[PATCH] D138508: [TargetLowering] Teach DemandedBits about VSCALE

[Benjamin Maxwell via Phabricator via llvm-commits]

benmxwl-arm created this revision.
benmxwl-arm added reviewers: c-rhodes, peterwaller-arm, dtemirbulatov, MattDevereau.
Herald added subscribers: ctetreau, hiraditya, kristof.beyls.
Herald added a project: All.
benmxwl-arm requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

This allows DemandedBits to see the result of VSCALE will be at most
VScaleMax * some compile-time constant. This relies on the vscale_range()
attribute being present on the function, with a max set. (This is done by
default when clang is targeting AArch64+SVE).

Using this various redundant operations (zexts, sexts, ands, ors, etc)
can be eliminated.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D138508

Files:
  llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
  llvm/test/CodeGen/AArch64/vscale-and-sve-cnt-demandedbits.ll


Index: llvm/test/CodeGen/AArch64/vscale-and-sve-cnt-demandedbits.ll
===================================================================
--- llvm/test/CodeGen/AArch64/vscale-and-sve-cnt-demandedbits.ll
+++ llvm/test/CodeGen/AArch64/vscale-and-sve-cnt-demandedbits.ll
@@ -14,9 +14,8 @@
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    rdvl x8, #1
 ; CHECK-NEXT:    lsr x8, x8, #4
-; CHECK-NEXT:    and w9, w8, #0x1f
-; CHECK-NEXT:    and w8, w8, #0xfffffffc
-; CHECK-NEXT:    add w0, w9, w8
+; CHECK-NEXT:    and w9, w8, #0x1c
+; CHECK-NEXT:    add w0, w8, w9
 ; CHECK-NEXT:    ret
   %vscale = call i32 @llvm.vscale.i32()
   %and_redundant = and i32 %vscale, 31
@@ -89,8 +88,7 @@
 ; CHECK-LABEL: vscale_trunc_zext:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    rdvl x8, #1
-; CHECK-NEXT:    lsr x8, x8, #4
-; CHECK-NEXT:    and x0, x8, #0xffffffff
+; CHECK-NEXT:    lsr x0, x8, #4
 ; CHECK-NEXT:    ret
   %vscale = call i32 @llvm.vscale.i32()
   %zext = zext i32 %vscale to i64
@@ -101,8 +99,7 @@
 ; CHECK-LABEL: vscale_trunc_sext:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    rdvl x8, #1
-; CHECK-NEXT:    lsr x8, x8, #4
-; CHECK-NEXT:    sxtw x0, w8
+; CHECK-NEXT:    lsr x0, x8, #4
 ; CHECK-NEXT:    ret
   %vscale = call i32 @llvm.vscale.i32()
   %sext = sext i32 %vscale to i64
@@ -212,9 +209,8 @@
 ; CHECK-NEXT:    mov w9, #5
 ; CHECK-NEXT:    lsr x8, x8, #4
 ; CHECK-NEXT:    mul x8, x8, x9
-; CHECK-NEXT:    and w9, w8, #0x7f
-; CHECK-NEXT:    and w8, w8, #0x3f
-; CHECK-NEXT:    add w0, w9, w8
+; CHECK-NEXT:    and w9, w8, #0x3f
+; CHECK-NEXT:    add w0, w8, w9
 ; CHECK-NEXT:    ret
   %vscale = call i32 @llvm.vscale.i32()
   %mul = mul i32 %vscale, 5
@@ -231,9 +227,8 @@
 ; CHECK-NEXT:    mov x9, #-5
 ; CHECK-NEXT:    lsr x8, x8, #4
 ; CHECK-NEXT:    mul x8, x8, x9
-; CHECK-NEXT:    orr w9, w8, #0xffffff80
-; CHECK-NEXT:    and w8, w8, #0xffffffc0
-; CHECK-NEXT:    add w0, w9, w8
+; CHECK-NEXT:    and w9, w8, #0xffffffc0
+; CHECK-NEXT:    add w0, w8, w9
 ; CHECK-NEXT:    ret
   %vscale = call i32 @llvm.vscale.i32()
   %mul = mul i32 %vscale, -5
Index: llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
===================================================================
--- llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -1129,6 +1129,28 @@
 
   KnownBits Known2;
   switch (Op.getOpcode()) {
+  case ISD::VSCALE: {
+    Function const &F = TLO.DAG.getMachineFunction().getFunction();
+    Attribute const &Attr = F.getFnAttribute(Attribute::VScaleRange);
+    if (!Attr.isValid())
+      return false;
+    Optional<unsigned> MaxVScale = Attr.getVScaleRangeMax();
+    if (!MaxVScale.has_value())
+      return false;
+    int64_t VScaleResultUpperbound = *MaxVScale;
+    if (auto *MulImm = dyn_cast<ConstantSDNode>(Op.getOperand(0))) {
+      VScaleResultUpperbound *= MulImm->getSExtValue();
+    } else {
+      return false;
+    }
+    bool Negative = VScaleResultUpperbound < 0;
+    if (Negative)
+      VScaleResultUpperbound = -VScaleResultUpperbound;
+    unsigned RequiredBits = Log2_64(VScaleResultUpperbound) + 1;
+    if (RequiredBits < BitWidth)
+      (Negative ? Known.One : Known.Zero).setHighBits(BitWidth - RequiredBits);
+    return false;
+  }
   case ISD::SCALAR_TO_VECTOR: {
     if (!DemandedElts[0])
       return TLO.CombineTo(Op, TLO.DAG.getUNDEF(VT));


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D138508.477230.patch
Type: text/x-patch
Size: 3352 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221122/1f4e717d/attachment.bin>