[PATCH] D104836: [PowerPC] Combine 64-bit bswap(load) without LDBRX

[Nemanja Ivanovic via Phabricator via llvm-commits]

nemanjai created this revision.
nemanjai added reviewers: spatel, nickdesaulniers, nathanchance, PowerPC.
Herald added subscribers: shchenz, kbarton, hiraditya.
nemanjai requested review of this revision.
Herald added a project: LLVM.

When targeting CPUs that don't have LDBRX, we end up producing code that is very inefficient and large for this common idiom. This patch just optimizes it two 32-bit LWBRX instructions along with a merge.

This fixes https://bugs.llvm.org/show_bug.cgi?id=49610


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D104836

Files:
  llvm/lib/Target/PowerPC/PPCISelLowering.cpp
  llvm/test/CodeGen/PowerPC/ld-bswap64-no-ldbrx.ll


Index: llvm/test/CodeGen/PowerPC/ld-bswap64-no-ldbrx.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/PowerPC/ld-bswap64-no-ldbrx.ll
@@ -0,0 +1,17 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=powerpc64-- -mcpu=pwr5 -verify-machineinstrs | FileCheck %s
+define void @bs(i64* %p) {
+; CHECK-LABEL: bs:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li 4, 4
+; CHECK-NEXT:    lwbrx 5, 0, 3
+; CHECK-NEXT:    lwbrx 4, 3, 4
+; CHECK-NEXT:    rldimi 4, 5, 32, 0
+; CHECK-NEXT:    std 4, 0(3)
+; CHECK-NEXT:    blr
+  %x = load i64, i64* %p, align 8
+  %b = call i64 @llvm.bswap.i64(i64 %x)
+  store i64 %b, i64* %p, align 8
+  ret void
+}
+declare i64 @llvm.bswap.i64(i64) #2
Index: llvm/lib/Target/PowerPC/PPCISelLowering.cpp
===================================================================
--- llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -15202,13 +15202,17 @@
       }
     }
     break;
-  case ISD::BSWAP:
+  case ISD::BSWAP: {
     // Turn BSWAP (LOAD) -> lhbrx/lwbrx.
-    if (ISD::isNON_EXTLoad(N->getOperand(0).getNode()) &&
-        N->getOperand(0).hasOneUse() &&
+    // For subtargets without LDBRX, we can still do better than the default
+    // expansion even for 64-bit BSWAP (LOAD).
+    bool is64BitBswapOn64BitTgt =
+        Subtarget.isPPC64() && N->getValueType(0) == MVT::i64;
+    bool isSingleUseNonExtLd = ISD::isNON_EXTLoad(N->getOperand(0).getNode()) &&
+                               N->getOperand(0).hasOneUse();
+    if (isSingleUseNonExtLd &&
         (N->getValueType(0) == MVT::i32 || N->getValueType(0) == MVT::i16 ||
-         (Subtarget.hasLDBRX() && Subtarget.isPPC64() &&
-          N->getValueType(0) == MVT::i64))) {
+         (Subtarget.hasLDBRX() && is64BitBswapOn64BitTgt))) {
       SDValue Load = N->getOperand(0);
       LoadSDNode *LD = cast<LoadSDNode>(Load);
       // Create the byte-swapping load.
@@ -15238,8 +15242,30 @@
 
       // Return N so it doesn't get rechecked!
       return SDValue(N, 0);
+    } else if (is64BitBswapOn64BitTgt && isSingleUseNonExtLd) {
+      // Convert this to two 32-bit bswap loads and a BUILD_PAIR. Do this only
+      // before legalization so that the BUILD_PAIR is handled correctly.
+      if (!DCI.isBeforeLegalize())
+        return SDValue();
+      LoadSDNode *LD = cast<LoadSDNode>(N->getOperand(0));
+      SDValue BasePtr = LD->getBasePtr();
+      SDValue Lo = DAG.getLoad(MVT::i32, dl, LD->getChain(), BasePtr,
+                               LD->getPointerInfo(), LD->getAlignment());
+      Lo = DAG.getNode(ISD::BSWAP, dl, MVT::i32, Lo);
+      BasePtr = DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), BasePtr,
+                            DAG.getIntPtrConstant(4, dl));
+      SDValue Hi = DAG.getLoad(MVT::i32, dl, LD->getChain(), BasePtr,
+                               LD->getPointerInfo(), LD->getAlignment());
+      Hi = DAG.getNode(ISD::BSWAP, dl, MVT::i32, Hi);
+      SDValue Res = DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Hi, Lo);
+      SDValue TF = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
+                               Hi.getOperand(0).getValue(1),
+                               Lo.getOperand(0).getValue(1));
+      DAG.ReplaceAllUsesOfValueWith(SDValue(LD, 1), TF);
+      return Res;
     }
     break;
+  }
   case PPCISD::VCMP:
     // If a VCMP_rec node already exists with exactly the same operands as this
     // node, use its result instead of this node (VCMP_rec computes both a CR6


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D104836.354145.patch
Type: text/x-patch
Size: 3587 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210624/68cad4f5/attachment.bin>