| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py |
| ; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v,+f,+d,+zvfh \ |
| ; RUN: -target-abi=ilp32d -verify-machineinstrs | FileCheck %s --check-prefix=RV32 |
| ; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv64 -mattr=+v,+f,+d,+zvfh \ |
| ; RUN: -target-abi=lp64d -verify-machineinstrs | FileCheck %s --check-prefix=RV64-i32 |
| ; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v,+f,+d,+zvfh \ |
| ; RUN: -target-abi=lp64d -verify-machineinstrs | FileCheck %s --check-prefix=RV64-i64 |
| |
| define <1 x iXLen> @lround_v1f16(<1 x half> %x) nounwind { |
| ; RV32-LABEL: lround_v1f16: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV32-NEXT: vmv.s.x v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v1f16: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vmv.s.x v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v1f16: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vsetvli zero, zero, e64, m4, ta, ma |
| ; RV64-i64-NEXT: vmv.s.x v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <1 x iXLen> @llvm.lround.v1iXLen.v1f16(<1 x half> %x) |
| ret <1 x iXLen> %a |
| } |
| declare <1 x iXLen> @llvm.lround.v1iXLen.v1f16(<1 x half>) |
| |
| define <2 x iXLen> @lround_v2f16(<2 x half> %x) nounwind { |
| ; RV32-LABEL: lround_v2f16: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e16, mf4, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vsetivli zero, 2, e32, mf2, ta, ma |
| ; RV32-NEXT: vmv.v.x v8, a0 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v2f16: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, mf4, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vsetivli zero, 2, e32, mf2, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v8, a0 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v2f16: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, mf4, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vsetivli zero, 2, e64, m1, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: ret |
| %a = call <2 x iXLen> @llvm.lround.v2iXLen.v2f16(<2 x half> %x) |
| ret <2 x iXLen> %a |
| } |
| declare <2 x iXLen> @llvm.lround.v2iXLen.v2f16(<2 x half>) |
| |
| define <3 x iXLen> @lround_v3f16(<3 x half> %x) nounwind { |
| ; RV32-LABEL: lround_v3f16: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e16, mf2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV32-NEXT: vmv.v.x v9, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, mf2, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v3f16: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, mf2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v9, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, mf2, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v3f16: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, mf2, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 3 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: vsetivli zero, 4, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vsetvli zero, zero, e16, mf2, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <3 x iXLen> @llvm.lround.v3iXLen.v3f16(<3 x half> %x) |
| ret <3 x iXLen> %a |
| } |
| declare <3 x iXLen> @llvm.lround.v3iXLen.v3f16(<3 x half>) |
| |
| define <4 x iXLen> @lround_v4f16(<4 x half> %x) nounwind { |
| ; RV32-LABEL: lround_v4f16: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e16, mf2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV32-NEXT: vmv.v.x v9, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, mf2, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v4f16: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, mf2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v9, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, mf2, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v4f16: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, mf2, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 3 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: vsetivli zero, 4, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vsetvli zero, zero, e16, mf2, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <4 x iXLen> @llvm.lround.v4iXLen.v4f16(<4 x half> %x) |
| ret <4 x iXLen> %a |
| } |
| declare <4 x iXLen> @llvm.lround.v4iXLen.v4f16(<4 x half>) |
| |
| define <8 x iXLen> @lround_v8f16(<8 x half> %x) nounwind { |
| ; RV32-LABEL: lround_v8f16: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma |
| ; RV32-NEXT: vmv.v.x v10, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v9, v8, 4 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV32-NEXT: vslide1down.vx v10, v10, a1 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: vslidedown.vi v9, v8, 5 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV32-NEXT: vslide1down.vx v10, v10, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 6 |
| ; RV32-NEXT: vslidedown.vi v13, v8, 7 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV32-NEXT: vslide1down.vx v8, v10, a1 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v13 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v8f16: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetivli zero, 8, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v10, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 4 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v10, v10, a1 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 5 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v10, v10, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 6 |
