[AArch64][SVE2p1] Add SVE2.1 fclamp intrinsic

[david-arm]

Summary

Adds an intrinsic for the following instruction:

• fclamp

[kmclaughlin]

llvm/test/CodeGen/AArch64/sve2p1-intrinsics-fclamp.ll

7

I think this should be f16? Similarly, the tests below use i32 and i64 in the names.

[david-arm]

• Fixed the silly test names

[david-arm]

llvm/test/CodeGen/AArch64/sve2p1-intrinsics-fclamp.ll

7

Doh! :face_palm

llvm/include/llvm/IR/IntrinsicsAArch64.td

let TargetPrefix = "aarch64" in {
   def int_aarch64_sme_za_disable
       : DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>;
 
   // Clamp
   //
 
   def int_aarch64_sve_sclamp : AdvSIMD_3VectorArg_Intrinsic;
   def int_aarch64_sve_uclamp : AdvSIMD_3VectorArg_Intrinsic;
+  def int_aarch64_sve_fclamp : AdvSIMD_3VectorArg_Intrinsic;
+
 
   //
   // Reversal
   //
 
   def int_aarch64_sve_revd : AdvSIMD_Merged1VectorArg_Intrinsic;
 
   //

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

 defm PSEL_PPPRI : sve2_int_perm_sel_p<"psel", int_aarch64_sve_psel>;
 } // End HasSVE2p1_or_HasSME
 
 //===----------------------------------------------------------------------===//
 // SME2 or SVE2.1 instructions
 //===----------------------------------------------------------------------===//
 
 let Predicates = [HasSVE2p1_or_HasSME2] in {
-defm FCLAMP_ZZZ : sve2p1_fclamp<"fclamp">;
+defm FCLAMP_ZZZ : sve2p1_fclamp<"fclamp", int_aarch64_sve_fclamp>;
 def FDOT_ZZZ_S  : sve_float_dot<0b0, "fdot">;
 def FDOT_ZZZI_S : sve_float_dot_indexed<0b0, "fdot">;
 def BFMLSLB_ZZZ_S : sve2_fp_mla_long<0b110, "bfmlslb">;
 def BFMLSLT_ZZZ_S : sve2_fp_mla_long<0b111, "bfmlslt">;
 def BFMLSLB_ZZZI_S : sve2_fp_mla_long_by_indexed_elem<0b110, "bfmlslb">;
 def BFMLSLT_ZZZI_S : sve2_fp_mla_long_by_indexed_elem<0b111, "bfmlslt">;
 
 def SDOT_ZZZ_HtoS  : sve2p1_two_way_dot_vv<"sdot", 0b0>;

llvm/lib/Target/AArch64/SVEInstrFormats.td

class sve2p1_fclamp<string asm, bits<2> sz, ZPRRegOp zpr_ty>
   let Inst{9-5}   = Zn;
   let Inst{4-0}   = Zd;
 
   let Constraints = "$Zd = $_Zd";
   let DestructiveInstType = DestructiveOther;
   let ElementSize = zpr_ty.ElementSize;
 }
 
-multiclass sve2p1_fclamp<string asm> {
+multiclass sve2p1_fclamp<string asm, SDPatternOperator op> {
   def _H : sve2p1_fclamp<asm, 0b01, ZPR16>;
   def _S : sve2p1_fclamp<asm, 0b10, ZPR32>;
   def _D : sve2p1_fclamp<asm, 0b11, ZPR64>;
+
+  def : SVE_3_Op_Pat<nxv8f16, op, nxv8f16, nxv8f16, nxv8f16, !cast<Instruction>(NAME # _H)>;
+  def : SVE_3_Op_Pat<nxv4f32, op, nxv4f32, nxv4f32, nxv4f32, !cast<Instruction>(NAME # _S)>;
+  def : SVE_3_Op_Pat<nxv2f64, op, nxv2f64, nxv2f64, nxv2f64, !cast<Instruction>(NAME # _D)>;
 }
 
 // SVE two-way dot product
 class sve2p1_two_way_dot_vv<string mnemonic, bit u>
     : I<(outs ZPR32:$Zda), (ins ZPR32:$_Zda, ZPR16:$Zn, ZPR16:$Zm),
         mnemonic, "\t$Zda, $Zn, $Zm",
         "", []>, Sched<[]> {
   bits<5> Zda;

llvm/test/CodeGen/AArch64/sve2p1-intrinsics-fclamp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -verify-machineinstrs < %s | FileCheck %s
+
+target triple = "aarch64-linux-gnu"
+
+define <vscale x 8 x half> @test_fclamp_f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) #0 {
+; CHECK-LABEL: test_fclamp_f16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fclamp z0.h, z1.h, z2.h
+; CHECK-NEXT:    ret
+  %res = call <vscale x 8 x half> @llvm.aarch64.sve.fclamp.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c)
+  ret <vscale x 8 x half> %res
+}
+
+define <vscale x 4 x float> @test_fclamp_f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) #0 {
+; CHECK-LABEL: test_fclamp_f32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fclamp z0.s, z1.s, z2.s
+; CHECK-NEXT:    ret
+  %res = call <vscale x 4 x float> @llvm.aarch64.sve.fclamp.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c)
+  ret <vscale x 4 x float> %res
+}
+
+define <vscale x 2 x double> @test_fclamp_f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) #0 {
+; CHECK-LABEL: test_fclamp_f64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    fclamp z0.d, z1.d, z2.d
+; CHECK-NEXT:    ret
+  %res = call <vscale x 2 x double> @llvm.aarch64.sve.fclamp.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c)
+  ret <vscale x 2 x double> %res
+}
+
+attributes #0 = { "target-features"="+sve2p1" }
+
+declare <vscale x 8 x half> @llvm.aarch64.sve.fclamp.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 4 x float> @llvm.aarch64.sve.fclamp.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 2 x double> @llvm.aarch64.sve.fclamp.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)