[PATCH] D19713: [mips][microMIPS] Implement CFC*, CTC* and LDC* instructions

Hrvoje Varga via llvm-commits llvm-commits at lists.llvm.org
Tue May 10 23:21:36 PDT 2016


hvarga added inline comments.

================
Comment at: test/CodeGen/Mips/cconv/callee-saved-float.ll:20
@@ -19,2 +19,3 @@
 ; RUN: llc -march=mips64el -target-abi n64 < %s | FileCheck --check-prefix=ALL --check-prefix=ALL-INV --check-prefix=N64-INV %s
+; RUN: llc -march=mips -mcpu=mips32r6 -mattr=micromips -filetype=obj < %s -o - | llvm-objdump -no-show-raw-insn -arch mips -mcpu=mips32r6 -mattr=micromips -d - | FileCheck --check-prefix=MM32R6 %s
 
----------------
sdardis wrote:
> 
> Add
> 
>   ; CHECK-NOT: LDC164
> 
> along with -asm-show-inst
I can not check that `LDC164` is not occurring in this test file because it will occur. :) So, I can not use  `-asm-show-inst` to check for correct mapping.

But this is not a problem since `LDC164` is indeed correctly mapped to `LDC1_64_MMR6`. This is tested by generating the object file with `llc` and afterwards dumping it using `llvm-objdump`. This output is then checked for the occurrence of `ldc1`. From my perspective, since the object file is compiled and decompiled as microMIPS32r6 (containing `ldc1` in its output), I think that this also proves that the correct implementation of `ldc1` is selected.

Do you agree, can I commit this patch into mainline?


http://reviews.llvm.org/D19713





More information about the llvm-commits mailing list