.text
	.allow_optimizations
	.file	"3"
	.section	.text.__runCodelet_LayerNormMoments,"ax",@progbits
	.globl	__runCodelet_LayerNormMoments   # -- Begin function __runCodelet_LayerNormMoments
	.p2align	2
	.type	__runCodelet_LayerNormMoments,@function
__runCodelet_LayerNormMoments:          # @__runCodelet_LayerNormMoments
	.worker
.Lfunc_begin0:
# %bb.0:                                # %entry
	get $m0, 1
	ld32 $m3, $m13, $m15, 4
	and $m1, $m0, 7
	ld32 $m0, $m13, $m15, 3
	ld32 $m2, $m13, $m15, 0
	brz $m3, .LBB0_5
# %bb.1:                                # %for.body11.preheader.i
	ld64 $ap0, $m0, $m15, 0
	{
		ld64 $ap1, $m0, $m15, 1
		f32v2add $ap0, $ap0, $apzero
	}
	{
		ld64 $ap1, $m0, $m15, 2
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m0, $m15, 3
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m0, $m15, 4
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m0, $m15, 5
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld32 $a1, $m13, $m15, 2
		f32v2add $ap1, $ap0, $ap1
	}
	{
		cmpeq $m3, $m3, 1
		f32add $a0, $a2, $a3
	}
	f32fromui32 $a2, $a1
	{
		brz $m3, .LBB0_15
		f32div $a0, $a0, $a2
	}
# %bb.2:                                # %for.cond25.preheader.i
	mov	$m3, $a1
	shr $m3, $m3, 1
	cmpult $m4, $m1, $m3
	brz $m4, .LBB0_3
# %bb.11:                               # %for.body30.lr.ph.i
	{
		xnor $m4, $m1, $m15
		zero	$ap1
	}
	add $m3, $m4, $m3
	setzi $m4, 43691
	sort4x16lo $m5, $m3, $m15
	shr $m3, $m3, 16
	mul $m6, $m5, $m4
	mul $m4, $m3, $m4
	shr $m6, $m6, 16
	add $m4, $m4, $m6
	setzi $m6, 43690
	sort4x16lo $m7, $m4, $m15
	mul $m5, $m5, $m6
	shr $m4, $m4, 16
	mul $m3, $m3, $m6
	add $m5, $m5, $m7
	add $m3, $m3, $m4
	shr $m4, $m5, 16
	add $m3, $m3, $m4
	shl $m4, $m1, 2
	{
		shr $m3, $m3, 2
		mov	$a1, $a0
	}
	add $m2, $m2, $m4
	add $m3, $m3, 1
	ld32step $a4, $m15, $m2+=, 6
	{
		setzi $m4, 1
		f16v2tof32 $ap2, $a4
	}
	{
		cmpslt $m4, $m4, $m3
		f32v2sub $ap2, $ap2, $ap0
	}
	{
		add $m3, $m3, -1
		f32v2mul $ap2, $ap2, $ap2
	}
	brz $m4, .LBB0_14
# %bb.12:                               # %for.body30.i
	add $m3, $m3, -1
.LBB0_13:                               # %for.body30.i
                                        # =>This Inner Loop Header: Depth=1
	{
		ld32step $a6, $m15, $m2+=, 6
		f32v2add $ap1, $ap2, $ap1
	}
	f16v2tof32 $ap2, $a6
	f32v2sub $ap2, $ap2, $ap0
	{
		brnzdec $m3, .LBB0_13
		f32v2mul $ap2, $ap2, $ap2
	}
.LBB0_14:
	{
		bri .LBB0_4
		f32v2add $ap0, $ap2, $ap1
	}
.LBB0_5:                                # %for.cond.preheader.i
	ld32 $m3, $m13, $m15, 2
	{
		shr $m3, $m3, 1
		zero	$ap0
	}
	cmpult $m4, $m1, $m3
	brz $m4, .LBB0_6
# %bb.7:                                # %for.body.i.preheader
	xnor $m4, $m1, $m15
	add $m3, $m4, $m3
	setzi $m4, 43691
	sort4x16lo $m5, $m3, $m15
	shr $m3, $m3, 16
	mul $m6, $m5, $m4
	mul $m4, $m3, $m4
	shr $m6, $m6, 16
	add $m4, $m4, $m6
	setzi $m6, 43690
	sort4x16lo $m7, $m4, $m15
	mul $m5, $m5, $m6
	shr $m4, $m4, 16
	mul $m3, $m3, $m6
	add $m5, $m5, $m7
	add $m3, $m3, $m4
	shr $m4, $m5, 16
	add $m3, $m3, $m4
	shl $m4, $m1, 2
	shr $m3, $m3, 2
	add $m2, $m2, $m4
	add $m3, $m3, 1
	setzi $m4, 1
	ld32step $a2, $m15, $m2+=, 6
	{
		cmpslt $m4, $m4, $m3
		f16v2tof32 $ap1, $a2
	}
	add $m3, $m3, -1
	brz $m4, .LBB0_10
# %bb.8:                                # %for.body.i
	add $m3, $m3, -1
.LBB0_9:                                # %for.body.i
                                        # =>This Inner Loop Header: Depth=1
	{
		ld32step $a4, $m15, $m2+=, 6
		f32v2add $ap0, $ap0, $ap1
	}
	{
		brnzdec $m3, .LBB0_9
		f16v2tof32 $ap1, $a4
	}
.LBB0_10:
	f32v2add $ap0, $ap0, $ap1
	st64 $ap0, $m0, $m15, $m1
	exitz $m15
.LBB0_15:                               # %if.else.i
	brz $m1, .LBB0_16
# %bb.17:                               # %_ZN16LayerNormMoments7computeEj.exit
	exitz $m15
.LBB0_6:                                # %for.cond.cleanup.i
	st64 $ap0, $m0, $m15, $m1
	exitz $m15
.LBB0_3:
	zero	$ap0
.LBB0_4:                                # %for.cond.cleanup29.i
	shl $m1, $m1, 3
	st64 $ap0, $m1, $m0, 6
	exitz $m15
.LBB0_16:                               # %for.body45.preheader.i
	ld64 $ap1, $m0, $m15, 6
	{
		ld64 $ap2, $m0, $m15, 7
		f32v2add $ap1, $ap1, $apzero
	}
	{
		ld64 $ap2, $m0, $m15, 8
		f32v2add $ap1, $ap1, $ap2
	}
	{
		ld64 $ap2, $m0, $m15, 9
		f32v2add $ap1, $ap1, $ap2
	}
	{
		ld64 $ap2, $m0, $m15, 10
		f32v2add $ap1, $ap1, $ap2
	}
	{
		ld64 $ap2, $m0, $m15, 11
		f32v2add $ap1, $ap1, $ap2
	}
	{
		ld32 $m0, $m13, $m15, 1
		f32v2add $ap1, $ap1, $ap2
	}
	f32add $a1, $a2, $a3
	st64 $ap0, $m0, $m15, 0
	exitz $m15
.Lfunc_end0:
	.size	__runCodelet_LayerNormMoments, .Lfunc_end0-__runCodelet_LayerNormMoments
	.section	.stack_sizes,"o",@progbits,.text.__runCodelet_LayerNormMoments
	.long	.Lfunc_begin0
	.byte	0
	.section	.text.__runCodelet_LayerNormMoments,"ax",@progbits
                                        # -- End function
	.ident	"clang version 16.0.0 (git@github.com:Software-GCAI/llvm-project.git 6343a843cd5af31429c62a83b6ba249a74d36969)"
	.section	".note.GNU-stack","",@progbits
	.addrsig