Commit c103a79f authored by Pedro Gonnet's avatar Pedro Gonnet
Browse files

add memswap.h, which provides in-place swapping.

parent 4f696e9b
......@@ -44,7 +44,7 @@ include_HEADERS = space.h runner.h queue.h task.h lock.h cell.h part.h const.h \
common_io.h single_io.h multipole.h map.h tools.h partition.h clocks.h parser.h \
physical_constants.h physical_constants_cgs.h potential.h version.h \
hydro_properties.h riemann.h threadpool.h cooling.h cooling_struct.h sourceterms.h \
sourceterms_struct.h statistics.h
sourceterms_struct.h statistics.h memswap.h
# Common source files
......@@ -83,7 +83,8 @@ nobase_noinst_HEADERS = align.h approx_math.h atomic.h cycle.h error.h inline.h
potential/softened_isothermal/potential.h \
cooling/none/cooling.h cooling/none/cooling_struct.h \
cooling/const_du/cooling.h cooling/const_du/cooling_struct.h \
cooling/const_lambda/cooling.h cooling/const_lambda/cooling_struct.h
cooling/const_lambda/cooling.h cooling/const_lambda/cooling_struct.h \
# Sources and flags for regular library
* This file is part of SWIFT.
* Copyright (c) 2012 Pedro Gonnet (
* Matthieu Schaller (
* 2015 Peter W. Draper (
* 2016 John A. Regan (
* Tom Theuns (
* This program 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.
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* GNU General Public License for more details.
* You should have received a copy of the GNU Lesser General Public License
* along with this program. If not, see <>.
/* Config parameters. */
#include "../config.h"
/* Include the header file with the intrinsics for Intel architecture. */
#include <immintrin.h>
/* Include the header file with the intrinsics for Intel architecture. */
#include <altivec.h>
/* Macro for in-place swap of two values a and b of type t. */
#define swap_loop(t, a, b, c) \
while (c >= sizeof(t)) { \
register t temp = *(t *)a; \
*(t *)a = *(t *)b; \
*(t *)b = temp; \
a += sizeof(t); \
b += sizeof(t); \
bytes -= sizeof(t); \
* @brief Swap the contents of two elements in-place.
* Keep in mind that this function works best when the underlying data
* is aligned to the vector length, e.g. with the @c __attribute__((aligned(32)))
* syntax, and the code is compiled with @c -funroll-loops.
* @param a Pointer to the first element.
* @param b Pointer to the second element.
* @param bytes Size, in bytes, of the data pointed to by @c a and @c b.
__attribute__((always_inline)) inline void memswap(void *void_a, void *void_b,
size_t bytes) {
char *a = (char *)void_a, *b = (char *)void_b;
#ifdef __AVX512F__
swap_loop(__m512i, a, b, bytes)
#ifdef __AVX__
swap_loop(__m256i, a, b, bytes)
#ifdef __SSE2__
swap_loop(__m128i, a, b, bytes)
#ifdef __ALTIVEC__
swap_loop(vector int, a, b, bytes)
swap_loop(size_t, a, b, bytes)
swap_loop(int, a, b, bytes)
swap_loop(short, a, b, bytes)
swap_loop(char, a, b, bytes)
#endif /* SWIFT_MEMSWAP_H */
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment