[SystemZ] Automatically detect zEC12 and z196 hosts
As on other hosts, the CPU identification instruction is priveleged,
so we need to look through /proc/cpuinfo.  I copied the PowerPC way of
handling "generic".
Several tests were implicitly assuming z10 and so failed on z196.
llvm-svn: 193742
diff --git a/llvm/lib/Support/Host.cpp b/llvm/lib/Support/Host.cpp
index 12df08c..6a5d4d2 100644
--- a/llvm/lib/Support/Host.cpp
+++ b/llvm/lib/Support/Host.cpp
@@ -535,6 +535,48 @@
 
   return "generic";
 }
+#elif defined(__linux__) && defined(__s390x__)
+std::string sys::getHostCPUName() {
+  // STIDP is a privileged operation, so use /proc/cpuinfo instead.
+  // Note: We cannot mmap /proc/cpuinfo here and then process the resulting
+  // memory buffer because the 'file' has 0 size (it can be read from only
+  // as a stream).
+
+  std::string Err;
+  DataStreamer *DS = getDataFileStreamer("/proc/cpuinfo", &Err);
+  if (!DS) {
+    DEBUG(dbgs() << "Unable to open /proc/cpuinfo: " << Err << "\n");
+    return "generic";
+  }
+
+  // The "processor 0:" line comes after a fair amount of other information,
+  // including a cache breakdown, but this should be plenty.
+  char buffer[2048];
+  size_t CPUInfoSize = DS->GetBytes((unsigned char*) buffer, sizeof(buffer));
+  delete DS;
+
+  StringRef Str(buffer, CPUInfoSize);
+  SmallVector<StringRef, 32> Lines;
+  Str.split(Lines, "\n");
+  for (unsigned I = 0, E = Lines.size(); I != E; ++I) {
+    if (Lines[I].startswith("processor ")) {
+      size_t Pos = Lines[I].find("machine = ");
+      if (Pos != StringRef::npos) {
+        Pos += sizeof("machine = ") - 1;
+        unsigned int Id;
+        if (!Lines[I].drop_front(Pos).getAsInteger(10, Id)) {
+          if (Id >= 2827)
+            return "zEC12";
+          if (Id >= 2817)
+            return "z196";
+        }
+      }
+      break;
+    }
+  }
+  
+  return "generic";
+}
 #else
 std::string sys::getHostCPUName() {
   return "generic";
diff --git a/llvm/lib/Target/SystemZ/SystemZProcessors.td b/llvm/lib/Target/SystemZ/SystemZProcessors.td
index 00d4338..f241fb0 100644
--- a/llvm/lib/Target/SystemZ/SystemZProcessors.td
+++ b/llvm/lib/Target/SystemZ/SystemZProcessors.td
@@ -36,8 +36,9 @@
   "Assume that the floating-point extension facility is installed"
 >;
 
-def : Processor<"z10",   NoItineraries, []>;
-def : Processor<"z196",  NoItineraries,
+def : Processor<"generic", NoItineraries, []>;
+def : Processor<"z10", NoItineraries, []>;
+def : Processor<"z196", NoItineraries,
                 [FeatureDistinctOps, FeatureLoadStoreOnCond, FeatureHighWord,
                  FeatureFPExtension]>;
 def : Processor<"zEC12", NoItineraries,
diff --git a/llvm/lib/Target/SystemZ/SystemZSubtarget.cpp b/llvm/lib/Target/SystemZ/SystemZSubtarget.cpp
index b6a63923..4741920 100644
--- a/llvm/lib/Target/SystemZ/SystemZSubtarget.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZSubtarget.cpp
@@ -9,6 +9,7 @@
 
 #include "SystemZSubtarget.h"
 #include "llvm/IR/GlobalValue.h"
+#include "llvm/Support/Host.h"
 #include "MCTargetDesc/SystemZMCTargetDesc.h"
 
 #define GET_SUBTARGETINFO_TARGET_DESC
@@ -25,7 +26,11 @@
     TargetTriple(TT) {
   std::string CPUName = CPU;
   if (CPUName.empty())
-    CPUName = "z10";
+    CPUName = "generic";
+#if defined(__linux__) && defined(__s390x__)
+  if (CPUName == "generic")
+    CPUName = sys::getHostCPUName();
+#endif
 
