From: Werner Saar Date: Sun, 27 Mar 2016 14:31:50 +0000 (+0200) Subject: added optimized zscal kernel for POWER8 X-Git-Tag: v0.2.18^2~12^2~1 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=35c98a355613b677134709cfc0ac648147397314;p=platform%2Fupstream%2Fopenblas.git added optimized zscal kernel for POWER8 --- diff --git a/kernel/power/KERNEL.POWER8 b/kernel/power/KERNEL.POWER8 index dc670d0..7b5e2c2 100644 --- a/kernel/power/KERNEL.POWER8 +++ b/kernel/power/KERNEL.POWER8 @@ -138,7 +138,7 @@ DROTKERNEL = drot.c SSCALKERNEL = sscal.c DSCALKERNEL = dscal.c #CSCALKERNEL = ../arm/zscal.c -#ZSCALKERNEL = zscal.c +ZSCALKERNEL = zscal.c # SSWAPKERNEL = sswap.c DSWAPKERNEL = dswap.c diff --git a/kernel/power/zscal.c b/kernel/power/zscal.c new file mode 100644 index 0000000..213839a --- /dev/null +++ b/kernel/power/zscal.c @@ -0,0 +1,176 @@ +/*************************************************************************** +Copyright (c) 2013-2016, 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. +*****************************************************************************/ + +/************************************************************************************** +* 2016/03/27 Werner Saar (wernsaar@googlemail.com) +* BLASTEST : OK +* CTEST : OK +* TEST : OK +* LAPACK-TEST : OK +**************************************************************************************/ + + +#include "common.h" + +#pragma GCC optimize "O1" + +#if defined(POWER8) +#include "zscal_microk_power8.c" +#endif + + +#ifndef HAVE_KERNEL_8 + +static void zscal_kernel_8(BLASLONG n, FLOAT *x, FLOAT *alpha) +{ + + BLASLONG i=0; + FLOAT *x1=x; + FLOAT alpha_r1=alpha[0]; + FLOAT alpha_r2=alpha[1]; + FLOAT alpha_i1=alpha[2]; + FLOAT alpha_i2=alpha[3]; + FLOAT temp00, temp01, temp10, temp11, temp20, temp21, temp30, temp31; + FLOAT x0_r, x0_i, x1_r, x1_i, x2_r, x2_i, x3_r, x3_i; + + while ( i 0 ) + { + alpha[0] = da_r; + alpha[1] = da_r; + alpha[2] = -da_i; + alpha[3] = da_i; + zscal_kernel_8(n1, x, alpha); + i=n1; + ip = n1 * 2; + + } + + while ( i < n ) + { + + temp = da_r * x[ip] - da_i * x[ip+1] ; + x[ip+1] = da_r * x[ip+1] + da_i * x[ip] ; + x[ip] = temp; + ip += 2; + i++; + } + + } + else + { + + inc_x2 = 2 * inc_x; + + while ( i < n ) + { + + temp = da_r * x[ip] - da_i * x[ip+1] ; + x[ip+1] = da_r * x[ip+1] + da_i * x[ip] ; + x[ip] = temp; + ip += inc_x2; + i++; + } + + + } + + return(0); + +} + + diff --git a/kernel/power/zscal_microk_power8.c b/kernel/power/zscal_microk_power8.c new file mode 100644 index 0000000..5e09d8d --- /dev/null +++ b/kernel/power/zscal_microk_power8.c @@ -0,0 +1,224 @@ +/*************************************************************************** +Copyright (c) 2013-2016, 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. +*****************************************************************************/ + +/************************************************************************************** +* 2016/03/25 Werner Saar (wernsaar@googlemail.com) +* +* I don't use fused multipy-add ( lapack precision problems ) +* +* BLASTEST : OK +* CTEST : OK +* TEST : OK +* LAPACK-TEST : OK +**************************************************************************************/ + +#define HAVE_KERNEL_8 1 + +static void zscal_kernel_8( BLASLONG n, FLOAT *x, FLOAT *alpha) __attribute__ ((noinline)); + +static void zscal_kernel_8( BLASLONG n, FLOAT *x, FLOAT *alpha) +{ + + + BLASLONG i = n; + BLASLONG o16 = 16; + BLASLONG o32 = 32; + BLASLONG o48 = 48; + BLASLONG o64 = 64; + BLASLONG o80 = 80; + BLASLONG o96 = 96; + BLASLONG o112 = 112; + FLOAT *x1=x; + FLOAT *x2=x+1; + BLASLONG pre = 384; + + __asm__ __volatile__ + ( + + "lxvd2x 32, 0, %3 \n\t" // alpha_r , alpha_r + "lxvd2x 33, %5, %3 \n\t" // -alpha_i , alpha_i + "addi %1, %1, -8 \n\t" + + "dcbt %2, %4 \n\t" + + "lxvd2x 40, 0, %2 \n\t" // x0_r, x0_i + "lxvd2x 41, %5, %2 \n\t" + "lxvd2x 42, %6, %2 \n\t" + "lxvd2x 43, %7, %2 \n\t" + "lxvd2x 44, %8, %2 \n\t" + "lxvd2x 45, %9, %2 \n\t" + "lxvd2x 46, %10, %2 \n\t" + "lxvd2x 47, %11, %2 \n\t" + + "addi %2, %2, 128 \n\t" + + "addic. %0 , %0 , -8 \n\t" + "ble 2f \n\t" + + ".align 5 \n\t" + "1: \n\t" + + "dcbt %2, %4 \n\t" + + "xvmuldp 48, 40, 32 \n\t" // x0_r * alpha_r, x0_i * alpha_r + "xvmuldp 49, 41, 32 \n\t" + "xvmuldp 50, 42, 32 \n\t" + "xvmuldp 51, 43, 32 \n\t" + "xvmuldp 52, 44, 32 \n\t" + "xvmuldp 53, 45, 32 \n\t" + "xvmuldp 54, 46, 32 \n\t" + "xvmuldp 55, 47, 32 \n\t" + + "xxswapd 56, 40 \n\t" + "xxswapd 57, 41 \n\t" + "xxswapd 58, 42 \n\t" + "xxswapd 59, 43 \n\t" + "xxswapd 60, 44 \n\t" + "xxswapd 61, 45 \n\t" + "xxswapd 62, 46 \n\t" + "xxswapd 63, 47 \n\t" + + "xvmuldp 56, 56, 33 \n\t" // x0_i * -alpha_i, x0_r * alpha_i + "xvmuldp 57, 57, 33 \n\t" + + "lxvd2x 40, 0, %2 \n\t" // x0_r, x0_i + "lxvd2x 41, %5, %2 \n\t" + + "xvmuldp 58, 58, 33 \n\t" + "xvmuldp 59, 59, 33 \n\t" + + "lxvd2x 42, %6, %2 \n\t" + "lxvd2x 43, %7, %2 \n\t" + + "xvmuldp 60, 60, 33 \n\t" + "xvmuldp 61, 61, 33 \n\t" + + "lxvd2x 44, %8, %2 \n\t" + "lxvd2x 45, %9, %2 \n\t" + + "xvmuldp 62, 62, 33 \n\t" + "xvmuldp 63, 63, 33 \n\t" + + "lxvd2x 46, %10, %2 \n\t" + "lxvd2x 47, %11, %2 \n\t" + + "xvadddp 48, 48 , 56 \n\t" + "xvadddp 49, 49 , 57 \n\t" + "xvadddp 50, 50 , 58 \n\t" + "xvadddp 51, 51 , 59 \n\t" + + "stxvd2x 48, 0, %1 \n\t" + "stxvd2x 49, %5, %1 \n\t" + + "xvadddp 52, 52 , 60 \n\t" + "xvadddp 53, 53 , 61 \n\t" + + "stxvd2x 50, %6, %1 \n\t" + "stxvd2x 51, %7, %1 \n\t" + + "xvadddp 54, 54 , 62 \n\t" + "xvadddp 55, 55 , 63 \n\t" + + "stxvd2x 52, %8, %1 \n\t" + "stxvd2x 53, %9, %1 \n\t" + "stxvd2x 54, %10, %1 \n\t" + "stxvd2x 55, %11, %1 \n\t" + + "addi %1, %1, 128 \n\t" + "addi %2, %2, 128 \n\t" + + "addic. %0 , %0 , -8 \n\t" + "bgt 1b \n\t" + + "2: \n\t" + + "xvmuldp 48, 40, 32 \n\t" // x0_r * alpha_r, x0_i * alpha_r + "xvmuldp 49, 41, 32 \n\t" + "xvmuldp 50, 42, 32 \n\t" + "xvmuldp 51, 43, 32 \n\t" + "xvmuldp 52, 44, 32 \n\t" + "xvmuldp 53, 45, 32 \n\t" + "xvmuldp 54, 46, 32 \n\t" + "xvmuldp 55, 47, 32 \n\t" + + "xxswapd 56, 40 \n\t" + "xxswapd 57, 41 \n\t" + "xxswapd 58, 42 \n\t" + "xxswapd 59, 43 \n\t" + "xxswapd 60, 44 \n\t" + "xxswapd 61, 45 \n\t" + "xxswapd 62, 46 \n\t" + "xxswapd 63, 47 \n\t" + + "xvmuldp 56, 56, 33 \n\t" // x0_i * -alpha_i, x0_r * alpha_i + "xvmuldp 57, 57, 33 \n\t" + "xvmuldp 58, 58, 33 \n\t" + "xvmuldp 59, 59, 33 \n\t" + "xvmuldp 60, 60, 33 \n\t" + "xvmuldp 61, 61, 33 \n\t" + "xvmuldp 62, 62, 33 \n\t" + "xvmuldp 63, 63, 33 \n\t" + + "xvadddp 48, 48 , 56 \n\t" + "xvadddp 49, 49 , 57 \n\t" + "xvadddp 50, 50 , 58 \n\t" + "xvadddp 51, 51 , 59 \n\t" + "xvadddp 52, 52 , 60 \n\t" + "xvadddp 53, 53 , 61 \n\t" + "xvadddp 54, 54 , 62 \n\t" + "xvadddp 55, 55 , 63 \n\t" + + "stxvd2x 48, 0, %1 \n\t" + "stxvd2x 49, %5, %1 \n\t" + "stxvd2x 50, %6, %1 \n\t" + "stxvd2x 51, %7, %1 \n\t" + "stxvd2x 52, %8, %1 \n\t" + "stxvd2x 53, %9, %1 \n\t" + "stxvd2x 54, %10, %1 \n\t" + "stxvd2x 55, %11, %1 \n\t" + + + : + : + "r" (i), // 0 + "r" (x2), // 1 + "r" (x1), // 2 + "r" (alpha), // 3 + "r" (pre), // 4 + "r" (o16), // 5 + "r" (o32), // 6 + "r" (o48), // 7 + "r" (o64), // 8 + "r" (o80), // 9 + "r" (o96), // 10 + "r" (o112) // 11 + : "cr0", "%0", "%2" , "%1", "memory" + ); + +} + +