[WebAssembly] Expand SIMD shifts while V8's implementation disagrees

Summary:
V8 currently implements SIMD shifts as taking an immediate operation,
which disagrees with the spec proposal and the toolchain
implementation. As a stopgap measure to get things working, unroll all
vector shifts. Since this is a temporary measure, there are no tests.

Reviewers: aheejin, dschuff

Subscribers: sbc100, jgravelle-google, sunfish, dmgreen, llvm-commits

Differential Revision: https://reviews.llvm.org/D56520

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@351151 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp b/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
index d0c8aacac7cde816318260b1a47958ab797216bd..003848e3422796850c7ea19a43fa9b7a3a036779 100644 (file)
--- a/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
+++ b/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
@@ -1155,6 +1155,31 @@ WebAssemblyTargetLowering::LowerAccessVectorElement(SDValue Op,
     return SDValue();
 }
 
+static SDValue UnrollVectorShift(SDValue Op, SelectionDAG &DAG) {
+  EVT LaneT = Op.getSimpleValueType().getVectorElementType();
+  // 32-bit and 64-bit unrolled shifts will have proper semantics
+  if (LaneT.bitsGE(MVT::i32))
+    return DAG.UnrollVectorOp(Op.getNode());
+  // Otherwise mask the shift value to get proper semantics from 32-bit shift
+  SDLoc DL(Op);
+  SDValue ShiftVal = Op.getOperand(1);
+  uint64_t MaskVal = LaneT.getSizeInBits() - 1;
+  SDValue MaskedShiftVal = DAG.getNode(
+      ISD::AND,                    // mask opcode
+      DL, ShiftVal.getValueType(), // masked value type
+      ShiftVal,                    // original shift value operand
+      DAG.getConstant(MaskVal, DL, ShiftVal.getValueType()) // mask operand
+  );
+
+  return DAG.UnrollVectorOp(
+      DAG.getNode(Op.getOpcode(),        // original shift opcode
+                  DL, Op.getValueType(), // original return type
+                  Op.getOperand(0),      // original vector operand,
+                  MaskedShiftVal         // new masked shift value operand
+                  )
+          .getNode());
+}
+
 SDValue WebAssemblyTargetLowering::LowerShift(SDValue Op,
                                               SelectionDAG &DAG) const {
   SDLoc DL(Op);
@@ -1162,12 +1187,17 @@ SDValue WebAssemblyTargetLowering::LowerShift(SDValue Op,
   // Only manually lower vector shifts
   assert(Op.getSimpleValueType().isVector());
 
+  // Expand all vector shifts until V8 fixes its implementation
+  // TODO: remove this once V8 is fixed
+  if (!Subtarget->hasUnimplementedSIMD128())
+    return UnrollVectorShift(Op, DAG);
+
   // Unroll non-splat vector shifts
   BuildVectorSDNode *ShiftVec;
   SDValue SplatVal;
   if (!(ShiftVec = dyn_cast<BuildVectorSDNode>(Op.getOperand(1).getNode())) ||
       !(SplatVal = ShiftVec->getSplatValue()))
-    return DAG.UnrollVectorOp(Op.getNode());
+    return UnrollVectorShift(Op, DAG);
 
   // All splats except i64x2 const splats are handled by patterns
   ConstantSDNode *SplatConst = dyn_cast<ConstantSDNode>(SplatVal);

