Browse Source
Merge pull request #725 from yuriks/remove-common-crap
Merge pull request #725 from yuriks/remove-common-crap
Remove unused hash and mem_arena from commonpull/15/merge
Yuri Kunde Schlesner
11 years ago
7 changed files with 31 additions and 1103 deletions
-
4src/common/CMakeLists.txt
-
524src/common/hash.cpp
-
17src/common/hash.h
-
394src/common/mem_arena.cpp
-
70src/common/mem_arena.h
-
113src/core/mem_map.cpp
-
12src/core/mem_map.h
@ -1,524 +0,0 @@ |
|||
// Copyright 2013 Dolphin Emulator Project / 2014 Citra Emulator Project
|
|||
// Licensed under GPLv2 or any later version
|
|||
// Refer to the license.txt file included.
|
|||
|
|||
#include <algorithm>
|
|||
|
|||
#include "common/common_funcs.h" // For rotl
|
|||
#include "common/hash.h"
|
|||
#include "common/platform.h"
|
|||
|
|||
#if _M_SSE >= 0x402
|
|||
#include "common/cpu_detect.h"
|
|||
#include <nmmintrin.h>
|
|||
#endif
|
|||
|
|||
static u64 (*ptrHashFunction)(const u8 *src, int len, u32 samples) = &GetMurmurHash3; |
|||
|
|||
// uint32_t
|
|||
// WARNING - may read one more byte!
|
|||
// Implementation from Wikipedia.
|
|||
u32 HashFletcher(const u8* data_u8, size_t length) |
|||
{ |
|||
const u16* data = (const u16*)data_u8; /* Pointer to the data to be summed */ |
|||
size_t len = (length + 1) / 2; /* Length in 16-bit words */ |
|||
u32 sum1 = 0xffff, sum2 = 0xffff; |
|||
|
|||
while (len) |
|||
{ |
|||
size_t tlen = len > 360 ? 360 : len; |
|||
len -= tlen; |
|||
|
|||
do { |
|||
sum1 += *data++; |
|||
sum2 += sum1; |
|||
} |
|||
while (--tlen); |
|||
|
|||
sum1 = (sum1 & 0xffff) + (sum1 >> 16); |
|||
sum2 = (sum2 & 0xffff) + (sum2 >> 16); |
|||
} |
|||
|
|||
// Second reduction step to reduce sums to 16 bits
|
|||
sum1 = (sum1 & 0xffff) + (sum1 >> 16); |
|||
sum2 = (sum2 & 0xffff) + (sum2 >> 16); |
|||
return(sum2 << 16 | sum1); |
|||
} |
|||
|
|||
|
|||
// Implementation from Wikipedia
|
|||
// Slightly slower than Fletcher above, but slightly more reliable.
|
|||
#define MOD_ADLER 65521
|
|||
// data: Pointer to the data to be summed; len is in bytes
|
|||
u32 HashAdler32(const u8* data, size_t len) |
|||
{ |
|||
u32 a = 1, b = 0; |
|||
|
|||
while (len) |
|||
{ |
|||
size_t tlen = len > 5550 ? 5550 : len; |
|||
len -= tlen; |
|||
|
|||
do |
|||
{ |
|||
a += *data++; |
|||
b += a; |
|||
} |
|||
while (--tlen); |
|||
|
|||
a = (a & 0xffff) + (a >> 16) * (65536 - MOD_ADLER); |
|||
b = (b & 0xffff) + (b >> 16) * (65536 - MOD_ADLER); |
|||
} |
|||
|
|||
// It can be shown that a <= 0x1013a here, so a single subtract will do.
|
|||
if (a >= MOD_ADLER) |
|||
{ |
|||
a -= MOD_ADLER; |
|||
} |
|||
|
|||
// It can be shown that b can reach 0xfff87 here.
|
|||
b = (b & 0xffff) + (b >> 16) * (65536 - MOD_ADLER); |
|||
|
|||
if (b >= MOD_ADLER) |
|||
{ |
|||
b -= MOD_ADLER; |
|||
} |
|||
|
|||
return((b << 16) | a); |
|||
} |
|||
|
|||
// Stupid hash - but can't go back now :)
|
|||
// Don't use for new things. At least it's reasonably fast.
