Add SIMD version with AVX (no AVX2) and XTC output

Signed-off-by: Rafael Ravedutti <rafaelravedutti@gmail.com>
This commit is contained in:
Rafael Ravedutti 2022-03-02 23:12:04 +01:00
parent 022aa75c75
commit af92800c64
8 changed files with 228 additions and 31 deletions

View File

@ -10,6 +10,7 @@ Q ?= @
include $(MAKE_DIR)/config.mk
include $(MAKE_DIR)/include_$(TAG).mk
include $(MAKE_DIR)/include_LIKWID.mk
include $(MAKE_DIR)/include_GROMACS.mk
INCLUDES += -I./$(SRC_DIR)/includes
ifeq ($(strip $(DATA_LAYOUT)),AOS)
@ -52,6 +53,10 @@ ifeq ($(strip $(COMPUTE_STATS)),true)
DEFINES += -DCOMPUTE_STATS
endif
ifeq ($(strip $(XTC_OUTPUT)),true)
DEFINES += -DXTC_OUTPUT
endif
ifeq ($(strip $(USE_REFERENCE_VERSION)),true)
DEFINES += -DUSE_REFERENCE_VERSION
endif
@ -64,6 +69,10 @@ ifneq ($(VECTOR_WIDTH),)
DEFINES += -DVECTOR_WIDTH=$(VECTOR_WIDTH)
endif
ifeq ($(strip $(NO_AVX2)),true)
DEFINES += -DNO_AVX2
endif
VPATH = $(SRC_DIR) $(ASM_DIR)
ASM = $(patsubst $(SRC_DIR)/%.c, $(BUILD_DIR)/%.s,$(wildcard $(SRC_DIR)/*.c))
OVERWRITE:= $(patsubst $(ASM_DIR)/%-new.s, $(BUILD_DIR)/%.o,$(wildcard $(ASM_DIR)/*-new.s))

View File

@ -1,9 +1,9 @@
# Compiler tag (GCC/CLANG/ICC)
TAG ?= ICC
# Optimization scheme (lammps/gromacs/clusters_per_bin)
OPT_SCHEME = gromacs
OPT_SCHEME ?= gromacs
# Enable likwid (true or false)
ENABLE_LIKWID ?= false
ENABLE_LIKWID ?= true
# SP or DP
DATA_TYPE ?= DP
# AOS or SOA
@ -24,7 +24,9 @@ MEM_TRACER ?= false
# Trace indexes and distances for gather-md (true or false)
INDEX_TRACER ?= false
# Vector width (elements) for index and distance tracer
VECTOR_WIDTH ?= 8
VECTOR_WIDTH ?= 4
# When vector width is 4 but AVX2 is not supported (AVX only), set this to true
NO_AVX2 ?= true
# Compute statistics
COMPUTE_STATS ?= true
@ -33,6 +35,8 @@ COMPUTE_STATS ?= true
CLUSTER_LAYOUT ?= SOA
# Use reference version
USE_REFERENCE_VERSION ?= false
# Enable XTC output
XTC_OUTPUT ?= false
#Feature options
OPTIONS = -DALIGNMENT=64

View File

@ -33,6 +33,7 @@ typedef struct {
int force_field;
char* input_file;
char* vtk_file;
char *xtc_file;
MD_FLOAT epsilon;
MD_FLOAT sigma6;
MD_FLOAT temp;
@ -41,8 +42,10 @@ typedef struct {
int ntypes;
int ntimes;
int nstat;
int every;
int reneigh_every;
int prune_every;
int x_out_every;
int v_out_every;
MD_FLOAT dt;
MD_FLOAT dtforce;
MD_FLOAT cutforce;

