added dgemm_ncopy_4_vfpv3.S

kernel/arm/dgemm_ncopy_4_vfpv3.S

@@ -0,0 +1,344 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+/**************************************************************************************
+* 2013/10/11 Saar
+* 	 BLASTEST 		: xOK
+* 	 CTEST			: xOK
+* 	 TEST			: xOK
+*
+**************************************************************************************/
+
+#define ASSEMBLER
+#include "common.h"
+
+#define STACKSIZE 256
+
+#define	OLD_M	r0
+#define	OLD_N	r1
+#define	OLD_A	r2
+#define	OLD_LDA	r3
+
+
+/******************************************************
+* [fp, #-128] - [fp, #-64] is reserved
+* for store and restore of floating point
+* registers
+*******************************************************/
+
+#define LDA	[fp, #-260 ]
+
+#define B	[fp, #4 ]
+
+#define M	r0
+#define N	r1
+#define A	r2
+
+#define	BO	r5
+
+#define	AO1	r6
+#define	AO2	r7
+#define	AO3	r8
+#define	AO4	r9
+
+#define I	r3
+#define	J	r12
+
+#define A_PRE	96
+
+/**************************************************************************************
+* Macro definitions
+**************************************************************************************/
+
+.macro COPY4x4
+
+	fldd	d0 , [ AO1, #0  ]
+	fldd	d1 , [ AO2, #0  ]
+	fldd	d2 , [ AO3, #0  ]
+	fldd	d3 , [ AO4, #0  ]
+
+	fldd	d4 , [ AO1, #8  ]
+	fldd	d8 , [ AO1, #16 ]
+	fldd	d12, [ AO1, #24 ]
+
+	fldd	d5 , [ AO2, #8  ]
+	add	AO1, AO1, #32
+	fldd	d9 , [ AO2, #16 ]
+	fldd	d13, [ AO2, #24 ]
+
+	fldd	d6 , [ AO3, #8  ]
+	add	AO2, AO2, #32
+	fldd	d10, [ AO3, #16 ]
+	fldd	d14, [ AO3, #24 ]
+
+	fldd	d7 , [ AO4, #8  ]
+	add	AO3, AO3, #32
+	fldd	d11, [ AO4, #16 ]
+	fldd	d15, [ AO4, #24 ]
+
+	fstmiad	BO!, { d0 - d3 }
+	add	AO4, AO4, #32
+	fstmiad	BO!, { d4 - d7 }
+	fstmiad	BO!, { d8 - d15 }
+
+.endm
+
+.macro COPY1x4
+
+	fldd	d0 , [ AO1, #0  ]
+	fldd	d1 , [ AO2, #0  ]
+	add	AO1, AO1, #8
+	fldd	d2 , [ AO3, #0  ]
+	add	AO2, AO2, #8
+	fldd	d3 , [ AO4, #0  ]
+
+	add	AO3, AO3, #8
+	fstmiad	BO!, { d0 - d3 }
+	add	AO4, AO4, #8
+
+.endm
+
+.macro COPY4x2
+
+	fldd	d0 , [ AO1, #0  ]
+	fldd	d2 , [ AO1, #8  ]
+	fldd	d4 , [ AO1, #16 ]
+	fldd	d6 , [ AO1, #24 ]
+
+	fldd	d1 , [ AO2, #0  ]
+	fldd	d3 , [ AO2, #8  ]
+	add	AO1, AO1, #32
+	fldd	d5 , [ AO2, #16 ]
+	fldd	d7 , [ AO2, #24 ]
+
+	fstmiad	BO!, { d0 - d7 }
+	add	AO2, AO2, #32
+
+.endm
+
+
+.macro COPY1x2
+
+	fldd	d0 , [ AO1, #0  ]
+	fldd	d1 , [ AO2, #0  ]
+	add	AO1, AO1, #8
+
+	fstmiad	BO!, { d0 - d1 }
+	add	AO2, AO2, #8
+
+.endm
+
+.macro COPY4x1
+
