gemv_n_sse2.s

Optimized GotoBLAS libraries

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L106

.L104:
	KERNELMACRO7( 0)
	KERNELMACRO7( 2)
	KERNELMACRO7( 4)
	KERNELMACRO7( 6)
	KERNELMACRO7( 8)
	KERNELMACRO7(10)
	KERNELMACRO7(12)
	KERNELMACRO7(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L104

.L106:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L108

.L107:
	KERNELMACRO7( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L107

.L108:
	movsd	  0 * SIZE(%ecx), %xmm0
	mulpd	%xmm3, %xmm4
	movl	INCY, %eax
	mulpd	%xmm3, %xmm5
	movl	%ecx, %edx
	mulpd	%xmm3, %xmm6
	addl	%eax, %edx
	mulsd	%xmm3, %xmm7
	cmpl	$8, %eax
	jne	.L109

	movhpd	  1 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm4

	movsd	  2 * SIZE(%ecx), %xmm0
	movhpd	  3 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm5

	movsd	  4 * SIZE(%ecx), %xmm0
	movhpd	  5 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm6

	movsd	  6 * SIZE(%ecx), %xmm0
	addsd	%xmm0, %xmm7

	movsd	%xmm4, 0 * SIZE(%ecx)
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 1 * SIZE(%ecx)

	movsd	%xmm5, 2 * SIZE(%ecx)
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 3 * SIZE(%ecx)

	movsd	%xmm6, 4 * SIZE(%ecx)
	unpckhpd %xmm6, %xmm6
	movsd	%xmm6, 5 * SIZE(%ecx)

	movsd	%xmm7, 6 * SIZE(%ecx)
	jmp	.L99

.L109:
	movhpd	  0 * SIZE(%edx), %xmm0
	addl	  %eax, %edx
	addpd	%xmm0, %xmm4

	movsd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	movhpd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	addpd	%xmm1, %xmm5

	movsd	  0 * SIZE(%edx), %xmm2
	addl	  %eax, %edx
	movhpd	  0 * SIZE(%edx), %xmm2
	addl	  %eax, %edx
	addpd	%xmm2, %xmm6

	movsd	  0 * SIZE(%edx), %xmm0
	addsd	%xmm0, %xmm7

	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm6, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm6, %xmm6
	movsd	%xmm6, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm7, 0 * SIZE(%ecx)
	jmp	.L99

.L110:
	cmpl	$6, %esi
	jne	.L120

.L111:
	pxor	%xmm4, %xmm4
	pxor	%xmm5, %xmm5
	pxor	%xmm6, %xmm6

	movl	A, %edx		# aoffset = a
	addl	$6 * SIZE, A	# a += 6

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L116

.L114:
	KERNELMACRO6( 0)
	KERNELMACRO6( 2)
	KERNELMACRO6( 4)
	KERNELMACRO6( 6)
	KERNELMACRO6( 8)
	KERNELMACRO6(10)
	KERNELMACRO6(12)
	KERNELMACRO6(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L114

.L116:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L118

.L117:
	KERNELMACRO6( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L117

.L118:
	movsd	  0 * SIZE(%ecx), %xmm0
	mulpd	%xmm3, %xmm4
	movl	INCY, %eax
	mulpd	%xmm3, %xmm5
	movl	%ecx, %edx
	mulpd	%xmm3, %xmm6
	addl	%eax, %edx
	cmpl	$8, %eax
	jne	.L119

	movhpd	  1 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm4

	movsd	  2 * SIZE(%ecx), %xmm0
	movhpd	  3 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm5

	movsd	  4 * SIZE(%ecx), %xmm0
	movhpd	  5 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm6

	movsd	%xmm4, 0 * SIZE(%ecx)
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 1 * SIZE(%ecx)

	movsd	%xmm5, 2 * SIZE(%ecx)
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 3 * SIZE(%ecx)

	movsd	%xmm6, 4 * SIZE(%ecx)
	unpckhpd %xmm6, %xmm6
	movsd	%xmm6, 5 * SIZE(%ecx)
	jmp	.L99

.L119:
	movhpd	  0 * SIZE(%edx), %xmm0
	addl	  %eax, %edx
	addpd	%xmm0, %xmm4

	movsd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	movhpd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	addpd	%xmm1, %xmm5

	movsd	  0 * SIZE(%edx), %xmm2
	addl	  %eax, %edx
	movhpd	  0 * SIZE(%edx), %xmm2
	addpd	%xmm2, %xmm6