   // Parse features string.
   ParseSubtargetFeatures(CPUName, FS);
diff --git a/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-03.ll b/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-03.ll
index 418f156..f2152c6 100644
--- a/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-03.ll
+++ b/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-03.ll
@@ -1,6 +1,7 @@
-; Test 32-bit atomic minimum and maximum.
+; Test 32-bit atomic minimum and maximum.  Here we match the z10 versions,
+; which can't use LOCR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Check signed minium.
 define i32 @f1(i32 %dummy, i32 *%src, i32 %b) {
diff --git a/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-04.ll b/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-04.ll
index 9d26d28..037eb1a 100644
--- a/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-04.ll
+++ b/llvm/test/CodeGen/SystemZ/atomicrmw-minmax-04.ll
@@ -1,6 +1,7 @@
-; Test 64-bit atomic minimum and maximum.
+; Test 64-bit atomic minimum and maximum.  Here we match the z10 versions,
+; which can't use LOCGR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Check signed minium.
 define i64 @f1(i64 %dummy, i64 *%src, i64 %b) {
diff --git a/llvm/test/CodeGen/SystemZ/cond-store-01.ll b/llvm/test/CodeGen/SystemZ/cond-store-01.ll
index 5b55934..d55ea21 100644
--- a/llvm/test/CodeGen/SystemZ/cond-store-01.ll
+++ b/llvm/test/CodeGen/SystemZ/cond-store-01.ll
@@ -1,6 +1,7 @@
-; Test 8-bit conditional stores that are presented as selects.
+; Test 8-bit conditional stores that are presented as selects.  The volatile
+; tests require z10, which use a branch instead of a LOCR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 declare void @foo(i8 *)
 
diff --git a/llvm/test/CodeGen/SystemZ/cond-store-02.ll b/llvm/test/CodeGen/SystemZ/cond-store-02.ll
index 9e18843..91bc486 100644
--- a/llvm/test/CodeGen/SystemZ/cond-store-02.ll
+++ b/llvm/test/CodeGen/SystemZ/cond-store-02.ll
@@ -1,6 +1,7 @@
-; Test 16-bit conditional stores that are presented as selects.
+; Test 16-bit conditional stores that are presented as selects.  The volatile
+; tests require z10, which use a branch instead of a LOCR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 declare void @foo(i16 *)
 
diff --git a/llvm/test/CodeGen/SystemZ/fp-cmp-01.ll b/llvm/test/CodeGen/SystemZ/fp-cmp-01.ll
index 7f19401..d7c0cce 100644
--- a/llvm/test/CodeGen/SystemZ/fp-cmp-01.ll
+++ b/llvm/test/CodeGen/SystemZ/fp-cmp-01.ll
@@ -1,6 +1,7 @@
-; Test 32-bit floating-point comparison.
+; Test 32-bit floating-point comparison.  The tests assume a z10 implementation
+; of select, using conditional branches rather than LOCGR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 declare float @foo()
 
diff --git a/llvm/test/CodeGen/SystemZ/fp-cmp-02.ll b/llvm/test/CodeGen/SystemZ/fp-cmp-02.ll
index 1cd6da8..c61f04e 100644
--- a/llvm/test/CodeGen/SystemZ/fp-cmp-02.ll
+++ b/llvm/test/CodeGen/SystemZ/fp-cmp-02.ll
@@ -1,6 +1,7 @@
-; Test 64-bit floating-point comparison.
+; Test 64-bit floating-point comparison.  The tests assume a z10 implementation
+; of select, using conditional branches rather than LOCGR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 declare double @foo()
 
diff --git a/llvm/test/CodeGen/SystemZ/fp-cmp-03.ll b/llvm/test/CodeGen/SystemZ/fp-cmp-03.ll
index 0f71f4e..e777d00 100644
--- a/llvm/test/CodeGen/SystemZ/fp-cmp-03.ll
+++ b/llvm/test/CodeGen/SystemZ/fp-cmp-03.ll
@@ -1,6 +1,7 @@
-; Test 128-bit floating-point comparison.
+; Test 128-bit floating-point comparison.  The tests assume a z10 implementation
+; of select, using conditional branches rather than LOCGR.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; There is no memory form of 128-bit comparison.
 define i64 @f1(i64 %a, i64 %b, fp128 *%ptr, float %f2) {
diff --git a/llvm/test/CodeGen/SystemZ/fp-move-02.ll b/llvm/test/CodeGen/SystemZ/fp-move-02.ll
index 6f9da9a..505ee8d 100644
--- a/llvm/test/CodeGen/SystemZ/fp-move-02.ll
+++ b/llvm/test/CodeGen/SystemZ/fp-move-02.ll
@@ -1,6 +1,7 @@
-; Test moves between FPRs and GPRs.
+; Test moves between FPRs and GPRs.  The 32-bit cases test the z10
+; implementation, which has no high-word support.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 declare i64 @foo()
 declare double @bar()
diff --git a/llvm/test/CodeGen/SystemZ/frame-13.ll b/llvm/test/CodeGen/SystemZ/frame-13.ll
index 60bff50..393850f 100644
--- a/llvm/test/CodeGen/SystemZ/frame-13.ll
+++ b/llvm/test/CodeGen/SystemZ/frame-13.ll
@@ -1,8 +1,11 @@
 ; Test the handling of base + 12-bit displacement addresses for large frames,
-; in cases where no 20-bit form exists.
+; in cases where no 20-bit form exists.  The tests here assume z10 register
+; pressure, without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck -check-prefix=CHECK-NOFP %s
-; RUN: llc < %s -mtriple=s390x-linux-gnu -disable-fp-elim | FileCheck -check-prefix=CHECK-FP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | \
+; RUN:   FileCheck -check-prefix=CHECK-NOFP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 -disable-fp-elim | \
+; RUN:   FileCheck -check-prefix=CHECK-FP %s
 
