[RISCV] Add isel pattern for (and X, 0xffffffff << C) with Zba.

We can use SRLI by C to clear LSBs followed by a SLLI_UW.

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td b/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
index 90ae2ac..61476dd 100644 (file)
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
@@ -242,6 +242,16 @@ def non_imm12 : PatLeaf<(XLenVT GPR:$a), [{
   return !C || !isInt<12>(C->getSExtValue());
 }]>;
 
+def Shifted32OnesMask : PatLeaf<(imm), [{
+  uint64_t Imm = N->getZExtValue();
+  if (!isShiftedMask_64(Imm))
+    return false;
+
+  unsigned TrailingZeros = countTrailingZeros(Imm);
+  return TrailingZeros > 0 && TrailingZeros < 32 &&
+         Imm == UINT64_C(0xFFFFFFFF) << TrailingZeros;
+}], TrailingZeros>;
+
 def sh1add_op : ComplexPattern<XLenVT, 1, "selectSH1ADDOp", [], [], 6>;
 def sh2add_op : ComplexPattern<XLenVT, 1, "selectSH2ADDOp", [], [], 6>;
 def sh3add_op : ComplexPattern<XLenVT, 1, "selectSH3ADDOp", [], [], 6>;
@@ -731,6 +741,12 @@ def : Pat<(mul_const_oneuse GPR:$r, (XLenVT 81)),
 let Predicates = [HasStdExtZba, IsRV64] in {
 def : Pat<(i64 (shl (and GPR:$rs1, 0xFFFFFFFF), uimm5:$shamt)),
           (SLLI_UW GPR:$rs1, uimm5:$shamt)>;
+// Match a shifted 0xffffffff mask. Use SRLI to clear the LSBs and SLLI_UW to
+// mask and shift.
+def : Pat<(i64 (and GPR:$rs1, Shifted32OnesMask:$mask)),
+          (SLLI_UW (SRLI GPR:$rs1, Shifted32OnesMask:$mask),
+                   Shifted32OnesMask:$mask)>;
+
 def : Pat<(i64 (add (and GPR:$rs1, 0xFFFFFFFF), non_imm12:$rs2)),
           (ADD_UW GPR:$rs1, GPR:$rs2)>;
 def : Pat<(i64 (and GPR:$rs, 0xFFFFFFFF)), (ADD_UW GPR:$rs, X0)>;

diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index e9161ae..6fa4e37 100644 (file)
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -1408,6 +1408,33 @@ define signext i16 @srliw_1_sh1add(i16* %0, i32 signext %1) {
   ret i16 %6
 }
 
+define i128 @slliuw_ptrdiff(i64 %diff, i128* %baseptr) {
+; RV64I-LABEL: slliuw_ptrdiff:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a2, 1
+; RV64I-NEXT:    slli a2, a2, 36
+; RV64I-NEXT:    addi a2, a2, -16
+; RV64I-NEXT:    and a0, a0, a2
+; RV64I-NEXT:    add a1, a1, a0
+; RV64I-NEXT:    ld a0, 0(a1)
+; RV64I-NEXT:    ld a1, 8(a1)
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: slliuw_ptrdiff:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    srli a0, a0, 4
+; RV64ZBA-NEXT:    slli.uw a0, a0, 4
+; RV64ZBA-NEXT:    add a1, a1, a0
+; RV64ZBA-NEXT:    ld a0, 0(a1)
+; RV64ZBA-NEXT:    ld a1, 8(a1)
+; RV64ZBA-NEXT:    ret
+  %ptrdiff = lshr exact i64 %diff, 4
+  %cast = and i64 %ptrdiff, 4294967295
+  %ptr = getelementptr inbounds i128, i128* %baseptr, i64 %cast
+  %res = load i128, i128* %ptr
+  ret i128 %res
+}
+
 define signext i32 @srliw_2_sh2add(i32* %0, i32 signext %1) {
 ; RV64I-LABEL: srliw_2_sh2add:
 ; RV64I:       # %bb.0: