#include "/srv/home/gc-sdk/ipu-stack/device/static_runtime.S"
#include "/srv/home/gc-sdk/ipu-stack/device/worker_support.S"
	.text
	.allow_optimizations
	.file	"3"
	.section	.text.__runCodelet_AddF16,"ax",@progbits
	.globl	__runCodelet_AddF16             # -- Begin function __runCodelet_AddF16
	.p2align	2
	.type	__runCodelet_AddF16,@function
__runCodelet_AddF16:                    # @__runCodelet_AddF16
	.worker
.Lcode0_func_begin0:
# %bb.0:                                # %entry
	setzi $m6, 8
	cmpult $m6, $m6, __worker_stack_max_growth_plus_one
	brz $m6, _stackoverflow
	add $m11, $m12, -8
	get $m0, 1
	and $m0, $m0, 7
	ld32 $m8, $m13, $m15, 3
	shr $m6, $m8, 1
	brz $m0, .Lcode0_BB0_2
# %bb.1:                                # %entry
	cmpult $m1, $m0, $m6
	brnz $m1, .Lcode0_BB0_2
.Lcode0_BB0_53:                               # %_ZN6AddF167computeEj.exit
	exitz $m15
.Lcode0_BB0_2:                                # %if.end.i
	ld32 $m1, $m13, $m15, 4
	or $m2, $m8, $m1
	and $m2, $m2, 1
	brnz $m2, .Lcode0_BB0_31
# %bb.3:                                # %land.lhs.true7.i
	ld32 $m7, $m13, $m15, 5
	and $m2, $m7, 1
	brnz $m2, .Lcode0_BB0_31
# %bb.4:                                # %if.then10.i
	cmpne $m2, $m1, $m8
	ld32 $m3, $m13, $m15, 0
	ld32 $m5, $m13, $m15, 1
	ld32 $m10, $m13, $m15, 2
	cmpne $m9, $m7, $m8
	brnz $m2, .Lcode0_BB0_9
# %bb.5:                                # %if.then10.i
	brnz $m9, .Lcode0_BB0_9
# %bb.6:                                # %for.cond.preheader.i
	cmpult $m1, $m0, $m6
	brz $m1, .Lcode0_BB0_53
# %bb.7:                                # %for.body.i.preheader
	shl $m6, $m0, 2
	add $m1, $m10, $m6
	add $m2, $m5, $m6
	add $m3, $m3, $m6
.Lcode0_BB0_8:                                # %for.body.i
                                        # =>This Inner Loop Header: Depth=1
	ld32step $a0, $m15, $m3+=, 6
	ld32step $a1, $m15, $m2+=, 6
	f16v2add $a0, $a0, $a1
	st32step $a0, $m15, $m1+=, 6
	ld32 $m4, $m13, $m15, 3
	add $m0, $m0, 6
	shr $m4, $m4, 1
	cmpult $m4, $m0, $m4
	brnz $m4, .Lcode0_BB0_8
	bri .Lcode0_BB0_53
.Lcode0_BB0_31:                               # %if.end121.i
	shl $m4, $m0, 1
	and $m7, $m8, 1
	add $m5, $m1, $m4
	mov	$m2, $m4
.Lcode0_BB0_32:                               # %while.cond.i290.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m3, $m2, $m1
	sub $m2, $m2, $m1
	sub $m5, $m5, $m1
	brz $m3, .Lcode0_BB0_32
# %bb.33:                               # %_ZN6AddF164wrapEjj.exit291.i
	ld32 $m3, $m13, $m15, 5
	add $m6, $m3, $m4
	mov	$m2, $m4
.Lcode0_BB0_34:                               # %while.cond.i295.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m9, $m2, $m3
	sub $m2, $m2, $m3
	sub $m6, $m6, $m3
	brz $m9, .Lcode0_BB0_34
# %bb.35:                               # %for.cond131.preheader.i
	or $m2, $m4, 1
	cmpult $m2, $m2, $m8
	brz $m2, .Lcode0_BB0_46
.Lcode0_BB0_36:                               # %for.body136.i
                                        # =>This Loop Header: Depth=1
                                        #     Child Loop BB0_37 Depth 2
                                        #     Child Loop BB0_39 Depth 2
                                        #     Child Loop BB0_41 Depth 2
                                        #     Child Loop BB0_43 Depth 2
	shl $m8, $m5, 1
	shl $m2, $m1, 1
	ld32 $m7, $m13, $m15, 0
	add $m8, $m8, $m2
	ldb16 $a0, $m7, $m15, $m5
	add $m9, $m5, 1
	add $m8, $m8, 2
.Lcode0_BB0_37:                               # %while.cond.i300.i
                                        #   Parent Loop BB0_36 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	cmpult $m10, $m9, $m1
	sub $m9, $m9, $m1
	sub $m8, $m8, $m2
	brz $m10, .Lcode0_BB0_37
# %bb.38:                               # %_ZN6AddF164wrapEjj.exit301.i
                                        #   in Loop: Header=BB0_36 Depth=1
	shl $m1, $m6, 1
	shl $m2, $m3, 1
	add $m9, $m1, $m2
	ldb16 $a1, $m7, $m8, 0
	ld32 $m1, $m13, $m15, 1
	ldb16 $a2, $m1, $m15, $m6
	add $m8, $m6, 1
	add $m7, $m9, 2
.Lcode0_BB0_39:                               # %while.cond.i305.i
                                        #   Parent Loop BB0_36 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	cmpult $m9, $m8, $m3
	sub $m8, $m8, $m3
	sub $m7, $m7, $m2
	brz $m9, .Lcode0_BB0_39
# %bb.40:                               # %_ZN6AddF164wrapEjj.exit306.i
                                        #   in Loop: Header=BB0_36 Depth=1
	{
		ldb16 $a1, $m1, $m7, 0
		sort4x16lo $a0, $a0, $a1
	}
	{
		ld32 $m1, $m13, $m15, 2
		sort4x16lo $a1, $a2, $a1
	}
	{
		shl $m2, $m4, 1
		f16v2add $a0, $a0, $a1
	}
	st32 $a0, $m1, $m2, 0
	add $m2, $m5, 12
	ld32 $m1, $m13, $m15, 4
	add $m5, $m2, $m1
.Lcode0_BB0_41:                               # %while.cond.i310.i
                                        #   Parent Loop BB0_36 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	cmpult $m3, $m2, $m1
	sub $m2, $m2, $m1
	sub $m5, $m5, $m1
	brz $m3, .Lcode0_BB0_41
# %bb.42:                               # %_ZN6AddF164wrapEjj.exit311.i
                                        #   in Loop: Header=BB0_36 Depth=1
	add $m2, $m6, 12
	ld32 $m3, $m13, $m15, 5
	add $m6, $m2, $m3
.Lcode0_BB0_43:                               # %while.cond.i315.i
                                        #   Parent Loop BB0_36 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	cmpult $m7, $m2, $m3
	sub $m2, $m2, $m3
	sub $m6, $m6, $m3
	brz $m7, .Lcode0_BB0_43
# %bb.44:                               # %_ZN6AddF164wrapEjj.exit316.i
                                        #   in Loop: Header=BB0_36 Depth=1
	add $m4, $m4, 12
	or $m2, $m4, 1
	ld32 $m8, $m13, $m15, 3
	cmpult $m2, $m2, $m8
	brnz $m2, .Lcode0_BB0_36
# %bb.45:                               # %for.cond.cleanup135.loopexit.i
	and $m7, $m8, 1
.Lcode0_BB0_46:                               # %for.cond.cleanup135.i
	brnz $m0, .Lcode0_BB0_53
# %bb.47:                               # %for.cond.cleanup135.i
	brz $m7, .Lcode0_BB0_53
# %bb.48:                               # %if.then173.i
	shl $m4, $m8, 1
	shl $m5, $m1, 1
	add $m0, $m8, -1
	add $m2, $m4, $m5
	add $m6, $m2, -2
	mov	$m2, $m0
.Lcode0_BB0_49:                               # %while.cond.i320.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m7, $m2, $m1
	sub $m2, $m2, $m1
	sub $m6, $m6, $m5
	brz $m7, .Lcode0_BB0_49
# %bb.50:                               # %_ZN6AddF164wrapEjj.exit321.i
	shl $m2, $m3, 1
	add $m1, $m4, $m2
	add $m1, $m1, -2
	ld32 $m4, $m13, $m15, 0
	ldb16 $a0, $m4, $m6, 0
	mov	$m4, $m0
.Lcode0_BB0_51:                               # %while.cond.i325.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m5, $m4, $m3
	sub $m4, $m4, $m3
	sub $m1, $m1, $m2
	brz $m5, .Lcode0_BB0_51
# %bb.52:                               # %_ZN6AddF164wrapEjj.exit326.i
	ld32 $m2, $m13, $m15, 1
	shl $m0, $m0, 1
	ldb16 $a1, $m2, $m1, 0
	{
		ld32 $m1, $m13, $m15, 2
		f16v2add $a0, $a0, $a1
	}
	{
		andc $m0, $m0, 3
		sort4x16lo $a0, $a0, $azero
	}
	ld32 $m2, $m1, $m0, 0
	setzi $m3, 65535
	andc $m2, $m2, $m3
	mov	$m3, $a0
	or $m2, $m2, $m3
	st32 $m2, $m1, $m0, 0
	exitz $m15
.Lcode0_BB0_9:                                # %if.else.i
	cmpeq $m2, $m1, $m8
	brnz $m2, .Lcode0_BB0_11
# %bb.10:                               # %if.else.i
	brnz $m9, .Lcode0_BB0_21
.Lcode0_BB0_11:                               # %land.lhs.true33.i
	st32 $m2, $m11, $m15, 1                 # 4-byte Folded Spill
	st32 $m10, $m11, $m15, 0                # 4-byte Folded Spill
	cmpult $m2, $m1, $m7
	mov	$m9, $m7
	movnz	$m9, $m2, $m1
	sub $m10, 0, $m9
	mov	$m2, $m8
.Lcode0_BB0_12:                               # %while.cond.i.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m4, $m2, $m9
	sub $m2, $m2, $m9
	add $m10, $m10, $m9
	brz $m4, .Lcode0_BB0_12
# %bb.13:                               # %_ZN6AddF164wrapEjj.exit.i
	cmpne $m2, $m8, $m10
	ld32 $m10, $m11, $m15, 0                # 4-byte Folded Reload
	ld32 $m4, $m11, $m15, 1                 # 4-byte Folded Reload
	brnz $m2, .Lcode0_BB0_21
# %bb.14:                               # %if.then42.i
	cmpult $m2, $m8, 2
	brnz $m2, .Lcode0_BB0_53
# %bb.15:                               # %for.cond74.preheader.lr.ph.i
	mov	$m6, $m15
	movnz	$m1, $m4, $m7
	shr $m1, $m1, 1
	xnor $m2, $m0, $m15
	add $m2, $m2, $m1
	setzi $m4, 43691
	sort4x16lo $m7, $m2, $m15
	shr $m2, $m2, 16
	mul $m9, $m7, $m4
	mul $m4, $m2, $m4
	shr $m9, $m9, 16
	add $m4, $m4, $m9
	setzi $m9, 43690
	mul $m7, $m7, $m9
	mul $m2, $m2, $m9
	sort4x16lo $m9, $m4, $m15
	shr $m4, $m4, 16
	add $m7, $m7, $m9
	add $m2, $m2, $m4
	shr $m4, $m7, 16
	mov	$m7, $m5
	add $m2, $m2, $m4
	ld32 $m4, $m11, $m15, 1                 # 4-byte Folded Reload
	movnz	$m7, $m4, $m3
	movnz	$m3, $m4, $m5
	shl $m9, $m0, 2
	shr $m2, $m2, 2
	add $m3, $m3, $m9
	st32 $m3, $m11, $m15, 1                 # 4-byte Folded Spill
	add $m4, $m10, $m9
	shl $m5, $m1, 2
	add $m7, $m7, $m9
	add $m2, $m2, 1
	st32 $m2, $m11, $m15, 0                 # 4-byte Folded Spill
	bri .Lcode0_BB0_16
.Lcode0_BB0_20:                               # %for.cond.cleanup76.i
                                        #   in Loop: Header=BB0_16 Depth=1
	add $m6, $m6, $m1
	shr $m2, $m8, 1
	cmpult $m2, $m6, $m2
	add $m4, $m4, $m5
	add $m7, $m7, $m5
	brz $m2, .Lcode0_BB0_53
.Lcode0_BB0_16:                               # %for.cond74.preheader.i
                                        # =>This Loop Header: Depth=1
                                        #     Child Loop BB0_18 Depth 2
	cmpult $m2, $m0, $m1
	brz $m2, .Lcode0_BB0_20
# %bb.17:                               # %for.body77.i.preheader
                                        #   in Loop: Header=BB0_16 Depth=1
	ld32 $m2, $m11, $m15, 0                 # 4-byte Folded Reload
	mov	$m9, $m7
	mov	$m10, $m4
	ld32 $m3, $m11, $m15, 1                 # 4-byte Folded Reload
	add $m2, $m2, -1
