Recommit "[RISCV] Strengthen register usage validation for XTheadMemPair loads (#136241)"

With test fix.

Closes #136087

https://github.com/XUANTIE-RV/thead-extension-spec/blob/master/xtheadmempair/lwd.adoc

Author: el-ev

llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp

@@ -3656,9 +3656,9 @@ bool RISCVAsmParser::validateInstruction(MCInst &Inst,
     MCRegister Rd2 = Inst.getOperand(1).getReg();
     MCRegister Rs1 = Inst.getOperand(2).getReg();
     // The encoding with rd1 == rd2 == rs1 is reserved for XTHead load pair.
-    if (Rs1 == Rd1 && Rs1 == Rd2) {
+    if (Rs1 == Rd1 || Rs1 == Rd2 || Rd1 == Rd2) {
       SMLoc Loc = Operands[1]->getStartLoc();
-      return Error(Loc, "rs1, rd1, and rd2 cannot all be the same");
+      return Error(Loc, "rs1, rd1, and rd2 cannot overlap");
     }
   }
 

llvm/test/MC/RISCV/rv32xtheadmempair-invalid.s

@@ -15,6 +15,7 @@ th.lwd a3, a4, (a5), 3, 5   # CHECK: [[@LINE]]:25: error: operand must be consta
 th.swd t3, t4, (t5), 5, 4   # CHECK: [[@LINE]]:22: error: immediate must be an integer in the range [0, 3]
 th.swd t3, t4, (t5)         # CHECK: [[@LINE]]:1: error: too few operands for instruction
 th.swd t3, t4, (t5), 3, 5   # CHECK: [[@LINE]]:25: error: operand must be constant 3
-th.lwud x6, x6, (x6), 2, 3  # CHECK: [[@LINE]]:9: error: rs1, rd1, and rd2 cannot all be the same
+th.lwd x6, x7, (x7), 2, 3   # CHECK: [[@LINE]]:8: error: rs1, rd1, and rd2 cannot overlap
+th.lwud x6, x6, (x6), 2, 3  # CHECK: [[@LINE]]:9: error: rs1, rd1, and rd2 cannot overlap
 th.ldd t0, t1, (t2), 2, 4   # CHECK: [[@LINE]]:1: error: instruction requires the following: RV64I Base Instruction Set{{$}}
 th.sdd t0, t1, (t2), 2, 4   # CHECK: [[@LINE]]:1: error: instruction requires the following: RV64I Base Instruction Set{{$}}

llvm/test/MC/RISCV/rv64xtheadmempair-invalid.s

@@ -15,4 +15,5 @@ th.lwd a3, a4, (a5), 3, 5   # CHECK: [[@LINE]]:25: error: operand must be consta
 th.swd t3, t4, (t5), 5, 4   # CHECK: [[@LINE]]:22: error: immediate must be an integer in the range [0, 3]
 th.swd t3, t4, (t5)         # CHECK: [[@LINE]]:1: error: too few operands for instruction
 th.swd t3, t4, (t5), 3, 5   # CHECK: [[@LINE]]:25: error: operand must be constant 3
-th.lwud x6, x6, (x6), 2, 3  # CHECK: [[@LINE]]:9: error: rs1, rd1, and rd2 cannot all be the same
+th.ldd x6, x6, (x7), 2, 4   # CHECK: [[@LINE]]:8: error: rs1, rd1, and rd2 cannot overlap
+th.lwud x6, x6, (x6), 2, 3  # CHECK: [[@LINE]]:9: error: rs1, rd1, and rd2 cannot overlap