| ; RV64-i32-NEXT: vslidedown.vi v13, v8, 7 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v8, v10, a1 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v13 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v8f16: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: addi sp, sp, -128 |
| ; RV64-i64-NEXT: sd ra, 120(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: sd s0, 112(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: addi s0, sp, 128 |
| ; RV64-i64-NEXT: andi sp, sp, -64 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 7 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 6 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 5 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 4 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 3 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a2, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v8, v8, 1 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a3, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a4, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a5, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a6, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: sd a4, 32(sp) |
| ; RV64-i64-NEXT: sd a3, 40(sp) |
| ; RV64-i64-NEXT: sd a2, 48(sp) |
| ; RV64-i64-NEXT: sd a1, 56(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: sd a0, 0(sp) |
| ; RV64-i64-NEXT: sd a1, 8(sp) |
| ; RV64-i64-NEXT: sd a6, 16(sp) |
| ; RV64-i64-NEXT: sd a5, 24(sp) |
| ; RV64-i64-NEXT: mv a0, sp |
| ; RV64-i64-NEXT: vsetivli zero, 8, e64, m4, ta, ma |
| ; RV64-i64-NEXT: vle64.v v8, (a0) |
| ; RV64-i64-NEXT: addi sp, s0, -128 |
| ; RV64-i64-NEXT: ld ra, 120(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: ld s0, 112(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: addi sp, sp, 128 |
| ; RV64-i64-NEXT: ret |
| %a = call <8 x iXLen> @llvm.lround.v8iXLen.v8f16(<8 x half> %x) |
| ret <8 x iXLen> %a |
| } |
| declare <8 x iXLen> @llvm.lround.v8iXLen.v8f16(<8 x half>) |
| |
| define <16 x iXLen> @lround_v16f16(<16 x half> %x) nounwind { |
| ; RV32-LABEL: lround_v16f16: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: addi sp, sp, -128 |
| ; RV32-NEXT: sw ra, 124(sp) # 4-byte Folded Spill |
| ; RV32-NEXT: sw s0, 120(sp) # 4-byte Folded Spill |
| ; RV32-NEXT: addi s0, sp, 128 |
| ; RV32-NEXT: andi sp, sp, -64 |
| ; RV32-NEXT: vsetivli zero, 1, e16, m2, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 15 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 14 |
| ; RV32-NEXT: vslidedown.vi v14, v8, 13 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 12 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 11 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a2, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v14 |
| ; RV32-NEXT: vslidedown.vi v14, v8, 10 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a3, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 9 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a5, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 8 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a4, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v14 |
| ; RV32-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 7 |
| ; RV32-NEXT: vslidedown.vi v11, v8, 6 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a6, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV32-NEXT: vslidedown.vi v13, v8, 4 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s a7, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV32-NEXT: vslidedown.vi v14, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 1 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s t0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s t1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v11 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s t2, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s t3, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v13 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: fcvt.w.s t4, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: sw a5, 48(sp) |
| ; RV32-NEXT: sw a3, 52(sp) |
| ; RV32-NEXT: sw a2, 56(sp) |
| ; RV32-NEXT: sw a1, 60(sp) |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v14 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: sw t0, 32(sp) |
| ; RV32-NEXT: sw a7, 36(sp) |
| ; RV32-NEXT: sw a6, 40(sp) |
| ; RV32-NEXT: sw a4, 44(sp) |
| ; RV32-NEXT: fcvt.w.s a2, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV32-NEXT: sw t4, 16(sp) |
| ; RV32-NEXT: sw t3, 20(sp) |
| ; RV32-NEXT: sw t2, 24(sp) |
| ; RV32-NEXT: sw t1, 28(sp) |
| ; RV32-NEXT: fcvt.w.s a3, fa5, rmm |
| ; RV32-NEXT: sw a0, 0(sp) |
| ; RV32-NEXT: sw a3, 4(sp) |
| ; RV32-NEXT: sw a2, 8(sp) |
| ; RV32-NEXT: sw a1, 12(sp) |
| ; RV32-NEXT: mv a0, sp |
| ; RV32-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV32-NEXT: vle32.v v8, (a0) |
| ; RV32-NEXT: addi sp, s0, -128 |
| ; RV32-NEXT: lw ra, 124(sp) # 4-byte Folded Reload |
| ; RV32-NEXT: lw s0, 120(sp) # 4-byte Folded Reload |