.Lcode0_BB0_18:                               # %for.body77.i
                                        #   Parent Loop BB0_16 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	ld32step $a0, $m15, $m9+=, 6
	ld32step $a1, $m15, $m3+=, 6
	f16v2add $a0, $a0, $a1
	st32step $a0, $m15, $m10+=, 6
	brnzdec $m2, .Lcode0_BB0_18
# %bb.19:                               # %for.cond.cleanup76.loopexit.i
                                        #   in Loop: Header=BB0_16 Depth=1
	ld32 $m8, $m13, $m15, 3
	bri .Lcode0_BB0_20
.Lcode0_BB0_21:                               # %if.else90.i
	shr $m2, $m1, 1
	add $m1, $m2, $m0
	mov	$m8, $m0
.Lcode0_BB0_22:                               # %while.cond.i270.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m4, $m8, $m2
	sub $m8, $m8, $m2
	sub $m1, $m1, $m2
	brz $m4, .Lcode0_BB0_22
# %bb.23:                               # %_ZN6AddF164wrapEjj.exit271.i
	shr $m7, $m7, 1
	add $m2, $m7, $m0
	mov	$m8, $m0
.Lcode0_BB0_24:                               # %while.cond.i275.i
                                        # =>This Inner Loop Header: Depth=1
	cmpult $m4, $m8, $m7
	sub $m8, $m8, $m7
	sub $m2, $m2, $m7
	brz $m4, .Lcode0_BB0_24
# %bb.25:                               # %for.cond98.preheader.i
	cmpult $m4, $m0, $m6
	brz $m4, .Lcode0_BB0_53
.Lcode0_BB0_26:                               # %for.body103.i
                                        # =>This Loop Header: Depth=1
                                        #     Child Loop BB0_27 Depth 2
                                        #     Child Loop BB0_29 Depth 2
	ld32 $a0, $m3, $m15, $m1
	ld32 $a1, $m5, $m15, $m2
	f16v2add $a0, $a0, $a1
	st32 $a0, $m10, $m15, $m0
	ld32 $m4, $m13, $m15, 4
	add $m7, $m1, 6
	shr $m6, $m4, 1
	add $m1, $m7, $m6
.Lcode0_BB0_27:                               # %while.cond.i280.i
                                        #   Parent Loop BB0_26 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	cmpult $m4, $m7, $m6
	sub $m7, $m7, $m6
	sub $m1, $m1, $m6
	brz $m4, .Lcode0_BB0_27
# %bb.28:                               # %_ZN6AddF164wrapEjj.exit281.i
                                        #   in Loop: Header=BB0_26 Depth=1
	ld32 $m4, $m13, $m15, 5
	add $m7, $m2, 6
	shr $m6, $m4, 1
	add $m2, $m7, $m6
.Lcode0_BB0_29:                               # %while.cond.i285.i
                                        #   Parent Loop BB0_26 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	cmpult $m4, $m7, $m6
	sub $m7, $m7, $m6
	sub $m2, $m2, $m6
	brz $m4, .Lcode0_BB0_29
# %bb.30:                               # %_ZN6AddF164wrapEjj.exit286.i
                                        #   in Loop: Header=BB0_26 Depth=1
	ld32 $m4, $m13, $m15, 3
	add $m0, $m0, 6
	shr $m4, $m4, 1
	cmpult $m4, $m0, $m4
	brnz $m4, .Lcode0_BB0_26
	bri .Lcode0_BB0_53
.Lcode0_func_end0:
	.size	__runCodelet_AddF16, .Lcode0_func_end0-__runCodelet_AddF16
	.section	.stack_sizes,"o",@progbits,.text.__runCodelet_AddF16
	.long	.Lcode0_func_begin0
	.byte	8
	.section	.text.__runCodelet_AddF16,"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

#define WORKER_CALL_SYMBOL add_f16
#define WORKER_CODELET_SYMBOL __runCodelet_AddF16
#define WORKER_ARGUMENTS $m3,$m4,$m2,$m5,$m6,$m7
#define WORKER_FRAME_BYTES 48
#include "/srv/home/gc-sdk/ipu-stack/device/worker_call.S"
#undef WORKER_CALL_SYMBOL
#undef WORKER_CODELET_SYMBOL
#undef WORKER_ARGUMENTS
#undef WORKER_FRAME_BYTES
	.text
	.allow_optimizations
	.file	"3"
	.section	.text.__runCodelet_LayerNormF16,"ax",@progbits
	.globl	__runCodelet_LayerNormF16       # -- Begin function __runCodelet_LayerNormF16
	.p2align	2
	.type	__runCodelet_LayerNormF16,@function
__runCodelet_LayerNormF16:              # @__runCodelet_LayerNormF16
	.worker
.Lcode1_func_begin0:
# %bb.0:                                # %entry
	setzi $m6, 8
	cmpult $m6, $m6, __worker_stack_max_growth_plus_one
	brz $m6, _stackoverflow
	add $m11, $m12, -8
	get $m0, 1
	ld32 $m2, $m13, $m15, 7
	and $m0, $m0, 7
	ld32 $m1, $m13, $m15, 6
	brz $m2, .Lcode1_BB0_6
# %bb.1:                                # %for.body10.preheader.i
	ld64 $ap0, $m1, $m15, 0
	{
		ld64 $ap1, $m1, $m15, 1
		f32v2add $ap0, $ap0, $apzero
	}
	{
		ld64 $ap1, $m1, $m15, 2
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m1, $m15, 3
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m1, $m15, 4
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m1, $m15, 5
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld32 $a3, $m13, $m15, 5
		f32v2add $ap0, $ap0, $ap1
	}
	{
		cmpeq $m3, $m2, 1
		f32add $a0, $a0, $a1
	}
	{
		mov	$m2, $a3
		f32fromui32 $a2, $a3
	}
	{
		shr $m2, $m2, 1
		f32div $a0, $a0, $a2
	}
	brz $m3, .Lcode1_BB0_2
# %bb.11:                               # %for.cond24.preheader.i
	cmpult $m3, $m0, $m2
	brz $m3, .Lcode1_BB0_12
# %bb.15:                               # %for.body29.lr.ph.i
	{
		xnor $m3, $m0, $m15
		zero	$ap1
	}
	add $m2, $m3, $m2
	setzi $m3, 43691
	sort4x16lo $m4, $m2, $m15
	shr $m2, $m2, 16
	mul $m5, $m4, $m3
	mul $m3, $m2, $m3
	shr $m5, $m5, 16
	add $m3, $m3, $m5
	setzi $m5, 43690
	sort4x16lo $m6, $m3, $m15
	mul $m4, $m4, $m5
	shr $m3, $m3, 16
	mul $m2, $m2, $m5
	add $m4, $m4, $m6
	add $m2, $m2, $m3
	shr $m3, $m4, 16
	add $m2, $m2, $m3
	ld32 $m3, $m13, $m15, 0
	shl $m4, $m0, 2
	{
		shr $m5, $m2, 2
		mov	$a1, $a0
	}
	add $m2, $m3, $m4
	add $m3, $m5, 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, .Lcode1_BB0_18
# %bb.16:                               # %for.body29.i
	add $m3, $m3, -1
.Lcode1_BB0_17:                               # %for.body29.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, .Lcode1_BB0_17
		f32v2mul $ap2, $ap2, $ap2
	}
.Lcode1_BB0_18:
	{
		bri .Lcode1_BB0_13
		f32v2add $ap0, $ap2, $ap1
	}
.Lcode1_BB0_6:                                # %for.cond.preheader.i
	ld32 $m2, $m13, $m15, 5
	{
		shr $m2, $m2, 1
		zero	$ap0
	}
	cmpult $m3, $m0, $m2
	brz $m3, .Lcode1_BB0_14
# %bb.7:                                # %for.body.lr.ph.i
	xnor $m3, $m0, $m15
	add $m2, $m3, $m2
	setzi $m3, 43691
	sort4x16lo $m4, $m2, $m15
	shr $m2, $m2, 16
	mul $m5, $m4, $m3
	mul $m3, $m2, $m3
	shr $m5, $m5, 16
	add $m3, $m3, $m5
	setzi $m5, 43690
	sort4x16lo $m6, $m3, $m15
	mul $m4, $m4, $m5
	shr $m3, $m3, 16
	mul $m2, $m2, $m5
	add $m4, $m4, $m6
	add $m2, $m2, $m3
	shr $m3, $m4, 16
	add $m2, $m2, $m3
	ld32 $m3, $m13, $m15, 0
	shl $m4, $m0, 2
	shr $m5, $m2, 2
	add $m2, $m3, $m4
	add $m3, $m5, 1
	setzi $m4, 1
	ld32step $a2, $m15, $m2+=, 6
	{
		cmpslt $m4, $m4, $m3
		f16v2tof32 $ap1, $a2
	}
	add $m3, $m3, -1
	brz $m4, .Lcode1_BB0_10
# %bb.8:                                # %for.body.i
	add $m3, $m3, -1
.Lcode1_BB0_9:                                # %for.body.i
                                        # =>This Inner Loop Header: Depth=1
	{
		ld32step $a4, $m15, $m2+=, 6
		f32v2add $ap0, $ap0, $ap1
	}
	{
		brnzdec $m3, .Lcode1_BB0_9
		f16v2tof32 $ap1, $a4
	}
.Lcode1_BB0_10:
	f32v2add $ap0, $ap0, $ap1
	st64 $ap0, $m1, $m15, $m0
	exitz $m15
.Lcode1_BB0_2:                                # %for.body43.preheader.i
	cmpult $m2, $m0, $m2
	brz $m2, .Lcode1_BB0_5
# %bb.3:                                # %for.body67.lr.ph.i
	{
		ld32 $m2, $m13, $m15, 1
		setzi $a1, 407485
	}
	{
		ld32 $m3, $m13, $m15, 2
		or $a3, $a1, 897581056
	}
	{
		ld32 $m4, $m13, $m15, 3
		mov	$a1, $a0
	}
	shl $m5, $m0, 2
	ld64 $ap2, $m1, $m15, 6
	{
		ld64 $ap3, $m1, $m15, 7
		f32v2add $ap2, $ap2, $apzero
	}
	{
		ld64 $ap3, $m1, $m15, 8
		f32v2add $ap2, $ap2, $ap3
	}
	{
		ld64 $ap3, $m1, $m15, 9
		f32v2add $ap2, $ap2, $ap3
	}
	{
		ld64 $ap3, $m1, $m15, 10
		f32v2add $ap2, $ap2, $ap3
	}
	{
		ld64 $ap3, $m1, $m15, 11
		f32v2add $ap2, $ap2, $ap3
	}
	f32v2add $ap2, $ap2, $ap3
	f32add $a4, $a4, $a5
	f32div $a2, $a4, $a2
	f32add $a2, $a2, $a3
	f32oorx $a2, $a2
	mov	$a3, $a2
	st64 $ap1, $m11, $m15, 0                # 8-byte Folded Spill
.Lcode1_BB0_4:                                # %for.body67.i
                                        # =>This Inner Loop Header: Depth=1
	ld32 $m1, $m13, $m15, 0
	add $m0, $m0, 6
	ld32 $a4, $m1, $m5, 0
	{
		ld32 $a6, $m2, $m5, 0
		f16v2tof32 $ap2, $a4
	}
	{
		ld32 $a7, $m3, $m5, 0
		f32v2sub $ap2, $ap2, $ap0
	}
	ld64 $ap1, $m11, $m15, 0                # 8-byte Folded Reload
	f32v2mul $ap2, $ap1, $ap2
	mov	$ap1, $ap0
	f16v2tof32 $ap0, $a6
	f32v2mul $ap0, $ap2, $ap0
	f16v2tof32 $ap2, $a7
	f32v2add $ap0, $ap0, $ap2
	f32v2tof16 $a0, $ap0
	{
		st32step $a0, $m4, $m5+=, 6
		mov	$ap0, $ap1
	}
	ld32 $m1, $m13, $m15, 5
	shr $m1, $m1, 1
	cmpult $m1, $m0, $m1
	brnz $m1, .Lcode1_BB0_4
.Lcode1_BB0_5:                                # %_ZN12LayerNormF167computeEj.exit
	exitz $m15
.Lcode1_BB0_12:
	zero	$ap0
.Lcode1_BB0_13:                               # %for.cond.cleanup28.i
	add $m0, $m0, 6
.Lcode1_BB0_14:                               # %cleanup86.sink.split.i
	st64 $ap0, $m1, $m15, $m0
	exitz $m15
.Lcode1_func_end0:
	.size	__runCodelet_LayerNormF16, .Lcode1_func_end0-__runCodelet_LayerNormF16
	.section	.stack_sizes,"o",@progbits,.text.__runCodelet_LayerNormF16
	.long	.Lcode1_func_begin0
	.byte	8
	.section	.text.__runCodelet_LayerNormF16,"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

#include "arch/gc_tile_defines.h"

#ifdef NORM_WITH_ADD
#define NORM_SYMBOL add_layer_norm_f16
#define NORM_VERTEX __runCodelet_AddLayerNormF16
#define NORM_ROWS $m7
#define NORM_WIDTH $m8
#define NORM_GAMMA $m5
#define NORM_BETA $m6
#define NORM_SHIFT 1
#define NORM_STACK 160
#define NORM_SCRATCH 64
#else
#define NORM_SYMBOL layer_norm_f16
#define NORM_VERTEX __runCodelet_LayerNormF16
#define NORM_ROWS $m6
#define NORM_WIDTH $m7
#define NORM_GAMMA $m4
#define NORM_BETA $m5
#define NORM_SHIFT 0
#define NORM_STACK 128
#define NORM_SCRATCH 32
#endif

// Share each row across all six workers. Separate sum and centered-variance
// buffers avoid read/write races; local sync separates the three passes.