+	fldd	d0 , [ AO1, #0  ]
+	fldd	d1 , [ AO1, #8  ]
+	fldd	d2 , [ AO1, #16 ]
+	fldd	d3 , [ AO1, #24 ]
+
+	fstmiad	BO!, { d0 - d3 }
+	add	AO1, AO1, #32
+
+.endm
+
+
+.macro COPY1x1
+
+	fldd	d0 , [ AO1, #0  ]
+
+	fstmiad	BO!, { d0 }
+	add	AO1, AO1, #8
+
+.endm
+
+
+
+
+
+/**************************************************************************************
+* End of macro definitions
+**************************************************************************************/
+
+	PROLOGUE
+
+	.align 5
+
+	push	{r4 - r9, fp}
+	add	fp, sp, #24
+	sub	sp, sp, #STACKSIZE				// reserve stack
+
+
+	lsl	r3, r3, #3					// lda = lda * 8
+	str	r3, LDA
+
+	sub	r4, fp, #128
+	vstm	r4, { d8 - d15} 				// store floating point registers
+
+	ldr	BO, B	
+
+_L4_BEGIN:
+
+	asrs	J, N, #2					// J = N / 4
+	ble	_L2_BEGIN
+
+_L4_M4_BEGIN:
+
+	mov	AO1, A						// AO1 = A
+	ldr	r4 , LDA
+	add	AO2, AO1, r4
+	add	AO3, AO2, r4
+	add	AO4, AO3, r4
+	add	A  , AO4, r4					// A = A + 4 * LDA
+
+	asrs	I, M, #2					// I = M / 4
+	ble	_L4_M4_40
+
+_L4_M4_20:
+
+	COPY4x4
+
+	subs	I , I , #1
+	bne	_L4_M4_20
+	
+	
+_L4_M4_40:
+
+	ands	I, M , #3
+	ble	_L4_M4_END
+
+_L4_M4_60:
+
+	COPY1x4
+
+	subs	I , I , #1
+	bne	_L4_M4_60
+	
+
+_L4_M4_END:
+
+	subs	J , J, #1						// j--
+	bne	_L4_M4_BEGIN
+
+
+
+/*********************************************************************************************/
+
+_L2_BEGIN:
+
+	tst	N, #3
+	ble	_L999
+
+	tst	N, #2
+	ble	_L1_BEGIN
+
+_L2_M4_BEGIN:
+
+	mov	AO1, A						// AO1 = A
+	ldr	r4 , LDA
+	add	AO2, AO1, r4
+	add	A  , AO2, r4 					// A = A + 2 * LDA
+
+	asrs	I, M, #2					// I = M / 4
+	ble	_L2_M4_40
+
+_L2_M4_20:
+
+	COPY4x2
+
+	subs	I , I , #1
+	bne	_L2_M4_20
+	
+	
+_L2_M4_40:
+
+	ands	I, M , #3
+	ble	_L2_M4_END
+
+_L2_M4_60:
+
+	COPY1x2
+
+	subs	I , I , #1
+	bne	_L2_M4_60
+	
+
+_L2_M4_END:
+
+
+/*********************************************************************************************/
+
+_L1_BEGIN:
+
+	tst	N, #1
+	ble	_L999
+
+
+_L1_M4_BEGIN:
+
+	mov	AO1, A						// AO1 = A
+	ldr	r4 , LDA
+	add	A  , AO1, r4 					// A = A + 1 * LDA
+
+	asrs	I, M, #2					// I = M / 4
+	ble	_L1_M4_40
+
+_L1_M4_20:
+
+	COPY4x1
+
+	subs	I , I , #1
+	bne	_L1_M4_20
+	
+	
+_L1_M4_40:
+
+	ands	I, M , #3
+	ble	_L1_M4_END
+
+_L1_M4_60:
+
+	COPY1x1
+
+	subs	I , I , #1
+	bne	_L1_M4_60
+	
+
+_L1_M4_END:
+
+
+
+_L999:
+
+	sub	r3, fp, #128
+	vldm	r3, { d8 - d15}					// restore floating point registers
+
+	movs	r0, #0						// set return value
+	sub	sp, fp, #24
+	pop	{r4 - r9, fp}
+	bx	lr
+
+	EPILOGUE
+