	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm6, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm6, %xmm6
	movsd	%xmm6, 0 * SIZE(%ecx)
	jmp	.L99

.L120:
	cmpl	$5, %esi
	jne	.L130

.L121:
	pxor	%xmm4, %xmm4
	pxor	%xmm5, %xmm5
	pxor	%xmm6, %xmm6

	movl	A, %edx		# aoffset = a
	addl	$5 * SIZE, A	# a += 5

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L126

.L124:
	KERNELMACRO5( 0)
	KERNELMACRO5( 2)
	KERNELMACRO5( 4)
	KERNELMACRO5( 6)
	KERNELMACRO5( 8)
	KERNELMACRO5(10)
	KERNELMACRO5(12)
	KERNELMACRO5(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L124

.L126:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L128

.L127:
	KERNELMACRO5( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L127

.L128:
	movsd	  0 * SIZE(%ecx), %xmm0
	mulpd	%xmm3, %xmm4
	movl	INCY, %eax
	mulpd	%xmm3, %xmm5
	movl	%ecx, %edx
	mulsd	%xmm3, %xmm6
	addl	%eax, %edx
	cmpl	$8, %eax
	jne	.L129

	movsd	  0 * SIZE(%ecx), %xmm0
	movhpd	  1 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm4

	movsd	  2 * SIZE(%ecx), %xmm0
	movhpd	  3 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm5

	movsd	  4 * SIZE(%ecx), %xmm0
	addsd	%xmm0, %xmm6

	movsd	%xmm4, 0 * SIZE(%ecx)
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 1 * SIZE(%ecx)

	movsd	%xmm5, 2 * SIZE(%ecx)
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 3 * SIZE(%ecx)

	movsd	%xmm6, 4 * SIZE(%ecx)
	jmp	.L99

.L129:
	movhpd	  0 * SIZE(%edx), %xmm0
	addl	  %eax, %edx
	addpd	%xmm0, %xmm4

	movsd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	movhpd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	addpd	%xmm1, %xmm5

	movsd	  0 * SIZE(%edx), %xmm2
	addsd	%xmm2, %xmm6

	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm6, 0 * SIZE(%ecx)
	jmp	.L99

.L130:
	cmpl	$4, %esi
	jne	.L140

.L131:
	pxor	%xmm4, %xmm4
	pxor	%xmm5, %xmm5

	movl	A, %edx		# aoffset = a
	addl	$4 * SIZE, A	# a += 5

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L136

.L134:
	KERNELMACRO4( 0)
	KERNELMACRO4( 2)
	KERNELMACRO4( 4)
	KERNELMACRO4( 6)
	KERNELMACRO4( 8)
	KERNELMACRO4(10)
	KERNELMACRO4(12)
	KERNELMACRO4(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L134

.L136:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L138

.L137:
	KERNELMACRO4( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L137

.L138:
	movsd	  0 * SIZE(%ecx), %xmm0
	mulpd	%xmm3, %xmm4
	movl	INCY, %eax
	mulpd	%xmm3, %xmm5
	movl	%ecx, %edx
	addl	%eax, %edx
	cmpl	$8, %eax
	jne	.L139

	movhpd	  1 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm4

	movsd	  2 * SIZE(%ecx), %xmm0
	movhpd	  3 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm5

	movsd	%xmm4, 0 * SIZE(%ecx)
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 1 * SIZE(%ecx)

	movsd	%xmm5, 2 * SIZE(%ecx)
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 3 * SIZE(%ecx)
	jmp	.L99

.L139:
	movhpd	  0 * SIZE(%edx), %xmm0
	addl	  %eax, %edx
	addpd	%xmm0, %xmm4

	movsd	  0 * SIZE(%edx), %xmm1
	addl	  %eax, %edx
	movhpd	  0 * SIZE(%edx), %xmm1
	addpd	%xmm1, %xmm5

	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm5, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm5, %xmm5
	movsd	%xmm5, 0 * SIZE(%ecx)
	jmp	.L99