	.text
	.allow_optimizations
	.section .text.NORM_SYMBOL,"ax",@progbits
	.globl NORM_SYMBOL
	.p2align 2
	.type NORM_SYMBOL,@function
NORM_SYMBOL:
	.supervisor
	add $m11, $m11, -NORM_STACK
	st32 $m3, $m11, $m15, 0
#ifdef NORM_WITH_ADD
	st32 $m4, $m11, $m15, 1
#endif
	st32 NORM_GAMMA, $m11, $m15, (1 + NORM_SHIFT)
	st32 NORM_BETA, $m11, $m15, (2 + NORM_SHIFT)
	st32 $m2, $m11, $m15, (3 + NORM_SHIFT)
	st32 NORM_ROWS, $m11, $m15, (4 + NORM_SHIFT)
	st32 NORM_WIDTH, $m11, $m15, (5 + NORM_SHIFT)
	add $m0, $m11, NORM_SCRATCH
	st32 $m0, $m11, $m15, (6 + NORM_SHIFT)
	setzi $m0, NORM_VERTEX
.Lwrap1_row:
	setzi $m1, 0
.Lwrap1_stage:
	st32 $m1, $m11, $m15, (7 + NORM_SHIFT)
	runall $m0, $m11, 0
	sync TEXCH_SYNCZONE_LOCAL
	add $m1, $m1, 1
	cmpeq $m2, $m1, 3
	brz $m2, .Lwrap1_stage
	shl $m2, NORM_WIDTH, 1
	ld32 $m3, $m11, $m15, 0
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, 0
	ld32 $m3, $m11, $m15, (3 + NORM_SHIFT)
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, (3 + NORM_SHIFT)
#ifdef NORM_WITH_ADD
	ld32 $m3, $m11, $m15, 1
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, 1
#endif
	sub NORM_ROWS, NORM_ROWS, 1
	brnz NORM_ROWS, .Lwrap1_row
	add $m11, $m11, NORM_STACK
	br $m10
	.size NORM_SYMBOL, .-NORM_SYMBOL

#undef NORM_WITH_ADD
#undef NORM_SYMBOL
#undef NORM_VERTEX
#undef NORM_ROWS
#undef NORM_WIDTH
#undef NORM_GAMMA
#undef NORM_BETA
#undef NORM_SHIFT
#undef NORM_STACK
#undef NORM_SCRATCH
	.text
	.allow_optimizations
	.file	"3"
	.section	.text.__runCodelet_AddLayerNormF16,"ax",@progbits
	.globl	__runCodelet_AddLayerNormF16    # -- Begin function __runCodelet_AddLayerNormF16
	.p2align	2
	.type	__runCodelet_AddLayerNormF16,@function
__runCodelet_AddLayerNormF16:           # @__runCodelet_AddLayerNormF16
	.worker
.Lcode2_func_begin0:
# %bb.0:                                # %entry
	setzi $m6, 8
	cmpult $m6, $m6, __worker_stack_max_growth_plus_one
	brz $m6, _stackoverflow
	add $m11, $m12, -8
	get $m0, 1
	ld32 $m2, $m13, $m15, 8
	and $m0, $m0, 7
	ld32 $m1, $m13, $m15, 7
	brz $m2, .Lcode2_BB0_6
# %bb.1:                                # %for.body10.preheader.i
	ld64 $ap0, $m1, $m15, 0
	{
		ld64 $ap1, $m1, $m15, 1
		f32v2add $ap0, $ap0, $apzero
	}
	{
		ld64 $ap1, $m1, $m15, 2
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m1, $m15, 3
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m1, $m15, 4
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld64 $ap1, $m1, $m15, 5
		f32v2add $ap0, $ap0, $ap1
	}
	{
		ld32 $a3, $m13, $m15, 6
		f32v2add $ap0, $ap0, $ap1
	}
	{
		cmpeq $m3, $m2, 1
		f32add $a0, $a0, $a1
	}
	{
		mov	$m2, $a3
		f32fromui32 $a2, $a3
	}
	{
		shr $m2, $m2, 1
		f32div $a0, $a0, $a2
	}
	brz $m3, .Lcode2_BB0_2
# %bb.11:                               # %for.cond24.preheader.i
	cmpult $m3, $m0, $m2
	brz $m3, .Lcode2_BB0_12
# %bb.15:                               # %for.body29.lr.ph.i
	{
		xnor $m3, $m0, $m15
		zero	$ap1
	}
	add $m2, $m3, $m2
	setzi $m3, 43691
	sort4x16lo $m4, $m2, $m15
	shr $m2, $m2, 16
	mul $m5, $m4, $m3
	mul $m3, $m2, $m3
	shr $m5, $m5, 16
	add $m3, $m3, $m5
	setzi $m5, 43690
	sort4x16lo $m6, $m3, $m15
	mul $m4, $m4, $m5
	shr $m3, $m3, 16
	mul $m2, $m2, $m5
	add $m4, $m4, $m6
	add $m2, $m2, $m3
	shr $m3, $m4, 16
	add $m2, $m2, $m3
	ld32 $m3, $m13, $m15, 0
	ld32 $m4, $m13, $m15, 1
	shl $m5, $m0, 2
	{
		shr $m6, $m2, 2
		mov	$a1, $a0
	}
	add $m2, $m4, $m5
	add $m3, $m3, $m5
	add $m4, $m6, 1
	ld32step $a4, $m15, $m3+=, 6
	ld32step $a5, $m15, $m2+=, 6
	{
		setzi $m5, 1
		f16v2add $a4, $a4, $a5
	}
	{
		cmpslt $m5, $m5, $m4
		f16v2tof32 $ap2, $a4
	}
	{
		add $m4, $m4, -1
		f32v2sub $ap2, $ap2, $ap0
	}
	brz $m5, .Lcode2_BB0_18
# %bb.16:                               # %for.body29.i
	add $m4, $m4, -1
.Lcode2_BB0_17:                               # %for.body29.i
                                        # =>This Inner Loop Header: Depth=1
	{
		ld32step $a6, $m15, $m3+=, 6
		f32v2mul $ap2, $ap2, $ap2
	}
	ld32step $a7, $m15, $m2+=, 6
	f16v2add $a6, $a6, $a7
	f16v2tof32 $ap3, $a6
	f32v2add $ap1, $ap2, $ap1
	{
		brnzdec $m4, .Lcode2_BB0_17
		f32v2sub $ap2, $ap3, $ap0
	}
.Lcode2_BB0_18:
	f32v2mul $ap0, $ap2, $ap2
	{
		bri .Lcode2_BB0_13
		f32v2add $ap0, $ap0, $ap1
	}
.Lcode2_BB0_6:                                # %for.cond.preheader.i
	ld32 $m2, $m13, $m15, 6
	{
