[ARM] Add more CPUs to host detection

[efriedma]

Summary

This returns "cortex-a73" for second-generation Kryo; not precisely correct, but close enough.

[kristof.beyls]

Hi Eli,

Thanks for adding this.
I expect you to have a much better understanding of Kryo cores, so I can't add much comments on the specific constants you added in the code for those.
For Cortex-A73, the constants in this patch seem incorrect, looking at the Cortex-A73 TRM, see my detailed comments.

Thanks!

Kristof

lib/Support/Host.cpp

211 ↗

(On Diff #114735)

The summary of this patch states returning cortex-a73 for second-generation Kryo, but this patch also introduces returning "kryo" for "0x211". Just checking if this was intentional.

unittests/Support/Host.cpp

108–110 ↗

(On Diff #114735)

I don't have easy access to a Cortex-A73 system at the moment, but I'd expect the CPU implementer for a Cortex-A73 be 0x41, the same as for other Arm-designed cores.
Also, according to http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.100048_0002_05_en/jfa1406793219362.html, the part number of Cortex-A73 is 0xD09, not 0x800.

[efriedma]

lib/Support/Host.cpp

211 ↗

(On Diff #114735)

Yes, intentional; 0x211 is first-generation Kryo.

unittests/Support/Host.cpp

108–110 ↗

(On Diff #114735)

Kryo second-generation is not a Cortex-A73. It's just close enough from the compiler's perspective that it isn't worth distinguishing here. (getHostCPUNameForARM already has separate detection code for actual A73 CPUs.)

[kristof.beyls]

unittests/Support/Host.cpp

108–110 ↗

(On Diff #114735)

ah - apologies, I didn't look at enough context code of this change.
I was confused by adding 2 "CPU parts" 0x800 and 0x801 both mapping to Cortex-A73, and didn't catch that both represent second-generation Kryo cores.
It now makes sense to me.

llvm/trunk/lib/Support/Host.cpp

for (unsigned I = 0, E = Lines.size(); I != E; ++I)
       if (Lines[I].startswith("CPU part"))
         // The CPU part is a 3 digit hexadecimal number with a 0x prefix. The
         // values correspond to the "Part number" in the CP15/c0 register. The
         // contents are specified in the various processor manuals.
         return StringSwitch<const char *>(Lines[I].substr(8).ltrim("\t :"))
             .Case("0x06f", "krait") // APQ8064
             .Case("0x201", "kryo")
             .Case("0x205", "kryo")
+            .Case("0x211", "kryo")
+            .Case("0x800", "cortex-a73")
+            .Case("0x801", "cortex-a73")
             .Default("generic");
 
   return "generic";
 }
 
 StringRef sys::detail::getHostCPUNameForS390x(
     const StringRef &ProcCpuinfoContent) {
   // STIDP is a privileged operation, so use /proc/cpuinfo instead.

llvm/trunk/unittests/Support/Host.cpp

EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x41\n"
             "cortex-a53");
   // Verify that both CPU implementer and CPU part are checked:
   EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x40\n"
                                               "CPU part        : 0xd03"),
             "generic");
   EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x51\n"
                                               "CPU part        : 0x201"),
             "kryo");
+  EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x51\n"
+                                              "CPU part        : 0x800"),
+            "cortex-a73");
+  EXPECT_EQ(sys::detail::getHostCPUNameForARM("CPU implementer : 0x51\n"
+                                              "CPU part        : 0x801"),
+            "cortex-a73");
 
   // MSM8992/4 weirdness
   StringRef MSM8992ProcCpuInfo = R"(
 Processor       : AArch64 Processor rev 3 (aarch64)
 processor       : 0
 processor       : 1
 processor       : 2
 processor       : 3