[llvm] [RISCV] Select and/or/xor with certain constants to Zbb ANDN/ORN/XNOR (PR #120221)

Piotr Fusik via llvm-commits llvm-commits at lists.llvm.org
Thu Dec 19 10:06:21 PST 2024


https://github.com/pfusik updated https://github.com/llvm/llvm-project/pull/120221

>From 81e01d128ca484e9c82d94d4b40a1310c6944fa3 Mon Sep 17 00:00:00 2001
From: Piotr Fusik <p.fusik at samsung.com>
Date: Thu, 19 Dec 2024 17:53:32 +0100
Subject: [PATCH] [RISCV] Select and/or/xor with certain constants to Zbb
 ANDN/ORN/XNOR

    (and X, (C<<12|0xfff)) -> (ANDN X, ~C<<12)
    (or  X, (C<<12|0xfff)) -> (ORN  X, ~C<<12)
    (xor X, (C<<12|0xfff)) -> (XNOR X, ~C<<12)

Emits better code, typically by avoiding an `ADDI HI, -1` instruction.

Co-authored-by: Craig Topper <craig.topper at sifive.com>
---
 llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp  |  29 ++++
 llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h    |   2 +
 llvm/lib/Target/RISCV/RISCVInstrInfoZb.td    |   6 +
 llvm/test/CodeGen/RISCV/pr84653_pr85190.ll   |   6 +-
 llvm/test/CodeGen/RISCV/zbb-logic-neg-imm.ll | 151 +++++++------------
 5 files changed, 92 insertions(+), 102 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
index fa3357f3eacc3c..b33d58d177457e 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
@@ -3236,6 +3236,35 @@ bool RISCVDAGToDAGISel::selectSHXADD_UWOp(SDValue N, unsigned ShAmt,
   return false;
 }
 
+bool RISCVDAGToDAGISel::selectInvLogicImm(SDValue N, SDValue &Val) {
+  if (!isa<ConstantSDNode>(N))
+    return false;
+
+  int64_t Imm = cast<ConstantSDNode>(N)->getSExtValue();
+  if ((Imm & 0xfff) != 0xfff || Imm == -1)
+    return false;
+
+  for (const SDNode *U : N->users()) {
+    if (!ISD::isBitwiseLogicOp(U->getOpcode()))
+      return false;
+  }
+
+  // For 32-bit signed constants we already know it's a win: LUI+ADDI vs LUI.
+  // For 64-bit constants, the instruction sequences get complex,
+  // so we select inverted only if it's cheaper.
+  if (!isInt<32>(Imm)) {
+    int OrigImmCost = RISCVMatInt::getIntMatCost(APInt(64, Imm), 64, *Subtarget,
+                                                 /*CompressionCost=*/true);
+    int NegImmCost = RISCVMatInt::getIntMatCost(APInt(64, ~Imm), 64, *Subtarget,
+                                                /*CompressionCost=*/true);
+    if (OrigImmCost <= NegImmCost)
+      return false;
+  }
+
+  Val = selectImm(CurDAG, SDLoc(N), N->getSimpleValueType(0), ~Imm, *Subtarget);
+  return true;
+}
+
 static bool vectorPseudoHasAllNBitUsers(SDNode *User, unsigned UserOpNo,
                                         unsigned Bits,
                                         const TargetInstrInfo *TII) {
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h
index 2e738d8d25a6dc..e75aff7eda9938 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h
@@ -119,6 +119,8 @@ class RISCVDAGToDAGISel : public SelectionDAGISel {
     return selectSHXADD_UWOp(N, ShAmt, Val);
   }
 
+  bool selectInvLogicImm(SDValue N, SDValue &Val);
+
   bool hasAllNBitUsers(SDNode *Node, unsigned Bits,
                        const unsigned Depth = 0) const;
   bool hasAllBUsers(SDNode *Node) const { return hasAllNBitUsers(Node, 8); }
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td b/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
index a78091cd02a35f..124caa3b69d31e 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
@@ -475,10 +475,16 @@ def : InstAlias<"zext.h $rd, $rs", (PACKW GPR:$rd, GPR:$rs, X0)>;
 // Codegen patterns
 //===----------------------------------------------------------------------===//
 
