[X86] When checking the bits in cpu_features for function multiversioning dispatcher in the resolver, make sure all the required bits are set. Not just one of them

[craig.topper]

Summary

The multiversioning code repurposed the code from __builtin_cpu_supports for checking if a single feature is enabled. That code essentially performed (_cpu_features & (1 << C)) != 0. But with the multiversioning path, the mask is no longer guaranteed to be a power of 2. So we return true anytime any one of the bits in the mask is set not just all of the bits.

The correct check is (_cpu_features & mask) == mask

[erichkeane]

Yikes, good catch!

cfe/trunk/lib/CodeGen/CGBuiltin.cpp

llvm::Value *CodeGenFunction::EmitX86CpuSupports(uint32_t FeaturesMask) {
   // global in the struct STy.
   Value *Idxs[] = {ConstantInt::get(Int32Ty, 0), ConstantInt::get(Int32Ty, 3),
                    ConstantInt::get(Int32Ty, 0)};
   Value *CpuFeatures = Builder.CreateGEP(STy, CpuModel, Idxs);
   Value *Features =
       Builder.CreateAlignedLoad(CpuFeatures, CharUnits::fromQuantity(4));
 
   // Check the value of the bit corresponding to the feature requested.
-  Value *Bitset = Builder.CreateAnd(
-      Features, llvm::ConstantInt::get(Int32Ty, FeaturesMask));
-  return Builder.CreateICmpNE(Bitset, llvm::ConstantInt::get(Int32Ty, 0));
+  Value *Mask = Builder.getInt32(FeaturesMask);
+  Value *Bitset = Builder.CreateAnd(Features, Mask);
+  return Builder.CreateICmpEQ(Bitset, Mask);
 }
 
 Value *CodeGenFunction::EmitX86CpuInit() {
   llvm::FunctionType *FTy = llvm::FunctionType::get(VoidTy,
                                                     /*Variadic*/ false);
   llvm::Constant *Func = CGM.CreateRuntimeFunction(FTy, "__cpu_indicator_init");
   return Builder.CreateCall(Func);
 }

cfe/trunk/test/CodeGen/attr-target-mv.c

 
 // CHECK: define void @bar3()
 // CHECK: call void @foo_decls.ifunc()
 
 // CHECK: define void ()* @foo_decls.resolver() comdat
 // CHECK: ret void ()* @foo_decls.sse4.2
 // CHECK: ret void ()* @foo_decls
 
+// CHECK: define void @bar4()
+// CHECK: call void @foo_multi.ifunc()
+
+// CHECK: define void ()* @foo_multi.resolver() comdat
+// CHECK: and i32 %{{.*}}, 4352
+// CHECK: icmp eq i32 %{{.*}}, 4352
+// CHECK: ret void ()* @foo_multi.fma4_sse4.2
+// CHECK: icmp eq i32 %{{.*}}, 12
+// CHECK: and i32 %{{.*}}, 4352
+// CHECK: icmp eq i32 %{{.*}}, 4352
+// CHECK: ret void ()* @foo_multi.arch_ivybridge_fma4_sse4.2
+// CHECK: and i32 %{{.*}}, 768
+// CHECK: icmp eq i32 %{{.*}}, 768
+// CHECK: ret void ()* @foo_multi.avx_sse4.2
+// CHECK: ret void ()* @foo_multi
+
 // CHECK: declare i32 @foo.arch_sandybridge()
 
 // CHECK: define available_externally i32 @foo_inline.sse4.2()
 // CHECK: ret i32 0
 
 // CHECK: declare i32 @foo_inline.arch_sandybridge()
 //
 // CHECK: define available_externally i32 @foo_inline.arch_ivybridge()
 // CHECK: ret i32 1
 // CHECK: define available_externally i32 @foo_inline()
 // CHECK: ret i32 2
 
 // CHECK: define available_externally void @foo_decls()
 // CHECK: define available_externally void @foo_decls.sse4.2()
 
 // CHECK: define available_externally void @foo_multi.avx_sse4.2()
 // CHECK: define available_externally void @foo_multi.fma4_sse4.2()
 // CHECK: define available_externally void @foo_multi.arch_ivybridge_fma4_sse4.2()
-

cfe/trunk/test/CodeGen/builtin-cpu-supports.c

 // RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -emit-llvm < %s| FileCheck %s
 
 // Test that we have the structure definition, the gep offsets, the name of the
 // global, the bit grab, and the icmp correct.
 extern void a(const char *);
 
 int main() {
   __builtin_cpu_init();
 
   // CHECK: call void @__cpu_indicator_init
 
   if (__builtin_cpu_supports("sse4.2"))
     a("sse4.2");
 
   // CHECK: [[LOAD:%[^ ]+]] = load i32, i32* getelementptr inbounds ({ i32, i32, i32, [1 x i32] }, { i32, i32, i32, [1 x i32] }* @__cpu_model, i32 0, i32 3, i32 0)
   // CHECK: [[AND:%[^ ]+]] = and i32 [[LOAD]], 256
-  // CHECK: = icmp ne i32 [[AND]], 0
+  // CHECK: = icmp eq i32 [[AND]], 256
 
   return 0;
 }