| ; RV32-NEXT: addi sp, sp, 128 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v16f16: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: addi sp, sp, -128 |
| ; RV64-i32-NEXT: sd ra, 120(sp) # 8-byte Folded Spill |
| ; RV64-i32-NEXT: sd s0, 112(sp) # 8-byte Folded Spill |
| ; RV64-i32-NEXT: addi s0, sp, 128 |
| ; RV64-i32-NEXT: andi sp, sp, -64 |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, m2, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 15 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 14 |
| ; RV64-i32-NEXT: vslidedown.vi v14, v8, 13 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 12 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 11 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a2, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v14 |
| ; RV64-i32-NEXT: vslidedown.vi v14, v8, 10 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a3, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 9 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a5, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 8 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a4, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v14 |
| ; RV64-i32-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 7 |
| ; RV64-i32-NEXT: vslidedown.vi v11, v8, 6 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a6, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV64-i32-NEXT: vslidedown.vi v13, v8, 4 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s a7, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV64-i32-NEXT: vslidedown.vi v14, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 1 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s t0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s t1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s t2, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s t3, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v13 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: fcvt.w.s t4, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: sw a5, 48(sp) |
| ; RV64-i32-NEXT: sw a3, 52(sp) |
| ; RV64-i32-NEXT: sw a2, 56(sp) |
| ; RV64-i32-NEXT: sw a1, 60(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v14 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: sw t0, 32(sp) |
| ; RV64-i32-NEXT: sw a7, 36(sp) |
| ; RV64-i32-NEXT: sw a6, 40(sp) |
| ; RV64-i32-NEXT: sw a4, 44(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a2, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i32-NEXT: sw t4, 16(sp) |
| ; RV64-i32-NEXT: sw t3, 20(sp) |
| ; RV64-i32-NEXT: sw t2, 24(sp) |
| ; RV64-i32-NEXT: sw t1, 28(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a3, fa5, rmm |
| ; RV64-i32-NEXT: sw a0, 0(sp) |
| ; RV64-i32-NEXT: sw a3, 4(sp) |
| ; RV64-i32-NEXT: sw a2, 8(sp) |
| ; RV64-i32-NEXT: sw a1, 12(sp) |
| ; RV64-i32-NEXT: mv a0, sp |
| ; RV64-i32-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV64-i32-NEXT: vle32.v v8, (a0) |
| ; RV64-i32-NEXT: addi sp, s0, -128 |
| ; RV64-i32-NEXT: ld ra, 120(sp) # 8-byte Folded Reload |
| ; RV64-i32-NEXT: ld s0, 112(sp) # 8-byte Folded Reload |
| ; RV64-i32-NEXT: addi sp, sp, 128 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v16f16: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: addi sp, sp, -256 |
| ; RV64-i64-NEXT: sd ra, 248(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: sd s0, 240(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: addi s0, sp, 256 |
| ; RV64-i64-NEXT: andi sp, sp, -128 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, m2, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 15 |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 14 |
| ; RV64-i64-NEXT: vslidedown.vi v14, v8, 13 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 12 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 11 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a2, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v14 |
| ; RV64-i64-NEXT: vslidedown.vi v14, v8, 10 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a3, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 9 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a5, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 8 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a4, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v14 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e16, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 7 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 6 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a6, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV64-i64-NEXT: vslidedown.vi v13, v8, 4 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s a7, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV64-i64-NEXT: vslidedown.vi v14, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v8, v8, 1 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s t0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s t1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s t2, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s t3, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v13 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: fcvt.l.s t4, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: sd a5, 96(sp) |
| ; RV64-i64-NEXT: sd a3, 104(sp) |
| ; RV64-i64-NEXT: sd a2, 112(sp) |