		shr $m2, $m2, 1
		zero	$ap0
	}
	cmpult $m3, $m0, $m2
	brz $m3, .Lcode2_BB0_14
# %bb.7:                                # %for.body.lr.ph.i
	xnor $m3, $m0, $m15
	add $m2, $m3, $m2
	setzi $m3, 43691
	sort4x16lo $m4, $m2, $m15
	shr $m2, $m2, 16
	mul $m5, $m4, $m3
	mul $m3, $m2, $m3
	shr $m5, $m5, 16
	add $m3, $m3, $m5
	setzi $m5, 43690
	sort4x16lo $m6, $m3, $m15
	mul $m4, $m4, $m5
	shr $m3, $m3, 16
	mul $m2, $m2, $m5
	add $m4, $m4, $m6
	add $m2, $m2, $m3
	shr $m3, $m4, 16
	add $m2, $m2, $m3
	ld32 $m3, $m13, $m15, 0
	ld32 $m4, $m13, $m15, 1
	shl $m5, $m0, 2
	shr $m6, $m2, 2
	add $m2, $m4, $m5
	add $m3, $m3, $m5
	add $m4, $m6, 1
	setzi $m5, 1
	ld32step $a2, $m15, $m3+=, 6
	ld32step $a3, $m15, $m2+=, 6
	{
		cmpslt $m5, $m5, $m4
		f16v2add $a2, $a2, $a3
	}
	add $m4, $m4, -1
	brz $m5, .Lcode2_BB0_10
# %bb.8:                                # %for.body.i
	add $m4, $m4, -1
.Lcode2_BB0_9:                                # %for.body.i
                                        # =>This Inner Loop Header: Depth=1
	{
		ld32step $a4, $m15, $m3+=, 6
		f16v2tof32 $ap1, $a2
	}
	{
		ld32step $a2, $m15, $m2+=, 6
		f32v2add $ap0, $ap0, $ap1
	}
	{
		brnzdec $m4, .Lcode2_BB0_9
		f16v2add $a2, $a4, $a2
	}
.Lcode2_BB0_10:
	f16v2tof32 $ap1, $a2
	f32v2add $ap0, $ap0, $ap1
	st64 $ap0, $m1, $m15, $m0
	exitz $m15
.Lcode2_BB0_2:                                # %for.body43.preheader.i
	cmpult $m2, $m0, $m2
	brz $m2, .Lcode2_BB0_5
# %bb.3:                                # %for.body67.lr.ph.i
	{
		ld32 $m2, $m13, $m15, 2
		setzi $a1, 407485
	}
	{
		ld32 $m3, $m13, $m15, 3
		or $a3, $a1, 897581056
	}
	{
		ld32 $m4, $m13, $m15, 4
		mov	$a1, $a0
	}
	shl $m5, $m0, 2
	ld64 $ap2, $m1, $m15, 6
	{
		ld64 $ap3, $m1, $m15, 7
		f32v2add $ap2, $ap2, $apzero
	}
	{
		ld64 $ap3, $m1, $m15, 8
		f32v2add $ap2, $ap2, $ap3
	}
	{
		ld64 $ap3, $m1, $m15, 9
		f32v2add $ap2, $ap2, $ap3
	}
	{
		ld64 $ap3, $m1, $m15, 10
		f32v2add $ap2, $ap2, $ap3
	}
	{
		ld64 $ap3, $m1, $m15, 11
		f32v2add $ap2, $ap2, $ap3
	}
	f32v2add $ap2, $ap2, $ap3
	f32add $a4, $a4, $a5
	f32div $a2, $a4, $a2
	f32add $a2, $a2, $a3
	f32oorx $a2, $a2
	mov	$a3, $a2
	st64 $ap1, $m11, $m15, 0                # 8-byte Folded Spill
.Lcode2_BB0_4:                                # %for.body67.i
                                        # =>This Inner Loop Header: Depth=1
	ld32 $m1, $m13, $m15, 0
	ld32 $m6, $m13, $m15, 1
	add $m0, $m0, 6
	ld32 $a4, $m1, $m5, 0
	ld32 $a5, $m6, $m5, 0
	{
		ld32 $a5, $m2, $m5, 0
		f16v2add $a4, $a4, $a5
	}
	{
		ld32 $a4, $m3, $m5, 0
		f16v2tof32 $ap3, $a4
	}
	{
		ld64 $ap1, $m11, $m15, 0
		f32v2sub $ap3, $ap3, $ap0
	}
	f32v2mul $ap3, $ap1, $ap3
	mov	$ap1, $ap0
	f16v2tof32 $ap0, $a5
	f32v2mul $ap0, $ap3, $ap0
	f16v2tof32 $ap2, $a4
	f32v2add $ap0, $ap0, $ap2
	f32v2tof16 $a0, $ap0
	{
		st32step $a0, $m4, $m5+=, 6
		mov	$ap0, $ap1
	}
	ld32 $m1, $m13, $m15, 6
	shr $m1, $m1, 1
	cmpult $m1, $m0, $m1
	brnz $m1, .Lcode2_BB0_4
.Lcode2_BB0_5:                                # %_ZN15AddLayerNormF167computeEj.exit
	exitz $m15
.Lcode2_BB0_12:
	zero	$ap0
.Lcode2_BB0_13:                               # %for.cond.cleanup28.i
	add $m0, $m0, 6
.Lcode2_BB0_14:                               # %cleanup86.sink.split.i
	st64 $ap0, $m1, $m15, $m0
	exitz $m15
.Lcode2_func_end0:
	.size	__runCodelet_AddLayerNormF16, .Lcode2_func_end0-__runCodelet_AddLayerNormF16
	.section	.stack_sizes,"o",@progbits,.text.__runCodelet_AddLayerNormF16
	.long	.Lcode2_func_begin0
	.byte	8
	.section	.text.__runCodelet_AddLayerNormF16,"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

#define NORM_WITH_ADD
#include "arch/gc_tile_defines.h"

#ifdef NORM_WITH_ADD
#define NORM_SYMBOL add_layer_norm_f16
#define NORM_VERTEX __runCodelet_AddLayerNormF16
#define NORM_ROWS $m7
#define NORM_WIDTH $m8
#define NORM_GAMMA $m5
#define NORM_BETA $m6
#define NORM_SHIFT 1
#define NORM_STACK 160
#define NORM_SCRATCH 64
#else
#define NORM_SYMBOL layer_norm_f16
#define NORM_VERTEX __runCodelet_LayerNormF16
#define NORM_ROWS $m6
#define NORM_WIDTH $m7
#define NORM_GAMMA $m4
#define NORM_BETA $m5
#define NORM_SHIFT 0
#define NORM_STACK 128
#define NORM_SCRATCH 32
#endif

// Share each row across all six workers. Separate sum and centered-variance
// buffers avoid read/write races; local sync separates the three passes.
