[PATCH] D32685: [X86] Adding tests for the memory folding tables generated by the TableGen backend.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: test/CodeGen/X86/memoryFolding-avx512Other.mir:4
+--- |
+  attributes #0 = { "target-features" = "+avx512f,+avx512bw,+avx512dq,+avx512vbmi,+avx512er,+avx512cd,+avx512pf,+avx512vl,+avx512ifma" }
+  define void @stack_fold_vextractf32x8zmr_vextractf32x8zrr_test() #0 { ret void }
----------------
Split these more by target type, don't just dump them all them all together.


================
Comment at: test/CodeGen/X86/memoryFolding-other.mir:4
+--- |
+  attributes #0 = { "target-features" = "+3dnow,+adx,+bmi,+bmi2,+cmov,+f16c,+fma,+fma4,+lzcnt,+mmx,+mpx,+popcnt,+sha,+tbm,+xop" }
+  define void @stack_fold_bndmovmr32mr_bndmovmrrr_test() #0 { ret void }
----------------
Probably best to split these at least into targets with unusual/larger numbers of intrinsics: 3dnow, adx, bmi, bmi2, f16c, fma, fma4, mmx, mpx, sha, tbm, xop 


================
Comment at: test/CodeGen/X86/memoryFolding-x86.mir:5
+  attributes #0 = { "target-features" = "+avx512f,+avx512bw,+avx512dq,+avx512vbmi" }
+  attributes #1 = { optsize "target-features" = "+avx512f,+avx512bw,+avx512dq,+avx512vbmi" }
+  define void @stack_fold_adc16mi_adc16ri_test() #0 { ret void }
----------------
Why are you testing with avx512+ enabled?


https://reviews.llvm.org/D32685