.L140:
	cmpl	$3, %esi
	jne	.L150

.L141:
	pxor	%xmm4, %xmm4
	pxor	%xmm5, %xmm5

	movl	A, %edx		# aoffset = a
	addl	$3 * SIZE, A	# a += 5

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L146

.L144:
	KERNELMACRO3( 0)
	KERNELMACRO3( 2)
	KERNELMACRO3( 4)
	KERNELMACRO3( 6)
	KERNELMACRO3( 8)
	KERNELMACRO3(10)
	KERNELMACRO3(12)
	KERNELMACRO3(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L144

.L146:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L148

.L147:
	KERNELMACRO3( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L147

.L148:
	movsd	  0 * SIZE(%ecx), %xmm0
	mulpd	%xmm3, %xmm4
	movl	INCY, %eax
	mulsd	%xmm3, %xmm5
	movl	%ecx, %edx
	addl	%eax, %edx
	cmpl	$8, %eax
	jne	.L149

	movsd	  0 * SIZE(%ecx), %xmm0
	movhpd	  1 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm4

	movsd	  2 * SIZE(%ecx), %xmm0
	addsd	%xmm0, %xmm5

	movsd	%xmm4, 0 * SIZE(%ecx)
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 1 * SIZE(%ecx)

	movsd	%xmm5, 2 * SIZE(%ecx)
	jmp	.L99

.L149:
	movhpd	  0 * SIZE(%edx), %xmm0
	addl	  %eax, %edx
	addpd	%xmm0, %xmm4

	movsd	  0 * SIZE(%edx), %xmm1
	addsd	%xmm1, %xmm5

	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx

	movsd	%xmm5, 0 * SIZE(%ecx)
	jmp	.L99

.L150:
	cmpl	$2, %esi
	jne	.L160

.L151:
	pxor	%xmm4, %xmm4

	movl	A, %edx		# aoffset = a
	addl	$2 * SIZE, A	# a += 5

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L156

.L154:
	KERNELMACRO2( 0)
	KERNELMACRO2( 2)
	KERNELMACRO2( 4)
	KERNELMACRO2( 6)
	KERNELMACRO2( 8)
	KERNELMACRO2(10)
	KERNELMACRO2(12)
	KERNELMACRO2(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L154

.L156:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L158

.L157:
	KERNELMACRO2( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L157

.L158:
	movsd	  0 * SIZE(%ecx), %xmm0
	mulpd	%xmm3, %xmm4
	movl	INCY, %eax
	movl	%ecx, %edx
	addl	%eax, %edx
	cmpl	$8, %eax
	jne	.L159

	movsd	  0 * SIZE(%ecx), %xmm0
	movhpd	  1 * SIZE(%ecx), %xmm0
	addpd	%xmm0, %xmm4

	movsd	%xmm4, 0 * SIZE(%ecx)
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 1 * SIZE(%ecx)
	jmp	.L99

.L159:
	movhpd	  0 * SIZE(%edx), %xmm0
	addl	  %eax, %edx
	addpd	%xmm0, %xmm4

	movsd	%xmm4, 0 * SIZE(%ecx)
	addl	  %eax, %ecx
	unpckhpd %xmm4, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	jmp	.L99

.L160:
	pxor	%xmm4, %xmm4

	movl	A, %edx		# aoffset = a
	addl	$1 * SIZE, A	# a += 5

	movl	BUFFER, %ebx	# boffset = buffer
	movapd	  0 * SIZE(%ebx), %xmm0

	movl	%edi, %eax	# i = min_n
	sarl	$3, %eax
	jle	.L166

.L164:
	KERNELMACRO1( 0)
	KERNELMACRO1( 2)
	KERNELMACRO1( 4)
	KERNELMACRO1( 6)
	KERNELMACRO1( 8)
	KERNELMACRO1(10)
	KERNELMACRO1(12)
	KERNELMACRO1(14)
	addl	$16 * SIZE, %ebx
	decl	%eax
	jg	.L164

.L166:
	movl	%edi, %eax	# i = min_n
	andl	$7, %eax
	jle	.L168

.L167:
	KERNELMACRO1( 0)
	addl	$2 * SIZE, %ebx
	decl	%eax
	jg	.L167

.L168:
	mulsd	%xmm3, %xmm4
	movsd	  0 * SIZE(%ecx), %xmm0
	addsd	%xmm0, %xmm4
	movsd	%xmm4, 0 * SIZE(%ecx)
	jmp	.L99	

	EPILOGUE