[PATCH] D41819: [AArch64][SVE] Asm: PTRUE and PTRUES instructions

[Sander de Smalen via Phabricator via llvm-commits]

sdesmalen added inline comments.


================
Comment at: test/MC/AArch64/SVE/ptrue-diagnostics.s:8
+ptrue p0.s, vl512
+// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operand
+// CHECK-NEXT: ptrue p0.s, vl512
----------------
samparker wrote:
> Can we not give the user a more specific error message?
Thanks for the suggestion, I've created a new diagnostic for this.


================
Comment at: test/MC/AArch64/SVE/ptrue.s:175
+// ---------------------------------------------------------------------------//
+
+ptrue   p7.s, #23
----------------
samparker wrote:
> Maybe these should test the corner cases? If I understand the document, 14 to 28 should be legal?
Agreed, I've added these now.


================
Comment at: test/MC/AArch64/SVE/ptrues.s:6
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve < %s \
+// RUN:        | llvm-objdump -d -mattr=+sve - | FileCheck %s --check-prefix=CHECK-INST
+// RUN: llvm-mc -triple=aarch64 -filetype=obj -mattr=+sve < %s \
----------------
samparker wrote:
> I can see a few examples of using llvm-objdump for disassembly, but why not just use llvm-mc?
I've sort of copied this approach from other MC tests and the effect should be the same. Also, I wasn't able to pipe the output from llvm-mc back into llvm-mc, similar to how this is done for llvm-objdump.


https://reviews.llvm.org/D41819