| ; RV64-i64-NEXT: sd a1, 120(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v14 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: sd t0, 64(sp) |
| ; RV64-i64-NEXT: sd a7, 72(sp) |
| ; RV64-i64-NEXT: sd a6, 80(sp) |
| ; RV64-i64-NEXT: sd a4, 88(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a2, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.s.h fa5, fa5 |
| ; RV64-i64-NEXT: sd t4, 32(sp) |
| ; RV64-i64-NEXT: sd t3, 40(sp) |
| ; RV64-i64-NEXT: sd t2, 48(sp) |
| ; RV64-i64-NEXT: sd t1, 56(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a3, fa5, rmm |
| ; RV64-i64-NEXT: sd a0, 0(sp) |
| ; RV64-i64-NEXT: sd a3, 8(sp) |
| ; RV64-i64-NEXT: sd a2, 16(sp) |
| ; RV64-i64-NEXT: sd a1, 24(sp) |
| ; RV64-i64-NEXT: mv a0, sp |
| ; RV64-i64-NEXT: vsetivli zero, 16, e64, m8, ta, ma |
| ; RV64-i64-NEXT: vle64.v v8, (a0) |
| ; RV64-i64-NEXT: addi sp, s0, -256 |
| ; RV64-i64-NEXT: ld ra, 248(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: ld s0, 240(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: addi sp, sp, 256 |
| ; RV64-i64-NEXT: ret |
| %a = call <16 x iXLen> @llvm.lround.v16iXLen.v16f16(<16 x half> %x) |
| ret <16 x iXLen> %a |
| } |
| declare <16 x iXLen> @llvm.lround.v16iXLen.v16f16(<16 x half>) |
| |
| define <1 x iXLen> @lround_v1f32(<1 x float> %x) nounwind { |
| ; RV32-LABEL: lround_v1f32: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vmv.s.x v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v1f32: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vmv.s.x v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v1f32: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vmv.s.x v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <1 x iXLen> @llvm.lround.v1iXLen.v1f32(<1 x float> %x) |
| ret <1 x iXLen> %a |
| } |
| declare <1 x iXLen> @llvm.lround.v1iXLen.v1f32(<1 x float>) |
| |
| define <2 x iXLen> @lround_v2f32(<2 x float> %x) nounwind { |
| ; RV32-LABEL: lround_v2f32: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 2, e32, mf2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vmv.v.x v8, a0 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v2f32: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 2, e32, mf2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vmv.v.x v8, a0 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v2f32: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, mf2, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vsetivli zero, 2, e64, m1, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: ret |
| %a = call <2 x iXLen> @llvm.lround.v2iXLen.v2f32(<2 x float> %x) |
| ret <2 x iXLen> %a |
| } |
| declare <2 x iXLen> @llvm.lround.v2iXLen.v2f32(<2 x float>) |
| |
| define <3 x iXLen> @lround_v3f32(<3 x float> %x) nounwind { |
| ; RV32-LABEL: lround_v3f32: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: vmv.v.x v9, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v3f32: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: vmv.v.x v9, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v3f32: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 3 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vsetivli zero, 4, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <3 x iXLen> @llvm.lround.v3iXLen.v3f32(<3 x float> %x) |
| ret <3 x iXLen> %a |
| } |
| declare <3 x iXLen> @llvm.lround.v3iXLen.v3f32(<3 x float>) |
| |
| define <4 x iXLen> @lround_v4f32(<4 x float> %x) nounwind { |
| ; RV32-LABEL: lround_v4f32: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: vmv.v.x v9, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v4f32: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: vmv.v.x v9, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v4f32: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v11, v8, 3 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vsetivli zero, 4, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i64-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <4 x iXLen> @llvm.lround.v4iXLen.v4f32(<4 x float> %x) |
| ret <4 x iXLen> %a |
| } |
| declare <4 x iXLen> @llvm.lround.v4iXLen.v4f32(<4 x float>) |
| |
| define <8 x iXLen> @lround_v8f32(<8 x float> %x) nounwind { |
| ; RV32-LABEL: lround_v8f32: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v10, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslidedown.vi v11, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v11 |
| ; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma |
| ; RV32-NEXT: vmv.v.x v10, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 4 |
| ; RV32-NEXT: vslide1down.vx v10, v10, a1 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 5 |
| ; RV32-NEXT: vslide1down.vx v10, v10, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslidedown.vi v12, v8, 6 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 7 |
| ; RV32-NEXT: vslide1down.vx v10, v10, a1 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vslide1down.vx v10, v10, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vslide1down.vx v8, v10, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v8f32: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslidedown.vi v11, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 3 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v11 |
