MVT::v2i64, Custom);
void AArch64TargetLowering::addTypeForFixedLengthSVE(MVT VT) {
   setOperationAction(ISD::SRL, VT, Custom);
   setOperationAction(ISD::STORE, VT, Custom);
   setOperationAction(ISD::SUB, VT, Custom);
   setOperationAction(ISD::TRUNCATE, VT, Custom);
   setOperationAction(ISD::UMAX, VT, Custom);
   setOperationAction(ISD::UMIN, VT, Custom);
   setOperationAction(ISD::XOR, VT, Custom);
   setOperationAction(ISD::ZERO_EXTEND, VT, Custom);
+
+  if (VT.getVectorElementType() == MVT::i32 ||
+      VT.getVectorElementType() == MVT::i64)
+    setOperationAction(ISD::SDIV, VT, Custom);
 }
 
 void AArch64TargetLowering::addDRTypeForNEON(MVT VT) {
   addRegisterClass(VT, &AArch64::FPR64RegClass);
   addTypeForNEON(VT, MVT::v2i32);
 }
 
 void AArch64TargetLowering::addQRTypeForNEON(MVT VT) {
   addRegisterClass(VT, &AArch64::FPR128RegClass);
SDValue AArch64TargetLowering::LowerINSERT_SUBVECTOR(SDValue Op,
 
   return SDValue();
 }
 
 SDValue AArch64TargetLowering::LowerDIV(SDValue Op, SelectionDAG &DAG) const {
   EVT VT = Op.getValueType();
   bool Signed = Op.getOpcode() == ISD::SDIV;
   unsigned PredOpcode = Signed ? AArch64ISD::SDIV_PRED : AArch64ISD::UDIV_PRED;
+
+  if (useSVEForFixedLengthVectorVT(Op.getValueType(), /*OverrideNEON=*/true) &&
+      (VT.getVectorElementType() == MVT::i32 ||
+       VT.getVectorElementType() == MVT::i64))
+    return LowerToPredicatedOp(Op, DAG, PredOpcode, /*OverrideNEON=*/true);
+
   if (VT == MVT::nxv4i32 || VT == MVT::nxv2i64)
     return LowerToPredicatedOp(Op, DAG, PredOpcode);
 
   // SVE doesn't have i8 and i16 DIV operations; widen them to 32-bit
   // operations, and truncate the result.
   EVT WidenedVT;
   if (VT == MVT::nxv16i8)
     WidenedVT = MVT::nxv8i16;

llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll

 }
 
 ;
 ; NOTE: Tests beyond this point only have CHECK lines to validate the first
 ; VBYTES because the add tests already validate the legalisation code paths.
 ;
 
 ;
+; SDIV
+;
+
+; Vector v2i32 sdiv are not legal for NEON so use SVE when available.
+define <2 x i32> @sdiv_v2i32(<2 x i32> %op1, <2 x i32> %op2) #0 {
+; CHECK-LABEL: sdiv_v2i32:
+; CHECK: ptrue [[PG:p[0-9]+]].s, vl[[#min(div(VBYTES,4),2)]]
+; CHECK: sdiv z0.s, [[PG]]/m, z0.s, z1.s
+; CHECK: ret
+  %res = sdiv <2 x i32> %op1, %op2
+  ret <2 x i32> %res
+}
+
+; Vector v4i32 sdiv are not legal for NEON so use SVE when available.
+define <4 x i32> @sdiv_v4i32(<4 x i32> %op1, <4 x i32> %op2) #0 {
+; CHECK-LABEL: sdiv_v4i32:
+; CHECK: ptrue [[PG:p[0-9]+]].s, vl[[#min(div(VBYTES,4),4)]]
+; CHECK: sdiv z0.s, [[PG]]/m, z0.s, z1.s
+; CHECK: ret
+  %res = sdiv <4 x i32> %op1, %op2
+  ret <4 x i32> %res
+}
+
+define void @sdiv_v8i32(<8 x i32>* %a, <8 x i32>* %b) #0 {
+; CHECK-LABEL: sdiv_v8i32:
+; CHECK: ptrue [[PG:p[0-9]+]].s, vl[[#min(div(VBYTES,4),8)]]
+; CHECK-DAG: ld1w { [[OP1:z[0-9]+]].s }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1w { [[OP2:z[0-9]+]].s }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].s, [[PG]]/m, [[OP1]].s, [[OP2]].s
+; CHECK: st1w { [[RES]].s }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <8 x i32>, <8 x i32>* %a
+  %op2 = load <8 x i32>, <8 x i32>* %b
+  %res = sdiv <8 x i32> %op1, %op2
+  store <8 x i32> %res, <8 x i32>* %a
+  ret void
+}
+
+define void @sdiv_v16i32(<16 x i32>* %a, <16 x i32>* %b) #0 {
+; CHECK-LABEL: sdiv_v16i32:
+; CHECK: ptrue [[PG:p[0-9]+]].s, vl[[#min(div(VBYTES,4),16)]]
+; CHECK-DAG: ld1w { [[OP1:z[0-9]+]].s }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1w { [[OP2:z[0-9]+]].s }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].s, [[PG]]/m, [[OP1]].s, [[OP2]].s
+; CHECK: st1w { [[RES]].s }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <16 x i32>, <16 x i32>* %a
+  %op2 = load <16 x i32>, <16 x i32>* %b
+  %res = sdiv <16 x i32> %op1, %op2
+  store <16 x i32> %res, <16 x i32>* %a
+  ret void
+}
+
+define void @sdiv_v32i32(<32 x i32>* %a, <32 x i32>* %b) #0 {
+; CHECK-LABEL: sdiv_v32i32:
+; CHECK: ptrue [[PG:p[0-9]+]].s, vl[[#min(div(VBYTES,4),32)]]
+; CHECK-DAG: ld1w { [[OP1:z[0-9]+]].s }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1w { [[OP2:z[0-9]+]].s }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].s, [[PG]]/m, [[OP1]].s, [[OP2]].s
+; CHECK: st1w { [[RES]].s }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <32 x i32>, <32 x i32>* %a
+  %op2 = load <32 x i32>, <32 x i32>* %b
+  %res = sdiv <32 x i32> %op1, %op2
+  store <32 x i32> %res, <32 x i32>* %a
+  ret void
+}
+
+define void @sdiv_v64i32(<64 x i32>* %a, <64 x i32>* %b) #0 {
+; CHECK-LABEL: sdiv_v64i32:
+; CHECK: ptrue [[PG:p[0-9]+]].s, vl[[#min(div(VBYTES,4),64)]]
+; CHECK-DAG: ld1w { [[OP1:z[0-9]+]].s }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1w { [[OP2:z[0-9]+]].s }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].s, [[PG]]/m, [[OP1]].s, [[OP2]].s
+; CHECK: st1w { [[RES]].s }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <64 x i32>, <64 x i32>* %a
+  %op2 = load <64 x i32>, <64 x i32>* %b
+  %res = sdiv <64 x i32> %op1, %op2
+  store <64 x i32> %res, <64 x i32>* %a
+  ret void
+}
+
+; Vector i64 sdiv are not legal for NEON so use SVE when available.