 ; This file tests what happens when a displacement is converted from
 ; being relative to the start of a frame object to being relative to
diff --git a/llvm/test/CodeGen/SystemZ/frame-14.ll b/llvm/test/CodeGen/SystemZ/frame-14.ll
index 22a45ee..3b48179 100644
--- a/llvm/test/CodeGen/SystemZ/frame-14.ll
+++ b/llvm/test/CodeGen/SystemZ/frame-14.ll
@@ -1,9 +1,13 @@
 ; Test the handling of base + displacement addresses for large frames,
 ; in cases where both 12-bit and 20-bit displacements are allowed.
+; The tests here assume z10 register pressure, without the high words
+; being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck -check-prefix=CHECK-NOFP %s
-; RUN: llc < %s -mtriple=s390x-linux-gnu -disable-fp-elim | FileCheck -check-prefix=CHECK-FP %s
-
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | \
+; RUN:   FileCheck -check-prefix=CHECK-NOFP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 -disable-fp-elim | \
+; RUN:   FileCheck -check-prefix=CHECK-FP %s
+;
 ; This file tests what happens when a displacement is converted from
 ; being relative to the start of a frame object to being relative to
 ; the frame itself.  In some cases the test is only possible if two
diff --git a/llvm/test/CodeGen/SystemZ/frame-15.ll b/llvm/test/CodeGen/SystemZ/frame-15.ll
index d8b291d..b3c95e7 100644
--- a/llvm/test/CodeGen/SystemZ/frame-15.ll
+++ b/llvm/test/CodeGen/SystemZ/frame-15.ll
@@ -1,8 +1,11 @@
 ; Test the handling of base + index + 12-bit displacement addresses for
-; large frames, in cases where no 20-bit form exists.
+; large frames, in cases where no 20-bit form exists.  The tests here
+; assume z10 register pressure, without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck -check-prefix=CHECK-NOFP %s
-; RUN: llc < %s -mtriple=s390x-linux-gnu -disable-fp-elim | FileCheck -check-prefix=CHECK-FP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | \
+; RUN:   FileCheck -check-prefix=CHECK-NOFP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 -disable-fp-elim | \
+; RUN:   FileCheck -check-prefix=CHECK-FP %s
 
 declare void @foo(float *%ptr1, float *%ptr2)
 
diff --git a/llvm/test/CodeGen/SystemZ/frame-16.ll b/llvm/test/CodeGen/SystemZ/frame-16.ll
index 9f43b49..f7e2dfa 100644
--- a/llvm/test/CodeGen/SystemZ/frame-16.ll
+++ b/llvm/test/CodeGen/SystemZ/frame-16.ll
@@ -1,8 +1,12 @@
 ; Test the handling of base + index + displacement addresses for large frames,
 ; in cases where both 12-bit and 20-bit displacements are allowed.
+; The tests here assume z10 register pressure, without the high words
+; being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck -check-prefix=CHECK-NOFP %s
-; RUN: llc < %s -mtriple=s390x-linux-gnu -disable-fp-elim | FileCheck -check-prefix=CHECK-FP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | \
+; RUN:   FileCheck -check-prefix=CHECK-NOFP %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 -disable-fp-elim | \
+; RUN:   FileCheck -check-prefix=CHECK-FP %s
 
