zgemv_t.s

Optimized GotoBLAS libraries

	STFD [CST2] = f37
	FMA	f41 = ALPHA_R, f17, f41
	}
	{ .mmf
	add  CST1 = CST1, INCYM1
	add  CST2 = CST2, INCYM1
	FMA	f45 = ALPHA_R, f21, f45
	}
	;;
	{ .mmf
	STFD [CST1] = f34, SIZE
	STFD [CST2] = f38, SIZE
	FNMA	f42 = ALPHA_I, f19, f42
	}
	{ .mmf
	nop	__LINE__
	nop	__LINE__
	FNMA	f46 = ALPHA_I, f23, f46
	}
	;;
	{ .mmf
	STFD [CST1] = f35
	STFD [CST2] = f39
	FMA	f43 = ALPHA_R, f19, f43
	}
	{ .mmf
	add  CST1 = CST1, INCY3M1
	add  CST2 = CST2, INCY3M1
	FMA	f47 = ALPHA_R, f23, f47
	}
	;;
	{ .mmi
	STFD [CST1] = f40, SIZE
	STFD [CST2] = f44, SIZE
	adds J = -1, J
	}
	;;
	{ .mmi
	STFD [CST1] = f41
	STFD [CST2] = f45
	add  CST1 = CST1, INCYM1
	}
	{ .mmi
	nop  __LINE__
	nop  __LINE__
	add  CST2 = CST2, INCYM1
	}
	;;
	{ .mmi
	STFD [CST1] = f42, SIZE
	STFD [CST2] = f46, SIZE
	cmp.lt p6, p0 = 0, J
	}
	;;
	{ .mmi
	STFD [CST1] = f43
	STFD [CST2] = f47
	add  CST1 = CST1, INCY3M1
	}
	{ .mmb
	add  CST2 = CST2, INCY3M1
	(p6) br.cond.dptk .L11
	}
	;;
	.align 16

.L20:
	{ .mfi
	mov	AO1 = A
	mov	f8  = f0
	mov	pr.rot= 0
	}
	{ .mfi
	add	AO2 = LDA, A
	mov	f10 = f0
	tbit.z	p6, p0  = N, 2
	}
	;;
	{ .mmf
	shladd	AO3 = LDA, 1, A
	shladd	AO4 = LDA, 1, AO2
	mov	f12 = f0
	}
	{ .mfb
	mov	BO  = BUFFER
	mov	f14 = f0
	(p6) br.cond.dpnt .L30
	}
	;;
	{ .mfi
	adds	RPRE1  = (RPREFETCH +  0) * SIZE, AO1
	mov	f9  = f0
	mov	ar.ec= 5
	}
	{ .mmf
	adds	RPRE2  = (RPREFETCH +  2) * SIZE, AO2
	adds	I = -1, MIN_M
	mov	f11 = f0
	}
	;;
	{ .mmf
	adds	RPRE3  = (RPREFETCH +  4) * SIZE, AO3
	adds	RPRE4  = (RPREFETCH +  6) * SIZE, AO4
	mov	f13 = f0
	}
	{ .mmf
	cmp.eq	p16, p0 = r0, r0
	shladd	A   = LDA, 2, A
	mov	f15 = f0
	}
	;;
	{ .mmi
	lfetch.excl.nt1	[WPRE]
	adds	PREB   = RPREFETCH * SIZE, BO
	mov	ar.lc = I
	}
	{ .mmi
	adds	WPRE = 16 * SIZE, CLD1
	cmp.eq  p12, p0 = r0, r0
	mov	I = 0
	}
	;;
	.align 16

