From dc3c5a78f20f5d5e075453754d74021844e780a5 Mon Sep 17 00:00:00 2001
From: gonglingqin <gonglingqin@loongson.cn>
Date: Thu, 29 Sep 2022 10:05:58 +0800
Subject: [PATCH] [LoongArch] Add fp_to_sint support for soft floating point

Differential Revision: https://reviews.llvm.org/D134692
---
 .../lib/Target/LoongArch/LoongArchISelLowering.cpp |  20 ++-
 llvm/test/CodeGen/LoongArch/soft-fp-to-int.ll      | 155 +++++++++++++++++++++
 2 files changed, 173 insertions(+), 2 deletions(-)
 create mode 100644 llvm/test/CodeGen/LoongArch/soft-fp-to-int.ll

diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index e157ed8..9c443ba 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -544,8 +544,24 @@ void LoongArchTargetLowering::ReplaceNodeResults(
            "Unexpected custom legalisation");
     SDValue Src = N->getOperand(0);
     EVT VT = EVT::getFloatingPointVT(N->getValueSizeInBits(0));
-    SDValue Dst = DAG.getNode(LoongArchISD::FTINT, DL, VT, Src);
-    Results.push_back(DAG.getNode(ISD::BITCAST, DL, N->getValueType(0), Dst));
+    if (getTypeAction(*DAG.getContext(), Src.getValueType()) !=
+        TargetLowering::TypeSoftenFloat) {
+      SDValue Dst = DAG.getNode(LoongArchISD::FTINT, DL, VT, Src);
+      Results.push_back(DAG.getNode(ISD::BITCAST, DL, N->getValueType(0), Dst));
+      return;
+    }
+    // If the FP type needs to be softened, emit a library call using the 'si'
+    // version. If we left it to default legalization we'd end up with 'di'.
+    RTLIB::Libcall LC;
+    LC = RTLIB::getFPTOSINT(Src.getValueType(), N->getValueType(0));
+    MakeLibCallOptions CallOptions;
+    EVT OpVT = Src.getValueType();
+    CallOptions.setTypeListBeforeSoften(OpVT, N->getValueType(0), true);
+    SDValue Chain = SDValue();
+    SDValue Result;
+    std::tie(Result, Chain) =
+        makeLibCall(DAG, LC, N->getValueType(0), Src, CallOptions, DL, Chain);
+    Results.push_back(Result);
     break;
   }
   case ISD::BITCAST: {
diff --git a/llvm/test/CodeGen/LoongArch/soft-fp-to-int.ll b/llvm/test/CodeGen/LoongArch/soft-fp-to-int.ll
new file mode 100644
index 0000000..d12cbaa
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/soft-fp-to-int.ll
@@ -0,0 +1,155 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch32 < %s | FileCheck %s --check-prefix=LA32
+; RUN: llc --mtriple=loongarch64 < %s | FileCheck %s --check-prefix=LA64
+
+define i32 @fptosi_i32_fp128(fp128 %X) nounwind {
+; LA32-LABEL: fptosi_i32_fp128:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -32
+; LA32-NEXT:    st.w $ra, $sp, 28 # 4-byte Folded Spill
+; LA32-NEXT:    ld.w $a1, $a0, 12
+; LA32-NEXT:    st.w $a1, $sp, 20
+; LA32-NEXT:    ld.w $a1, $a0, 8
+; LA32-NEXT:    st.w $a1, $sp, 16
+; LA32-NEXT:    ld.w $a1, $a0, 4
+; LA32-NEXT:    st.w $a1, $sp, 12
+; LA32-NEXT:    ld.w $a0, $a0, 0
+; LA32-NEXT:    st.w $a0, $sp, 8
+; LA32-NEXT:    addi.w $a0, $sp, 8
+; LA32-NEXT:    bl %plt(__fixtfsi)
+; LA32-NEXT:    ld.w $ra, $sp, 28 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 32
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: fptosi_i32_fp128:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__fixtfsi)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %tmp = fptosi fp128 %X to i32
+  ret i32 %tmp
+}
+
+define i32 @fptosi_i32_double(double %X) nounwind {
+; LA32-LABEL: fptosi_i32_double:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__fixdfsi)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: fptosi_i32_double:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__fixdfsi)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %tmp = fptosi double %X to i32
+  ret i32 %tmp
+}
+
+define i32 @fptosi_i32_float(float %X) nounwind {
+; LA32-LABEL: fptosi_i32_float:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__fixsfsi)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: fptosi_i32_float:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bstrpick.d $a0, $a0, 31, 0
+; LA64-NEXT:    bl %plt(__fixsfsi)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %tmp = fptosi float %X to i32
+  ret i32 %tmp
+}
+
+define i64 @fptosi_i64_fp128(fp128 %X) nounwind {
+; LA32-LABEL: fptosi_i64_fp128:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -32
+; LA32-NEXT:    st.w $ra, $sp, 28 # 4-byte Folded Spill
+; LA32-NEXT:    ld.w $a1, $a0, 12
+; LA32-NEXT:    st.w $a1, $sp, 12
+; LA32-NEXT:    ld.w $a1, $a0, 8
+; LA32-NEXT:    st.w $a1, $sp, 8
+; LA32-NEXT:    ld.w $a1, $a0, 4
+; LA32-NEXT:    st.w $a1, $sp, 4
+; LA32-NEXT:    ld.w $a0, $a0, 0
+; LA32-NEXT:    st.w $a0, $sp, 0
+; LA32-NEXT:    addi.w $a0, $sp, 0
+; LA32-NEXT:    bl %plt(__fixtfdi)
+; LA32-NEXT:    ld.w $ra, $sp, 28 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 32
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: fptosi_i64_fp128:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__fixtfdi)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %tmp = fptosi fp128 %X to i64
+  ret i64 %tmp
+}
+
+define i64 @fptosi_i64_double(double %X) nounwind {
+; LA32-LABEL: fptosi_i64_double:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__fixdfdi)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: fptosi_i64_double:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__fixdfdi)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %tmp = fptosi double %X to i64
+  ret i64 %tmp
+}
+
+define i64 @fptosi_i64_float(float %X) nounwind {
+; LA32-LABEL: fptosi_i64_float:
+; LA32:       # %bb.0:
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__fixsfdi)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: fptosi_i64_float:
+; LA64:       # %bb.0:
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bstrpick.d $a0, $a0, 31, 0
+; LA64-NEXT:    bl %plt(__fixsfdi)
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+  %tmp = fptosi float %X to i64
+  ret i64 %tmp
+}
-- 
2.7.4