 ; This file tests what happens when a displacement is converted from
 ; being relative to the start of a frame object to being relative to
diff --git a/llvm/test/CodeGen/SystemZ/frame-18.ll b/llvm/test/CodeGen/SystemZ/frame-18.ll
index 57d6f7d..21dfc12 100644
--- a/llvm/test/CodeGen/SystemZ/frame-18.ll
+++ b/llvm/test/CodeGen/SystemZ/frame-18.ll
@@ -1,6 +1,7 @@
-; Test spilling of GPRs.
+; Test spilling of GPRs.  The tests here assume z10 register pressure,
+; without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; We need to allocate a 4-byte spill slot, rounded to 8 bytes.  The frame
 ; size should be exactly 160 + 8 = 168.
diff --git a/llvm/test/CodeGen/SystemZ/int-add-11.ll b/llvm/test/CodeGen/SystemZ/int-add-11.ll
index 212334e..679c206 100644
--- a/llvm/test/CodeGen/SystemZ/int-add-11.ll
+++ b/llvm/test/CodeGen/SystemZ/int-add-11.ll
@@ -1,6 +1,7 @@
-; Test 32-bit additions of constants to memory.
+; Test 32-bit additions of constants to memory.  The tests here
+; assume z10 register pressure, without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Check additions of 1.
 define void @f1(i32 *%ptr) {
diff --git a/llvm/test/CodeGen/SystemZ/int-conv-02.ll b/llvm/test/CodeGen/SystemZ/int-conv-02.ll
index 18cfd4a..dd7760d 100644
--- a/llvm/test/CodeGen/SystemZ/int-conv-02.ll
+++ b/llvm/test/CodeGen/SystemZ/int-conv-02.ll
@@ -1,6 +1,7 @@
-; Test zero extensions from a byte to an i32.
+; Test zero extensions from a byte to an i32.    The tests here
+; assume z10 register pressure, without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Test register extension, starting with an i32.
 define i32 @f1(i32 %a) {
diff --git a/llvm/test/CodeGen/SystemZ/int-conv-06.ll b/llvm/test/CodeGen/SystemZ/int-conv-06.ll
index 9c95bad..33860d1 100644
--- a/llvm/test/CodeGen/SystemZ/int-conv-06.ll
+++ b/llvm/test/CodeGen/SystemZ/int-conv-06.ll
@@ -1,6 +1,7 @@
-; Test zero extensions from a halfword to an i32.
+; Test zero extensions from a halfword to an i32.  The tests here
+; assume z10 register pressure, without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Test register extension, starting with an i32.
 define i32 @f1(i32 %a) {
diff --git a/llvm/test/CodeGen/SystemZ/risbg-01.ll b/llvm/test/CodeGen/SystemZ/risbg-01.ll
index 85de6dc..8a5d487 100644
--- a/llvm/test/CodeGen/SystemZ/risbg-01.ll
+++ b/llvm/test/CodeGen/SystemZ/risbg-01.ll
@@ -1,6 +1,7 @@
 ; Test sequences that can use RISBG with a zeroed first operand.
+; The tests here assume that RISBLG isn't available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Test an extraction of bit 0 from a right-shifted value.
 define i32 @f1(i32 %foo) {
diff --git a/llvm/test/CodeGen/SystemZ/setcc-01.ll b/llvm/test/CodeGen/SystemZ/setcc-01.ll
index 5313215..4626760 100644
--- a/llvm/test/CodeGen/SystemZ/setcc-01.ll
+++ b/llvm/test/CodeGen/SystemZ/setcc-01.ll
@@ -1,6 +1,7 @@
-; Test SETCC for every integer condition.
+; Test SETCC for every integer condition.  The tests here assume that
+; RISBLG isn't available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Test CC in { 0 }, with 3 don't care.
 define i32 @f1(i32 %a, i32 %b) {
diff --git a/llvm/test/CodeGen/SystemZ/setcc-02.ll b/llvm/test/CodeGen/SystemZ/setcc-02.ll
index 1788222..6a7be47 100644
--- a/llvm/test/CodeGen/SystemZ/setcc-02.ll
+++ b/llvm/test/CodeGen/SystemZ/setcc-02.ll
@@ -1,6 +1,7 @@
-; Test SETCC for every floating-point condition.
+; Test SETCC for every floating-point condition.  The tests here assume that
+; RISBLG isn't available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 ; Test CC in { 0 }
 define i32 @f1(float %a, float %b) {
diff --git a/llvm/test/CodeGen/SystemZ/spill-01.ll b/llvm/test/CodeGen/SystemZ/spill-01.ll
index 9de89d6..ca64a88 100644
--- a/llvm/test/CodeGen/SystemZ/spill-01.ll
+++ b/llvm/test/CodeGen/SystemZ/spill-01.ll
@@ -1,6 +1,7 @@
-; Test spilling using MVC.
+; Test spilling using MVC.  The tests here assume z10 register pressure,
+; without the high words being available.
 ;
-; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | FileCheck %s
 
 declare void @foo()