| ; RV64-i32-NEXT: vsetivli zero, 8, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v10, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 4 |
| ; RV64-i32-NEXT: vslide1down.vx v10, v10, a1 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 5 |
| ; RV64-i32-NEXT: vslide1down.vx v10, v10, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 6 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 7 |
| ; RV64-i32-NEXT: vslide1down.vx v10, v10, a1 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vslide1down.vx v10, v10, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v10, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v8f32: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: addi sp, sp, -128 |
| ; RV64-i64-NEXT: sd ra, 120(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: sd s0, 112(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: addi s0, sp, 128 |
| ; RV64-i64-NEXT: andi sp, sp, -64 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m2, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 7 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 6 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV64-i64-NEXT: fcvt.l.s a2, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 4 |
| ; RV64-i64-NEXT: fcvt.l.s a3, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 3 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v8, v8, 1 |
| ; RV64-i64-NEXT: fcvt.l.s a4, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.l.s a5, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: fcvt.l.s a6, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: sd a4, 32(sp) |
| ; RV64-i64-NEXT: sd a3, 40(sp) |
| ; RV64-i64-NEXT: sd a2, 48(sp) |
| ; RV64-i64-NEXT: sd a1, 56(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: sd a0, 0(sp) |
| ; RV64-i64-NEXT: sd a1, 8(sp) |
| ; RV64-i64-NEXT: sd a6, 16(sp) |
| ; RV64-i64-NEXT: sd a5, 24(sp) |
| ; RV64-i64-NEXT: mv a0, sp |
| ; RV64-i64-NEXT: vsetivli zero, 8, e64, m4, ta, ma |
| ; RV64-i64-NEXT: vle64.v v8, (a0) |
| ; RV64-i64-NEXT: addi sp, s0, -128 |
| ; RV64-i64-NEXT: ld ra, 120(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: ld s0, 112(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: addi sp, sp, 128 |
| ; RV64-i64-NEXT: ret |
| %a = call <8 x iXLen> @llvm.lround.v8iXLen.v8f32(<8 x float> %x) |
| ret <8 x iXLen> %a |
| } |
| declare <8 x iXLen> @llvm.lround.v8iXLen.v8f32(<8 x float>) |
| |
| define <16 x iXLen> @lround_v16f32(<16 x float> %x) nounwind { |
| ; RV32-LABEL: lround_v16f32: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: addi sp, sp, -192 |
| ; RV32-NEXT: sw ra, 188(sp) # 4-byte Folded Spill |
| ; RV32-NEXT: sw s0, 184(sp) # 4-byte Folded Spill |
| ; RV32-NEXT: addi s0, sp, 192 |
| ; RV32-NEXT: andi sp, sp, -64 |
| ; RV32-NEXT: mv a0, sp |
| ; RV32-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV32-NEXT: vse32.v v8, (a0) |
| ; RV32-NEXT: flw fa5, 60(sp) |
| ; RV32-NEXT: vfmv.f.s fa4, v8 |
| ; RV32-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v10, v8, 3 |
| ; RV32-NEXT: vslidedown.vi v11, v8, 2 |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: sw a0, 124(sp) |
| ; RV32-NEXT: flw fa5, 56(sp) |
| ; RV32-NEXT: fcvt.w.s a0, fa4, rmm |
| ; RV32-NEXT: vfmv.f.s fa4, v10 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 1 |
| ; RV32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV32-NEXT: sw a1, 120(sp) |
| ; RV32-NEXT: flw fa5, 52(sp) |
| ; RV32-NEXT: fcvt.w.s a1, fa4, rmm |
| ; RV32-NEXT: vfmv.f.s fa4, v11 |
| ; RV32-NEXT: fcvt.w.s a2, fa4, rmm |
| ; RV32-NEXT: fcvt.w.s a3, fa5, rmm |
| ; RV32-NEXT: sw a3, 116(sp) |
| ; RV32-NEXT: flw fa5, 48(sp) |
| ; RV32-NEXT: vfmv.f.s fa4, v10 |
| ; RV32-NEXT: vsetivli zero, 1, e32, m2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v10, v8, 7 |
| ; RV32-NEXT: fcvt.w.s a3, fa4, rmm |
| ; RV32-NEXT: fcvt.w.s a4, fa5, rmm |
| ; RV32-NEXT: sw a4, 112(sp) |
| ; RV32-NEXT: flw fa5, 44(sp) |
| ; RV32-NEXT: vfmv.f.s fa4, v10 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 6 |
| ; RV32-NEXT: fcvt.w.s a4, fa4, rmm |
| ; RV32-NEXT: fcvt.w.s a5, fa5, rmm |
| ; RV32-NEXT: sw a5, 108(sp) |
| ; RV32-NEXT: flw fa5, 40(sp) |
| ; RV32-NEXT: vfmv.f.s fa4, v10 |
| ; RV32-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV32-NEXT: fcvt.w.s a5, fa4, rmm |
| ; RV32-NEXT: fcvt.w.s a6, fa5, rmm |
| ; RV32-NEXT: sw a6, 104(sp) |
| ; RV32-NEXT: flw fa5, 36(sp) |
| ; RV32-NEXT: vfmv.f.s fa4, v10 |
| ; RV32-NEXT: fcvt.w.s a6, fa4, rmm |
| ; RV32-NEXT: vslidedown.vi v8, v8, 4 |
| ; RV32-NEXT: fcvt.w.s a7, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: sw a7, 100(sp) |
| ; RV32-NEXT: fcvt.w.s a7, fa5, rmm |
| ; RV32-NEXT: flw fa5, 32(sp) |
| ; RV32-NEXT: sw a0, 64(sp) |
| ; RV32-NEXT: sw a3, 68(sp) |
| ; RV32-NEXT: sw a2, 72(sp) |
| ; RV32-NEXT: sw a1, 76(sp) |
| ; RV32-NEXT: sw a7, 80(sp) |
| ; RV32-NEXT: sw a6, 84(sp) |
| ; RV32-NEXT: sw a5, 88(sp) |
| ; RV32-NEXT: sw a4, 92(sp) |
| ; RV32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV32-NEXT: sw a0, 96(sp) |
| ; RV32-NEXT: addi a0, sp, 64 |
| ; RV32-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV32-NEXT: vle32.v v8, (a0) |
| ; RV32-NEXT: addi sp, s0, -192 |
| ; RV32-NEXT: lw ra, 188(sp) # 4-byte Folded Reload |
| ; RV32-NEXT: lw s0, 184(sp) # 4-byte Folded Reload |