.L26:
	{ .mmf
	(p12) PREFETCH [RPRE1], 16 * SIZE
	(p16) LDFPD	f32,  f37  = [AO1], 2 * SIZE
	(p20) ADD1	f8  = f116, f36, f8
	}
	{ .mmf
	(p16) cmp.eq.unc p13, p0 = 2, I
	nop   __LINE__
	(p20) ADD2	f9  = f121, f36, f9
	}
	;;
	{ .mmf
	(p12) PREFETCH [PREB], 16 * SIZE
	(p16) LDFPD	f112, f117 = [BO], 2 * SIZE
	(p20) ADD1	f10 = f116, f46, f10
	}
	{ .mmf
	(p16) cmp.eq.unc p14, p0 = 4, I
	(p16) cmp.eq.unc p15, p0 = 6, I
	(p20) ADD2	f11 = f121, f46, f11
	}
	;;
	{ .mmf
	(p16) LDFPD	f42,  f47  = [AO2], 2 * SIZE
	nop   __LINE__
	(p20) ADD1	f12 = f116, f56, f12
	}
	{ .mmf
	nop   __LINE__
	nop   __LINE__
	(p20) ADD2	f13 = f121, f56, f13
	}
	;;
	{ .mmf
	(p13) PREFETCH [RPRE2], 16 * SIZE
	nop   __LINE__
	(p20) ADD1	f14 = f116, f66, f14
	}
	{ .mmf
	nop   __LINE__
	nop   __LINE__
	(p20) ADD2	f15 = f121, f66, f15
	}
	;;
	{ .mmf
	(p16) LDFPD	f52,  f57  = [AO3], 2 * SIZE
	nop   __LINE__
	(p20) ADD3	f8  = f121, f41, f8
	}
	{ .mmf
	(p16) adds I = 1, I
	nop   __LINE__
	(p20) ADD4	f9  = f116, f41, f9
	}
	;;
	{ .mmf
	(p14) PREFETCH [RPRE3], 16 * SIZE
	nop   __LINE__
	(p20) ADD3	f10 = f121, f51, f10
	}
	{ .mmf
	(p16) cmp.eq.unc p15, p0 = 8, I
	nop   __LINE__
	(p20) ADD4	f11 = f116, f51, f11
	}
	;;
	{ .mmf
	(p16) LDFPD	f62,  f67  = [AO4], 2 * SIZE
	nop   __LINE__
	(p20) ADD3	f12 = f121, f61, f12
	}
	{ .mmf
	(p15) mov I = 0
	nop   __LINE__
	(p20) ADD4	f13 = f116, f61, f13
	}
	;;
	{ .mmf
	(p15) PREFETCH [RPRE4], 16 * SIZE
	nop   __LINE__
	(p20) ADD3	f14 = f121, f71, f14
	}
	{ .mfb
	(p16) cmp.eq.unc p12, p0 = 0, I
	(p20) ADD4	f15 = f116, f71, f15
	br.ctop.sptk.few .L26
	}
	;;
.L28:
	LDFD	f32 = [CLD1], SIZE
	LDFD	f36 = [CLD2], SIZE
	shladd	CST2  = INCY, 1, CST1
	;;
	LDFD	f33 = [CLD1], INCYM1
	LDFD	f37 = [CLD2], INCYM1
	;;
	LDFD	f34 = [CLD1], SIZE
	LDFD	f38 = [CLD2], SIZE
	;;
	LDFD	f35 = [CLD1], INCY3M1
	LDFD	f39 = [CLD2], INCY3M1
	;;
	FMA	f32 = ALPHA_R, f8,  f32
	FMA	f36 = ALPHA_R, f12, f36
	FMA	f33 = ALPHA_I, f8,  f33
	FMA	f37 = ALPHA_I, f12, f37
	FMA	f34 = ALPHA_R, f10, f34
	FMA	f38 = ALPHA_R, f14, f38
	FMA	f35 = ALPHA_I, f10, f35
	FMA	f39 = ALPHA_I, f14, f39
	;;
	FNMA	f32 = ALPHA_I, f9,  f32
	FNMA	f36 = ALPHA_I, f13, f36
	FMA	f33 = ALPHA_R, f9,  f33
	FMA	f37 = ALPHA_R, f13, f37
	FNMA	f34 = ALPHA_I, f11, f34
	FNMA	f38 = ALPHA_I, f15, f38
	FMA	f35 = ALPHA_R, f11, f35
	FMA	f39 = ALPHA_R, f15, f39
	;;
	STFD [CST1] = f32, SIZE
	STFD [CST2] = f36, SIZE
	;;
	STFD [CST1] = f33
	STFD [CST2] = f37
	add  CST1 = CST1, INCYM1
	add  CST2 = CST2, INCYM1
	;;
	STFD [CST1] = f34, SIZE
	STFD [CST2] = f38, SIZE
	;;
	STFD [CST1] = f35
	STFD [CST2] = f39
	add  CST1 = CST1, INCY3M1
	add  CST2 = CST2, INCY3M1
	;;
	.align 16

