[NVPTX] Switch nvptx-use-infer-addrspace to true.

[jlebar]

Summary

This switches us to use a different, more powerful algorithm for address
space inference. I've tested this locally and it seems to work great.
Once we're more confident in it, we can remove the old pass altogether.

[jlebar]

Remove out-of-date comment.

[jlebar]

Update tests.

[jingyue]

Yay! Thanks for testing this!

llvm/trunk/lib/Target/NVPTX/NVPTXInferAddressSpaces.cpp

 // yet, we let `I` temporarily use `undef` and fix all the uses of undef later.
 // For instance, our algorithm first converts %y to
 //   %y' = phi float addrspace(3)* [ %input, undef ]
 // Then, it converts %y2 to
 //   %y2' = getelementptr %y', 1
 // Finally, it fixes the undef in %y' so that
 //   %y' = phi float addrspace(3)* [ %input, %y2' ]
 //
-// TODO: This pass is experimental and not enabled by default. Users can turn it
-// on by setting the -nvptx-use-infer-addrspace flag of llc. We plan to replace
-// NVPTXNonFavorGenericAddrSpaces with this pass shortly.
 //===----------------------------------------------------------------------===//
 
 #define DEBUG_TYPE "nvptx-infer-addrspace"
 
 #include "NVPTX.h"
 #include "MCTargetDesc/NVPTXBaseInfo.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/Optional.h"

llvm/trunk/lib/Target/NVPTX/NVPTXTargetMachine.cpp

 #include "llvm/Target/TargetSubtargetInfo.h"
 #include "llvm/Transforms/Scalar.h"
 #include "llvm/Transforms/Scalar/GVN.h"
 #include "llvm/Transforms/Vectorize.h"
 
 using namespace llvm;
 
 static cl::opt<bool> UseInferAddressSpaces(
-    "nvptx-use-infer-addrspace", cl::init(false), cl::Hidden,
+    "nvptx-use-infer-addrspace", cl::init(true), cl::Hidden,
     cl::desc("Optimize address spaces using NVPTXInferAddressSpaces instead of "
              "NVPTXFavorNonGenericAddrSpaces"));
 
 // LSV is still relatively new; this switch lets us turn it off in case we
 // encounter (or suspect) a bug.
 static cl::opt<bool>
     DisableLoadStoreVectorizer("disable-nvptx-load-store-vectorizer",
                                cl::desc("Disable load/store vectorizer"),

llvm/trunk/test/CodeGen/NVPTX/access-non-generic.ll

 ; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s --check-prefix PTX
 ; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s --check-prefix PTX
-; RUN: llc < %s  -march=nvptx64 -mcpu=sm_20 -nvptx-use-infer-addrspace | FileCheck %s --check-prefix PTX
+; RUN: llc < %s  -march=nvptx64 -mcpu=sm_20 -nvptx-use-infer-addrspace=true | FileCheck %s --check-prefix PTX
+; RUN: llc < %s  -march=nvptx64 -mcpu=sm_20 -nvptx-use-infer-addrspace=false | FileCheck %s --check-prefix PTX
 ; RUN: opt < %s -S -nvptx-favor-non-generic -dce | FileCheck %s --check-prefix IR
 ; RUN: opt < %s -S -nvptx-infer-addrspace | FileCheck %s --check-prefix IR --check-prefix IR-WITH-LOOP
 
 @array = internal addrspace(3) global [10 x float] zeroinitializer, align 4
 @scalar = internal addrspace(3) global float 0.000000e+00, align 4
 @generic_scalar = internal global float 0.000000e+00, align 4
 
 define float @ld_from_shared() {

llvm/trunk/test/CodeGen/NVPTX/addrspacecast.ll

-; RUN: llc < %s -march=nvptx -mcpu=sm_20 -disable-nvptx-favor-non-generic | FileCheck %s -check-prefix=PTX32
-; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 -disable-nvptx-favor-non-generic | FileCheck %s -check-prefix=PTX64
+; RUN: llc < %s -march=nvptx -mcpu=sm_20 -disable-nvptx-favor-non-generic \
+; RUN:   -nvptx-use-infer-addrspace=false | FileCheck %s -check-prefix=PTX32
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 -disable-nvptx-favor-non-generic \
+; RUN:   -nvptx-use-infer-addrspace=false | FileCheck %s -check-prefix=PTX64
 
 
 define i32 @conv1(i32 addrspace(1)* %ptr) {
 ; PTX32: conv1
 ; PTX32: cvta.global.u32
 ; PTX32: ld.u32
 ; PTX64: conv1
 ; PTX64: cvta.global.u64