| ; RV32-NEXT: addi sp, sp, 192 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v16f32: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: addi sp, sp, -192 |
| ; RV64-i32-NEXT: sd ra, 184(sp) # 8-byte Folded Spill |
| ; RV64-i32-NEXT: sd s0, 176(sp) # 8-byte Folded Spill |
| ; RV64-i32-NEXT: addi s0, sp, 192 |
| ; RV64-i32-NEXT: andi sp, sp, -64 |
| ; RV64-i32-NEXT: mv a0, sp |
| ; RV64-i32-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV64-i32-NEXT: vse32.v v8, (a0) |
| ; RV64-i32-NEXT: flw fa5, 60(sp) |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v8 |
| ; RV64-i32-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 3 |
| ; RV64-i32-NEXT: vslidedown.vi v11, v8, 2 |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: sw a0, 124(sp) |
| ; RV64-i32-NEXT: flw fa5, 56(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa4, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 1 |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa5, rmm |
| ; RV64-i32-NEXT: sw a1, 120(sp) |
| ; RV64-i32-NEXT: flw fa5, 52(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a1, fa4, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v11 |
| ; RV64-i32-NEXT: fcvt.w.s a2, fa4, rmm |
| ; RV64-i32-NEXT: fcvt.w.s a3, fa5, rmm |
| ; RV64-i32-NEXT: sw a3, 116(sp) |
| ; RV64-i32-NEXT: flw fa5, 48(sp) |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i32-NEXT: vsetivli zero, 1, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 7 |
| ; RV64-i32-NEXT: fcvt.w.s a3, fa4, rmm |
| ; RV64-i32-NEXT: fcvt.w.s a4, fa5, rmm |
| ; RV64-i32-NEXT: sw a4, 112(sp) |
| ; RV64-i32-NEXT: flw fa5, 44(sp) |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 6 |
| ; RV64-i32-NEXT: fcvt.w.s a4, fa4, rmm |
| ; RV64-i32-NEXT: fcvt.w.s a5, fa5, rmm |
| ; RV64-i32-NEXT: sw a5, 108(sp) |
| ; RV64-i32-NEXT: flw fa5, 40(sp) |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV64-i32-NEXT: fcvt.w.s a5, fa4, rmm |
| ; RV64-i32-NEXT: fcvt.w.s a6, fa5, rmm |
| ; RV64-i32-NEXT: sw a6, 104(sp) |
| ; RV64-i32-NEXT: flw fa5, 36(sp) |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i32-NEXT: fcvt.w.s a6, fa4, rmm |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 4 |
| ; RV64-i32-NEXT: fcvt.w.s a7, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: sw a7, 100(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a7, fa5, rmm |
| ; RV64-i32-NEXT: flw fa5, 32(sp) |
| ; RV64-i32-NEXT: sw a0, 64(sp) |
| ; RV64-i32-NEXT: sw a3, 68(sp) |
| ; RV64-i32-NEXT: sw a2, 72(sp) |
| ; RV64-i32-NEXT: sw a1, 76(sp) |
| ; RV64-i32-NEXT: sw a7, 80(sp) |
| ; RV64-i32-NEXT: sw a6, 84(sp) |
| ; RV64-i32-NEXT: sw a5, 88(sp) |
| ; RV64-i32-NEXT: sw a4, 92(sp) |
| ; RV64-i32-NEXT: fcvt.w.s a0, fa5, rmm |
| ; RV64-i32-NEXT: sw a0, 96(sp) |
| ; RV64-i32-NEXT: addi a0, sp, 64 |
| ; RV64-i32-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV64-i32-NEXT: vle32.v v8, (a0) |
| ; RV64-i32-NEXT: addi sp, s0, -192 |
| ; RV64-i32-NEXT: ld ra, 184(sp) # 8-byte Folded Reload |
| ; RV64-i32-NEXT: ld s0, 176(sp) # 8-byte Folded Reload |
| ; RV64-i32-NEXT: addi sp, sp, 192 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v16f32: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: addi sp, sp, -384 |
| ; RV64-i64-NEXT: sd ra, 376(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: sd s0, 368(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: addi s0, sp, 384 |
| ; RV64-i64-NEXT: andi sp, sp, -128 |
| ; RV64-i64-NEXT: addi a0, sp, 64 |
| ; RV64-i64-NEXT: vsetivli zero, 16, e32, m4, ta, ma |
| ; RV64-i64-NEXT: vse32.v v8, (a0) |
| ; RV64-i64-NEXT: flw fa5, 124(sp) |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v8 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 3 |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 2 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: sd a0, 248(sp) |
| ; RV64-i64-NEXT: flw fa5, 120(sp) |
| ; RV64-i64-NEXT: vslidedown.vi v13, v8, 1 |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa4, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa5, rmm |
| ; RV64-i64-NEXT: sd a1, 240(sp) |
| ; RV64-i64-NEXT: flw fa5, 116(sp) |
| ; RV64-i64-NEXT: vsetivli zero, 1, e32, m2, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 7 |
| ; RV64-i64-NEXT: fcvt.l.s a1, fa4, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v12 |
| ; RV64-i64-NEXT: fcvt.l.s a2, fa5, rmm |
| ; RV64-i64-NEXT: sd a2, 232(sp) |
| ; RV64-i64-NEXT: flw fa5, 112(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a2, fa4, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v13 |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 6 |
| ; RV64-i64-NEXT: fcvt.l.s a3, fa5, rmm |
| ; RV64-i64-NEXT: sd a3, 224(sp) |
| ; RV64-i64-NEXT: flw fa5, 108(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a3, fa4, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 5 |
| ; RV64-i64-NEXT: fcvt.l.s a4, fa5, rmm |
| ; RV64-i64-NEXT: sd a4, 216(sp) |
| ; RV64-i64-NEXT: flw fa5, 104(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a4, fa4, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v12 |
| ; RV64-i64-NEXT: fcvt.l.s a5, fa4, rmm |
| ; RV64-i64-NEXT: fcvt.l.s a6, fa5, rmm |
| ; RV64-i64-NEXT: sd a6, 208(sp) |
| ; RV64-i64-NEXT: flw fa5, 100(sp) |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i64-NEXT: fcvt.l.s a6, fa4, rmm |
| ; RV64-i64-NEXT: vslidedown.vi v8, v8, 4 |