.L30:
	{ .mfi
	mov	AO1 = A
	mov	f8  = f0
	mov	pr.rot= 0
	}
	{ .mfi
	add	AO2 = LDA, A
	mov	f10 = f0
	tbit.z	p6, p0  = N, 1
	}
	;;
	{ .mmf
	adds	RPRE1  = (RPREFETCH +  0) * SIZE, AO1
	adds	RPRE2  = (RPREFETCH +  2) * SIZE, AO2
	mov	f12 = f0
	}
	{ .mfb
	adds	I = -1, MIN_M
	mov	f14 = f0
	(p6) br.cond.dpnt .L40
	}
	;;
	{ .mfi
	mov	BO  = BUFFER
	mov	f9  = f0
	mov	ar.ec= 5
	}
	{ .mmf
	cmp.eq	p16, p0 = r0, r0
	shladd	A   = LDA, 1, A
	mov	f11 = f0
	}
	;;
	{ .mfi
	adds	WPRE = 16 * SIZE, CLD1
	mov	f13 = f0
	mov	ar.lc = I
	}
	{ .mmf
	adds	PREB   = RPREFETCH * SIZE, BO
	nop	__LINE__
	mov	f15 = f0
	}
	;;
	{ .mmi
	lfetch.excl.nt1	[WPRE]
	cmp.eq  p12, p0 = r0, r0
	mov	I = 0
	}
	;;
	.align 16

.L36:
	{ .mmf
	(p12) PREFETCH [RPRE1], 16 * SIZE
	(p16) LDFPD	f32,  f37  = [AO1], 2 * SIZE
	(p20) ADD1	f8  = f116, f36, f8
	}
	{ .mmf
	(p16) cmp.eq.unc p13, p0 = 4, I
	(p16) adds I = 1, I
	(p20) ADD2	f9  = f121, f36, f9
	}
	;;
	{ .mmf
	(p12) PREFETCH [PREB], 16 * SIZE
	(p16) LDFPD	f112, f117 = [BO], 2 * SIZE
	(p20) ADD1	f10 = f116, f46, f10
	}
	{ .mmf
	(p16) cmp.eq.unc p12, p0 = 8, I
	(p20) ADD2	f11 = f121, f46, f11
	}
	;;
	{ .mmf
	(p13) PREFETCH [RPRE2], 16 * SIZE
	(p16) LDFPD	f42,  f47  = [AO2], 2 * SIZE
	(p20) ADD3	f12 = f121, f41, f12
	}
	{ .mmf
	(p12) mov I = 0
	(p20) ADD4	f13 = f116, f41, f13
	}
	;;
	{ .mmf
	(p20) ADD3	f14 = f121, f51, f14
	}
	{ .mfb
	nop	__LINE__
	(p20) ADD4	f15 = f116, f51, f15
	br.ctop.sptk.few .L36
	}
	;;

.L38:
	LDFD	f32 = [CLD1], SIZE
	FADD	f8  = f8,  f12
	shladd	CST2  = INCY, 1, CST1
	;;
	LDFD	f33 = [CLD1], INCYM1
	FADD	f10 = f10, f14
	;;
	LDFD	f34 = [CLD1], SIZE
	FADD	f9  = f9,  f13
	;;
	LDFD	f35 = [CLD1], INCYM1
	FADD	f11 = f11, f15
	;;
	FMA	f32 = ALPHA_R, f8,  f32
	FMA	f33 = ALPHA_I, f8,  f33
	FMA	f34 = ALPHA_R, f10, f34
	FMA	f35 = ALPHA_I, f10, f35
	;;
	FNMA	f32 = ALPHA_I, f9,  f32
	FMA	f33 = ALPHA_R, f9,  f33
	FNMA	f34 = ALPHA_I, f11, f34
	FMA	f35 = ALPHA_R, f11, f35
	;;
	STFD [CST1] = f32, SIZE
	;;
	STFD [CST1] = f33
	add  CST1 = CST1, INCYM1
	;;
	STFD [CST1] = f34, SIZE
	;;
	STFD [CST1] = f35
	add  CST1 = CST1, INCYM1
	;;
	.align 16