View File

@ -41,10 +41,10 @@ static inline MD_SIMD_FLOAT simd_mul(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return
static inline MD_SIMD_FLOAT simd_fma(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_FLOAT c) { return _mm512_fmadd_pd(a, b, c); }
static inline MD_SIMD_FLOAT simd_reciprocal(MD_SIMD_FLOAT a) { return _mm512_rcp14_pd(a); }
static inline MD_SIMD_FLOAT simd_masked_add(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_MASK m) { return _mm512_mask_add_pd(a, m, a, b); }
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask8(a); }
static inline MD_SIMD_MASK simd_mask_to_u32(unsigned int a) { return _cvtmask8_u32(a); }
static inline MD_SIMD_MASK simd_mask_and(MD_SIMD_MASK a, MD_SIMD_MASK b) { return _kand_mask8(a, b); }
static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm512_cmp_pd_mask(a, b, _CMP_LT_OQ); }
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask8(a); }
static inline unsigned int simd_mask_to_u32(MD_SIMD_MASK a) { return _cvtmask8_u32(a); }
static MD_SIMD_FLOAT simd_load2(MD_FLOAT *c0, MD_FLOAT *c1, int d) {
MD_SIMD_FLOAT x;
@ -64,39 +64,55 @@ static inline MD_FLOAT simd_horizontal_sum(MD_SIMD_FLOAT a) {
MD_SIMD_FLOAT x = _mm512_add_pd(a, _mm512_shuffle_f64x2(a, a, 0xee));
x = _mm512_add_pd(x, _mm512_shuffle_f64x2(x, x, 0x11));
x = _mm512_add_pd(x, _mm512_permute_pd(x, 0x01));
return *((double *) &x);
return *((MD_FLOAT *) &x);
}
#else // AVX2
#else // AVX or AVX2
#define MD_SIMD_FLOAT __m256d
#ifdef NO_AVX2
#define MD_SIMD_MASK __m256d
#else
#define MD_SIMD_MASK __mmask8
#endif
static inline MD_SIMD_FLOAT simd_broadcast(double scalar) { return _mm256_set1_pd(scalar); }
static inline MD_SIMD_FLOAT simd_zero() { return _mm256_set1_pd(0.0); }
static inline MD_SIMD_FLOAT simd_add(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm256_add_pd(a, b); }
static inline MD_SIMD_FLOAT simd_sub(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm256_sub_pd(a, b); }
static inline MD_SIMD_FLOAT simd_mul(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm256_mul_pd(a, b); }
static inline MD_SIMD_FLOAT simd_fma(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_FLOAT c) { return _mm256_fmadd_pd(a, b, c); }
static inline MD_SIMD_FLOAT simd_reciprocal(MD_SIMD_FLOAT a) { return _mm256_rcp14_pd(a); }
static inline MD_SIMD_FLOAT simd_masked_add(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_MASK m) { return _mm256_mask_add_pd(a, m, a, b); }
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask8(a); }
static inline MD_SIMD_MASK simd_mask_to_u32(unsigned int a) { return _cvtmask8_u32(a); }
static inline MD_SIMD_MASK simd_mask_and(MD_SIMD_MASK a, MD_SIMD_MASK b) { return _kand_mask8(a, b); }
static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm256_cmp_pd_mask(a, b, _CMP_LT_OQ); }
static MD_SIMD_FLOAT simd_load(MD_FLOAT *c0, int d) {
MD_SIMD_FLOAT x;
#ifdef CLUSTER_AOS
__m128i aos_gather_vindex = _mm128_set_epi32(9, 6, 3, 0);
__m128i vindex = _mm128_add_epi32(aos_gather_vindex, _mm128_set1_epi32(d));
x = _mm256_i32gather_pd(c0, vindex, sizeof(double));
#else
x = _mm256_load_pd(&c0[d * CLUSTER_DIM_M]);
#endif
return x;
#ifdef NO_AVX2
static inline MD_SIMD_FLOAT simd_reciprocal(MD_SIMD_FLOAT a) { return _mm256_cvtps_pd(_mm_rcp_ps(_mm256_cvtpd_ps(a))); }
static inline MD_SIMD_FLOAT simd_fma(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_FLOAT c) { return simd_add(simd_mul(a, b), c); }
static inline MD_SIMD_FLOAT simd_masked_add(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_MASK m) { return simd_add(a, _mm256_and_pd(b, m)); }
static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm256_cmp_pd(a, b, _CMP_LT_OQ); }
static inline MD_SIMD_MASK simd_mask_and(MD_SIMD_MASK a, MD_SIMD_MASK b) { return _mm256_and_pd(a, b); }
// TODO: Initialize all diagonal cases and just select the proper one (all bits set or diagonal) based on cond0
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) {
const unsigned long long int all = 0xFFFFFFFFFFFFFFFF;
const unsigned long long int none = 0x0;
return _mm256_castsi256_pd(_mm256_set_epi64x((a & 0x8) ? all : none, (a & 0x4) ? all : none, (a & 0x2) ? all : none, (a & 0x1) ? all : none));
}
// TODO: Implement this, althrough it is just required for debugging
static inline int simd_mask_to_u32(MD_SIMD_MASK a) { return 0; }
static inline MD_FLOAT simd_horizontal_sum(MD_SIMD_FLOAT a) {
__m128d a0, a1;
a = _mm256_add_pd(a, _mm256_permute_pd(a, 0b0101));
a0 = _mm256_castpd256_pd128(a);
a1 = _mm256_extractf128_pd(a, 0x1);
a0 = _mm_add_sd(a0, a1);
return *((MD_FLOAT *) &a0);
}
#else
static inline MD_SIMD_FLOAT simd_reciprocal(MD_SIMD_FLOAT a) { return _mm256_rcp14_pd(a); }
static inline MD_SIMD_FLOAT simd_fma(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_FLOAT c) { return _mm256_fmadd_pd(a, b, c); }
static inline MD_SIMD_FLOAT simd_masked_add(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_MASK m) { return _mm256_mask_add_pd(a, m, a, b); }
static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm256_cmp_pd_mask(a, b, _CMP_LT_OQ); }
static inline MD_SIMD_MASK simd_mask_and(MD_SIMD_MASK a, MD_SIMD_MASK b) { return _kand_mask8(a, b); }
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask8(a); }
static inline unsigned int simd_mask_to_u32(MD_SIMD_MASK a) { return _cvtmask8_u32(a); }
static inline MD_FLOAT simd_horizontal_sum(MD_SIMD_FLOAT a) {
__m128d a0, a1;
// test with shuffle & add as an alternative to hadd later
@ -104,7 +120,23 @@ static inline MD_FLOAT simd_horizontal_sum(MD_SIMD_FLOAT a) {
a0 = _mm256_castpd256_pd128(a);
a1 = _mm256_extractf128_pd(a, 0x1);
a0 = _mm_add_sd(a0, a1);
return *((double *) &a0);
return *((MD_FLOAT *) &a0);
}
#endif
static MD_SIMD_FLOAT simd_load(MD_FLOAT *c0, int d) {
MD_SIMD_FLOAT x;
#ifdef CLUSTER_AOS
#ifdef NO_AVX2
#error "Not possible to use AoS cluster layout without AVX2 support!"
#endif
__m128i aos_gather_vindex = _mm128_set_epi32(9, 6, 3, 0);
__m128i vindex = _mm128_add_epi32(aos_gather_vindex, _mm128_set1_epi32(d));
x = _mm256_i32gather_pd(c0, vindex, sizeof(double));
#else
x = _mm256_load_pd(&c0[d * CLUSTER_DIM_M]);
#endif
return x;
}
#endif

