[llvm] [RISCV] coalesce between VSETVLI and SF_VSETTNT (PR #203438)

[Craig Topper via llvm-commits]

================
@@ -933,6 +936,98 @@ void RISCVInsertVSETVLI::coalesceVSETVLIs(MachineBasicBlock &MBB) const {
   }
 }
 
+// When twiden != 0, LMUL, tail policy, and mask policy from the user are
+// ignored. The tail policy and mask policy are always treated as agnostic. The
+// normal RVV instruction will ignore the twiden parameter. This observation
+// could allow the RVV instruction and xsfmm instruction to share the same
+// configuration instruction.
+//
+// We need to make sure the AVL, SEW, and AltFmt is same between VSETVL and
+// VSETVLTN.
+//
+// For example:
+//
+// %avl = SETTM or SETTK
+// ...
+// VSETVL %avl, type1
+// VSETVLTNT %avl, type2
+//
+// ->
+//
+// %avl = SETTM or SETTK
+// ...
+// VSETVLTNT %avl, type2
+//
+bool RISCVInsertVSETVLI::canMutatePriorConfigWithTWiden(
+    const MachineInstr &PrevMI, const MachineInstr &MI) const {
+
+  if (PrevMI.getOpcode() != RISCV::PseudoVSETVLI)
+    return false;
+
+  if (MI.getOpcode() != RISCV::PseudoSF_VSETTNT)
+    return false;
+
+  auto PrevInfo = VIA.getInfoForVSETVLI(PrevMI);
+  auto CurrInfo = VIA.getInfoForVSETVLI(MI);
+
+  auto AVLReg = CurrInfo.getAVLReg();
+
+  auto *AVLRegDefMI = MRI->getUniqueVRegDef(AVLReg);
+
+  if (!AVLRegDefMI)
+    return false;
+
+  if (!RISCVInstrInfo::isXSfmmVectorConfigTMTKInstr(*AVLRegDefMI))
+    return false;
+
+  auto AVLRegDefMIInfo = VIA.computeInfoForInstr(*AVLRegDefMI);
+  if (AVLRegDefMIInfo.getTWiden() != CurrInfo.getTWiden())
+    return false;
+
+  if (AVLRegDefMIInfo.getSEW() != PrevInfo.getSEW())
+    return false;
+
+  // CurrInfo twiden != 0, so TailAgnostic and MaskAgnostic bit default to 1
+  if (!PrevInfo.getTailAgnostic() || !PrevInfo.getMaskAgnostic())
+    return false;
+
+  if (!PrevInfo.hasSameAVL(CurrInfo))
+    return false;
+
+  if (PrevInfo.getSEW() != CurrInfo.getSEW())
+    return false;
+
+  if (PrevInfo.getAltFmt() != CurrInfo.getAltFmt())
+    return false;
+
----------------
topperc wrote:

>  I think the LMUL of PrevMI can't be larger than 8/KMAX as well because the vsetvli may have larger VLMAX that can't be clipped when removing vsetvli?

We know the AVL came from a XSfmmVectorConfigTMTKInstr which has LMUL = min(8/KMAX, 8/TWIDEN, ceil(ETE/EVE)). If the LMUL of the vsetvli is at least 8/KMAX or 8/TWIDEN it shouldn't need to be clipped any further because it was already clipped by the  XSfmmVectorConfigTMTKInstr instruction.

https://github.com/llvm/llvm-project/pull/203438