[PATCH] D41168: [X86][avx512] Lowering X86 avx512 sqrt intrinsics to IR

[Craig Topper via Phabricator via cfe-commits]

craig.topper added inline comments.


================
Comment at: lib/CodeGen/CGBuiltin.cpp:8145
+  case X86::BI__builtin_ia32_sqrtpd512_mask_nr:
+  case X86::BI__builtin_ia32_sqrtps512_mask_nr: {
+    Function *F = CGM.getIntrinsic(Intrinsic::sqrt, Ops[0]->getType());
----------------
I would suggest just using __builtin_i32_sqrtpd512_mask here and select different intrinsics based on the value of the rounding mode operand. The user should get the same behavior if they use the explicit rounding mode intrinsic, but pass CUR_DIRECTION.


https://reviews.llvm.org/D41168