|
|||
u32 HashEctor(const u8* ptr, int length) |
|||
{ |
|||
u32 crc = 0; |
|||
|
|||
for (int i = 0; i < length; i++) |
|||
{ |
|||
crc ^= ptr[i]; |
|||
crc = (crc << 3) | (crc >> 29); |
|||
} |
|||
|
|||
return(crc); |
|||
} |
|||
|
|||
|
|||
#if EMU_ARCH_BITS == 64
|
|||
|
|||
//-----------------------------------------------------------------------------
|
|||
// Block read - if your platform needs to do endian-swapping or can only
|
|||
// handle aligned reads, do the conversion here
|
|||
|
|||
inline u64 getblock(const u64 * p, int i) |
|||
{ |
|||
return p[i]; |
|||
} |
|||
|
|||
//----------
|
|||
// Block mix - combine the key bits with the hash bits and scramble everything
|
|||
|
|||
inline void bmix64(u64 & h1, u64 & h2, u64 & k1, u64 & k2, u64 & c1, u64 & c2) |
|||
{ |
|||
k1 *= c1; |
|||
k1 = _rotl64(k1,23); |
|||
k1 *= c2; |
|||
h1 ^= k1; |
|||
h1 += h2; |
|||
|
|||
h2 = _rotl64(h2,41); |
|||
|
|||
k2 *= c2; |
|||
k2 = _rotl64(k2,23); |
|||
k2 *= c1; |
|||
h2 ^= k2; |
|||
h2 += h1; |
|||
|
|||
h1 = h1*3+0x52dce729; |
|||
h2 = h2*3+0x38495ab5; |
|||
|
|||
c1 = c1*5+0x7b7d159c; |
|||
c2 = c2*5+0x6bce6396; |
|||
} |
|||
|
|||
//----------
|
|||
// Finalization mix - avalanches all bits to within 0.05% bias
|
|||
|
|||
inline u64 fmix64(u64 k) |
|||
{ |
|||
k ^= k >> 33; |
|||
k *= 0xff51afd7ed558ccd; |
|||
k ^= k >> 33; |
|||
k *= 0xc4ceb9fe1a85ec53; |
|||
k ^= k >> 33; |
|||
|
|||
return k; |
|||
} |
|||
|
|||
u64 GetMurmurHash3(const u8 *src, int len, u32 samples) |
|||
{ |
|||
const u8 * data = (const u8*)src; |
|||
const int nblocks = len / 16; |
|||
u32 Step = (len / 8); |
|||
if(samples == 0) samples = std::max(Step, 1u); |
|||
Step = Step / samples; |
|||
if(Step < 1) Step = 1; |
|||
|
|||
u64 h1 = 0x9368e53c2f6af274; |
|||
u64 h2 = 0x586dcd208f7cd3fd; |
|||
|
|||
u64 c1 = 0x87c37b91114253d5; |
|||
u64 c2 = 0x4cf5ad432745937f; |
|||
|
|||
|
|||
//----------
|
|||
// body
|
|||
|
|||
const u64 * blocks = (const u64 *)(data); |
|||
|
|||
for(int i = 0; i < nblocks; i+=Step) |
|||
{ |
|||
u64 k1 = getblock(blocks,i*2+0); |
|||
u64 k2 = getblock(blocks,i*2+1); |
|||
|
|||
bmix64(h1,h2,k1,k2,c1,c2); |
|||
} |
|||
|
|||
//----------
|
|||
// tail
|
|||
|
|||
const u8 * tail = (const u8*)(data + nblocks*16); |
|||
|
|||
u64 k1 = 0; |
|||
u64 k2 = 0; |
|||
|
|||
switch(len & 15) |
|||
{ |
|||
case 15: k2 ^= u64(tail[14]) << 48; |
|||
case 14: k2 ^= u64(tail[13]) << 40; |
|||
case 13: k2 ^= u64(tail[12]) << 32; |
|||
case 12: k2 ^= u64(tail[11]) << 24; |
|||
case 11: k2 ^= u64(tail[10]) << 16; |
|||
case 10: k2 ^= u64(tail[ 9]) << 8; |
|||
case 9: k2 ^= u64(tail[ 8]) << 0; |
|||
|
|||
case 8: k1 ^= u64(tail[ 7]) << 56; |
|||
case 7: k1 ^= u64(tail[ 6]) << 48; |
|||
case 6: k1 ^= u64(tail[ 5]) << 40; |
|||
case 5: k1 ^= u64(tail[ 4]) << 32; |
|||
case 4: k1 ^= u64(tail[ 3]) << 24; |
|||
case 3: k1 ^= u64(tail[ 2]) << 16; |
|||
case 2: k1 ^= u64(tail[ 1]) << 8; |
|||
case 1: k1 ^= u64(tail[ 0]) << 0; |
|||
bmix64(h1,h2,k1,k2,c1,c2); |
|||
}; |
|||
|
|||
//----------
|
|||
// finalization
|
|||
|
|||
h2 ^= len; |
|||
|
|||
h1 += h2; |
|||
h2 += h1; |
|||
|
|||
h1 = fmix64(h1); |
|||
h2 = fmix64(h2); |
|||
|
|||
h1 += h2; |
|||
|
|||
return h1; |
|||
} |
|||
|
|||
|
|||
// CRC32 hash using the SSE4.2 instruction
|
|||
u64 GetCRC32(const u8 *src, int len, u32 samples) |
|||
{ |
|||
#if _M_SSE >= 0x402
|
|||
u64 h = len; |
|||
u32 Step = (len / 8); |
|||
const u64 *data = (const u64 *)src; |
|||
const u64 *end = data + Step; |
|||
