From 59083e3ce14e0cd6539329bfcdb87484a84c81fc Mon Sep 17 00:00:00 2001 From: Werner Saar Date: Mon, 18 May 2015 07:33:52 +0200 Subject: [PATCH] added optimized cscal kernel for bulldozer --- kernel/x86_64/KERNEL.BULLDOZER | 1 + kernel/x86_64/cscal.c | 2 + kernel/x86_64/cscal_microk_bulldozer-2.c | 348 +++++++++++++++++++++++ 3 files changed, 351 insertions(+) create mode 100644 kernel/x86_64/cscal_microk_bulldozer-2.c diff --git a/kernel/x86_64/KERNEL.BULLDOZER b/kernel/x86_64/KERNEL.BULLDOZER index 941ac8d4a..9f124c97f 100644 --- a/kernel/x86_64/KERNEL.BULLDOZER +++ b/kernel/x86_64/KERNEL.BULLDOZER @@ -1,4 +1,5 @@ DSCALKERNEL = dscal.c +CSCALKERNEL = cscal.c ZSCALKERNEL = zscal.c DAXPYKERNEL = daxpy.c diff --git a/kernel/x86_64/cscal.c b/kernel/x86_64/cscal.c index 3b530f5b4..4785702fd 100644 --- a/kernel/x86_64/cscal.c +++ b/kernel/x86_64/cscal.c @@ -30,6 +30,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. #if defined(HASWELL) #include "cscal_microk_haswell-2.c" +#elif defined(BULLDOZER) +#include "cscal_microk_bulldozer-2.c" #endif diff --git a/kernel/x86_64/cscal_microk_bulldozer-2.c b/kernel/x86_64/cscal_microk_bulldozer-2.c new file mode 100644 index 000000000..f470cf843 --- /dev/null +++ b/kernel/x86_64/cscal_microk_bulldozer-2.c @@ -0,0 +1,348 @@ +/*************************************************************************** +Copyright (c) 2014-2015, 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. +*****************************************************************************/ + + +#define HAVE_KERNEL_16 1 + +static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline)); + +static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) +{ + + + __asm__ __volatile__ + ( + "vbroadcastss (%2), %%xmm0 \n\t" // da_r + "vbroadcastss 4(%2), %%xmm1 \n\t" // da_i + + "addq $128, %1 \n\t" + + "vmovups -128(%1), %%xmm4 \n\t" + "vmovups -112(%1), %%xmm5 \n\t" + "vmovups -96(%1), %%xmm6 \n\t" + "vmovups -80(%1), %%xmm7 \n\t" + + "vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t" + "vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t" + "vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t" + "vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t" + + "subq $8 , %0 \n\t" + "jz 2f \n\t" + + ".align 16 \n\t" + "1: \n\t" + + "prefetcht0 320(%1) \n\t" + // ".align 2 \n\t" + + "vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1 + "vmovups -64(%1), %%xmm4 \n\t" + "vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t" + "vmovups -48(%1), %%xmm5 \n\t" + "vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t" + "vmovups -32(%1), %%xmm6 \n\t" + "vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t" + "vmovups -16(%1), %%xmm7 \n\t" + + "vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0 + "vaddsubps %%xmm12 , %%xmm8 , %%xmm8 \n\t" + "vmulps %%xmm1, %%xmm13, %%xmm13 \n\t" + "vaddsubps %%xmm13 , %%xmm9 , %%xmm9 \n\t" + "vmulps %%xmm1, %%xmm14, %%xmm14 \n\t" + "vaddsubps %%xmm14 , %%xmm10, %%xmm10 \n\t" + "vmulps %%xmm1, %%xmm15, %%xmm15 \n\t" + "vaddsubps %%xmm15 , %%xmm11, %%xmm11 \n\t" + + "vmovups %%xmm8 , -128(%1) \n\t" + "vmovups %%xmm9 , -112(%1) \n\t" + "vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t" + "vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t" + "vmovups %%xmm10, -96(%1) \n\t" + "vmovups %%xmm11, -80(%1) \n\t" + "vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t" + "vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t" + + "addq $64 ,%1 \n\t" + "subq $8 , %0 \n\t" + "jnz 1b \n\t" + + "2: \n\t" + + + "vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1 + "vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t" + "vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t" + "vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t" + + "vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0 + "vaddsubps %%xmm12 , %%xmm8 , %%xmm8 \n\t" + "vmulps %%xmm1, %%xmm13, %%xmm13 \n\t" + "vaddsubps %%xmm13 , %%xmm9 , %%xmm9 \n\t" + "vmulps %%xmm1, %%xmm14, %%xmm14 \n\t" + "vaddsubps %%xmm14 , %%xmm10, %%xmm10 \n\t" + "vmulps %%xmm1, %%xmm15, %%xmm15 \n\t" + "vaddsubps %%xmm15 , %%xmm11, %%xmm11 \n\t" + + "vmovups %%xmm8 , -128(%1) \n\t" + "vmovups %%xmm9 , -112(%1) \n\t" + "vmovups %%xmm10, -96(%1) \n\t" + "vmovups %%xmm11, -80(%1) \n\t" + + "vzeroupper \n\t" + + : + : + "r" (n), // 0 + "r" (x), // 1 + "r" (alpha) // 2 + : "cc", "%0", "%1", + "%xmm0", "%xmm1", "%xmm2", "%xmm3", + "%xmm4", "%xmm5", "%xmm6", "%xmm7", + "%xmm8", "%xmm9", "%xmm10", "%xmm11", + "%xmm12", "%xmm13", "%xmm14", "%xmm15", + "memory" + ); + +} + +static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline)); + +static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) +{ + + + __asm__ __volatile__ + ( + "vxorps %%xmm0, %%xmm0, %%xmm0 \n\t" + "vbroadcastss 4(%2), %%xmm1 \n\t" // da_i + + "addq $128, %1 \n\t" + + "vmovups -128(%1), %%xmm4 \n\t" + "vmovups -112(%1), %%xmm5 \n\t" + "vmovups -96(%1), %%xmm6 \n\t" + "vmovups -80(%1), %%xmm7 \n\t" + + "vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t" + "vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t" + "vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t" + "vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t" + + "subq $8 , %0 \n\t" + "jz 2f \n\t" + + ".align 16 \n\t" + "1: \n\t" + + //"prefetcht0 128(%1) \n\t" + // ".align 2 \n\t" + + "vmovups -64(%1), %%xmm4 \n\t" + "vmovups -48(%1), %%xmm5 \n\t" + "vmovups -32(%1), %%xmm6 \n\t" + "vmovups -16(%1), %%xmm7 \n\t" + + "vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0 + "vaddsubps %%xmm12 , %%xmm0 , %%xmm8 \n\t" + "vmulps %%xmm1, %%xmm13, %%xmm13 \n\t" + "vaddsubps %%xmm13 , %%xmm0 , %%xmm9 \n\t" + "vmulps %%xmm1, %%xmm14, %%xmm14 \n\t" + "vaddsubps %%xmm14 , %%xmm0 , %%xmm10 \n\t" + "vmulps %%xmm1, %%xmm15, %%xmm15 \n\t" + "vaddsubps %%xmm15 , %%xmm0 , %%xmm11 \n\t" + + "vmovups %%xmm8 , -128(%1) \n\t" + "vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t" + "vmovups %%xmm9 , -112(%1) \n\t" + "vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t" + "vmovups %%xmm10, -96(%1) \n\t" + "vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t" + "vmovups %%xmm11, -80(%1) \n\t" + "vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t" + + "addq $64 ,%1 \n\t" + "subq $8 , %0 \n\t" + "jnz 1b \n\t" + + "2: \n\t" + + "vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0 + "vaddsubps %%xmm12 , %%xmm0 , %%xmm8 \n\t" + "vmulps %%xmm1, %%xmm13, %%xmm13 \n\t" + "vaddsubps %%xmm13 , %%xmm0 , %%xmm9 \n\t" + "vmulps %%xmm1, %%xmm14, %%xmm14 \n\t" + "vaddsubps %%xmm14 , %%xmm0 , %%xmm10 \n\t" + "vmulps %%xmm1, %%xmm15, %%xmm15 \n\t" + "vaddsubps %%xmm15 , %%xmm0 , %%xmm11 \n\t" + + "vmovups %%xmm8 , -128(%1) \n\t" + "vmovups %%xmm9 , -112(%1) \n\t" + "vmovups %%xmm10, -96(%1) \n\t" + "vmovups %%xmm11, -80(%1) \n\t" + + "vzeroupper \n\t" + + : + : + "r" (n), // 0 + "r" (x), // 1 + "r" (alpha) // 2 + : "cc", "%0", "%1", + "%xmm0", "%xmm1", "%xmm2", "%xmm3", + "%xmm4", "%xmm5", "%xmm6", "%xmm7", + "%xmm8", "%xmm9", "%xmm10", "%xmm11", + "%xmm12", "%xmm13", "%xmm14", "%xmm15", + "memory" + ); + +} + + + +static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline)); + +static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) +{ + + + __asm__ __volatile__ + ( + "vbroadcastss (%2), %%xmm0 \n\t" // da_r + + "addq $128, %1 \n\t" + + "vmovups -128(%1), %%xmm4 \n\t" + "vmovups -112(%1), %%xmm5 \n\t" + "vmovups -96(%1), %%xmm6 \n\t" + "vmovups -80(%1), %%xmm7 \n\t" + + + "subq $8 , %0 \n\t" + "jz 2f \n\t" + + ".align 16 \n\t" + "1: \n\t" + + //"prefetcht0 128(%1) \n\t" + // ".align 2 \n\t" + + "vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1 + "vmovups -64(%1), %%xmm4 \n\t" + "vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t" + "vmovups -48(%1), %%xmm5 \n\t" + "vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t" + "vmovups -32(%1), %%xmm6 \n\t" + "vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t" + "vmovups -16(%1), %%xmm7 \n\t" + + "vmovups %%xmm8 , -128(%1) \n\t" + "vmovups %%xmm9 , -112(%1) \n\t" + "vmovups %%xmm10, -96(%1) \n\t" + "vmovups %%xmm11, -80(%1) \n\t" + + "addq $64 ,%1 \n\t" + "subq $8 , %0 \n\t" + "jnz 1b \n\t" + + "2: \n\t" + + + "vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1 + "vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t" + "vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t" + "vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t" + + "vmovups %%xmm8 , -128(%1) \n\t" + "vmovups %%xmm9 , -112(%1) \n\t" + "vmovups %%xmm10, -96(%1) \n\t" + "vmovups %%xmm11, -80(%1) \n\t" + + "vzeroupper \n\t" + + : + : + "r" (n), // 0 + "r" (x), // 1 + "r" (alpha) // 2 + : "cc", "%0", "%1", + "%xmm0", "%xmm1", "%xmm2", "%xmm3", + "%xmm4", "%xmm5", "%xmm6", "%xmm7", + "%xmm8", "%xmm9", "%xmm10", "%xmm11", + "%xmm12", "%xmm13", "%xmm14", "%xmm15", + "memory" + ); + +} + + +static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline)); + +static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) +{ + + + __asm__ __volatile__ + ( + "vxorps %%xmm0, %%xmm0, %%xmm0 \n\t" + + "addq $128, %1 \n\t" + + ".align 16 \n\t" + "1: \n\t" + + //"prefetcht0 128(%1) \n\t" + // ".align 2 \n\t" + + "vmovups %%xmm0 , -128(%1) \n\t" + "vmovups %%xmm0 , -112(%1) \n\t" + "vmovups %%xmm0 , -96(%1) \n\t" + "vmovups %%xmm0 , -80(%1) \n\t" + + "addq $64 ,%1 \n\t" + "subq $8 , %0 \n\t" + "jnz 1b \n\t" + + "vzeroupper \n\t" + + : + : + "r" (n), // 0 + "r" (x), // 1 + "r" (alpha) // 2 + : "cc", "%0", "%1", + "%xmm0", "%xmm1", "%xmm2", "%xmm3", + "%xmm4", "%xmm5", "%xmm6", "%xmm7", + "%xmm8", "%xmm9", "%xmm10", "%xmm11", + "%xmm12", "%xmm13", "%xmm14", "%xmm15", + "memory" + ); + +} + + +