+def invLogicImm : ComplexPattern<XLenVT, 1, "selectInvLogicImm", [], [], 0>;
+
 let Predicates = [HasStdExtZbbOrZbkb] in {
 def : Pat<(XLenVT (and GPR:$rs1, (not GPR:$rs2))), (ANDN GPR:$rs1, GPR:$rs2)>;
 def : Pat<(XLenVT (or  GPR:$rs1, (not GPR:$rs2))), (ORN  GPR:$rs1, GPR:$rs2)>;
 def : Pat<(XLenVT (xor GPR:$rs1, (not GPR:$rs2))), (XNOR GPR:$rs1, GPR:$rs2)>;
+
+def : Pat<(XLenVT (and GPR:$rs1, invLogicImm:$rs2)), (ANDN GPR:$rs1, invLogicImm:$rs2)>;
+def : Pat<(XLenVT (or  GPR:$rs1, invLogicImm:$rs2)), (ORN  GPR:$rs1, invLogicImm:$rs2)>;
+def : Pat<(XLenVT (xor GPR:$rs1, invLogicImm:$rs2)), (XNOR GPR:$rs1, invLogicImm:$rs2)>;
 } // Predicates = [HasStdExtZbbOrZbkb]
 
 let Predicates = [HasStdExtZbbOrZbkb] in {
diff --git a/llvm/test/CodeGen/RISCV/pr84653_pr85190.ll b/llvm/test/CodeGen/RISCV/pr84653_pr85190.ll
index b1bba5fdc92116..30a93557347727 100644
--- a/llvm/test/CodeGen/RISCV/pr84653_pr85190.ll
+++ b/llvm/test/CodeGen/RISCV/pr84653_pr85190.ll
@@ -21,8 +21,7 @@ define i1 @pr84653(i32 %x) {
 ; CHECK-ZBB:       # %bb.0:
 ; CHECK-ZBB-NEXT:    sext.w a1, a0
 ; CHECK-ZBB-NEXT:    lui a2, 524288
-; CHECK-ZBB-NEXT:    addi a2, a2, -1
-; CHECK-ZBB-NEXT:    xor a0, a0, a2
+; CHECK-ZBB-NEXT:    xnor a0, a0, a2
 ; CHECK-ZBB-NEXT:    sext.w a0, a0
 ; CHECK-ZBB-NEXT:    max a0, a0, zero
 ; CHECK-ZBB-NEXT:    slt a0, a0, a1
@@ -82,8 +81,7 @@ define i1 @select_to_or(i32 %x) {
 ; CHECK-ZBB:       # %bb.0:
 ; CHECK-ZBB-NEXT:    sext.w a1, a0
 ; CHECK-ZBB-NEXT:    lui a2, 524288
-; CHECK-ZBB-NEXT:    addi a2, a2, -1
-; CHECK-ZBB-NEXT:    xor a0, a0, a2
+; CHECK-ZBB-NEXT:    xnor a0, a0, a2
 ; CHECK-ZBB-NEXT:    sext.w a0, a0
 ; CHECK-ZBB-NEXT:    min a0, a0, zero
 ; CHECK-ZBB-NEXT:    slt a0, a0, a1
diff --git a/llvm/test/CodeGen/RISCV/zbb-logic-neg-imm.ll b/llvm/test/CodeGen/RISCV/zbb-logic-neg-imm.ll
index 87e72c28a99658..f1e4bd09fcb928 100644
--- a/llvm/test/CodeGen/RISCV/zbb-logic-neg-imm.ll
+++ b/llvm/test/CodeGen/RISCV/zbb-logic-neg-imm.ll
@@ -9,37 +9,21 @@
 ; RUN:   | FileCheck %s --check-prefixes=CHECK,RV64,ZBS
 
 define i32 @and0xabcdefff(i32 %x) {
-; RV32-LABEL: and0xabcdefff:
-; RV32:       # %bb.0:
-; RV32-NEXT:    lui a1, 703711
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    and a0, a0, a1
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: and0xabcdefff:
-; RV64:       # %bb.0:
-; RV64-NEXT:    lui a1, 703711
-; RV64-NEXT:    addiw a1, a1, -1
-; RV64-NEXT:    and a0, a0, a1
-; RV64-NEXT:    ret
+; CHECK-LABEL: and0xabcdefff:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    lui a1, 344865
+; CHECK-NEXT:    andn a0, a0, a1
+; CHECK-NEXT:    ret
   %and = and i32 %x, -1412567041
   ret i32 %and
 }
 
 define i32 @orlow13(i32 %x) {
-; RV32-LABEL: orlow13:
-; RV32:       # %bb.0:
-; RV32-NEXT:    lui a1, 2
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    or a0, a0, a1
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: orlow13:
-; RV64:       # %bb.0:
-; RV64-NEXT:    lui a1, 2
-; RV64-NEXT:    addiw a1, a1, -1
-; RV64-NEXT:    or a0, a0, a1
-; RV64-NEXT:    ret
+; CHECK-LABEL: orlow13:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    lui a1, 1048574
+; CHECK-NEXT:    orn a0, a0, a1
+; CHECK-NEXT:    ret
   %or = or i32 %x, 8191
   ret i32 %or
 }
@@ -47,53 +31,35 @@ define i32 @orlow13(i32 %x) {
 define i64 @orlow24(i64 %x) {
 ; RV32-LABEL: orlow24:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    lui a2, 4096
-; RV32-NEXT:    addi a2, a2, -1
-; RV32-NEXT:    or a0, a0, a2
+; RV32-NEXT:    lui a2, 1044480
+; RV32-NEXT:    orn a0, a0, a2
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: orlow24:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    lui a1, 4096
-; RV64-NEXT:    addiw a1, a1, -1
-; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    lui a1, 1044480
+; RV64-NEXT:    orn a0, a0, a1
 ; RV64-NEXT:    ret
   %or = or i64 %x, 16777215
   ret i64 %or
 }
 
