aoqi@0: /*
aoqi@0:  * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
aoqi@0:  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
aoqi@0:  *
aoqi@0:  * This code is free software; you can redistribute it and/or modify it
aoqi@0:  * under the terms of the GNU General Public License version 2 only, as
aoqi@0:  * published by the Free Software Foundation.
aoqi@0:  *
aoqi@0:  * This code is distributed in the hope that it will be useful, but WITHOUT
aoqi@0:  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
aoqi@0:  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
aoqi@0:  * version 2 for more details (a copy is included in the LICENSE file that
aoqi@0:  * accompanied this code).
aoqi@0:  *
aoqi@0:  * You should have received a copy of the GNU General Public License version
aoqi@0:  * 2 along with this work; if not, write to the Free Software Foundation,
aoqi@0:  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
aoqi@0:  *
aoqi@0:  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
aoqi@0:  * or visit www.oracle.com if you need additional information or have any
aoqi@0:  * questions.
aoqi@0:  *
aoqi@0:  */
aoqi@0: 
aoqi@0: #include "precompiled.hpp"
aoqi@0: #include "runtime/sharedRuntime.hpp"
aoqi@0: #include "utilities/copy.hpp"
aoqi@0: 
aoqi@0: 
aoqi@0: // Copy bytes; larger units are filled atomically if everything is aligned.
aoqi@0: void Copy::conjoint_memory_atomic(void* from, void* to, size_t size) {
aoqi@0:   address src = (address) from;
aoqi@0:   address dst = (address) to;
aoqi@0:   uintptr_t bits = (uintptr_t) src | (uintptr_t) dst | (uintptr_t) size;
aoqi@0: 
aoqi@0:   // (Note:  We could improve performance by ignoring the low bits of size,
aoqi@0:   // and putting a short cleanup loop after each bulk copy loop.
aoqi@0:   // There are plenty of other ways to make this faster also,
aoqi@0:   // and it's a slippery slope.  For now, let's keep this code simple
aoqi@0:   // since the simplicity helps clarify the atomicity semantics of
aoqi@0:   // this operation.  There are also CPU-specific assembly versions
aoqi@0:   // which may or may not want to include such optimizations.)
aoqi@0: 
aoqi@0:   if (bits % sizeof(jlong) == 0) {
aoqi@0:     Copy::conjoint_jlongs_atomic((jlong*) src, (jlong*) dst, size / sizeof(jlong));
aoqi@0:   } else if (bits % sizeof(jint) == 0) {
aoqi@0:     Copy::conjoint_jints_atomic((jint*) src, (jint*) dst, size / sizeof(jint));
aoqi@0:   } else if (bits % sizeof(jshort) == 0) {
aoqi@0:     Copy::conjoint_jshorts_atomic((jshort*) src, (jshort*) dst, size / sizeof(jshort));
aoqi@0:   } else {
aoqi@0:     // Not aligned, so no need to be atomic.
aoqi@0:     Copy::conjoint_jbytes((void*) src, (void*) dst, size);
aoqi@0:   }
aoqi@0: }
aoqi@0: 
aoqi@0: 
aoqi@0: // Fill bytes; larger units are filled atomically if everything is aligned.
aoqi@0: void Copy::fill_to_memory_atomic(void* to, size_t size, jubyte value) {
aoqi@0:   address dst = (address) to;
aoqi@0:   uintptr_t bits = (uintptr_t) to | (uintptr_t) size;
aoqi@0:   if (bits % sizeof(jlong) == 0) {
aoqi@0:     jlong fill = (julong)( (jubyte)value ); // zero-extend
aoqi@0:     if (fill != 0) {
aoqi@0:       fill += fill << 8;
aoqi@0:       fill += fill << 16;
aoqi@0:       fill += fill << 32;
aoqi@0:     }
aoqi@0:     //Copy::fill_to_jlongs_atomic((jlong*) dst, size / sizeof(jlong));
aoqi@0:     for (uintptr_t off = 0; off < size; off += sizeof(jlong)) {
aoqi@0:       *(jlong*)(dst + off) = fill;
aoqi@0:     }
aoqi@0:   } else if (bits % sizeof(jint) == 0) {
aoqi@0:     jint fill = (juint)( (jubyte)value ); // zero-extend
aoqi@0:     if (fill != 0) {
aoqi@0:       fill += fill << 8;
aoqi@0:       fill += fill << 16;
aoqi@0:     }
aoqi@0:     //Copy::fill_to_jints_atomic((jint*) dst, size / sizeof(jint));
aoqi@0:     for (uintptr_t off = 0; off < size; off += sizeof(jint)) {
aoqi@0:       *(jint*)(dst + off) = fill;
aoqi@0:     }
aoqi@0:   } else if (bits % sizeof(jshort) == 0) {
aoqi@0:     jshort fill = (jushort)( (jubyte)value ); // zero-extend
aoqi@0:     fill += fill << 8;
aoqi@0:     //Copy::fill_to_jshorts_atomic((jshort*) dst, size / sizeof(jshort));
aoqi@0:     for (uintptr_t off = 0; off < size; off += sizeof(jshort)) {
aoqi@0:       *(jshort*)(dst + off) = fill;
aoqi@0:     }
aoqi@0:   } else {
aoqi@0:     // Not aligned, so no need to be atomic.
aoqi@0:     Copy::fill_to_bytes(dst, size, value);
aoqi@0:   }
aoqi@0: }