[LoongArch] Enable LoopDataPrefetch pass

[XiaodongLoong]

Summary

Keep EnableLoopDataPrefetch option off for now because
we need a few more TTIs and ISels.

This patch is inspired by http://reviews.llvm.org/D17943.

[SixWeining]

LGTM except a nit.

llvm/test/Transforms/LoopDataPrefetch/LoongArch/basic.ll

2

According to previous review comments from other guys, it's better to use ;; for non-CHECK and non-RUN lines although this is not a coding standard.

[XiaodongLoong]

Fix a comment nit.

[sdesmalen]

Hi I'm getting build failures with this patch due to a linker error where it says:

ld.lld: error: undefined symbol: llvm::createLoopDataPrefetchPass()
>>> referenced by LoongArchTargetMachine.cpp
>>>               lib/Target/LoongArch/CMakeFiles/LLVMLoongArchCodeGen.dir/LoongArchTargetMachine.cpp.o:((anonymous namespace)::LoongArchPassConfig::addIRPasses())
collect2: error: ld returned 1 exit status`

I think this is missing a dependence on ScalarOpts in lib/Target/LoongArch/CMakeLists.txt

[XiaodongLoong]

OK, I'll reproduce this issue and submit a new patch. Thanks.

[XiaodongLoong]

I think this is missing a dependence on ScalarOpts in lib/Target/LoongArch/CMakeLists.txt

I submit a new patch to fix this issue: https://reviews.llvm.org/D146806. Thanks.

llvm/lib/Target/LoongArch/LoongArchTargetMachine.cpp

 #include "LoongArch.h"
 #include "LoongArchMachineFunctionInfo.h"
 #include "MCTargetDesc/LoongArchBaseInfo.h"
 #include "TargetInfo/LoongArchTargetInfo.h"
 #include "llvm/CodeGen/Passes.h"
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
 #include "llvm/CodeGen/TargetPassConfig.h"
 #include "llvm/MC/TargetRegistry.h"
+#include "llvm/Transforms/Scalar.h"
 #include <optional>
 
 using namespace llvm;
 
 #define DEBUG_TYPE "loongarch"
 
 extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeLoongArchTarget() {
   // Register the target.
   RegisterTargetMachine<LoongArchTargetMachine> X(getTheLoongArch32Target());
   RegisterTargetMachine<LoongArchTargetMachine> Y(getTheLoongArch64Target());
   auto *PR = PassRegistry::getPassRegistry();
   initializeLoongArchPreRAExpandPseudoPass(*PR);
   initializeLoongArchDAGToDAGISelPass(*PR);
 }
 
+static cl::opt<bool>
+    EnableLoopDataPrefetch("loongarch-enable-loop-data-prefetch", cl::Hidden,
+                           cl::desc("Enable the loop data prefetch pass"),
+                           cl::init(false));
+
 static std::string computeDataLayout(const Triple &TT) {
   if (TT.isArch64Bit())
     return "e-m:e-p:64:64-i64:64-i128:128-n64-S128";
   assert(TT.isArch32Bit() && "only LA32 and LA64 are currently supported");
   return "e-m:e-p:32:32-i64:64-n32-S128";
 }
 
 static Reloc::Model getEffectiveRelocModel(const Triple &TT,

 } // end namespace
 
 TargetPassConfig *
 LoongArchTargetMachine::createPassConfig(PassManagerBase &PM) {
   return new LoongArchPassConfig(*this, PM);
 }
 
 void LoongArchPassConfig::addIRPasses() {
+  // Run LoopDataPrefetch
+  //
+  // Run this before LSR to remove the multiplies involved in computing the
+  // pointer values N iterations ahead.
+  if (TM->getOptLevel() != CodeGenOpt::None && EnableLoopDataPrefetch)
+    addPass(createLoopDataPrefetchPass());
   addPass(createAtomicExpandPass());
 
   TargetPassConfig::addIRPasses();
 }
 
 bool LoongArchPassConfig::addInstSelector() {
   addPass(createLoongArchISelDag(getLoongArchTargetMachine()));
 

llvm/test/Transforms/LoopDataPrefetch/LoongArch/basic.ll

+;; Tag this 'XFAIL' because we need a few more TTIs and ISels.
+; XFAIL: *
+; RUN: opt --mtriple=loongarch64 --passes=loop-data-prefetch -loongarch-enable-loop-data-prefetch -S < %s | FileCheck %s
+
+define void @foo(ptr %a, ptr %b) {
+entry:
+  br label %for.body
+
+; CHECK: for.body:
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds double, ptr %b, i64 %indvars.iv
+; CHECK: call void @llvm.prefetch
+  %0 = load double, ptr %arrayidx, align 8
+  %add = fadd double %0, 1.000000e+00
+  %arrayidx2 = getelementptr inbounds double, ptr %a, i64 %indvars.iv
+  store double %add, ptr %arrayidx2, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp eq i64 %indvars.iv.next, 1600
+  br i1 %exitcond, label %for.end, label %for.body
+
+; CHECK: for.end:
+for.end:                                          ; preds = %for.body
+  ret void
+}

llvm/test/Transforms/LoopDataPrefetch/LoongArch/lit.local.cfg

+if not 'LoongArch' in config.root.targets:
+    config.unsupported = True