diff --git a/test/CodeGen/WebAssembly/simd-arith.ll b/test/CodeGen/WebAssembly/simd-arith.ll
index f0355febd51d78e79bf7d119ce9ad8d8e6630936..8d7f0205415b466c28d7cec0656bf4d8fcca4c67 100644 (file)
--- a/test/CodeGen/WebAssembly/simd-arith.ll
+++ b/test/CodeGen/WebAssembly/simd-arith.ll
@@ -90,7 +90,11 @@ define <16 x i8> @shl_const_v16i8(<16 x i8> %v) {
 ; NO-SIMD128-NOT: i8x16
 ; SIMD128-NEXT: .functype shl_vec_v16i8 (v128, v128) -> (v128){{$}}
 ; SIMD128-NEXT: i8x16.extract_lane_s $push[[L0:[0-9]+]]=, $0, 0{{$}}
-; SIMD128-NEXT: i8x16.extract_lane_u $push[[L1:[0-9]+]]=, $1, 0{{$}}
+; SIMD128-NEXT: i32.const $push[[M0:[0-9]+]]=, 7{{$}}
+; SIMD128-NEXT: i8x16.splat $push[[M1:[0-9]+]]=, $pop[[M0]]{{$}}
+; SIMD128-NEXT: v128.and $push[[M2:[0-9]+]]=, $1, $pop[[M1]]{{$}}
+; SIMD128-NEXT: local.tee $push[[M:[0-9]+]]=, $1=, $pop[[M2]]{{$}}
+; SIMD128-NEXT: i8x16.extract_lane_u $push[[L1:[0-9]+]]=, $pop[[M]], 0{{$}}
 ; SIMD128-NEXT: i32.shl $push[[L2:[0-9]+]]=, $pop[[L0]], $pop[[L1]]{{$}}
 ; SIMD128-NEXT: i8x16.splat $push[[L3:[0-9]+]]=, $pop[[L2]]{{$}}
 ; Skip 14 lanes
@@ -122,7 +126,11 @@ define <16 x i8> @shr_s_v16i8(<16 x i8> %v, i8 %x) {
 ; NO-SIMD128-NOT: i8x16
 ; SIMD128-NEXT: .functype shr_s_vec_v16i8 (v128, v128) -> (v128){{$}}
 ; SIMD128-NEXT: i8x16.extract_lane_s $push[[L0:[0-9]+]]=, $0, 0{{$}}
-; SIMD128-NEXT: i8x16.extract_lane_u $push[[L1:[0-9]+]]=, $1, 0{{$}}
+; SIMD128-NEXT: i32.const $push[[M0:[0-9]+]]=, 7{{$}}
+; SIMD128-NEXT: i8x16.splat $push[[M1:[0-9]+]]=, $pop[[M0]]{{$}}
+; SIMD128-NEXT: v128.and $push[[M2:[0-9]+]]=, $1, $pop[[M1]]{{$}}
+; SIMD128-NEXT: local.tee $push[[M:[0-9]+]]=, $1=, $pop[[M2]]{{$}}
+; SIMD128-NEXT: i8x16.extract_lane_u $push[[L1:[0-9]+]]=, $pop[[M]], 0{{$}}
 ; SIMD128-NEXT: i32.shr_s $push[[L2:[0-9]+]]=, $pop[[L0]], $pop[[L1]]{{$}}
 ; SIMD128-NEXT: i8x16.splat $push[[L3:[0-9]+]]=, $pop[[L2]]{{$}}
 ; Skip 14 lanes
@@ -154,7 +162,11 @@ define <16 x i8> @shr_u_v16i8(<16 x i8> %v, i8 %x) {
 ; NO-SIMD128-NOT: i8x16
 ; SIMD128-NEXT: .functype shr_u_vec_v16i8 (v128, v128) -> (v128){{$}}
 ; SIMD128-NEXT: i8x16.extract_lane_u $push[[L0:[0-9]+]]=, $0, 0{{$}}
-; SIMD128-NEXT: i8x16.extract_lane_u $push[[L1:[0-9]+]]=, $1, 0{{$}}
+; SIMD128-NEXT: i32.const $push[[M0:[0-9]+]]=, 7{{$}}
+; SIMD128-NEXT: i8x16.splat $push[[M1:[0-9]+]]=, $pop[[M0]]{{$}}
+; SIMD128-NEXT: v128.and $push[[M2:[0-9]+]]=, $1, $pop[[M1]]{{$}}
+; SIMD128-NEXT: local.tee $push[[M:[0-9]+]]=, $1=, $pop[[M2]]{{$}}
+; SIMD128-NEXT: i8x16.extract_lane_u $push[[L1:[0-9]+]]=, $pop[[M]], 0{{$}}