| ; RV64-i64-NEXT: fcvt.l.s a7, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: sd a7, 200(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a7, fa5, rmm |
| ; RV64-i64-NEXT: flw fa5, 96(sp) |
| ; RV64-i64-NEXT: sd a0, 128(sp) |
| ; RV64-i64-NEXT: sd a3, 136(sp) |
| ; RV64-i64-NEXT: sd a2, 144(sp) |
| ; RV64-i64-NEXT: sd a1, 152(sp) |
| ; RV64-i64-NEXT: sd a7, 160(sp) |
| ; RV64-i64-NEXT: sd a6, 168(sp) |
| ; RV64-i64-NEXT: sd a5, 176(sp) |
| ; RV64-i64-NEXT: sd a4, 184(sp) |
| ; RV64-i64-NEXT: fcvt.l.s a0, fa5, rmm |
| ; RV64-i64-NEXT: sd a0, 192(sp) |
| ; RV64-i64-NEXT: addi a0, sp, 128 |
| ; RV64-i64-NEXT: vsetivli zero, 16, e64, m8, ta, ma |
| ; RV64-i64-NEXT: vle64.v v8, (a0) |
| ; RV64-i64-NEXT: addi sp, s0, -384 |
| ; RV64-i64-NEXT: ld ra, 376(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: ld s0, 368(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: addi sp, sp, 384 |
| ; RV64-i64-NEXT: ret |
| %a = call <16 x iXLen> @llvm.lround.v16iXLen.v16f32(<16 x float> %x) |
| ret <16 x iXLen> %a |
| } |
| declare <16 x iXLen> @llvm.lround.v16iXLen.v16f32(<16 x float>) |
| |
| define <1 x iXLen> @lround_v1f64(<1 x double> %x) nounwind { |
| ; RV32-LABEL: lround_v1f64: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV32-NEXT: vmv.s.x v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v1f64: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV64-i32-NEXT: vmv.s.x v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v1f64: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa5, rmm |
| ; RV64-i64-NEXT: vmv.s.x v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <1 x iXLen> @llvm.lround.v1iXLen.v1f64(<1 x double> %x) |
| ret <1 x iXLen> %a |
| } |
| declare <1 x iXLen> @llvm.lround.v1iXLen.v1f64(<1 x double>) |
| |
| define <2 x iXLen> @lround_v2f64(<2 x double> %x) nounwind { |
| ; RV32-LABEL: lround_v2f64: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v9 |
| ; RV32-NEXT: fcvt.w.d a1, fa5, rmm |
| ; RV32-NEXT: vsetivli zero, 2, e32, mf2, ta, ma |
| ; RV32-NEXT: vmv.v.x v8, a0 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v2f64: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i32-NEXT: fcvt.w.d a1, fa5, rmm |
| ; RV64-i32-NEXT: vsetivli zero, 2, e32, mf2, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v8, a0 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v2f64: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 2, e64, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v9, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v9 |
| ; RV64-i64-NEXT: fcvt.l.d a1, fa5, rmm |
| ; RV64-i64-NEXT: vmv.v.x v8, a0 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i64-NEXT: ret |
| %a = call <2 x iXLen> @llvm.lround.v2iXLen.v2f64(<2 x double> %x) |
| ret <2 x iXLen> %a |
| } |
| declare <2 x iXLen> @llvm.lround.v2iXLen.v2f64(<2 x double>) |
| |
| define <4 x iXLen> @lround_v4f64(<4 x double> %x) nounwind { |
| ; RV32-LABEL: lround_v4f64: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v12, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: fcvt.w.d a1, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v10 |
| ; RV32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV32-NEXT: vmv.v.x v9, a0 |
| ; RV32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV32-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v4f64: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: fcvt.w.d a1, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i32-NEXT: vsetivli zero, 4, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v9, a0 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV64-i32-NEXT: vsetvli zero, zero, e64, m2, ta, ma |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vsetvli zero, zero, e32, m1, ta, ma |
| ; RV64-i32-NEXT: vslide1down.vx v8, v9, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v4f64: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v12, v8, 1 |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 2 |
| ; RV64-i64-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i64-NEXT: fcvt.l.d a1, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v10 |
| ; RV64-i64-NEXT: vsetivli zero, 4, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vmv.v.x v10, a0 |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v10, a1 |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa5, rmm |
| ; RV64-i64-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i64-NEXT: ret |
| %a = call <4 x iXLen> @llvm.lround.v4iXLen.v4f64(<4 x double> %x) |
| ret <4 x iXLen> %a |
| } |
| declare <4 x iXLen> @llvm.lround.v4iXLen.v4f64(<4 x double>) |
| |
| define <8 x iXLen> @lround_v8f64(<8 x double> %x) nounwind { |
| ; RV32-LABEL: lround_v8f64: |
| ; RV32: # %bb.0: |
| ; RV32-NEXT: addi sp, sp, -128 |
| ; RV32-NEXT: sw ra, 124(sp) # 4-byte Folded Spill |
| ; RV32-NEXT: sw s0, 120(sp) # 4-byte Folded Spill |
| ; RV32-NEXT: addi s0, sp, 128 |
| ; RV32-NEXT: andi sp, sp, -64 |
| ; RV32-NEXT: mv a0, sp |
| ; RV32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV32-NEXT: vslidedown.vi v14, v8, 1 |
| ; RV32-NEXT: vfmv.f.s fa5, v8 |
| ; RV32-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v12, v8, 2 |
| ; RV32-NEXT: vsetivli zero, 8, e64, m4, ta, ma |
| ; RV32-NEXT: vse64.v v8, (a0) |
| ; RV32-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV32-NEXT: vfmv.f.s fa4, v14 |
