llvm · SixWeining · Jan 23, 2024 · Jan 22, 2024
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -265,8 +265,10 @@ LoongArchTargetLowering::LoongArchTargetLowering(const TargetMachine &TM,
           {ISD::SETNE, ISD::SETGE, ISD::SETGT, ISD::SETUGE, ISD::SETUGT}, VT,
           Expand);
     }
-    setOperationAction({ISD::SINT_TO_FP, ISD::UINT_TO_FP},
-                       {MVT::v4i32, MVT::v2i64}, Legal);
+    for (MVT VT : {MVT::v4i32, MVT::v2i64}) {
+      setOperationAction({ISD::SINT_TO_FP, ISD::UINT_TO_FP}, VT, Legal);
+      setOperationAction({ISD::FP_TO_SINT, ISD::FP_TO_UINT}, VT, Legal);
+    }
     for (MVT VT : {MVT::v4f32, MVT::v2f64}) {
       setOperationAction({ISD::FADD, ISD::FSUB}, VT, Legal);
       setOperationAction({ISD::FMUL, ISD::FDIV}, VT, Legal);
@@ -309,8 +311,10 @@ LoongArchTargetLowering::LoongArchTargetLowering(const TargetMachine &TM,
           {ISD::SETNE, ISD::SETGE, ISD::SETGT, ISD::SETUGE, ISD::SETUGT}, VT,
           Expand);
     }
-    setOperationAction({ISD::SINT_TO_FP, ISD::UINT_TO_FP},
-                       {MVT::v8i32, MVT::v4i32, MVT::v4i64}, Legal);
+    for (MVT VT : {MVT::v8i32, MVT::v4i32, MVT::v4i64}) {
+      setOperationAction({ISD::SINT_TO_FP, ISD::UINT_TO_FP}, VT, Legal);
+      setOperationAction({ISD::FP_TO_SINT, ISD::FP_TO_UINT}, VT, Legal);
+    }
     for (MVT VT : {MVT::v8f32, MVT::v4f64}) {
       setOperationAction({ISD::FADD, ISD::FSUB}, VT, Legal);
       setOperationAction({ISD::FMUL, ISD::FDIV}, VT, Legal);

diff --git a/llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td
@@ -1637,6 +1637,28 @@ def : Pat<(v4f32 (uint_to_fp v4i64:$vj)),
                                        (XVFFINT_D_LU v4i64:$vj)),
                           sub_128)>;
 
+// XVFTINTRZ_{W_S/L_D}
+def : Pat<(v8i32 (fp_to_sint v8f32:$vj)), (XVFTINTRZ_W_S v8f32:$vj)>;
+def : Pat<(v4i64 (fp_to_sint v4f64:$vj)), (XVFTINTRZ_L_D v4f64:$vj)>;
+def : Pat<(v4i64 (fp_to_sint v4f32:$vj)),
+          (VEXT2XV_D_W (SUBREG_TO_REG (i64 0), (VFTINTRZ_W_S v4f32:$vj),
+                                      sub_128))>;
+def : Pat<(v4i32 (fp_to_sint (v4f64 LASX256:$vj))),
+          (EXTRACT_SUBREG (XVFTINTRZ_W_S (XVFCVT_S_D (XVPERMI_D v4f64:$vj, 238),
+                                                     v4f64:$vj)),
+                          sub_128)>;
+
+// XVFTINTRZ_{W_SU/L_DU}
+def : Pat<(v8i32 (fp_to_uint v8f32:$vj)), (XVFTINTRZ_WU_S v8f32:$vj)>;
+def : Pat<(v4i64 (fp_to_uint v4f64:$vj)), (XVFTINTRZ_LU_D v4f64:$vj)>;
+def : Pat<(v4i64 (fp_to_uint v4f32:$vj)),
+          (VEXT2XV_DU_WU (SUBREG_TO_REG (i64 0), (VFTINTRZ_WU_S v4f32:$vj),
+                                        sub_128))>;
+def : Pat<(v4i32 (fp_to_uint (v4f64 LASX256:$vj))),
+          (EXTRACT_SUBREG (XVFTINTRZ_W_S (XVFCVT_S_D (XVPERMI_D v4f64:$vj, 238),
+                                                     v4f64:$vj)),
+                          sub_128)>;
+
 } // Predicates = [HasExtLASX]
 
