[AArch64][SVE] Add addsub carry long instrinsics

[dancgr]

Summary

Add intrinsics for the following instructions:

• adclb, adclt, sbclb, sbclt

[kmclaughlin]

Thanks @dancgr, LGTM!

llvm/include/llvm/IR/IntrinsicsAArch64.td

 
 def int_aarch64_sve_sqdmlalb    : SVE2_3VectorArg_Long_Intrinsic;
 def int_aarch64_sve_sqdmlalt    : SVE2_3VectorArg_Long_Intrinsic;
 def int_aarch64_sve_sqdmlslb    : SVE2_3VectorArg_Long_Intrinsic;
 def int_aarch64_sve_sqdmlslt    : SVE2_3VectorArg_Long_Intrinsic;
 def int_aarch64_sve_sqdmlalbt   : SVE2_3VectorArg_Long_Intrinsic;
 def int_aarch64_sve_sqdmlslbt   : SVE2_3VectorArg_Long_Intrinsic;
 
+// SVE2 ADDSUB Long Unpredicated.
+def int_aarch64_sve_adclb       : AdvSIMD_3VectorArg_Intrinsic;
+def int_aarch64_sve_adclt       : AdvSIMD_3VectorArg_Intrinsic;
+def int_aarch64_sve_sbclb       : AdvSIMD_3VectorArg_Intrinsic;
+def int_aarch64_sve_sbclt       : AdvSIMD_3VectorArg_Intrinsic;
 
 }

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

let Predicates = [HasSVE2] in {
 
   // SVE2 integer absolute difference and accumulate long
   defm SABALB_ZZZ : sve2_int_absdiff_accum_long<0b00, "sabalb", int_aarch64_sve_sabalb>;
   defm SABALT_ZZZ : sve2_int_absdiff_accum_long<0b01, "sabalt", int_aarch64_sve_sabalt>;
   defm UABALB_ZZZ : sve2_int_absdiff_accum_long<0b10, "uabalb", int_aarch64_sve_uabalb>;
   defm UABALT_ZZZ : sve2_int_absdiff_accum_long<0b11, "uabalt", int_aarch64_sve_uabalt>;
 
   // SVE2 integer add/subtract long with carry
-  defm ADCLB_ZZZ : sve2_int_addsub_long_carry<0b00, "adclb">;
-  defm ADCLT_ZZZ : sve2_int_addsub_long_carry<0b01, "adclt">;
-  defm SBCLB_ZZZ : sve2_int_addsub_long_carry<0b10, "sbclb">;
-  defm SBCLT_ZZZ : sve2_int_addsub_long_carry<0b11, "sbclt">;
+  defm ADCLB_ZZZ : sve2_int_addsub_long_carry<0b00, "adclb", int_aarch64_sve_adclb>;
+  defm ADCLT_ZZZ : sve2_int_addsub_long_carry<0b01, "adclt", int_aarch64_sve_adclt>;
+  defm SBCLB_ZZZ : sve2_int_addsub_long_carry<0b10, "sbclb", int_aarch64_sve_sbclb>;
+  defm SBCLT_ZZZ : sve2_int_addsub_long_carry<0b11, "sbclt", int_aarch64_sve_sbclt>;
 
   // SVE2 bitwise shift right narrow (bottom)
   defm SQSHRUNB_ZZI  : sve2_int_bin_shift_imm_right_narrow_bottom<0b000, "sqshrunb",  int_aarch64_sve_sqshrunb>;
   defm SQRSHRUNB_ZZI : sve2_int_bin_shift_imm_right_narrow_bottom<0b001, "sqrshrunb", int_aarch64_sve_sqrshrunb>;
   defm SHRNB_ZZI     : sve2_int_bin_shift_imm_right_narrow_bottom<0b010, "shrnb",     int_aarch64_sve_shrnb>;
   defm RSHRNB_ZZI    : sve2_int_bin_shift_imm_right_narrow_bottom<0b011, "rshrnb",    int_aarch64_sve_rshrnb>;
   defm SQSHRNB_ZZI   : sve2_int_bin_shift_imm_right_narrow_bottom<0b100, "sqshrnb",   int_aarch64_sve_sqshrnb>;
   defm SQRSHRNB_ZZI  : sve2_int_bin_shift_imm_right_narrow_bottom<0b101, "sqrshrnb",  int_aarch64_sve_sqrshrnb>;

