66 lines
2.1 KiB
LLVM
66 lines
2.1 KiB
LLVM
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
||
|
|
; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s
|
||
|
|
|
||
|
|
;; fsqrt
|
||
|
|
define void @sqrt_v4f32(ptr %res, ptr %a0) nounwind {
|
||
|
|
; CHECK-LABEL: sqrt_v4f32:
|
||
|
|
; CHECK: # %bb.0: # %entry
|
||
|
|
; CHECK-NEXT: vld $vr0, $a1, 0
|
||
|
|
; CHECK-NEXT: vfsqrt.s $vr0, $vr0
|
||
|
|
; CHECK-NEXT: vst $vr0, $a0, 0
|
||
|
|
; CHECK-NEXT: ret
|
||
|
|
entry:
|
||
|
|
%v0 = load <4 x float>, ptr %a0, align 16
|
||
|
|
%sqrt = call <4 x float> @llvm.sqrt.v4f32 (<4 x float> %v0)
|
||
|
|
store <4 x float> %sqrt, ptr %res, align 16
|
||
|
|
ret void
|
||
|
|
}
|
||
|
|
|
||
|
|
define void @sqrt_v2f64(ptr %res, ptr %a0) nounwind {
|
||
|
|
; CHECK-LABEL: sqrt_v2f64:
|
||
|
|
; CHECK: # %bb.0: # %entry
|
||
|
|
; CHECK-NEXT: vld $vr0, $a1, 0
|
||
|
|
; CHECK-NEXT: vfsqrt.d $vr0, $vr0
|
||
|
|
; CHECK-NEXT: vst $vr0, $a0, 0
|
||
|
|
; CHECK-NEXT: ret
|
||
|
|
entry:
|
||
|
|
%v0 = load <2 x double>, ptr %a0, align 16
|
||
|
|
%sqrt = call <2 x double> @llvm.sqrt.v2f64 (<2 x double> %v0)
|
||
|
|
store <2 x double> %sqrt, ptr %res, align 16
|
||
|
|
ret void
|
||
|
|
}
|
||
|
|
|
||
|
|
;; 1.0 / (fsqrt vec)
|
||
|
|
define void @one_div_sqrt_v4f32(ptr %res, ptr %a0) nounwind {
|
||
|
|
; CHECK-LABEL: one_div_sqrt_v4f32:
|
||
|
|
; CHECK: # %bb.0: # %entry
|
||
|
|
; CHECK-NEXT: vld $vr0, $a1, 0
|
||
|
|
; CHECK-NEXT: vfrsqrt.s $vr0, $vr0
|
||
|
|
; CHECK-NEXT: vst $vr0, $a0, 0
|
||
|
|
; CHECK-NEXT: ret
|
||
|
|
entry:
|
||
|
|
%v0 = load <4 x float>, ptr %a0, align 16
|
||
|
|
%sqrt = call <4 x float> @llvm.sqrt.v4f32 (<4 x float> %v0)
|
||
|
|
%div = fdiv <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
|
||
|
|
store <4 x float> %div, ptr %res, align 16
|
||
|
|
ret void
|
||
|
|
}
|
||
|
|
|
||
|
|
define void @one_div_sqrt_v2f64(ptr %res, ptr %a0) nounwind {
|
||
|
|
; CHECK-LABEL: one_div_sqrt_v2f64:
|
||
|
|
; CHECK: # %bb.0: # %entry
|
||
|
|
; CHECK-NEXT: vld $vr0, $a1, 0
|
||
|
|
; CHECK-NEXT: vfrsqrt.d $vr0, $vr0
|
||
|
|
; CHECK-NEXT: vst $vr0, $a0, 0
|
||
|
|
; CHECK-NEXT: ret
|
||
|
|
entry:
|
||
|
|
%v0 = load <2 x double>, ptr %a0, align 16
|
||
|
|
%sqrt = call <2 x double> @llvm.sqrt.v2f64 (<2 x double> %v0)
|
||
|
|
%div = fdiv <2 x double> <double 1.0, double 1.0>, %sqrt
|
||
|
|
store <2 x double> %div, ptr %res, align 16
|
||
|
|
ret void
|
||
|
|
}
|
||
|
|
|
||
|
|
declare <4 x float> @llvm.sqrt.v4f32(<4 x float>)
|
||
|
|
declare <2 x double> @llvm.sqrt.v2f64(<2 x double>)
|