[PATCH] D24661: More processors support under AARch64 state for auto detection.

[Renato Golin via llvm-commits]

rengolin added a comment.

As adding XGene1 goes, the change is reasonable, but there's an auto-detection left-over.

A couple of LLVM-specifics...

- Clang and LLVM are in different repositories, so we'll need two different reviews, one for each part.
- The Clang tests for xgene are good, but we'll also need TargetParser tests (see `./unittests/Support/TargetParserTest.cpp`)

A few other comments inline.

Thanks!
--renato


================
Comment at: include/llvm/Support/ARMTargetParser.def:237
@@ -236,1 +236,3 @@
+ARM_CPU_NAME("xgene1", AK_ARMV8A, FK_NEON_FP_ARMV8, false, 
+	ARM::AEK_FP | ARM::AEK_SIMD)
 // Non-standard Arch names.
----------------
@jgreenhalgh, is this the same defaults as GCC?

================
Comment at: include/llvm/Support/TargetParser.h:165
@@ -164,3 +164,3 @@
   AEK_PROFILE = 0x40,
-  AEK_RAS = 0x80
+  AEK_RAS = 0x80,
 };
----------------
this change now is unnecessary

================
Comment at: lib/Support/Host.cpp:1098
@@ -1097,1 +1097,3 @@
 }
+#elif defined(__linux__) && defined(__aarch64__)
+StringRef sys::getHostCPUName() {
----------------
As discussed over the mailing list, please discard this change.


https://reviews.llvm.org/D24661