llvm/lib/Target/AArch64/SVEInstrFormats.td

multiclass sve2_int_absdiff_accum_long<bits<2> opc, string asm,
   def _S : sve2_int_absdiff_accum<0b10, { 0b00, opc }, asm, ZPR32, ZPR16>;
   def _D : sve2_int_absdiff_accum<0b11, { 0b00, opc }, asm, ZPR64, ZPR32>;
 
   def : SVE_3_Op_Pat<nxv8i16, op, nxv8i16, nxv16i8, nxv16i8, !cast<Instruction>(NAME # _H)>;
   def : SVE_3_Op_Pat<nxv4i32, op, nxv4i32, nxv8i16, nxv8i16, !cast<Instruction>(NAME # _S)>;
   def : SVE_3_Op_Pat<nxv2i64, op, nxv2i64, nxv4i32, nxv4i32, !cast<Instruction>(NAME # _D)>;
 }
 
-multiclass sve2_int_addsub_long_carry<bits<2> opc, string asm> {
+multiclass sve2_int_addsub_long_carry<bits<2> opc, string asm, SDPatternOperator op> {
   def _S : sve2_int_absdiff_accum<{ opc{1}, 0b0 }, { 0b010, opc{0} }, asm,
                                   ZPR32, ZPR32>;
   def _D : sve2_int_absdiff_accum<{ opc{1}, 0b1 }, { 0b010, opc{0} }, asm,
                                   ZPR64, ZPR64>;
+
+  def : SVE_3_Op_Pat<nxv4i32, op, nxv4i32, nxv4i32, nxv4i32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_3_Op_Pat<nxv2i64, op, nxv2i64, nxv2i64, nxv2i64, !cast<Instruction>(NAME # _D)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SVE2 Narrowing Group
 //===----------------------------------------------------------------------===//
 
 class sve2_int_bin_shift_imm_narrow_bottom<bits<3> tsz8_64, bits<3> opc,
                                            string asm, ZPRRegOp zprty1,