.L40:
	{ .mfi
	mov	AO1 = A
	mov	f8  = f0
	mov	pr.rot= 0
	}
	{ .mfi
	mov	f9  = f0
	tbit.z	p6, p0  = N, 0
	}
	;;
	{ .mfi
	adds	RPRE1  = (RPREFETCH +  0) * SIZE, AO1
	mov	f10 = f0
	mov	ar.ec= 5
	}
	{ .mfb
	adds	I = -1, MIN_M
	mov	f11 = f0
	(p6) br.cond.dpnt .L99
	}
	;;
	{ .mmi
	cmp.eq	p16, p0 = r0, r0
	add	A   = LDA, A
	mov	ar.lc = I
	}
	{ .mmi
	adds	WPRE = 16 * SIZE, CLD1
	adds	PREB   = RPREFETCH * SIZE, BO
	mov	BO  = BUFFER
	}
	;;
	{ .mmi
	lfetch.excl.nt1	[WPRE]
	cmp.eq  p12, p0 = r0, r0
	mov	I = 0
	}
	;;
	.align 16

.L46:
	{ .mmf
	(p12) PREFETCH [RPRE1], 16 * SIZE
	(p16) LDFPD	f32,  f37  = [AO1], 2 * SIZE
	(p20) ADD1	f8  = f116, f36, f8
	}
	{ .mmf
	(p16) cmp.eq.unc p12, p0 = 7, I
	(p16) adds I = 1, I
	(p20) ADD2	f9  = f121, f36, f9
	}
	;;
	{ .mmf
	(p16) LDFPD	f112, f117 = [BO], 2 * SIZE
	(p20) ADD3	f10  = f121, f41, f10
	}
	{ .mfb
	(p12) mov I = 0
	(p20) ADD4	f11  = f116, f41, f11
	br.ctop.sptk.few .L46
	}
	;;

.L48:
	LDFD	f32 = [CLD1], SIZE
	FADD	f8  = f8,  f10
	shladd	CST2  = INCY, 1, CST1
	;;
	LDFD	f33 = [CLD1], INCYM1
	FADD	f9  = f9,  f11
	;;
	FMA	f32 = ALPHA_R, f8,  f32
	FMA	f33 = ALPHA_I, f8,  f33
	;;
	FNMA	f32 = ALPHA_I, f9,  f32
	FMA	f33 = ALPHA_R, f9,  f33
	;;
	STFD [CST1] = f32, SIZE
	;;
	STFD [CST1] = f33
	add  CST1 = CST1, INCYM1
	;;
	.align 16

.L99:
	adds	IS = P, IS
	shladd	A  = LDAP, ZBASE_SHIFT, A
	;;
	cmp.gt	p6, p0 = M, IS
	(p6) br.cond.dptk .LIs_loop
	br   .L999
	.align 16
	;;

.L100:
	{ .mmi
	mov	CLD1  = Y
	shladd	CLD2  = INCY, 1, Y
	shr	J   = N, 3
	}
	;;
	{ .mmb
	mov	CST1  = Y
	cmp.eq	p6, p0 = r0, J
	(p6) br.cond.dpnt .L120
	}
	;;
	.align 16