	.text
	.allow_optimizations
	.section .text.NORM_SYMBOL,"ax",@progbits
	.globl NORM_SYMBOL
	.p2align 2
	.type NORM_SYMBOL,@function
NORM_SYMBOL:
	.supervisor
	add $m11, $m11, -NORM_STACK
	st32 $m3, $m11, $m15, 0
#ifdef NORM_WITH_ADD
	st32 $m4, $m11, $m15, 1
#endif
	st32 NORM_GAMMA, $m11, $m15, (1 + NORM_SHIFT)
	st32 NORM_BETA, $m11, $m15, (2 + NORM_SHIFT)
	st32 $m2, $m11, $m15, (3 + NORM_SHIFT)
	st32 NORM_ROWS, $m11, $m15, (4 + NORM_SHIFT)
	st32 NORM_WIDTH, $m11, $m15, (5 + NORM_SHIFT)
	add $m0, $m11, NORM_SCRATCH
	st32 $m0, $m11, $m15, (6 + NORM_SHIFT)
	setzi $m0, NORM_VERTEX
.Lwrap2_row:
	setzi $m1, 0
.Lwrap2_stage:
	st32 $m1, $m11, $m15, (7 + NORM_SHIFT)
	runall $m0, $m11, 0
	sync TEXCH_SYNCZONE_LOCAL
	add $m1, $m1, 1
	cmpeq $m2, $m1, 3
	brz $m2, .Lwrap2_stage
	shl $m2, NORM_WIDTH, 1
	ld32 $m3, $m11, $m15, 0
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, 0
	ld32 $m3, $m11, $m15, (3 + NORM_SHIFT)
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, (3 + NORM_SHIFT)
#ifdef NORM_WITH_ADD
	ld32 $m3, $m11, $m15, 1
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, 1
#endif
	sub NORM_ROWS, NORM_ROWS, 1
	brnz NORM_ROWS, .Lwrap2_row
	add $m11, $m11, NORM_STACK
	br $m10
	.size NORM_SYMBOL, .-NORM_SYMBOL

#undef NORM_WITH_ADD
#undef NORM_SYMBOL
#undef NORM_VERTEX
#undef NORM_ROWS
#undef NORM_WIDTH
#undef NORM_GAMMA
#undef NORM_BETA
#undef NORM_SHIFT
#undef NORM_STACK
#undef NORM_SCRATCH
	.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
.Lcode3_func_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, .Lcode3_BB0_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, .Lcode3_BB0_15
		f32div $a0, $a0, $a2
	}
# %bb.2:                                # %for.cond25.preheader.i
	mov	$m3, $a1
	shr $m3, $m3, 1
	cmpult $m4, $m1, $m3
	brz $m4, .Lcode3_BB0_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, .Lcode3_BB0_14
# %bb.12:                               # %for.body30.i
	add $m3, $m3, -1
.Lcode3_BB0_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, .Lcode3_BB0_13
		f32v2mul $ap2, $ap2, $ap2
	}
.Lcode3_BB0_14:
	{
		bri .Lcode3_BB0_4
		f32v2add $ap0, $ap2, $ap1
	}
.Lcode3_BB0_5:                                # %for.cond.preheader.i
	ld32 $m3, $m13, $m15, 2
	{
		shr $m3, $m3, 1
		zero	$ap0
	}
	cmpult $m4, $m1, $m3
	brz $m4, .Lcode3_BB0_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, .Lcode3_BB0_10
# %bb.8:                                # %for.body.i
	add $m3, $m3, -1
.Lcode3_BB0_9:                                # %for.body.i
                                        # =>This Inner Loop Header: Depth=1
	{
		ld32step $a4, $m15, $m2+=, 6
		f32v2add $ap0, $ap0, $ap1
	}
	{
		brnzdec $m3, .Lcode3_BB0_9
		f16v2tof32 $ap1, $a4
	}
.Lcode3_BB0_10:
	f32v2add $ap0, $ap0, $ap1
	st64 $ap0, $m0, $m15, $m1
	exitz $m15
.Lcode3_BB0_15:                               # %if.else.i
	brz $m1, .Lcode3_BB0_16
# %bb.17:                               # %_ZN16LayerNormMoments7computeEj.exit
	exitz $m15
.Lcode3_BB0_6:                                # %for.cond.cleanup.i
	st64 $ap0, $m0, $m15, $m1
	exitz $m15
.Lcode3_BB0_3:
	zero	$ap0
.Lcode3_BB0_4:                                # %for.cond.cleanup29.i
	shl $m1, $m1, 3
	st64 $ap0, $m1, $m0, 6
	exitz $m15
.Lcode3_BB0_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
.Lcode3_func_end0:
	.size	__runCodelet_LayerNormMoments, .Lcode3_func_end0-__runCodelet_LayerNormMoments
	.section	.stack_sizes,"o",@progbits,.text.__runCodelet_LayerNormMoments
	.long	.Lcode3_func_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

#include "arch/gc_tile_defines.h"
// m2 output moments, m3 source, m4 rows, m5 local feature width.
	.text
	.allow_optimizations
	.section .text.layer_norm_moments,"ax",@progbits
	.globl layer_norm_moments
	.p2align 2
	.type layer_norm_moments,@function
layer_norm_moments:
	.supervisor
	add $m11, $m11, -128
	st32 $m3, $m11, $m15, 0
	st32 $m2, $m11, $m15, 1
	st32 $m5, $m11, $m15, 2
	add $m0, $m11, 32
	st32 $m0, $m11, $m15, 3
	setzi $m0, __runCodelet_LayerNormMoments
.Lwrap3_row:
	setzi $m1, 0
.Lwrap3_stage:
	st32 $m1, $m11, $m15, 4
	runall $m0, $m11, 0
	sync TEXCH_SYNCZONE_LOCAL
	add $m1, $m1, 1
	cmpeq $m2, $m1, 3
	brz $m2, .Lwrap3_stage
	ld32 $m3, $m11, $m15, 0
	shl $m2, $m5, 1
	add $m3, $m3, $m2
	st32 $m3, $m11, $m15, 0
	ld32 $m2, $m11, $m15, 1
	add $m2, $m2, 8
	st32 $m2, $m11, $m15, 1
	sub $m4, $m4, 1
	brnz $m4, .Lwrap3_row
	add $m11, $m11, 128
	br $m10
	.size layer_norm_moments, .-layer_norm_moments

#undef NORM_WITH_ADD
#undef NORM_SYMBOL
#undef NORM_VERTEX
#undef NORM_ROWS
#undef NORM_WIDTH
#undef NORM_GAMMA
#undef NORM_BETA
#undef NORM_SHIFT
#undef NORM_STACK
#undef NORM_SCRATCH
	.text
	.allow_optimizations
	.file	"3"
	.section	.text.__runCodelet_LayerNormApply,"ax",@progbits
	.globl	__runCodelet_LayerNormApply     # -- Begin function __runCodelet_LayerNormApply
	.p2align	2
	.type	__runCodelet_LayerNormApply,@function
__runCodelet_LayerNormApply:            # @__runCodelet_LayerNormApply
	.worker
.Lcode4_func_begin0:
# %bb.0:                                # %entry
	setzi $m6, 32
	cmpult $m6, $m6, __worker_stack_max_growth_plus_one
	brz $m6, _stackoverflow
	add $m11, $m12, -32
	get $m2, 1
	ld32 $m9, $m13, $m15, 5
	brz $m9, .Lcode4_BB0_16
# %bb.1:                                # %for.cond6.preheader.lr.ph.i
	mov	$m0, $m15
	mov	$m1, $m15
	{
		and $m2, $m2, 7
		setzi $a0, 407485
	}
	{
		ld32 $m6, $m13, $m15, 1
		or $a0, $a0, 897581056
	}
	st32 $a0, $m11, $m15, 3                 # 4-byte Folded Spill
	ld32 $m7, $m13, $m15, 2
	ld32 $m3, $m13, $m15, 3
	shl $m8, $m2, 2
	ld32 $m4, $m13, $m15, 0
	ld32 $m5, $m13, $m15, 4
	add $m7, $m7, $m8
	st32 $m7, $m11, $m15, 2                 # 4-byte Folded Spill