| ; RV32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV32-NEXT: vfmv.f.s fa5, v12 |
| ; RV32-NEXT: vfmv.f.s fa3, v8 |
| ; RV32-NEXT: fcvt.w.d a1, fa4, rmm |
| ; RV32-NEXT: fcvt.w.d a2, fa5, rmm |
| ; RV32-NEXT: fcvt.w.d a3, fa3, rmm |
| ; RV32-NEXT: fld fa5, 32(sp) |
| ; RV32-NEXT: fld fa4, 40(sp) |
| ; RV32-NEXT: fld fa3, 48(sp) |
| ; RV32-NEXT: fld fa2, 56(sp) |
| ; RV32-NEXT: fcvt.w.d a4, fa5, rmm |
| ; RV32-NEXT: fcvt.w.d a5, fa4, rmm |
| ; RV32-NEXT: fcvt.w.d a6, fa3, rmm |
| ; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma |
| ; RV32-NEXT: vmv.v.x v8, a0 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a2 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a3 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a4 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a5 |
| ; RV32-NEXT: vslide1down.vx v8, v8, a6 |
| ; RV32-NEXT: fcvt.w.d a0, fa2, rmm |
| ; RV32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV32-NEXT: addi sp, s0, -128 |
| ; RV32-NEXT: lw ra, 124(sp) # 4-byte Folded Reload |
| ; RV32-NEXT: lw s0, 120(sp) # 4-byte Folded Reload |
| ; RV32-NEXT: addi sp, sp, 128 |
| ; RV32-NEXT: ret |
| ; |
| ; RV64-i32-LABEL: lround_v8f64: |
| ; RV64-i32: # %bb.0: |
| ; RV64-i32-NEXT: addi sp, sp, -128 |
| ; RV64-i32-NEXT: sd ra, 120(sp) # 8-byte Folded Spill |
| ; RV64-i32-NEXT: sd s0, 112(sp) # 8-byte Folded Spill |
| ; RV64-i32-NEXT: addi s0, sp, 128 |
| ; RV64-i32-NEXT: andi sp, sp, -64 |
| ; RV64-i32-NEXT: mv a0, sp |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v14, v8, 1 |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v12, v8, 2 |
| ; RV64-i32-NEXT: vsetivli zero, 8, e64, m4, ta, ma |
| ; RV64-i32-NEXT: vse64.v v8, (a0) |
| ; RV64-i32-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV64-i32-NEXT: vslidedown.vi v8, v8, 3 |
| ; RV64-i32-NEXT: vfmv.f.s fa4, v14 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa5, rmm |
| ; RV64-i32-NEXT: vfmv.f.s fa5, v12 |
| ; RV64-i32-NEXT: vfmv.f.s fa3, v8 |
| ; RV64-i32-NEXT: fcvt.w.d a1, fa4, rmm |
| ; RV64-i32-NEXT: fcvt.w.d a2, fa5, rmm |
| ; RV64-i32-NEXT: fcvt.w.d a3, fa3, rmm |
| ; RV64-i32-NEXT: fld fa5, 32(sp) |
| ; RV64-i32-NEXT: fld fa4, 40(sp) |
| ; RV64-i32-NEXT: fld fa3, 48(sp) |
| ; RV64-i32-NEXT: fld fa2, 56(sp) |
| ; RV64-i32-NEXT: fcvt.w.d a4, fa5, rmm |
| ; RV64-i32-NEXT: fcvt.w.d a5, fa4, rmm |
| ; RV64-i32-NEXT: fcvt.w.d a6, fa3, rmm |
| ; RV64-i32-NEXT: vsetivli zero, 8, e32, m2, ta, ma |
| ; RV64-i32-NEXT: vmv.v.x v8, a0 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a1 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a2 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a3 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a4 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a5 |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a6 |
| ; RV64-i32-NEXT: fcvt.w.d a0, fa2, rmm |
| ; RV64-i32-NEXT: vslide1down.vx v8, v8, a0 |
| ; RV64-i32-NEXT: addi sp, s0, -128 |
| ; RV64-i32-NEXT: ld ra, 120(sp) # 8-byte Folded Reload |
| ; RV64-i32-NEXT: ld s0, 112(sp) # 8-byte Folded Reload |
| ; RV64-i32-NEXT: addi sp, sp, 128 |
| ; RV64-i32-NEXT: ret |
| ; |
| ; RV64-i64-LABEL: lround_v8f64: |
| ; RV64-i64: # %bb.0: |
| ; RV64-i64-NEXT: addi sp, sp, -192 |
| ; RV64-i64-NEXT: sd ra, 184(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: sd s0, 176(sp) # 8-byte Folded Spill |
| ; RV64-i64-NEXT: addi s0, sp, 192 |
| ; RV64-i64-NEXT: andi sp, sp, -64 |
| ; RV64-i64-NEXT: mv a0, sp |
| ; RV64-i64-NEXT: vsetivli zero, 8, e64, m4, ta, ma |
| ; RV64-i64-NEXT: vse64.v v8, (a0) |
| ; RV64-i64-NEXT: fld fa5, 56(sp) |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v8 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e64, m1, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 1 |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa4, rmm |
| ; RV64-i64-NEXT: fcvt.l.d a1, fa5, rmm |
| ; RV64-i64-NEXT: sd a1, 120(sp) |
| ; RV64-i64-NEXT: fld fa5, 48(sp) |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i64-NEXT: vsetivli zero, 1, e64, m2, ta, ma |
| ; RV64-i64-NEXT: vslidedown.vi v10, v8, 3 |
| ; RV64-i64-NEXT: fcvt.l.d a1, fa4, rmm |
| ; RV64-i64-NEXT: fcvt.l.d a2, fa5, rmm |
| ; RV64-i64-NEXT: sd a2, 112(sp) |
| ; RV64-i64-NEXT: fld fa5, 40(sp) |
| ; RV64-i64-NEXT: vfmv.f.s fa4, v10 |
| ; RV64-i64-NEXT: fcvt.l.d a2, fa4, rmm |
| ; RV64-i64-NEXT: vslidedown.vi v8, v8, 2 |
| ; RV64-i64-NEXT: fcvt.l.d a3, fa5, rmm |
| ; RV64-i64-NEXT: vfmv.f.s fa5, v8 |
| ; RV64-i64-NEXT: sd a3, 104(sp) |
| ; RV64-i64-NEXT: fcvt.l.d a3, fa5, rmm |
| ; RV64-i64-NEXT: fld fa5, 32(sp) |
| ; RV64-i64-NEXT: sd a0, 64(sp) |
| ; RV64-i64-NEXT: sd a1, 72(sp) |
| ; RV64-i64-NEXT: sd a3, 80(sp) |
| ; RV64-i64-NEXT: sd a2, 88(sp) |
| ; RV64-i64-NEXT: fcvt.l.d a0, fa5, rmm |
| ; RV64-i64-NEXT: sd a0, 96(sp) |
| ; RV64-i64-NEXT: addi a0, sp, 64 |
| ; RV64-i64-NEXT: vsetivli zero, 8, e64, m4, ta, ma |
| ; RV64-i64-NEXT: vle64.v v8, (a0) |
| ; RV64-i64-NEXT: addi sp, s0, -192 |
| ; RV64-i64-NEXT: ld ra, 184(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: ld s0, 176(sp) # 8-byte Folded Reload |
| ; RV64-i64-NEXT: addi sp, sp, 192 |
| ; RV64-i64-NEXT: ret |
| %a = call <8 x iXLen> @llvm.lround.v8iXLen.v8f64(<8 x double> %x) |
| ret <8 x iXLen> %a |
| } |
| declare <8 x iXLen> @llvm.lround.v8iXLen.v8f64(<8 x double>) |