if(samples == 0) samples = std::max(Step, 1u); |
|||
Step = Step / samples; |
|||
if(Step < 1) Step = 1; |
|||
while(data < end) |
|||
{ |
|||
h = _mm_crc32_u64(h, data[0]); |
|||
data += Step; |
|||
} |
|||
|
|||
const u8 *data2 = (const u8*)end; |
|||
return _mm_crc32_u64(h, u64(data2[0])); |
|||
#else
|
|||
return 0; |
|||
#endif
|
|||
} |
|||
|
|||
|
|||
/*
|
|||
* NOTE: This hash function is used for custom texture loading/dumping, so |
|||
* it should not be changed, which would require all custom textures to be |
|||
* recalculated for their new hash values. If the hashing function is |
|||
* changed, make sure this one is still used when the legacy parameter is |
|||
* true. |
|||
*/ |
|||
u64 GetHashHiresTexture(const u8 *src, int len, u32 samples) |
|||
{ |
|||
const u64 m = 0xc6a4a7935bd1e995; |
|||
u64 h = len * m; |
|||
const int r = 47; |
|||
u32 Step = (len / 8); |
|||
const u64 *data = (const u64 *)src; |
|||
const u64 *end = data + Step; |
|||
if(samples == 0) samples = std::max(Step, 1u); |
|||
Step = Step / samples; |
|||
if(Step < 1) Step = 1; |
|||
while(data < end) |
|||
{ |
|||
u64 k = data[0]; |
|||
data+=Step; |
|||
k *= m; |
|||
k ^= k >> r; |
|||
k *= m; |
|||
h ^= k; |
|||
h *= m; |
|||
} |
|||
|
|||
const u8 * data2 = (const u8*)end; |
|||
|
|||
switch(len & 7) |
|||
{ |
|||
case 7: h ^= u64(data2[6]) << 48; |
|||
case 6: h ^= u64(data2[5]) << 40; |
|||
case 5: h ^= u64(data2[4]) << 32; |
|||
case 4: h ^= u64(data2[3]) << 24; |
|||
case 3: h ^= u64(data2[2]) << 16; |
|||
case 2: h ^= u64(data2[1]) << 8; |
|||
case 1: h ^= u64(data2[0]); |
|||
h *= m; |
|||
}; |
|||
|
|||
h ^= h >> r; |
|||
h *= m; |
|||
h ^= h >> r; |
|||
|
|||
return h; |
|||
} |
|||
#else
|
|||
// CRC32 hash using the SSE4.2 instruction
|
|||
u64 GetCRC32(const u8 *src, int len, u32 samples) |
|||
{ |
|||
#if _M_SSE >= 0x402
|
|||
u32 h = len; |
|||
u32 Step = (len/4); |
|||
const u32 *data = (const u32 *)src; |
|||
const u32 *end = data + Step; |
|||
if(samples == 0) samples = std::max(Step, 1u); |
|||
Step = Step / samples; |
|||
if(Step < 1) Step = 1; |
|||
while(data < end) |
|||
{ |
|||
h = _mm_crc32_u32(h, data[0]); |
|||
data += Step; |
|||
} |
|||
|
|||
const u8 *data2 = (const u8*)end; |
|||
return (u64)_mm_crc32_u32(h, u32(data2[0])); |
|||
#else
|
|||
return 0; |
|||
#endif
|
|||
} |
|||
|
|||
//-----------------------------------------------------------------------------
|
|||
// Block read - if your platform needs to do endian-swapping or can only
|
|||
// handle aligned reads, do the conversion here
|
|||
|
|||
inline u32 getblock(const u32 * p, int i) |
|||
{ |
|||
return p[i]; |
|||
} |
|||
|
|||
//----------
|
|||
// Finalization mix - force all bits of a hash block to avalanche
|
|||
|
|||
// avalanches all bits to within 0.25% bias
|
|||
|
|||
inline u32 fmix32(u32 h) |
|||
{ |
|||
h ^= h >> 16; |
|||
h *= 0x85ebca6b; |
|||
h ^= h >> 13; |
|||
h *= 0xc2b2ae35; |
|||
h ^= h >> 16; |
|||
|
|||
return h; |
|||
} |
|||
|
|||
inline void bmix32(u32 & h1, u32 & h2, u32 & k1, u32 & k2, u32 & c1, u32 & c2) |
|||
{ |
|||
k1 *= c1; |
|||
k1 = _rotl(k1,11); |
|||
k1 *= c2; |
|||
h1 ^= k1; |
|||
h1 += h2; |
|||
|
|||
h2 = _rotl(h2,17); |
|||
|
|||
k2 *= c2; |
|||
k2 = _rotl(k2,11); |
|||
k2 *= c1; |
|||
h2 ^= k2; |
|||
h2 += h1; |
|||
|
|||
h1 = h1*3+0x52dce729; |
|||
h2 = h2*3+0x38495ab5; |
|||
|
|||
c1 = c1*5+0x7b7d159c; |
|||
c2 = c2*5+0x6bce6396; |
|||
} |
|||
|
|||
//----------
|
|||
|
|||
u64 GetMurmurHash3(const u8* src, int len, u32 samples) |
|||
{ |
