[LoongArch] Add codegen support for fcopysign

llvm/lib/Target/LoongArch/LoongArchFloat32InstrInfo.td

 let Predicates = [HasBasicF] in {
 
 /// Float arithmetic operations
 
 def : PatFprFpr<fadd, FADD_S, FPR32>;
 def : PatFprFpr<fsub, FSUB_S, FPR32>;
 def : PatFprFpr<fmul, FMUL_S, FPR32>;
 def : PatFprFpr<fdiv, FDIV_S, FPR32>;
+def : PatFprFpr<fcopysign, FCOPYSIGN_S, FPR32>;
 def : PatFpr<fneg, FNEG_S, FPR32>;
 def : PatFpr<fabs, FABS_S, FPR32>;
 def : PatFpr<fsqrt, FSQRT_S, FPR32>;
 
 def : Pat<(fdiv fpimm1, (fsqrt FPR32:$fj)), (FRSQRT_S FPR32:$fj)>;
 
 /// Setcc
 

llvm/lib/Target/LoongArch/LoongArchFloat64InstrInfo.td

 let Predicates = [HasBasicD] in {
 
 /// Float arithmetic operations
 
 def : PatFprFpr<fadd, FADD_D, FPR64>;
 def : PatFprFpr<fsub, FSUB_D, FPR64>;
 def : PatFprFpr<fmul, FMUL_D, FPR64>;
 def : PatFprFpr<fdiv, FDIV_D, FPR64>;
+def : PatFprFpr<fcopysign, FCOPYSIGN_D, FPR64>;
 def : PatFpr<fneg, FNEG_D, FPR64>;
 def : PatFpr<fabs, FABS_D, FPR64>;
 def : PatFpr<fsqrt, FSQRT_D, FPR64>;
 
 def : Pat<(fdiv fpimm1, (fsqrt FPR64:$fj)), (FRSQRT_D FPR64:$fj)>;
 
+def : Pat<(fcopysign FPR64:$fj, FPR32:$fk),
+          (FCOPYSIGN_D FPR64:$fj, (FCVT_D_S FPR32:$fk))>;
+def : Pat<(fcopysign FPR32:$fj, FPR64:$fk),
+          (FCOPYSIGN_S FPR32:$fj, (FCVT_S_D FPR64:$fk))>;
+
 /// Setcc
 
 // Match non-signaling comparison
 
 // TODO: Change setcc to any_fsetcc after call is supported because
 // we need to call llvm.experimental.constrained.fcmp.f64 in testcase.
 // See RISCV float-fcmp-strict.ll for reference.
 

llvm/test/CodeGen/LoongArch/fcopysign.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch32 --mattr=+f,-d < %s | FileCheck %s --check-prefix=LA32F
+; RUN: llc --mtriple=loongarch32 --mattr=+d < %s | FileCheck %s --check-prefix=LA32D
+; RUN: llc --mtriple=loongarch64 --mattr=+f,-d < %s | FileCheck %s --check-prefix=LA64F
+; RUN: llc --mtriple=loongarch64 --mattr=+d < %s | FileCheck %s --check-prefix=LA64D
+
+declare float @llvm.copysign.f32(float, float)
+declare double @llvm.copysign.f64(double, double)
+
+define float @fcopysign_s(float %a, float %b) nounwind {
+; LA32F-LABEL: fcopysign_s:
+; LA32F:       # %bb.0:
+; LA32F-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA32F-NEXT:    ret
+;
+; LA32D-LABEL: fcopysign_s:
+; LA32D:       # %bb.0:
+; LA32D-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA32D-NEXT:    ret
+;
+; LA64F-LABEL: fcopysign_s:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: fcopysign_s:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA64D-NEXT:    ret
+  %1 = call float @llvm.copysign.f32(float %a, float %b)
+  ret float %1
+}
+
+define double @fcopysign_d(double %a, double %b) nounwind {
+; LA32F-LABEL: fcopysign_d:
+; LA32F:       # %bb.0:
+; LA32F-NEXT:    srli.w $a2, $a3, 31
+; LA32F-NEXT:    bstrins.w $a1, $a2, 31, 31
+; LA32F-NEXT:    ret
+;
+; LA32D-LABEL: fcopysign_d:
+; LA32D:       # %bb.0:
+; LA32D-NEXT:    fcopysign.d $fa0, $fa0, $fa1
+; LA32D-NEXT:    ret
+;
+; LA64F-LABEL: fcopysign_d:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    srli.d $a1, $a1, 63
+; LA64F-NEXT:    bstrins.d $a0, $a1, 63, 63
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: fcopysign_d:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fcopysign.d $fa0, $fa0, $fa1
+; LA64D-NEXT:    ret
+  %1 = call double @llvm.copysign.f64(double %a, double %b)
+  ret double %1
+}
+
+define double @fold_promote_d_s(double %a, float %b) nounwind {
+; LA32F-LABEL: fold_promote_d_s:
+; LA32F:       # %bb.0:
+; LA32F-NEXT:    movfr2gr.s $a2, $fa0
+; LA32F-NEXT:    srli.w $a2, $a2, 31
+; LA32F-NEXT:    bstrins.w $a1, $a2, 31, 31
+; LA32F-NEXT:    ret
+;
+; LA32D-LABEL: fold_promote_d_s:
+; LA32D:       # %bb.0:
+; LA32D-NEXT:    fcvt.d.s $fa1, $fa1
+; LA32D-NEXT:    fcopysign.d $fa0, $fa0, $fa1
+; LA32D-NEXT:    ret
+;
+; LA64F-LABEL: fold_promote_d_s:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    lu12i.w $a1, -524288
+; LA64F-NEXT:    lu32i.d $a1, 0
+; LA64F-NEXT:    movfr2gr.s $a2, $fa0
+; LA64F-NEXT:    and $a1, $a2, $a1
+; LA64F-NEXT:    slli.d $a1, $a1, 32
+; LA64F-NEXT:    bstrins.d $a1, $a0, 62, 0
+; LA64F-NEXT:    move $a0, $a1
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: fold_promote_d_s:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fcvt.d.s $fa1, $fa1
+; LA64D-NEXT:    fcopysign.d $fa0, $fa0, $fa1
+; LA64D-NEXT:    ret
+  %c = fpext float %b to double
+  %t = call double @llvm.copysign.f64(double %a, double %c)
+  ret double %t
+}
+
+define float @fold_demote_s_d(float %a, double %b) nounwind {
+; LA32F-LABEL: fold_demote_s_d:
+; LA32F:       # %bb.0:
+; LA32F-NEXT:    movgr2fr.w $fa1, $a1
+; LA32F-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA32F-NEXT:    ret
+;
+; LA32D-LABEL: fold_demote_s_d:
+; LA32D:       # %bb.0:
+; LA32D-NEXT:    fcvt.s.d $fa1, $fa1
+; LA32D-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA32D-NEXT:    ret
+;
+; LA64F-LABEL: fold_demote_s_d:
+; LA64F:       # %bb.0:
+; LA64F-NEXT:    srli.d $a0, $a0, 32
+; LA64F-NEXT:    movgr2fr.w $fa1, $a0
+; LA64F-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA64F-NEXT:    ret
+;
+; LA64D-LABEL: fold_demote_s_d:
+; LA64D:       # %bb.0:
+; LA64D-NEXT:    fcvt.s.d $fa1, $fa1
+; LA64D-NEXT:    fcopysign.s $fa0, $fa0, $fa1
+; LA64D-NEXT:    ret
+  %c = fptrunc double %b to float
+  %t = call float @llvm.copysign.f32(float %a, float %c)
+  ret float %t
+}