[llvm] [RISCV] Enable prefetching writes (PR #130561)

[Pengcheng Wang via llvm-commits]

https://github.com/wangpc-pp created https://github.com/llvm/llvm-project/pull/130561

We should prefetch writes since `Zicbop` has `prefetch.w`.


>From 4724a8a5e903ff3521511f22d2ea2b10eb026bfc Mon Sep 17 00:00:00 2001
From: wangpc <wangpengcheng.pp at bytedance.com>
Date: Mon, 10 Mar 2025 15:44:45 +0800
Subject: [PATCH] [RISCV] Enable prefetching writes

We should prefetch writes since `Zicbop` has `prefetch.w`.
---
 llvm/lib/Target/RISCV/RISCVSubtarget.h               | 1 +
 llvm/test/Transforms/LoopDataPrefetch/RISCV/basic.ll | 6 +++++-
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVSubtarget.h b/llvm/lib/Target/RISCV/RISCVSubtarget.h
index cc9aef2d52556..035ce4745cfd9 100644
--- a/llvm/lib/Target/RISCV/RISCVSubtarget.h
+++ b/llvm/lib/Target/RISCV/RISCVSubtarget.h
@@ -354,6 +354,7 @@ class RISCVSubtarget : public RISCVGenSubtargetInfo {
   unsigned getMaxPrefetchIterationsAhead() const override {
     return TuneInfo->MaxPrefetchIterationsAhead;
   };
+  bool enableWritePrefetching() const override { return true; }
 
   unsigned getMinimumJumpTableEntries() const;
 
diff --git a/llvm/test/Transforms/LoopDataPrefetch/RISCV/basic.ll b/llvm/test/Transforms/LoopDataPrefetch/RISCV/basic.ll
index 71d9953a03da1..6edd0ff149b43 100644
--- a/llvm/test/Transforms/LoopDataPrefetch/RISCV/basic.ll
+++ b/llvm/test/Transforms/LoopDataPrefetch/RISCV/basic.ll
@@ -11,12 +11,16 @@ define void @foo(ptr nocapture %a, ptr nocapture readonly %b) {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = shl nuw nsw i64 [[INDVARS_IV]], 3
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[TMP0]], 64
-; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = shl nuw nsw i64 [[INDVARS_IV]], 3
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[TMP4]], 64
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    call void @llvm.prefetch.p0(ptr [[SCEVGEP]], i32 0, i32 3, i32 1)
 ; CHECK-NEXT:    [[TMP2:%.*]] = load double, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[ADD:%.*]] = fadd double [[TMP2]], 1.000000e+00
 ; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    call void @llvm.prefetch.p0(ptr [[SCEVGEP1]], i32 1, i32 3, i32 1)
 ; CHECK-NEXT:    store double [[ADD]], ptr [[ARRAYIDX2]], align 8
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1600