|||
const u8 * data = (const u8*)src; |
|||
u32 out[2]; |
|||
const int nblocks = len / 8; |
|||
u32 Step = (len / 4); |
|||
if(samples == 0) samples = std::max(Step, 1u); |
|||
Step = Step / samples; |
|||
if(Step < 1) Step = 1; |
|||
|
|||
u32 h1 = 0x8de1c3ac; |
|||
u32 h2 = 0xbab98226; |
|||
|
|||
u32 c1 = 0x95543787; |
|||
u32 c2 = 0x2ad7eb25; |
|||
|
|||
//----------
|
|||
// body
|
|||
|
|||
const u32 * blocks = (const u32 *)(data + nblocks*8); |
|||
|
|||
for(int i = -nblocks; i < 0; i+=Step) |
|||
{ |
|||
u32 k1 = getblock(blocks,i*2+0); |
|||
u32 k2 = getblock(blocks,i*2+1); |
|||
|
|||
bmix32(h1,h2,k1,k2,c1,c2); |
|||
} |
|||
|
|||
//----------
|
|||
// tail
|
|||
|
|||
const u8 * tail = (const u8*)(data + nblocks*8); |
|||
|
|||
u32 k1 = 0; |
|||
u32 k2 = 0; |
|||
|
|||
switch(len & 7) |
|||
{ |
|||
case 7: k2 ^= tail[6] << 16; |
|||
case 6: k2 ^= tail[5] << 8; |
|||
case 5: k2 ^= tail[4] << 0; |
|||
case 4: k1 ^= tail[3] << 24; |
|||
case 3: k1 ^= tail[2] << 16; |
|||
case 2: k1 ^= tail[1] << 8; |
|||
case 1: k1 ^= tail[0] << 0; |
|||
bmix32(h1,h2,k1,k2,c1,c2); |
|||
}; |
|||
|
|||
//----------
|
|||
// finalization
|
|||
|
|||
h2 ^= len; |
|||
|
|||
h1 += h2; |
|||
h2 += h1; |
|||
|
|||
h1 = fmix32(h1); |
|||
h2 = fmix32(h2); |
|||
|
|||
h1 += h2; |
|||
h2 += h1; |
|||
|
|||
out[0] = h1; |
|||
out[1] = h2; |
|||
|
|||
return *((u64 *)&out); |
|||
} |
|||
|
|||
/*
|
|||
* FIXME: The old 32-bit version of this hash made different hashes than the |
|||
* 64-bit version. Until someone can make a new version of the 32-bit one that |
|||
* makes identical hashes, this is just a c/p of the 64-bit one. |
|||
*/ |
|||
u64 GetHashHiresTexture(const u8 *src, int len, u32 samples) |
|||
{ |
|||
const u64 m = 0xc6a4a7935bd1e995ULL; |
|||
u64 h = len * m; |
|||
const int r = 47; |
|||
u32 Step = (len / 8); |
|||
const u64 *data = (const u64 *)src; |
|||
const u64 *end = data + Step; |
|||
if(samples == 0) samples = std::max(Step, 1u); |
|||
Step = Step / samples; |
|||
if(Step < 1) Step = 1; |
|||
while(data < end) |
|||
{ |
|||
u64 k = data[0]; |
|||
data+=Step; |
|||
k *= m; |
|||
k ^= k >> r; |
|||
k *= m; |
|||
h ^= k; |
|||
h *= m; |
|||
} |
|||
|
|||
const u8 * data2 = (const u8*)end; |
|||
|
|||
switch(len & 7) |
|||
{ |
|||
case 7: h ^= u64(data2[6]) << 48; |
|||
case 6: h ^= u64(data2[5]) << 40; |
|||
case 5: h ^= u64(data2[4]) << 32; |
|||
case 4: h ^= u64(data2[3]) << 24; |
|||
case 3: h ^= u64(data2[2]) << 16; |
|||
case 2: h ^= u64(data2[1]) << 8; |
|||
case 1: h ^= u64(data2[0]); |
|||
h *= m; |
|||
}; |
|||
|
|||
h ^= h >> r; |
|||
h *= m; |
|||
h ^= h >> r; |
|||
|
|||
return h; |
|||
} |
|||
#endif
|
|||
|
|||
u64 GetHash64(const u8 *src, int len, u32 samples) |
|||
{ |
|||
return ptrHashFunction(src, len, samples); |
|||
} |
|||
|
|||
// sets the hash function used for the texture cache
|
|||
void SetHash64Function(bool useHiresTextures) |
|||
{ |
|||
if (useHiresTextures) |
|||
{ |
|||
ptrHashFunction = &GetHashHiresTexture; |
|||
} |
|||
#if _M_SSE >= 0x402
|
|||
else if (cpu_info.bSSE4_2 && !useHiresTextures) // sse crc32 version
|
|||
{ |
|||
ptrHashFunction = &GetCRC32; |
|||
} |
|||
#endif
|
|||
else |
|||
{ |
|||
ptrHashFunction = &GetMurmurHash3; |
|||
} |
|||
} |
|||
|
|||
|
|||
|
|||
@ -1,17 +0,0 @@ |
|||
// Copyright 2013 Dolphin Emulator Project / 2014 Citra Emulator Project |
|||
// Licensed under GPLv2 or any later version |
|||
// Refer to the license.txt file included. |
|||
|
|||
#pragma once |
|||
|
|||
