[PATCH] D38714: [AVX512] Don't mark EXTLOAD as legal with AVX512. Continue using custom lowering.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: test/CodeGen/X86/vec_int_to_fp.ll:2648
 ; AVX512VL:       # BB#0:
-; AVX512VL-NEXT:    vpmovzxdq {{.*#+}} xmm0 = mem[0],zero,mem[1],zero
-; AVX512VL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512VL-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; AVX512VL-NEXT:    vcvtdq2pd %xmm0, %xmm0
----------------
Missing fold?


================
Comment at: test/CodeGen/X86/vec_int_to_fp.ll:2660
+; AVX512VLDQ-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; AVX512VLDQ-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX512VLDQ-NEXT:    retq
----------------
Missing fold?


================
Comment at: test/CodeGen/X86/vec_int_to_fp.ll:2984
+; AVX512VL-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; AVX512VL-NEXT:    vcvtudq2pd %xmm0, %xmm0
 ; AVX512VL-NEXT:    retq
----------------
Missing fold?


================
Comment at: test/CodeGen/X86/vec_int_to_fp.ll:2998
+; AVX512VLDQ-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; AVX512VLDQ-NEXT:    vcvtudq2pd %xmm0, %xmm0
 ; AVX512VLDQ-NEXT:    retq
----------------
Missing fold?


https://reviews.llvm.org/D38714