37
gromacs/includes/xtc.h Normal file
View File

@ -0,0 +1,37 @@
/*
* =======================================================================================
*
* Author: Jan Eitzinger (je), jan.eitzinger@fau.de
* Copyright (c) 2020 RRZE, University Erlangen-Nuremberg
*
* This file is part of MD-Bench.
*
* MD-Bench is free software: you can redistribute it and/or modify it
* under the terms of the GNU Lesser General Public License as published
* by the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* MD-Bench is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
* PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
* details.
*
* You should have received a copy of the GNU Lesser General Public License along
* with MD-Bench. If not, see <https://www.gnu.org/licenses/>.
* =======================================================================================
*/
#include <atom.h>
#ifndef __XTC_H_
#define __XTC_H_
#ifdef XTC_OUTPUT
void xtc_init(const char *, Atom*, int);
void xtc_write(Atom*, int, int, int);
void xtc_end();
#else
#define xtc_init(a,b,c)
#define xtc_write(a,b,c,d)
#define xtc_end()
#endif
#endif

View File

@ -41,6 +41,7 @@
#include <timers.h>
#include <eam.h>
#include <vtk.h>
#include <xtc.h>
#include <util.h>
#define HLINE "----------------------------------------------------------------------------\n"
@ -58,6 +59,7 @@ extern double computeForceEam(Eam*, Parameter*, Atom*, Neighbor*, Stats*);
void init(Parameter *param) {
param->input_file = NULL;
param->vtk_file = NULL;
param->xtc_file = NULL;
param->force_field = FF_LJ;
param->epsilon = 1.0;
param->sigma6 = 1.0;
@ -75,8 +77,10 @@ void init(Parameter *param) {
param->nstat = 100;
param->mass = 1.0;
param->dtforce = 0.5 * param->dt;
param->every = 20;
param->reneigh_every = 20;
param->prune_every = 1000;
param->x_out_every = 20;
param->v_out_every = 5;
param->proc_freq = 2.4;
}
@ -241,6 +245,15 @@ int main(int argc, char** argv) {
param.vtk_file = strdup(argv[++i]);
continue;
}
if((strcmp(argv[i], "--xtc") == 0)) {
#ifndef XTC_OUTPUT
fprintf(stderr, "XTC not available, set XTC_OUTPUT option in config.mk file and recompile MD-Bench!");
exit(-1);
#else
param.xtc_file = strdup(argv[++i]);
#endif
continue;
}
if((strcmp(argv[i], "-h") == 0) || (strcmp(argv[i], "--help") == 0)) {
printf("MD Bench: A minimalistic re-implementation of miniMD\n");
printf(HLINE);
@ -253,6 +266,7 @@ int main(int argc, char** argv) {
printf("-s / --skin <real>: set skin (verlet buffer)\n");
printf("--freq <real>: processor frequency (GHz)\n");
printf("--vtk <string>: VTK file for visualization\n");
printf("--xtc <string>: XTC file for visualization\n");
printf(HLINE);
exit(EXIT_SUCCESS);
}
@ -277,10 +291,14 @@ int main(int argc, char** argv) {
write_data_to_vtk_file(param.vtk_file, &atom, 0);
}
if(param.xtc_file != NULL) {
xtc_init(param.xtc_file, &atom, 0);
}
for(int n = 0; n < param.ntimes; n++) {
initialIntegrate(&param, &atom);
if((n + 1) % param.every) {
if((n + 1) % param.reneigh_every) {
if(!((n + 1) % param.prune_every)) {
pruneNeighbor(&param, &atom, &neighbor);
}
@ -306,14 +324,26 @@ int main(int argc, char** argv) {
computeThermo(n + 1, &param, &atom);
}
if(param.vtk_file != NULL) {
write_data_to_vtk_file(param.vtk_file, &atom, n + 1);
int write_pos = !((n + 1) % param.x_out_every);
int write_vel = !((n + 1) % param.v_out_every);
if(write_pos || write_vel) {
if(param.vtk_file != NULL) {
write_data_to_vtk_file(param.vtk_file, &atom, n + 1);
}
if(param.xtc_file != NULL) {
xtc_write(&atom, n + 1, write_pos, write_vel);
}
}
}
timer[TOTAL] = getTimeStamp() - timer[TOTAL];
computeThermo(-1, &param, &atom);
if(param.xtc_file != NULL) {
xtc_end();
}
printf(HLINE);
printf("Force field: %s\n", ff2str(param.force_field));
printf("Data layout for positions: %s\n", POS_DATA_LAYOUT);