#include "common/common_types.h" |
|||
|
|||
u32 HashFletcher(const u8* data_u8, size_t length); // FAST. Length & 1 == 0. |
|||
u32 HashAdler32(const u8* data, size_t len); // Fairly accurate, slightly slower |
|||
u32 HashFNV(const u8* ptr, int length); // Another fast and decent hash |
|||
u32 HashEctor(const u8* ptr, int length); // JUNK. DO NOT USE FOR NEW THINGS |
|||
u64 GetCRC32(const u8 *src, int len, u32 samples); // SSE4.2 version of CRC32 |
|||
u64 GetHashHiresTexture(const u8 *src, int len, u32 samples); |
|||
u64 GetMurmurHash3(const u8 *src, int len, u32 samples); |
|||
u64 GetHash64(const u8 *src, int len, u32 samples); |
|||
void SetHash64Function(bool useHiresTextures); |
|||
@ -1,394 +0,0 @@ |
|||
// Copyright (C) 2003 Dolphin Project.
|
|||
|
|||
// This program is free software: you can redistribute it and/or modify
|
|||
// it under the terms of the GNU General Public License as published by
|
|||
// the Free Software Foundation, version 2.0 or later versions.
|
|||
|
|||
// This program 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 General Public License 2.0 for more details.
|
|||
|
|||
// A copy of the GPL 2.0 should have been included with the program.
|
|||
// If not, see http://www.gnu.org/licenses/
|
|||
|
|||
// Official SVN repository and contact information can be found at
|
|||
// http://code.google.com/p/dolphin-emu/
|
|||
|
|||
#include <string>
|
|||
|
|||
#include "common/logging/log.h"
|
|||
#include "common/mem_arena.h"
|
|||
#include "common/memory_util.h"
|
|||
#include "common/platform.h"
|
|||
#include "common/string_util.h"
|
|||
|
|||
#ifndef _WIN32
|
|||
#include <fcntl.h>
|
|||
#include <string.h>
|
|||
#include <unistd.h>
|
|||
|
|||
#ifdef ANDROID
|
|||
#include <sys/ioctl.h>
|
|||
#include <linux/ashmem.h>
|
|||
#endif
|
|||
#endif
|
|||
|
|||
#ifdef ANDROID
|
|||
|
|||
// Hopefully this ABI will never change...
|
|||
|
|||
|
|||
#define ASHMEM_DEVICE "/dev/ashmem"
|
|||
|
|||
/*
|
|||
* ashmem_create_region - creates a new ashmem region and returns the file |
|||
* descriptor, or <0 on error |
|||
* |
|||
* `name' is an optional label to give the region (visible in /proc/pid/maps) |
|||
* `size' is the size of the region, in page-aligned bytes |
|||
*/ |
|||
int ashmem_create_region(const char *name, size_t size) |
|||
{ |
|||
int fd, ret; |
|||
|
|||
fd = open(ASHMEM_DEVICE, O_RDWR); |
|||
if (fd < 0) |
|||
return fd; |
|||
|
|||
if (name) { |
|||
char buf[ASHMEM_NAME_LEN]; |
|||
|
|||
strncpy(buf, name, sizeof(buf)); |
|||
ret = ioctl(fd, ASHMEM_SET_NAME, buf); |
|||
if (ret < 0) |
|||
goto error; |
|||
} |
|||
|
|||
ret = ioctl(fd, ASHMEM_SET_SIZE, size); |
|||
if (ret < 0) |
|||
goto error; |
|||
|
|||
return fd; |
|||
|
|||
error: |
|||
LOG_ERROR(Common_Memory, "NASTY ASHMEM ERROR: ret = %08x", ret); |
|||
close(fd); |
|||
return ret; |
|||
} |
|||
|
|||
int ashmem_set_prot_region(int fd, int prot) |
|||
{ |
|||
return ioctl(fd, ASHMEM_SET_PROT_MASK, prot); |
|||
} |
|||
|
|||
int ashmem_pin_region(int fd, size_t offset, size_t len) |
|||
{ |
|||
struct ashmem_pin pin = { offset, len }; |
|||
return ioctl(fd, ASHMEM_PIN, &pin); |
|||
} |
|||
|
|||
int ashmem_unpin_region(int fd, size_t offset, size_t len) |
|||
{ |
|||
struct ashmem_pin pin = { offset, len }; |
|||
return ioctl(fd, ASHMEM_UNPIN, &pin); |
|||
} |
|||
#endif // Android
|
|||
|
|||
|
|||
#if defined(_WIN32)
|
|||
SYSTEM_INFO sysInfo; |
|||
#endif
|
|||
|
|||
|
|||
// Windows mappings need to be on 64K boundaries, due to Alpha legacy.