 /// Intrinsic pattern

diff --git a/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
@@ -1754,6 +1754,14 @@ def : Pat<(v2f64 (sint_to_fp v2i64:$vj)), (VFFINT_D_L v2i64:$vj)>;
 def : Pat<(v4f32 (uint_to_fp v4i32:$vj)), (VFFINT_S_WU v4i32:$vj)>;
 def : Pat<(v2f64 (uint_to_fp v2i64:$vj)), (VFFINT_D_LU v2i64:$vj)>;
 
+// VFTINTRZ_{W_S/L_D}
+def : Pat<(v4i32 (fp_to_sint v4f32:$vj)), (VFTINTRZ_W_S v4f32:$vj)>;
+def : Pat<(v2i64 (fp_to_sint v2f64:$vj)), (VFTINTRZ_L_D v2f64:$vj)>;
+
+// VFTINTRZ_{W_SU/L_DU}
+def : Pat<(v4i32 (fp_to_uint v4f32:$vj)), (VFTINTRZ_WU_S v4f32:$vj)>;
+def : Pat<(v2i64 (fp_to_uint v2f64:$vj)), (VFTINTRZ_LU_D v2f64:$vj)>;
+
 } // Predicates = [HasExtLSX]
 
 /// Intrinsic pattern

diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/fptosi.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/fptosi.ll
@@ -0,0 +1,57 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s
+
+define void @fptosi_v8f32_v8i32(ptr %res, ptr %in){
+; CHECK-LABEL: fptosi_v8f32_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a1, 0
+; CHECK-NEXT:    xvftintrz.w.s $xr0, $xr0
+; CHECK-NEXT:    xvst $xr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <8 x float>, ptr %in
+  %v1 = fptosi <8 x float> %v0 to <8 x i32>
+  store <8 x i32> %v1, ptr %res
+  ret void
+}
+
+define void @fptosi_v4f64_v4i64(ptr %res, ptr %in){
+; CHECK-LABEL: fptosi_v4f64_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a1, 0
+; CHECK-NEXT:    xvftintrz.l.d $xr0, $xr0
+; CHECK-NEXT:    xvst $xr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x double>, ptr %in
+  %v1 = fptosi <4 x double> %v0 to <4 x i64>
+  store <4 x i64> %v1, ptr %res
+  ret void
+}
+
+define void @fptosi_v4f64_v4i32(ptr %res, ptr %in){
+; CHECK-LABEL: fptosi_v4f64_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a1, 0
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 238
+; CHECK-NEXT:    xvfcvt.s.d $xr0, $xr1, $xr0
+; CHECK-NEXT:    xvftintrz.w.s $xr0, $xr0
+; CHECK-NEXT:    vst $vr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x double>, ptr %in
+  %v1 = fptosi <4 x double> %v0 to <4 x i32>
+  store <4 x i32> %v1, ptr %res
+  ret void
+}
+
+define void @fptosi_v4f32_v4i64(ptr %res, ptr %in){
+; CHECK-LABEL: fptosi_v4f32_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a1, 0
+; CHECK-NEXT:    vftintrz.w.s $vr0, $vr0
+; CHECK-NEXT:    vext2xv.d.w $xr0, $xr0
+; CHECK-NEXT:    xvst $xr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x float>, ptr %in
+  %v1 = fptosi <4 x float> %v0 to <4 x i64>
+  store <4 x i64> %v1, ptr %res
+  ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/fptoui.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/fptoui.ll
@@ -0,0 +1,57 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s
+
+define void @fptoui_v8f32_v8i32(ptr %res, ptr %in){
+; CHECK-LABEL: fptoui_v8f32_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a1, 0
+; CHECK-NEXT:    xvftintrz.wu.s $xr0, $xr0
+; CHECK-NEXT:    xvst $xr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <8 x float>, ptr %in
+  %v1 = fptoui <8 x float> %v0 to <8 x i32>
+  store <8 x i32> %v1, ptr %res
+  ret void
+}
+