 ; SIMD128-NEXT: i32.shr_u $push[[L2:[0-9]+]]=, $pop[[L0]], $pop[[L1]]{{$}}
 ; SIMD128-NEXT: i8x16.splat $push[[L3:[0-9]+]]=, $pop[[L2]]{{$}}
 ; Skip 14 lanes
@@ -304,7 +316,11 @@ define <8 x i16> @shl_const_v8i16(<8 x i16> %v) {
 ; NO-SIMD128-NOT: i16x8
 ; SIMD128-NEXT: .functype shl_vec_v8i16 (v128, v128) -> (v128){{$}}
 ; SIMD128-NEXT: i16x8.extract_lane_s $push[[L0:[0-9]+]]=, $0, 0{{$}}
-; SIMD128-NEXT: i16x8.extract_lane_u $push[[L1:[0-9]+]]=, $1, 0{{$}}
+; SIMD128-NEXT: i32.const $push[[M0:[0-9]+]]=, 15{{$}}
+; SIMD128-NEXT: i16x8.splat $push[[M1:[0-9]+]]=, $pop[[M0]]{{$}}
+; SIMD128-NEXT: v128.and $push[[M2:[0-9]+]]=, $1, $pop[[M1]]{{$}}
+; SIMD128-NEXT: local.tee $push[[M:[0-9]+]]=, $1=, $pop[[M2]]{{$}}
+; SIMD128-NEXT: i16x8.extract_lane_u $push[[L1:[0-9]+]]=, $pop[[M]], 0{{$}}
 ; SIMD128-NEXT: i32.shl $push[[L2:[0-9]+]]=, $pop[[L0]], $pop[[L1]]{{$}}
 ; SIMD128-NEXT: i16x8.splat $push[[L3:[0-9]+]]=, $pop[[L2]]{{$}}
 ; Skip 6 lanes
@@ -335,7 +351,11 @@ define <8 x i16> @shr_s_v8i16(<8 x i16> %v, i16 %x) {
 ; NO-SIMD128-NOT: i16x8
 ; SIMD128-NEXT: .functype shr_s_vec_v8i16 (v128, v128) -> (v128){{$}}
 ; SIMD128-NEXT: i16x8.extract_lane_s $push[[L0:[0-9]+]]=, $0, 0{{$}}
-; SIMD128-NEXT: i16x8.extract_lane_u $push[[L1:[0-9]+]]=, $1, 0{{$}}
+; SIMD128-NEXT: i32.const $push[[M0:[0-9]+]]=, 15{{$}}
+; SIMD128-NEXT: i16x8.splat $push[[M1:[0-9]+]]=, $pop[[M0]]{{$}}
+; SIMD128-NEXT: v128.and $push[[M2:[0-9]+]]=, $1, $pop[[M1]]{{$}}
+; SIMD128-NEXT: local.tee $push[[M:[0-9]+]]=, $1=, $pop[[M2]]{{$}}
+; SIMD128-NEXT: i16x8.extract_lane_u $push[[L1:[0-9]+]]=, $pop[[M]], 0{{$}}
 ; SIMD128-NEXT: i32.shr_s $push[[L2:[0-9]+]]=, $pop[[L0]], $pop[[L1]]{{$}}
 ; SIMD128-NEXT: i16x8.splat $push[[L3:[0-9]+]]=, $pop[[L2]]{{$}}
 ; Skip 6 lanes
@@ -366,7 +386,11 @@ define <8 x i16> @shr_u_v8i16(<8 x i16> %v, i16 %x) {
 ; NO-SIMD128-NOT: i16x8
 ; SIMD128-NEXT: .functype shr_u_vec_v8i16 (v128, v128) -> (v128){{$}}
 ; SIMD128-NEXT: i16x8.extract_lane_u $push[[L0:[0-9]+]]=, $0, 0{{$}}
-; SIMD128-NEXT: i16x8.extract_lane_u $push[[L1:[0-9]+]]=, $1, 0{{$}}
+; SIMD128-NEXT: i32.const $push[[M0:[0-9]+]]=, 15{{$}}
+; SIMD128-NEXT: i16x8.splat $push[[M1:[0-9]+]]=, $pop[[M0]]{{$}}
+; SIMD128-NEXT: v128.and $push[[M2:[0-9]+]]=, $1, $pop[[M1]]{{$}}
+; SIMD128-NEXT: local.tee $push[[M:[0-9]+]]=, $1=, $pop[[M2]]{{$}}
+; SIMD128-NEXT: i16x8.extract_lane_u $push[[L1:[0-9]+]]=, $pop[[M]], 0{{$}}
 ; SIMD128-NEXT: i32.shr_u $push[[L2:[0-9]+]]=, $pop[[L0]], $pop[[L1]]{{$}}
 ; SIMD128-NEXT: i16x8.splat $push[[L3:[0-9]+]]=, $pop[[L2]]{{$}}
 ; Skip 6 lanes