|
|||
#ifdef _WIN32
|
|||
size_t roundup(size_t x) { |
|||
int gran = sysInfo.dwAllocationGranularity ? sysInfo.dwAllocationGranularity : 0x10000; |
|||
return (x + gran - 1) & ~(gran - 1); |
|||
} |
|||
#else
|
|||
size_t roundup(size_t x) { |
|||
return x; |
|||
} |
|||
#endif
|
|||
|
|||
|
|||
void MemArena::GrabLowMemSpace(size_t size) |
|||
{ |
|||
#ifdef _WIN32
|
|||
hMemoryMapping = CreateFileMapping(INVALID_HANDLE_VALUE, nullptr, PAGE_READWRITE, 0, (DWORD)(size), nullptr); |
|||
GetSystemInfo(&sysInfo); |
|||
#elif defined(ANDROID)
|
|||
// Use ashmem so we don't have to allocate a file on disk!
|
|||
fd = ashmem_create_region("Citra_RAM", size); |
|||
// Note that it appears that ashmem is pinned by default, so no need to pin.
|
|||
if (fd < 0) |
|||
{ |
|||
LOG_ERROR(Common_Memory, "Failed to grab ashmem space of size: %08x errno: %d", (int)size, (int)(errno)); |
|||
return; |
|||
} |
|||
#else
|
|||
// Try to find a non-existing filename for our shared memory.
|
|||
// In most cases the first one will be available, but it's nicer to search
|
|||
// a bit more.
|
|||
for (int i = 0; i < 10000; i++) |
|||
{ |
|||
std::string file_name = Common::StringFromFormat("/citramem.%d", i); |
|||
fd = shm_open(file_name.c_str(), O_RDWR | O_CREAT | O_EXCL, 0600); |
|||
if (fd != -1) |
|||
{ |
|||
shm_unlink(file_name.c_str()); |
|||
break; |
|||
} |
|||
else if (errno != EEXIST) |
|||
{ |
|||
LOG_ERROR(Common_Memory, "shm_open failed: %s", strerror(errno)); |
|||
return; |
|||
} |
|||
} |
|||
if (ftruncate(fd, size) < 0) |
|||
LOG_ERROR(Common_Memory, "Failed to allocate low memory space"); |
|||
#endif
|
|||
} |
|||
|
|||
|
|||
void MemArena::ReleaseSpace() |
|||
{ |
|||
#ifdef _WIN32
|
|||
CloseHandle(hMemoryMapping); |
|||
hMemoryMapping = 0; |
|||
#else
|
|||
close(fd); |
|||
#endif
|
|||
} |
|||
|
|||
|
|||
void *MemArena::CreateView(s64 offset, size_t size, void *base) |
|||
{ |
|||
#ifdef _WIN32
|
|||
size = roundup(size); |
|||
void *ptr = MapViewOfFileEx(hMemoryMapping, FILE_MAP_ALL_ACCESS, 0, (DWORD)((u64)offset), size, base); |
|||
return ptr; |
|||
#else
|
|||
void *retval = mmap(base, size, PROT_READ | PROT_WRITE, MAP_SHARED | |
|||
// Do not sync memory to underlying file. Linux has this by default.
|
|||
#ifdef __FreeBSD__
|
|||
MAP_NOSYNC | |
|||
#endif
|
|||
((base == nullptr) ? 0 : MAP_FIXED), fd, offset); |
|||
|
|||
if (retval == MAP_FAILED) |
|||
{ |
|||
LOG_ERROR(Common_Memory, "mmap failed"); |
|||
return nullptr; |
|||
} |
|||
return retval; |
|||
#endif
|
|||
} |
|||
|
|||
|
|||
void MemArena::ReleaseView(void* view, size_t size) |
|||
{ |
|||
#ifdef _WIN32
|
|||
UnmapViewOfFile(view); |
|||
#else
|
|||
munmap(view, size); |
|||
#endif
|
|||
} |
|||
|
|||
u8* MemArena::Find4GBBase() |
|||
{ |
|||
#if EMU_ARCH_BITS == 64
|
|||
#ifdef _WIN32
|
|||
// 64 bit
|
|||
u8* base = (u8*)VirtualAlloc(0, 0xE1000000, MEM_RESERVE, PAGE_READWRITE); |
|||
VirtualFree(base, 0, MEM_RELEASE); |
|||
return base; |
|||
#else
|
|||
// Very precarious - mmap cannot return an error when trying to map already used pages.
