[llvm] [RISCV][AsmParser] Allow parsing vset{i}vli omitting LMUL parameter (PR #115277)

[Pengcheng Wang via llvm-commits]

================
@@ -2320,7 +2326,7 @@ bool RISCVAsmParser::generateVTypeError(SMLoc ErrorLoc) {
   return Error(
       ErrorLoc,
       "operand must be "
-      "e[8|16|32|64],m[1|2|4|8|f2|f4|f8],[ta|tu],[ma|mu]");
+      "e{8|16|32|64},[m{1|2|4|8|f2|f4|f8},]{ta|tu},{ma|mu}");
----------------
wangpc-pp wrote:

Actually GCC(binutils) supports optional mask/tail policies: https://godbolt.org/z/GPEG8xzrf
It's a known issue for LLVM if I remember correctly, we may have some discussion before but I can't find the link.
Binutils supports such thing just for compatibility with old version (0.7.1, pity) of RVV.

https://github.com/llvm/llvm-project/pull/115277