Skip to content

[LoongArch] Support llvm.lround intrinsics with i32 return type. #114733

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Nov 6, 2024

Conversation

ylzsx
Copy link
Contributor

@ylzsx ylzsx commented Nov 4, 2024

This is needed by flang, similar to RISCV-64 in https://reviews.llvm.org/D147195.

@llvmbot
Copy link
Member

llvmbot commented Nov 4, 2024

@llvm/pr-subscribers-backend-loongarch

Author: Zhaoxin Yang (ylzsx)

Changes

This is needed by flang, similar to RISCV-64 in D147195.


Full diff: https://github.com/llvm/llvm-project/pull/114733.diff

3 Files Affected:

  • (modified) llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp (+13)
  • (added) llvm/test/CodeGen/LoongArch/double-lround.ll (+20)
  • (added) llvm/test/CodeGen/LoongArch/float-lround.ll (+20)
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index 6bee00d1ce3823..7bea9bdcfb8c1a 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -143,6 +143,7 @@ LoongArchTargetLowering::LoongArchTargetLowering(const TargetMachine &TM,
     setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
     setOperationAction(ISD::BSWAP, MVT::i32, Custom);
     setOperationAction({ISD::UDIV, ISD::UREM}, MVT::i32, Custom);
+    setOperationAction(ISD::LROUND, MVT::i32, Custom);
   }
 
   // Set operations for LA32 only.
@@ -3103,6 +3104,18 @@ void LoongArchTargetLowering::ReplaceNodeResults(
     replaceINTRINSIC_WO_CHAINResults(N, Results, DAG, Subtarget);
     break;
   }
+  case ISD::LROUND: {
+    SDValue Op0 = N->getOperand(0);
+    RTLIB::Libcall LC =
+        Op0.getValueType() == MVT::f64 ? RTLIB::LROUND_F64 : RTLIB::LROUND_F32;
+    MakeLibCallOptions CallOptions;
+    EVT OpVT = Op0.getValueType();
+    CallOptions.setTypeListBeforeSoften(OpVT, MVT::i64, true);
+    SDValue Result = makeLibCall(DAG, LC, MVT::i64, Op0, CallOptions, DL).first;
+    Result = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Result);
+    Results.push_back(Result);
+    break;
+  }
   }
 }
 
diff --git a/llvm/test/CodeGen/LoongArch/double-lround.ll b/llvm/test/CodeGen/LoongArch/double-lround.ll
new file mode 100644
index 00000000000000..e90a7ce29e5a0b
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/double-lround.ll
@@ -0,0 +1,20 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch64 --mattr=-f,-d %s -o - \
+; RUN:   | FileCheck -check-prefix=LA64S %s
+
+declare i32 @llvm.lround.i32.f64(double)
+
+;; We support lround with i32 as return type on LoongArch64. This is needed by flang.
+define i32 @lround_i32_f64(double %a) nounwind {
+; LA64S-LABEL: lround_i32_f64:
+; LA64S:       # %bb.0:
+; LA64S-NEXT:    addi.d $sp, $sp, -16
+; LA64S-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64S-NEXT:    bl %plt(lround)
+; LA64S-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64S-NEXT:    addi.d $sp, $sp, 16
+; LA64S-NEXT:    ret
+
+  %1 = call i32 @llvm.lround.i32.f64(double %a)
+  ret i32 %1
+}
diff --git a/llvm/test/CodeGen/LoongArch/float-lround.ll b/llvm/test/CodeGen/LoongArch/float-lround.ll
new file mode 100644
index 00000000000000..a17ba4e3e5d926
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/float-lround.ll
@@ -0,0 +1,20 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch64 --mattr=-f,-d %s -o - \
+; RUN:   | FileCheck -check-prefix=LA64S %s
+
+declare i32 @llvm.lround.i32.f32(float)
+
+; We support lround with i32 as return type on LoongArch64. This is needed by flang.
+define i32 @lround_i32_f32(float %a) nounwind {
+; LA64S-LABEL: lround_i32_f32:
+; LA64S:       # %bb.0:
+; LA64S-NEXT:    addi.d $sp, $sp, -16
+; LA64S-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64S-NEXT:    bl %plt(lroundf)
+; LA64S-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64S-NEXT:    addi.d $sp, $sp, 16
+; LA64S-NEXT:    ret
+
+  %1 = call i32 @llvm.lround.i32.f32(float %a)
+  ret i32 %1
+}

This is needed by flang, similar to RISCV-64 in D147195.
@SixWeining SixWeining merged commit 8c565de into llvm:main Nov 6, 2024
6 of 7 checks passed
@ylzsx ylzsx deleted the pr2 branch November 7, 2024 00:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants