[clang] [llvm] [RISCV][MC] Add experimental `Zvvmm` MC support (PR #193956)
via cfe-commits
cfe-commits at lists.llvm.org
Mon Apr 27 03:45:50 PDT 2026
================
@@ -0,0 +1,11 @@
+# RUN: llvm-mc -triple=riscv64 -show-encoding --mattr=+experimental-zvvmm %s \
+# RUN: | FileCheck %s --check-prefixes=CHECK-ENCODING,CHECK-INST
+
+# SEW=32, LMUL=1, ta, ma, lambda=4, altfmt_A=1, altfmt_B=1.
+li t1, 0x3c000000000000d0
+vsetvl t0, a0, t1
----------------
imkiva wrote:
You're right that these two MC tests don't really add coverage, since `li` + `vsetvl` is not IME-specific. I'll remove them and keep the bit-layout coverage in the TargetParser unit test instead.
> do we have a special syntax for vset(i)vli to specify the lambda field?
No, we don't have a dedicated `vset(i)vli` syntax for the IME fields. The new IME vtype fields live in the upeer bits of vtype (like `lambda[2:0]` at `vtype[XLEN-2:XLEN-4]`), while the existing vtype-immediate only encodes `vtype[10:0]` or `vtype[9:0]`, so the intended way here is to materialize the full vtype value and use `vsetvl`.
https://github.com/llvm/llvm-project/pull/193956
More information about the cfe-commits
mailing list