|
|||
// This makes the Windows approach above unusable on Linux, so we will simply pray...
|
|||
return reinterpret_cast<u8*>(0x2300000000ULL); |
|||
#endif
|
|||
|
|||
#else // 32 bit
|
|||
|
|||
#ifdef _WIN32
|
|||
u8* base = (u8*)VirtualAlloc(0, 0x10000000, MEM_RESERVE, PAGE_READWRITE); |
|||
if (base) { |
|||
VirtualFree(base, 0, MEM_RELEASE); |
|||
} |
|||
return base; |
|||
#else
|
|||
void* base = mmap(0, 0x10000000, PROT_READ | PROT_WRITE, |
|||
MAP_ANON | MAP_SHARED, -1, 0); |
|||
if (base == MAP_FAILED) { |
|||
LOG_ERROR(Common_Memory, "Failed to map 256 MB of memory space: %s", strerror(errno)); |
|||
return 0; |
|||
} |
|||
munmap(base, 0x10000000); |
|||
return static_cast<u8*>(base); |
|||
#endif
|
|||
#endif
|
|||
} |
|||
|
|||
|
|||
// yeah, this could also be done in like two bitwise ops...
|
|||
#define SKIP(a_flags, b_flags)
|
|||
//if (!(a_flags & MV_WII_ONLY) && (b_flags & MV_WII_ONLY))
|
|||
// continue;
|
|||
//if (!(a_flags & MV_FAKE_VMEM) && (b_flags & MV_FAKE_VMEM))
|
|||
// continue;
|
|||
|
|||
static bool Memory_TryBase(u8 *base, const MemoryView *views, int num_views, u32 flags, MemArena *arena) { |
|||
// OK, we know where to find free space. Now grab it!
|
|||
// We just mimic the popular BAT setup.
|
|||
size_t position = 0; |
|||
size_t last_position = 0; |
|||
|
|||
// Zero all the pointers to be sure.
|
|||
for (int i = 0; i < num_views; i++) |
|||
{ |
|||
if (views[i].out_ptr_low) |
|||
*views[i].out_ptr_low = 0; |
|||
if (views[i].out_ptr) |
|||
*views[i].out_ptr = 0; |
|||
} |
|||
|
|||
int i; |
|||
for (i = 0; i < num_views; i++) |
|||
{ |
|||
const MemoryView &view = views[i]; |
|||
if (view.size == 0) |
|||
continue; |
|||
SKIP(flags, view.flags); |
|||
if (view.flags & MV_MIRROR_PREVIOUS) { |
|||
position = last_position; |
|||
} |
|||
else { |
|||
*(view.out_ptr_low) = (u8*)arena->CreateView(position, view.size); |
|||
if (!*view.out_ptr_low) |
|||
goto bail; |
|||
} |
|||
#if EMU_ARCH_BITS == 64
|
|||
*view.out_ptr = (u8*)arena->CreateView( |
|||
position, view.size, base + view.virtual_address); |
|||
#else
|
|||
if (view.flags & MV_MIRROR_PREVIOUS) { // TODO: should check if the two & 0x3FFFFFFF are identical.
|
|||
// No need to create multiple identical views.
|
|||
*view.out_ptr = *views[i - 1].out_ptr; |
|||
} |
|||
else { |
|||
*view.out_ptr = (u8*)arena->CreateView( |
|||
position, view.size, base + (view.virtual_address & 0x3FFFFFFF)); |
|||
if (!*view.out_ptr) |
|||
goto bail; |
|||
} |
|||
#endif
|
|||
|
|||
last_position = position; |
|||
position += roundup(view.size); |
|||
} |
|||
|
|||
return true; |
|||
|
|||
bail: |
|||
// Argh! ERROR! Free what we grabbed so far so we can try again.
|
|||
for (int j = 0; j <= i; j++) |
|||
{ |
|||
if (views[i].size == 0) |
|||
continue; |
|||
SKIP(flags, views[i].flags); |
|||
if (views[j].out_ptr_low && *views[j].out_ptr_low) |
|||
{ |
|||
arena->ReleaseView(*views[j].out_ptr_low, views[j].size); |
|||
*views[j].out_ptr_low = nullptr; |
|||
} |
|||
if (*views[j].out_ptr) |
|||
{ |
|||
#if EMU_ARCH_BITS == 64
|
|||
arena->ReleaseView(*views[j].out_ptr, views[j].size); |
|||
#else
|
|||
if (!(views[j].flags & MV_MIRROR_PREVIOUS)) |
|||
{ |
|||
arena->ReleaseView(*views[j].out_ptr, views[j].size); |
|||
} |
|||
#endif
|
|||
*views[j].out_ptr = nullptr; |
|||
} |
|||
} |
|||
return false; |
|||
} |
|||
|
|||
u8 *MemoryMap_Setup(const MemoryView *views, int num_views, u32 flags, MemArena *arena) |
|||
{ |
|||
size_t total_mem = 0; |
|||
int base_attempts = 0; |
|||
|
|||
for (int i = 0; i < num_views; i++) |
|||
{ |
|||
if (views[i].size == 0) |
|||
continue; |
|||
SKIP(flags, views[i].flags); |
|||
if ((views[i].flags & MV_MIRROR_PREVIOUS) == 0) |
|||
total_mem += roundup(views[i].size); |
|||
} |
|||
// Grab some pagefile backed memory out of the void ...