 define i32 @xorlow16(i32 %x) {
-; RV32-LABEL: xorlow16:
-; RV32:       # %bb.0:
-; RV32-NEXT:    lui a1, 16
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    xor a0, a0, a1
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: xorlow16:
-; RV64:       # %bb.0:
-; RV64-NEXT:    lui a1, 16
-; RV64-NEXT:    addiw a1, a1, -1
-; RV64-NEXT:    xor a0, a0, a1
-; RV64-NEXT:    ret
+; CHECK-LABEL: xorlow16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    lui a1, 1048560
+; CHECK-NEXT:    xnor a0, a0, a1
+; CHECK-NEXT:    ret
   %xor = xor i32 %x, 65535
   ret i32 %xor
 }
 
 define i32 @xorlow31(i32 %x) {
-; RV32-LABEL: xorlow31:
-; RV32:       # %bb.0:
-; RV32-NEXT:    lui a1, 524288
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    xor a0, a0, a1
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: xorlow31:
-; RV64:       # %bb.0:
-; RV64-NEXT:    lui a1, 524288
-; RV64-NEXT:    addiw a1, a1, -1
-; RV64-NEXT:    xor a0, a0, a1
-; RV64-NEXT:    ret
+; CHECK-LABEL: xorlow31:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    lui a1, 524288
+; CHECK-NEXT:    xnor a0, a0, a1
+; CHECK-NEXT:    ret
   %xor = xor i32 %x, 2147483647
   ret i32 %xor
 }
@@ -164,8 +130,7 @@ define void @orarray100(ptr %a) {
 ; RV32:       # %bb.0: # %entry
 ; RV32-NEXT:    li a1, 0
 ; RV32-NEXT:    li a2, 0
-; RV32-NEXT:    lui a3, 16
-; RV32-NEXT:    addi a3, a3, -1
+; RV32-NEXT:    lui a3, 1048560
 ; RV32-NEXT:  .LBB8_1: # %for.body
 ; RV32-NEXT:    # =>This Inner Loop Header: Depth=1
 ; RV32-NEXT:    slli a4, a1, 2
@@ -175,7 +140,7 @@ define void @orarray100(ptr %a) {
 ; RV32-NEXT:    seqz a6, a1
 ; RV32-NEXT:    add a2, a2, a6
 ; RV32-NEXT:    xori a6, a1, 100
-; RV32-NEXT:    or a5, a5, a3
+; RV32-NEXT:    orn a5, a5, a3
 ; RV32-NEXT:    or a6, a6, a2
 ; RV32-NEXT:    sw a5, 0(a4)
 ; RV32-NEXT:    bnez a6, .LBB8_1
@@ -185,12 +150,11 @@ define void @orarray100(ptr %a) {
 ; RV64-LABEL: orarray100:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi a1, a0, 400
-; RV64-NEXT:    lui a2, 16
-; RV64-NEXT:    addi a2, a2, -1
+; RV64-NEXT:    lui a2, 1048560
 ; RV64-NEXT:  .LBB8_1: # %for.body
 ; RV64-NEXT:    # =>This Inner Loop Header: Depth=1
 ; RV64-NEXT:    lw a3, 0(a0)
-; RV64-NEXT:    or a3, a3, a2
+; RV64-NEXT:    orn a3, a3, a2
 ; RV64-NEXT:    sw a3, 0(a0)
 ; RV64-NEXT:    addi a0, a0, 4
 ; RV64-NEXT:    bne a0, a1, .LBB8_1
@@ -216,17 +180,16 @@ for.body:
 define void @orarray3(ptr %a) {
 ; CHECK-LABEL: orarray3:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    lui a1, 16
-; CHECK-NEXT:    lw a2, 0(a0)
-; CHECK-NEXT:    lw a3, 4(a0)
-; CHECK-NEXT:    lw a4, 8(a0)
-; CHECK-NEXT:    addi a1, a1, -1
-; CHECK-NEXT:    or a2, a2, a1
-; CHECK-NEXT:    or a3, a3, a1
-; CHECK-NEXT:    or a1, a4, a1
-; CHECK-NEXT:    sw a2, 0(a0)
-; CHECK-NEXT:    sw a3, 4(a0)
-; CHECK-NEXT:    sw a1, 8(a0)
+; CHECK-NEXT:    lw a1, 0(a0)
+; CHECK-NEXT:    lw a2, 4(a0)
+; CHECK-NEXT:    lw a3, 8(a0)