	add $m6, $m6, $m8
	st32 $m6, $m11, $m15, 1                 # 4-byte Folded Spill
	st32 $m3, $m11, $m15, 5                 # 4-byte Folded Spill
	add $m3, $m3, 4
	st32 $m3, $m11, $m15, 4                 # 4-byte Folded Spill
	bri .Lcode4_BB0_2
.Lcode4_BB0_15:                               # %for.cond.cleanup44.i
                                        #   in Loop: Header=BB0_2 Depth=1
	add $m1, $m1, 1
	cmpult $m3, $m1, $m9
	add $m0, $m0, 8
	brz $m3, .Lcode4_BB0_16
.Lcode4_BB0_2:                                # %for.cond6.preheader.i
                                        # =>This Loop Header: Depth=1
                                        #     Child Loop BB0_5 Depth 2
                                        #     Child Loop BB0_8 Depth 2
                                        #     Child Loop BB0_13 Depth 2
	ld32 $m10, $m13, $m15, 7
	brz $m10, .Lcode4_BB0_10
# %bb.3:                                # %for.body9.lr.ph.i
                                        #   in Loop: Header=BB0_2 Depth=1
	{
		mul $m6, $m10, $m0
		mov	$a1, $azero
	}
	ld32 $m3, $m11, $m15, 5                 # 4-byte Folded Reload
	add $m7, $m3, $m6
	setzi $m8, 1
	cmpslt $m3, $m8, $m10
	ld32step $a2, $m15, $m7+=, 2
	add $m8, $m10, -1
	brz $m3, .Lcode4_BB0_6
# %bb.4:                                # %for.body9.i
                                        #   in Loop: Header=BB0_2 Depth=1
	add $m8, $m8, -1
.Lcode4_BB0_5:                                # %for.body9.i
                                        #   Parent Loop BB0_2 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	{
		ld32step $a2, $m15, $m7+=, 2
		f32add $a1, $a1, $a2
	}
	brnzdec $m8, .Lcode4_BB0_5
.Lcode4_BB0_6:                                #   in Loop: Header=BB0_2 Depth=1
	f32add $a2, $a1, $a2
	{
		st32 $m10, $m11, $m15, 0
		mov	$a1, $azero
	}
	ld32 $a3, $m11, $m15, 0
	{
		ld32 $a4, $m13, $m15, 6
		f32fromui32 $a3, $a3
	}
	{
		ld32 $m3, $m11, $m15, 4
		f32div $a2, $a2, $a3
	}
	{
		add $m7, $m3, $m6
		f32fromui32 $a3, $a4
	}
	mov	$m6, $a4
	add $m3, $m15, -4
	{
		ld64 $ap2, $m7, $m3, 0
		mov	$a0, $a2
	}
	{
		setzi $m3, 1
		f32sub $a4, $a4, $a2
	}
	{
		cmpslt $m3, $m3, $m10
		f32mul $a6, $a4, $a3
	}
	{
		add $m7, $m7, 8
		f32mul $a4, $a6, $a4
	}
	add $m8, $m10, -1
	brz $m3, .Lcode4_BB0_9
# %bb.7:                                # %for.body18.i
                                        #   in Loop: Header=BB0_2 Depth=1
	add $m8, $m8, -1
.Lcode4_BB0_8:                                # %for.body18.i
                                        #   Parent Loop BB0_2 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	{
		add $m3, $m15, -4
		f32add $a6, $a4, $a5
	}
	ld64 $ap2, $m7, $m3, 0
	{
		add $m7, $m7, 8
		f32sub $a4, $a4, $a0
	}
	f32add $a1, $a1, $a6
	f32mul $a6, $a4, $a3
	{
		brnzdec $m8, .Lcode4_BB0_8
		f32mul $a4, $a6, $a4
	}
.Lcode4_BB0_9:                                #   in Loop: Header=BB0_2 Depth=1
	f32add $a3, $a4, $a5
	f32add $a1, $a1, $a3
	{
		bri .Lcode4_BB0_11
		mov	$a2, $a0
	}
.Lcode4_BB0_10:                               # %for.cond.cleanup8.for.cond.cleanup17_crit_edge.i
                                        #   in Loop: Header=BB0_2 Depth=1
	{
		ld32 $m6, $m13, $m15, 6
		mov	$a1, $azero
	}
	or $a2, $azero, 2143289344
.Lcode4_BB0_11:                               # %for.cond.cleanup17.i
                                        #   in Loop: Header=BB0_2 Depth=1
	shr $m7, $m6, 1
	cmpult $m3, $m2, $m7
	brz $m3, .Lcode4_BB0_15
# %bb.12:                               # %for.body45.lr.ph.i
                                        #   in Loop: Header=BB0_2 Depth=1
	mul $m3, $m6, $m10
	st32 $m3, $m11, $m15, 0
	ld32 $a3, $m11, $m15, 0
	f32fromui32 $a4, $a3
	{
		ld32 $m6, $m11, $m15, 1
		mov	$a3, $a2
	}
	ld32 $m9, $m11, $m15, 2                 # 4-byte Folded Reload
	{
		mov	$m10, $m2
		f32div $a1, $a1, $a4
	}
	ld32 $a0, $m11, $m15, 3                 # 4-byte Folded Reload
	f32add $a1, $a1, $a0
	f32oorx $a0, $a1
	mov	$a1, $a0
	{
		st64 $ap0, $m11, $m15, 3
		mov	$ap2, $ap1
	}
.Lcode4_BB0_13:                               # %for.body45.i
                                        #   Parent Loop BB0_2 Depth=1
                                        # =>  This Inner Loop Header: Depth=2
	mul $m3, $m1, $m7
	add $m3, $m10, $m3
	add $m10, $m10, 6
	ld32 $a1, $m4, $m15, $m3
	{
		ld32step $a1, $m15, $m6+=, 6
		f16v2tof32 $ap3, $a1
	}
	{
		ld32step $a0, $m15, $m9+=, 6
		f32v2sub $ap3, $ap3, $ap2
	}
	ld64 $ap1, $m11, $m15, 3                # 8-byte Folded Reload
	f32v2mul $ap3, $ap1, $ap3
	f16v2tof32 $ap1, $a1
	f32v2mul $ap1, $ap3, $ap1
	f16v2tof32 $ap0, $a0
	f32v2add $ap0, $ap1, $ap0
	f32v2tof16 $a0, $ap0
	st32 $a0, $m5, $m15, $m3
	ld32 $m3, $m13, $m15, 6
	shr $m7, $m3, 1
	cmpult $m3, $m10, $m7
	brnz $m3, .Lcode4_BB0_13
# %bb.14:                               # %for.cond.cleanup44.loopexit.i
                                        #   in Loop: Header=BB0_2 Depth=1
	ld32 $m9, $m13, $m15, 5
	bri .Lcode4_BB0_15
.Lcode4_BB0_16:                               # %_ZN14LayerNormApply7computeEj.exit
	exitz $m15
.Lcode4_func_end0:
	.size	__runCodelet_LayerNormApply, .Lcode4_func_end0-__runCodelet_LayerNormApply
	.section	.stack_sizes,"o",@progbits,.text.__runCodelet_LayerNormApply
	.long	.Lcode4_func_begin0
	.byte	32
	.section	.text.__runCodelet_LayerNormApply,"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

#define WORKER_CALL_SYMBOL layer_norm_apply
#define WORKER_CODELET_SYMBOL __runCodelet_LayerNormApply
#define WORKER_ARGUMENTS $m3,$m4,$m5,$m6,$m2,$m7,$m8,$m9
#define WORKER_FRAME_BYTES 48
#include "/srv/home/gc-sdk/ipu-stack/device/worker_call.S"
#undef WORKER_CALL_SYMBOL
#undef WORKER_CODELET_SYMBOL
#undef WORKER_ARGUMENTS
#undef WORKER_FRAME_BYTES