android: enable double-word CAS on x86_64

[compnerd]

Summary

The android target assumes that for the x86_64 target, the CPU supports SSE4.2 and popcnt. This implies that the CPU is Nehalem or newer. This should be sufficiently new to provide the double word compare and exchange instruction. This allows us to directly lower __sync_val_compare_and_swap_16 to a cmpxchg16b. It appears that the libatomic in android's NDK does not provide the implementation for lowering calls to the library function.

[lebedev.ri]

Is there some test for this area of code?

[srhines]

Craig, can you confirm that this is acceptable? I don't think there are any chips with SSE4.2 but without cx16, so this just seemed like an oversight. It might be a good idea to really audit the list of possible CPU features for other missing inclusions. Another idea would be to set a baseline minimum CPU for Android, which would cut down on having to specify so many features separately.

[compnerd]

@lebedev.ri - sure, I will add a driver test to ensure that the feature is set on the command line when invoked from the driver, however, I don't think that there is really much in terms of testing that you can do for this type of stuff other than throw a large corpus at it.

[compnerd]

add additional context and test case

[craig.topper]

test/Driver/android-default-x86_64.c

1 ↗

(On Diff #206508)

should we just update the android 64 case in clang-translation.c instead?

[craig.topper]

Another odd feature that's missing from early x86-64 CPUs is sahf.

[compnerd]

Move test case around

[compnerd]

@craig.topper, hmm, what happens in terms of CG when LAHF/SAHF are not available? I assume its just worse CG as you could spill AH onto the stack and do a load/store. This actually results in library calls which may not be possible to fulfill.

[craig.topper]

Yeah LAHF/SAHF just improves codegen. Lack of it won't result in any library calls.

This patch LGTM

[compnerd]

SVN r364352

lib/Driver/ToolChains/Arch/X86.cpp

   }
 
   const llvm::Triple::ArchType ArchType = Triple.getArch();
   // Add features to be compatible with gcc for Android.
   if (Triple.isAndroid()) {
     if (ArchType == llvm::Triple::x86_64) {
       Features.push_back("+sse4.2");
       Features.push_back("+popcnt");
+      Features.push_back("+mcx16");
     } else
       Features.push_back("+ssse3");
   }
 
   // Translate the high level `-mretpoline` flag to the specific target feature
   // flags. We also detect if the user asked for retpoline external thunks but
   // failed to ask for retpolines themselves (through any of the different
   // flags). This is a bit hacky but keeps existing usages working. We should

test/Driver/clang-translation.c

 
 // RUN: %clang -target x86_64-linux-android -### -S %s 2>&1 \
 // RUN:        --sysroot=%S/Inputs/basic_android_tree/sysroot \
 // RUN:   | FileCheck --check-prefix=ANDROID-X86_64 %s
 // ANDROID-X86_64: clang
 // ANDROID-X86_64: "-target-cpu" "x86-64"
 // ANDROID-X86_64: "-target-feature" "+sse4.2"
 // ANDROID-X86_64: "-target-feature" "+popcnt"
+// ANDROID-X86_64: "-target-feature" "+mcx16"
 
 // RUN: %clang -target mips-linux-gnu -### -S %s 2>&1 | \
 // RUN: FileCheck -check-prefix=MIPS %s
 // MIPS: clang
 // MIPS: "-cc1"
 // MIPS: "-target-cpu" "mips32r2"
 // MIPS: "-mfloat-abi" "hard"