71
gromacs/xtc.c Normal file
View File

@ -0,0 +1,71 @@
/*
* =======================================================================================
*
* Author: Jan Eitzinger (je), jan.eitzinger@fau.de
* Copyright (c) 2020 RRZE, University Erlangen-Nuremberg
*
* This file is part of MD-Bench.
*
* MD-Bench is free software: you can redistribute it and/or modify it
* under the terms of the GNU Lesser General Public License as published
* by the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* MD-Bench is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
* PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
* details.
*
* You should have received a copy of the GNU Lesser General Public License along
* with MD-Bench. If not, see <https://www.gnu.org/licenses/>.
* =======================================================================================
*/
#include <stdlib.h>
//---
#include <atom.h>
#include <allocate.h>
#include <xtc.h>
#ifdef XTC_OUTPUT
#include <gromacs/fileio/xtcio.h>
static struct t_fileio *xtc_file = NULL;
static rvec *x_buf = NULL;
static rvec basis[3];
void xtc_init(const char *filename, Atom *atom, int timestep) {
basis[0][XX] = 1.0;
basis[0][YY] = 0.0;
basis[0][ZZ] = 0.0;
basis[1][XX] = 0.0;
basis[1][YY] = 1.0;
basis[1][ZZ] = 0.0;
basis[2][XX] = 0.0;
basis[2][YY] = 0.0;
basis[2][ZZ] = 1.0;
xtc_file = open_xtc(filename, "w");
x_buf = (rvec *) allocate(ALIGNMENT, sizeof(rvec) * (atom->Nlocal + 1));
xtc_write(atom, timestep, 1, 1);
}
void xtc_write(Atom *atom, int timestep, int write_pos, int write_vel) {
int i = 0;
for(int ci = 0; ci < atom->Nclusters_local; ++ci) {
MD_FLOAT *cptr = cluster_pos_ptr(ci);
for(int cii = 0; cii < atom->clusters[ci].natoms; ++cii) {
x_buf[i][XX] = cluster_x(cptr, cii);
x_buf[i][YY] = cluster_y(cptr, cii);
x_buf[i][ZZ] = cluster_z(cptr, cii);
i++;
}
}
write_xtc(xtc_file, atom->Nlocal, timestep, 0.0, (const rvec *) basis, (const rvec *) x_buf, 1000);
}
void xtc_end() {
free(x_buf);
close_xtc(xtc_file);
}
#endif

11
include_GROMACS.mk Normal file
View File

@ -0,0 +1,11 @@
GROMACS_PATH=/apps/Gromacs/2018.1-mkl
GROMACS_INC ?= -I${GROMACS_PATH}/include
GROMACS_DEFINES ?=
GROMACS_LIB ?= -L${GROMACS_PATH}/lib64
ifeq ($(strip $(XTC_OUTPUT)),true)
INCLUDES += ${GROMACS_INC}
DEFINES += ${GROMACS_DEFINES}
LIBS += -lgromacs
LFLAGS += ${GROMACS_LIB}
endif