[AArch64][SVE] Add reg+imm addressing mode for unpredicated stores

[efriedma]

LGTM

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

multiclass sve_prefetch<SDPatternOperator prefetch, ValueType PredTy, Instruction RegImmInst, Instruction RegRegInst, int scale, ComplexPattern AddrCP> {
 
   defm : pred_store<nxv16i8, nxv16i1, non_temporal_store, STNT1B_ZRR, STNT1B_ZRI, am_sve_regreg_lsl0>;
   defm : pred_store<nxv8i16, nxv8i1,  non_temporal_store, STNT1H_ZRR, STNT1H_ZRI, am_sve_regreg_lsl1>;
   defm : pred_store<nxv4i32, nxv4i1,  non_temporal_store, STNT1W_ZRR, STNT1W_ZRI, am_sve_regreg_lsl2>;
   defm : pred_store<nxv2i64, nxv2i1,  non_temporal_store, STNT1D_ZRR, STNT1D_ZRI, am_sve_regreg_lsl3>;
 
   multiclass unpred_store<PatFrag Store, ValueType Ty, Instruction RegImmInst,
                           Instruction PTrue> {
-    def : Pat<(Store (Ty ZPR:$val), (am_sve_fi GPR64sp:$base, simm4s1:$offset)),
+    let AddedComplexity = 1 in {
+      def _imm : Pat<(Store (Ty ZPR:$val), (am_sve_indexed_s4 GPR64sp:$base, simm4s1:$offset)),
                      (RegImmInst ZPR:$val, (PTrue 31), GPR64sp:$base, simm4s1:$offset)>;
+    }
+    let AddedComplexity = 2 in {
+      def _fi : Pat<(Store (Ty ZPR:$val), (am_sve_fi GPR64sp:$base, simm4s1:$offset)),
+                    (RegImmInst ZPR:$val, (PTrue 31), GPR64sp:$base, simm4s1:$offset)>;
+    }
 
     def : Pat<(Store (Ty ZPR:$val), GPR64:$base),
               (RegImmInst ZPR:$val, (PTrue 31), GPR64:$base, (i64 0))>;
   }
 
   defm : unpred_store<         store, nxv16i8,   ST1B_IMM, PTRUE_B>;
   defm : unpred_store< truncstorevi8, nxv8i16, ST1B_H_IMM, PTRUE_H>;
   defm : unpred_store< truncstorevi8, nxv4i32, ST1B_S_IMM, PTRUE_S>;

llvm/test/CodeGen/AArch64/sve-st1-addressing-mode-reg-imm.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
+
+; ST1B
+
+define void @st1b_lower_bound(<vscale x 16 x i8> %data, <vscale x 16 x i8>* %a) {
+; CHECK-LABEL: st1b_lower_bound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    st1b { z0.b }, p0, [x0, #-8, mul vl]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %a, i64 -8
+  store <vscale x 16 x i8> %data, <vscale x 16 x i8>* %base
+  ret void
+}
+
+define void @st1b_inbound(<vscale x 16 x i8> %data, <vscale x 16 x i8>* %a) {
+; CHECK-LABEL: st1b_inbound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    st1b { z0.b }, p0, [x0, #1, mul vl]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %a, i64 1
+  store <vscale x 16 x i8> %data, <vscale x 16 x i8>* %base
+  ret void
+}
+
+define void @st1b_upper_bound(<vscale x 16 x i8> %data, <vscale x 16 x i8>* %a) {
+; CHECK-LABEL: st1b_upper_bound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    st1b { z0.b }, p0, [x0, #7, mul vl]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %a, i64 7
+  store <vscale x 16 x i8> %data, <vscale x 16 x i8>* %base
+  ret void
+}
+
+define void @st1b_out_of_upper_bound(<vscale x 16 x i8> %data, <vscale x 16 x i8>* %a) {
+; CHECK-LABEL: st1b_out_of_upper_bound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    rdvl x8, #8
+; CHECK-NEXT:    add x8, x0, x8
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    st1b { z0.b }, p0, [x8]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %a, i64 8
+  store <vscale x 16 x i8> %data, <vscale x 16 x i8>* %base
+  ret void
+}
+
+define void @st1b_out_of_lower_bound(<vscale x 16 x i8> %data, <vscale x 16 x i8>* %a) {
+; CHECK-LABEL: st1b_out_of_lower_bound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    rdvl x8, #-9
+; CHECK-NEXT:    add x8, x0, x8
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    st1b { z0.b }, p0, [x8]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %a, i64 -9
+  store <vscale x 16 x i8> %data, <vscale x 16 x i8>* %base
+  ret void
+}
+
+; ST1H
+
+define void @st1h_inbound(<vscale x 8 x i16> %data, <vscale x 8 x i16>* %a) {
+; CHECK-LABEL: st1h_inbound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    st1h { z0.h }, p0, [x0, #-6, mul vl]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %a, i64 -6
+  store <vscale x 8 x i16> %data, <vscale x 8 x i16>* %base
+  ret void
+}
+
+; ST1W
+
+define void @st1w_inbound(<vscale x 4 x i32> %data, <vscale x 4 x i32>* %a) {
+; CHECK-LABEL: st1w_inbound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    st1w { z0.s }, p0, [x0, #2, mul vl]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %a, i64 2
+  store <vscale x 4 x i32> %data, <vscale x 4 x i32>* %base
+  ret void
+}
+
+; ST1D
+
+define void @st1d_inbound(<vscale x 2 x i64> %data, <vscale x 2 x i64>* %a) {
+; CHECK-LABEL: st1d_inbound:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    st1d { z0.d }, p0, [x0, #5, mul vl]
+; CHECK-NEXT:    ret
+  %base = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %a, i64 5
+  store <vscale x 2 x i64> %data, <vscale x 2 x i64>* %base
+  ret void
+}