llvm/test/CodeGen/AArch64/sve2-int-addsub-long.ll

+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s | FileCheck %s
+
+;
+; ADCLB (vector, long, unpredicated)
+;
+define <vscale x 4 x i32> @adclb_i32(<vscale x 4 x i32> %a,
+                                     <vscale x 4 x i32> %b,
+                                     <vscale x 4 x i32> %c) {
+; CHECK-LABEL: adclb_i32
+; CHECK: adclb z0.s, z1.s, z2.s
+; CHECK-NEXT: ret
+  %res = call <vscale x 4 x i32> @llvm.aarch64.sve.adclb.nxv4i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c)
+  ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 2 x i64> @adclb_i64(<vscale x 2 x i64> %a,
+                                     <vscale x 2 x i64> %b,
+                                     <vscale x 2 x i64> %c) {
+; CHECK-LABEL: adclb_i64
+; CHECK: adclb z0.d, z1.d, z2.d
+; CHECK-NEXT: ret
+  %res = call <vscale x 2 x i64> @llvm.aarch64.sve.adclb.nxv2i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c)
+  ret <vscale x 2 x i64> %res
+}
+
+;
+; ADCLT (vector, long, unpredicated)
+;
+define <vscale x 4 x i32> @adclt_i32(<vscale x 4 x i32> %a,
+                                     <vscale x 4 x i32> %b,
+                                     <vscale x 4 x i32> %c) {
+; CHECK-LABEL: adclt_i32
+; CHECK: adclt z0.s, z1.s, z2.s
+; CHECK-NEXT: ret
+  %res = call <vscale x 4 x i32> @llvm.aarch64.sve.adclt.nxv4i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c)
+  ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 2 x i64> @adclt_i64(<vscale x 2 x i64> %a,
+                                     <vscale x 2 x i64> %b,
+                                     <vscale x 2 x i64> %c) {
+; CHECK-LABEL: adclt_i64
+; CHECK: adclt z0.d, z1.d, z2.d
+; CHECK-NEXT: ret
+  %res = call <vscale x 2 x i64> @llvm.aarch64.sve.adclt.nxv2i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c)
+  ret <vscale x 2 x i64> %res
+}
+
+;
+; SBCLB (vector, long, unpredicated)
+;
+define <vscale x 4 x i32> @sbclb_i32(<vscale x 4 x i32> %a,
+                                     <vscale x 4 x i32> %b,
+                                     <vscale x 4 x i32> %c) {
+; CHECK-LABEL: sbclb_i32
+; CHECK: sbclb z0.s, z1.s, z2.s
+; CHECK-NEXT: ret
+  %res = call <vscale x 4 x i32> @llvm.aarch64.sve.sbclb.nxv4i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c)
+  ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 2 x i64> @sbclb_i64(<vscale x 2 x i64> %a,
+                                     <vscale x 2 x i64> %b,
+                                     <vscale x 2 x i64> %c) {
+; CHECK-LABEL: sbclb_i64
+; CHECK: sbclb z0.d, z1.d, z2.d
+; CHECK-NEXT: ret
+  %res = call <vscale x 2 x i64> @llvm.aarch64.sve.sbclb.nxv2i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c)
+  ret <vscale x 2 x i64> %res
+}
+
+;
+; SBCLT (vector, long, unpredicated)
+;
+define <vscale x 4 x i32> @sbclt_i32(<vscale x 4 x i32> %a,
+                                     <vscale x 4 x i32> %b,
+                                     <vscale x 4 x i32> %c) {
+; CHECK-LABEL: sbclt_i32
+; CHECK: sbclt z0.s, z1.s, z2.s
+; CHECK-NEXT: ret
+  %res = call <vscale x 4 x i32> @llvm.aarch64.sve.sbclt.nxv4i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c)
+  ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 2 x i64> @sbclt_i64(<vscale x 2 x i64> %a,
+                                     <vscale x 2 x i64> %b,
+                                     <vscale x 2 x i64> %c) {
+; CHECK-LABEL: sbclt_i64
+; CHECK: sbclt z0.d, z1.d, z2.d
+; CHECK-NEXT: ret
+  %res = call <vscale x 2 x i64> @llvm.aarch64.sve.sbclt.nxv2i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c)
+  ret <vscale x 2 x i64> %res
+}
+
+declare <vscale x 4 x i32> @llvm.aarch64.sve.adclb.nxv4i32(<vscale x 4 x i32>,<vscale x 4 x i32>,<vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.adclb.nxv2i64(<vscale x 2 x i64>,<vscale x 2 x i64>,<vscale x 2 x i64>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.adclt.nxv4i32(<vscale x 4 x i32>,<vscale x 4 x i32>,<vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.adclt.nxv2i64(<vscale x 2 x i64>,<vscale x 2 x i64>,<vscale x 2 x i64>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sbclb.nxv4i32(<vscale x 4 x i32>,<vscale x 4 x i32>,<vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sbclb.nxv2i64(<vscale x 2 x i64>,<vscale x 2 x i64>,<vscale x 2 x i64>)
+declare <vscale x 4 x i32> @llvm.aarch64.sve.sbclt.nxv4i32(<vscale x 4 x i32>,<vscale x 4 x i32>,<vscale x 4 x i32>)
+declare <vscale x 2 x i64> @llvm.aarch64.sve.sbclt.nxv2i64(<vscale x 2 x i64>,<vscale x 2 x i64>,<vscale x 2 x i64>)