.L111:
	{ .mfi
	mov	AO1 = A
	mov	f8  = f0
	mov	pr.rot= 0
	}
	{ .mfi
	add	AO2 = LDA, A
	mov	f10 = f0
	mov	BO  = BUFFER
	}
	;;
	{ .mmf
	shladd	AO3 = LDA, 1, A
	shladd	AO4 = LDA, 1, AO2
	mov	f12 = f0
	}
	{ .mmf
	adds	RPRE1  = (RPREFETCH +  0) * SIZE, AO1
	adds	RPRE2  = (RPREFETCH +  2) * SIZE, AO2
	mov	f14 = f0
	}
	;;
	{ .mmf
	shladd	AO5 = LDA, 1, AO3
	shladd	AO6 = LDA, 1, AO4
	mov	f16 = f0
	}
	{ .mmf
	adds	RPRE3  = (RPREFETCH +  4) * SIZE, AO3
	adds	RPRE4  = (RPREFETCH +  6) * SIZE, AO4
	mov	f18 = f0
	}
	;;
	{ .mmf
	shladd	AO7 = LDA, 1, AO5
	shladd	AO8 = LDA, 1, AO6
	mov	f20 = f0
	}
	{ .mmf
	adds	RPRE5  = (RPREFETCH +  8) * SIZE, AO5
	adds	RPRE6  = (RPREFETCH + 10) * SIZE, AO6
	mov	f22 = f0
	}
	;;
	{ .mfi
	shladd	A   = LDA, 3, A
	mov	f9  = f0
	mov	ar.ec= 5
	}
	{ .mmf
	adds	RPRE7  = (RPREFETCH + 12) * SIZE, AO7
	adds	RPRE8  = (RPREFETCH + 14) * SIZE, AO8
	mov	f11 = f0
	}
	;;
	{ .mmf
	adds	WPRE = 16 * SIZE, CLD1
	adds	PREB   = RPREFETCH * SIZE, BO
	mov	f13 = f0
	}
	{ .mmf
	adds	I = -1, MIN_M
	cmp.eq	p16, p0 = r0, r0
	mov	f15 = f0
	}
	;;
	{ .mfi
	cmp.eq  p12, p0 = r0, r0
	mov	f17 = f0
	mov	ar.lc = I
	}
	{ .mmf
	nop	__LINE__
	nop	__LINE__
	mov	f19 = f0
	}
	;;
	{ .mmf
	lfetch.excl.nt1	[WPRE]
	nop	__LINE__
	mov	f21 = f0
	}
	{ .mmf
	mov	I = 0
	nop	__LINE__
	mov	f23 = f0
	}
	;;
	.align 16

.L116:
	{ .mmf
	(p12) PREFETCH [RPRE1], 16 * SIZE
	(p16) LDFD	f32 = [AO1], 1 * SIZE
	(p20) ADD1	f8  = f116, f36, f8
	}
	{ .mmf
	(p16) cmp.eq.unc p13, p0 = 1, I
	(p16) cmp.eq.unc p14, p0 = 2, I
	(p20) ADD2	f9  = f121, f36, f9
	}
	;;
	{ .mmf
	(p13) PREFETCH [PREB], 16 * SIZE
	(p16) LDFPD	f112, f117 = [BO], 2 * SIZE
	(p20) ADD1	f10 = f116, f46, f10
	}
	{ .mmf
	(p16) LDFD	f37 = [AO1], 1 * SIZE
	(p16) cmp.eq.unc p15, p0 = 3, I
	(p20) ADD2	f11 = f121, f46, f11
	}
	;;
	{ .mmf
	(p13) PREFETCH [RPRE2], 16 * SIZE
	(p16) LDFD	f42 = [AO2], 1 * SIZE
	(p20) ADD1	f12 = f116, f56, f12
	}
	{ .mmf
	nop   __LINE__
	nop   __LINE__
	(p20) ADD2	f13 = f121, f56, f13
	}
	;;
	{ .mmf
	(p16) LDFD	f47 = [AO2], 1 * SIZE
	nop   __LINE__
	(p20) ADD1	f14 = f116, f66, f14
	}
	{ .mmf
	nop   __LINE__
	nop   __LINE__
	(p20) ADD2	f15 = f121, f66, f15
	}
	;;
	{ .mmf
	(p14) PREFETCH [RPRE3], 16 * SIZE
	(p16) LDFD	f52 = [AO3], 1 * SIZE
	(p20) ADD3	f8  = f121, f41, f8
	}
	{ .mmf
	nop   __LINE__
	nop   __LINE__
	(p20) ADD4	f9  = f116, f41, f9
	}
	;;
	{ .mmf
	(p16) LDFD	f57 = [AO3], 1 * SIZE
	nop   __LINE__
	(p20) ADD3	f10 = f121, f51, f10
	}
	{ .mmf
	nop   __LINE__
	nop   __LINE__
	(p20) ADD4	f11 = f116, f51, f11
	}
	;;
	{ .mmf
	(p15) PREFETCH [RPRE4], 16 * SIZE
	(p16) LDFD	f62 = [AO4], 1 * SIZE
	(p20) ADD3	f12 = f121, f61, f12
	}
	{ .mmf
	(p16) cmp.eq.unc p12, p0 = 4, I
	(p16) cmp.eq.unc p13, p0 = 5, I
	(p20) ADD4	f13 = f116, f61, f13
	}
	;;
	{ .mmf
	(p16) LDFD	f67 = [AO4], 1 * SIZE
	nop   __LINE__
	(p20) ADD3	f14 = f121, f71, f14
	}