+; CHECK-NEXT:    lui a4, 1048560
+; CHECK-NEXT:    orn a1, a1, a4
+; CHECK-NEXT:    orn a2, a2, a4
+; CHECK-NEXT:    orn a3, a3, a4
+; CHECK-NEXT:    sw a1, 0(a0)
+; CHECK-NEXT:    sw a2, 4(a0)
+; CHECK-NEXT:    sw a3, 8(a0)
 ; CHECK-NEXT:    ret
   %1 = load i32, ptr %a, align 4
   %or = or i32 %1, 65535
@@ -279,16 +242,14 @@ define i32 @compl(i32 %x) {
 define i32 @orlow12(i32 %x) {
 ; NOZBS32-LABEL: orlow12:
 ; NOZBS32:       # %bb.0:
-; NOZBS32-NEXT:    lui a1, 1
-; NOZBS32-NEXT:    addi a1, a1, -1
-; NOZBS32-NEXT:    or a0, a0, a1
+; NOZBS32-NEXT:    lui a1, 1048575
+; NOZBS32-NEXT:    orn a0, a0, a1
 ; NOZBS32-NEXT:    ret
 ;
 ; NOZBS64-LABEL: orlow12:
 ; NOZBS64:       # %bb.0:
-; NOZBS64-NEXT:    lui a1, 1
-; NOZBS64-NEXT:    addiw a1, a1, -1
-; NOZBS64-NEXT:    or a0, a0, a1
+; NOZBS64-NEXT:    lui a1, 1048575
+; NOZBS64-NEXT:    orn a0, a0, a1
 ; NOZBS64-NEXT:    ret
 ;
 ; ZBS-LABEL: orlow12:
@@ -303,16 +264,14 @@ define i32 @orlow12(i32 %x) {
 define i32 @xorlow12(i32 %x) {
 ; NOZBS32-LABEL: xorlow12:
 ; NOZBS32:       # %bb.0:
-; NOZBS32-NEXT:    lui a1, 1
-; NOZBS32-NEXT:    addi a1, a1, -1
-; NOZBS32-NEXT:    xor a0, a0, a1
+; NOZBS32-NEXT:    lui a1, 1048575
+; NOZBS32-NEXT:    xnor a0, a0, a1
 ; NOZBS32-NEXT:    ret
 ;
 ; NOZBS64-LABEL: xorlow12:
 ; NOZBS64:       # %bb.0:
-; NOZBS64-NEXT:    lui a1, 1
-; NOZBS64-NEXT:    addiw a1, a1, -1
-; NOZBS64-NEXT:    xor a0, a0, a1
+; NOZBS64-NEXT:    lui a1, 1048575
+; NOZBS64-NEXT:    xnor a0, a0, a1
 ; NOZBS64-NEXT:    ret
 ;
 ; ZBS-LABEL: xorlow12:
@@ -327,18 +286,16 @@ define i32 @xorlow12(i32 %x) {
 define i64 @andimm64(i64 %x) {
 ; RV32-LABEL: andimm64:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    lui a1, 1044496
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    lui a1, 4080
+; RV32-NEXT:    andn a0, a0, a1
 ; RV32-NEXT:    li a1, 0
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: andimm64:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    lui a1, 65281
+; RV64-NEXT:    lui a1, 983295
 ; RV64-NEXT:    slli a1, a1, 4
-; RV64-NEXT:    addi a1, a1, -1
-; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    andn a0, a0, a1
 ; RV64-NEXT:    ret
   %and = and i64 %x, 4278255615
   ret i64 %and
@@ -347,19 +304,17 @@ define i64 @andimm64(i64 %x) {
 define i64 @andimm64srli(i64 %x) {
 ; RV32-LABEL: andimm64srli:
 ; RV32:       # %bb.0:
+; RV32-NEXT:    lui a2, 1040384
+; RV32-NEXT:    orn a0, a0, a2
 ; RV32-NEXT:    lui a2, 917504
 ; RV32-NEXT:    or a1, a1, a2
-; RV32-NEXT:    lui a2, 8192
-; RV32-NEXT:    addi a2, a2, -1
-; RV32-NEXT:    or a0, a0, a2
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: andimm64srli:
 ; RV64:       # %bb.0:
 ; RV64-NEXT:    lui a1, 983040
 ; RV64-NEXT:    srli a1, a1, 3
-; RV64-NEXT:    not a1, a1
-; RV64-NEXT:    or a0, a0, a1
+; RV64-NEXT:    orn a0, a0, a1
 ; RV64-NEXT:    ret
   %or = or i64 %x, -2305843009180139521
   ret i64 %or



More information about the llvm-commits mailing list