+define void @fptoui_v4f64_v4i64(ptr %res, ptr %in){
+; CHECK-LABEL: fptoui_v4f64_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a1, 0
+; CHECK-NEXT:    xvftintrz.lu.d $xr0, $xr0
+; CHECK-NEXT:    xvst $xr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x double>, ptr %in
+  %v1 = fptoui <4 x double> %v0 to <4 x i64>
+  store <4 x i64> %v1, ptr %res
+  ret void
+}
+
+define void @fptoui_v4f64_v4i32(ptr %res, ptr %in){
+; CHECK-LABEL: fptoui_v4f64_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a1, 0
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 238
+; CHECK-NEXT:    xvfcvt.s.d $xr0, $xr1, $xr0
+; CHECK-NEXT:    xvftintrz.w.s $xr0, $xr0
+; CHECK-NEXT:    vst $vr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x double>, ptr %in
+  %v1 = fptoui <4 x double> %v0 to <4 x i32>
+  store <4 x i32> %v1, ptr %res
+  ret void
+}
+
+define void @fptoui_v4f32_v4i64(ptr %res, ptr %in){
+; CHECK-LABEL: fptoui_v4f32_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a1, 0
+; CHECK-NEXT:    vftintrz.wu.s $vr0, $vr0
+; CHECK-NEXT:    vext2xv.du.wu $xr0, $xr0
+; CHECK-NEXT:    xvst $xr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x float>, ptr %in
+  %v1 = fptoui <4 x float> %v0 to <4 x i64>
+  store <4 x i64> %v1, ptr %res
+  ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/fptosi.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/fptosi.ll
@@ -0,0 +1,28 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s
+
+define void @fptosi_v4f32_v4i32(ptr %res, ptr %in){
+; CHECK-LABEL: fptosi_v4f32_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a1, 0
+; CHECK-NEXT:    vftintrz.w.s $vr0, $vr0
+; CHECK-NEXT:    vst $vr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x float>, ptr %in
+  %v1 = fptosi <4 x float> %v0 to <4 x i32>
+  store <4 x i32> %v1, ptr %res
+  ret void
+}
+
+define void @fptosi_v2f64_v2i64(ptr %res, ptr %in){
+; CHECK-LABEL: fptosi_v2f64_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a1, 0
+; CHECK-NEXT:    vftintrz.l.d $vr0, $vr0
+; CHECK-NEXT:    vst $vr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <2 x double>, ptr %in
+  %v1 = fptosi <2 x double> %v0 to <2 x i64>
+  store <2 x i64> %v1, ptr %res
+  ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/fptoui.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/fptoui.ll
@@ -0,0 +1,28 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s
+
+define void @fptoui_v4f32_v4i32(ptr %res, ptr %in){
+; CHECK-LABEL: fptoui_v4f32_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a1, 0
+; CHECK-NEXT:    vftintrz.wu.s $vr0, $vr0
+; CHECK-NEXT:    vst $vr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <4 x float>, ptr %in
+  %v1 = fptoui <4 x float> %v0 to <4 x i32>
+  store <4 x i32> %v1, ptr %res
+  ret void
+}
+
+define void @fptoui_v2f64_v2i64(ptr %res, ptr %in){
+; CHECK-LABEL: fptoui_v2f64_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a1, 0
+; CHECK-NEXT:    vftintrz.lu.d $vr0, $vr0
+; CHECK-NEXT:    vst $vr0, $a0, 0
+; CHECK-NEXT:    ret
+  %v0 = load <2 x double>, ptr %in
+  %v1 = fptoui <2 x double> %v0 to <2 x i64>
+  store <2 x i64> %v1, ptr %res
+  ret void
+}