+define <1 x i64> @sdiv_v1i64(<1 x i64> %op1, <1 x i64> %op2) #0 {
+; CHECK-LABEL: sdiv_v1i64:
+; CHECK: ptrue [[PG:p[0-9]+]].d, vl1
+; CHECK: sdiv z0.d, [[PG]]/m, z0.d, z1.d
+; CHECK: ret
+  %res = sdiv <1 x i64> %op1, %op2
+  ret <1 x i64> %res
+}
+
+; Vector i64 sdiv are not legal for NEON so use SVE when available.
+define <2 x i64> @sdiv_v2i64(<2 x i64> %op1, <2 x i64> %op2) #0 {
+; CHECK-LABEL: sdiv_v2i64:
+; CHECK: ptrue [[PG:p[0-9]+]].d, vl2
+; CHECK: sdiv z0.d, [[PG]]/m, z0.d, z1.d
+; CHECK: ret
+  %res = sdiv <2 x i64> %op1, %op2
+  ret <2 x i64> %res
+}
+
+define void @sdiv_v4i64(<4 x i64>* %a, <4 x i64>* %b) #0 {
+; CHECK-LABEL: sdiv_v4i64:
+; CHECK: ptrue [[PG:p[0-9]+]].d, vl[[#min(div(VBYTES,8),4)]]
+; CHECK-DAG: ld1d { [[OP1:z[0-9]+]].d }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1d { [[OP2:z[0-9]+]].d }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].d, [[PG]]/m, [[OP1]].d, [[OP2]].d
+; CHECK: st1d { [[RES]].d }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <4 x i64>, <4 x i64>* %a
+  %op2 = load <4 x i64>, <4 x i64>* %b
+  %res = sdiv <4 x i64> %op1, %op2
+  store <4 x i64> %res, <4 x i64>* %a
+  ret void
+}
+
+define void @sdiv_v8i64(<8 x i64>* %a, <8 x i64>* %b) #0 {
+; CHECK-LABEL: sdiv_v8i64:
+; CHECK: ptrue [[PG:p[0-9]+]].d, vl[[#min(div(VBYTES,8),8)]]
+; CHECK-DAG: ld1d { [[OP1:z[0-9]+]].d }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1d { [[OP2:z[0-9]+]].d }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].d, [[PG]]/m, [[OP1]].d, [[OP2]].d
+; CHECK: st1d { [[RES]].d }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <8 x i64>, <8 x i64>* %a
+  %op2 = load <8 x i64>, <8 x i64>* %b
+  %res = sdiv <8 x i64> %op1, %op2
+  store <8 x i64> %res, <8 x i64>* %a
+  ret void
+}
+
+define void @sdiv_v16i64(<16 x i64>* %a, <16 x i64>* %b) #0 {
+; CHECK-LABEL: sdiv_v16i64:
+; CHECK: ptrue [[PG:p[0-9]+]].d, vl[[#min(div(VBYTES,8),16)]]
+; CHECK-DAG: ld1d { [[OP1:z[0-9]+]].d }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1d { [[OP2:z[0-9]+]].d }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].d, [[PG]]/m, [[OP1]].d, [[OP2]].d
+; CHECK: st1d { [[RES]].d }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <16 x i64>, <16 x i64>* %a
+  %op2 = load <16 x i64>, <16 x i64>* %b
+  %res = sdiv <16 x i64> %op1, %op2
+  store <16 x i64> %res, <16 x i64>* %a
+  ret void
+}
+
+define void @sdiv_v32i64(<32 x i64>* %a, <32 x i64>* %b) #0 {
+; CHECK-LABEL: sdiv_v32i64:
+; CHECK: ptrue [[PG:p[0-9]+]].d, vl[[#min(div(VBYTES,8),32)]]
+; CHECK-DAG: ld1d { [[OP1:z[0-9]+]].d }, [[PG]]/z, [x0]
+; CHECK-DAG: ld1d { [[OP2:z[0-9]+]].d }, [[PG]]/z, [x1]
+; CHECK: sdiv [[RES:z[0-9]+]].d, [[PG]]/m, [[OP1]].d, [[OP2]].d
+; CHECK: st1d { [[RES]].d }, [[PG]], [x0]
+; CHECK: ret
+  %op1 = load <32 x i64>, <32 x i64>* %a
+  %op2 = load <32 x i64>, <32 x i64>* %b
+  %res = sdiv <32 x i64> %op1, %op2
+  store <32 x i64> %res, <32 x i64>* %a
+  ret void
+}
+
+;
 ; MUL
 ;
 
 ; Don't use SVE for 64-bit vectors.
 define <8 x i8> @mul_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 ; CHECK-LABEL: mul_v8i8:
 ; CHECK: mul v0.8b, v0.8b, v1.8b
 ; CHECK: ret