|
|||
arena->GrabLowMemSpace(total_mem); |
|||
|
|||
// Now, create views in high memory where there's plenty of space.
|
|||
#if EMU_ARCH_BITS == 64
|
|||
u8 *base = MemArena::Find4GBBase(); |
|||
// This really shouldn't fail - in 64-bit, there will always be enough
|
|||
// address space.
|
|||
if (!Memory_TryBase(base, views, num_views, flags, arena)) |
|||
{ |
|||
LOG_ERROR(Common_Memory, "MemoryMap_Setup: Failed finding a memory base."); |
|||
return 0; |
|||
} |
|||
#elif defined(_WIN32)
|
|||
// Try a whole range of possible bases. Return once we got a valid one.
|
|||
u32 max_base_addr = 0x7FFF0000 - 0x10000000; |
|||
u8 *base = nullptr; |
|||
|
|||
for (u32 base_addr = 0x01000000; base_addr < max_base_addr; base_addr += 0x400000) |
|||
{ |
|||
base_attempts++; |
|||
base = (u8 *)base_addr; |
|||
if (Memory_TryBase(base, views, num_views, flags, arena)) |
|||
{ |
|||
LOG_DEBUG(Common_Memory, "Found valid memory base at %p after %i tries.", base, base_attempts); |
|||
base_attempts = 0; |
|||
break; |
|||
} |
|||
} |
|||
#else
|
|||
// Linux32 is fine with the x64 method, although limited to 32-bit with no automirrors.
|
|||
u8 *base = MemArena::Find4GBBase(); |
|||
if (!Memory_TryBase(base, views, num_views, flags, arena)) |
|||
{ |
|||
LOG_ERROR(Common_Memory, "MemoryMap_Setup: Failed finding a memory base."); |
|||
return 0; |
|||
} |
|||
#endif
|
|||
if (base_attempts) |
|||
LOG_ERROR(Common_Memory, "No possible memory base pointer found!"); |
|||
return base; |
|||
} |
|||
|
|||
void MemoryMap_Shutdown(const MemoryView *views, int num_views, u32 flags, MemArena *arena) |
|||
{ |
|||
for (int i = 0; i < num_views; i++) |
|||
{ |
|||
if (views[i].size == 0) |
|||
continue; |
|||
SKIP(flags, views[i].flags); |
|||
if (views[i].out_ptr_low && *views[i].out_ptr_low) |
|||
arena->ReleaseView(*views[i].out_ptr_low, views[i].size); |
|||
if (*views[i].out_ptr && (views[i].out_ptr_low && *views[i].out_ptr != *views[i].out_ptr_low)) |
|||
arena->ReleaseView(*views[i].out_ptr, views[i].size); |
|||
*views[i].out_ptr = nullptr; |
|||
if (views[i].out_ptr_low) |
|||
*views[i].out_ptr_low = nullptr; |
|||
} |
|||
} |
|||
@ -1,70 +0,0 @@ |
|||
// Copyright (C) 2003 Dolphin Project. |
|||
|
|||
// This program is free software: you can redistribute it and/or modify |
|||
// it under the terms of the GNU General Public License as published by |
|||
// the Free Software Foundation, version 2.0 or later versions. |
|||
|
|||
// This program 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 General Public License 2.0 for more details. |
|||
|
|||
// A copy of the GPL 2.0 should have been included with the program. |
|||
// If not, see http://www.gnu.org/licenses/ |
|||
|
|||
// Official SVN repository and contact information can be found at |
|||
// http://code.google.com/p/dolphin-emu/ |
|||
|
|||
#pragma once |
|||
|
|||
#ifdef _WIN32 |
|||
#include <windows.h> |
|||
#endif |
|||
|
|||
#include "common/common_types.h" |
|||
|
|||
// This class lets you create a block of anonymous RAM, and then arbitrarily map views into it. |
|||
// Multiple views can mirror the same section of the block, which makes it very convient for emulating |
|||
// memory mirrors. |
|||
|
|||
class MemArena |
|||
{ |
|||
public: |
|||
void GrabLowMemSpace(size_t size); |
|||
void ReleaseSpace(); |
|||
void *CreateView(s64 offset, size_t size, void *base = 0); |
|||
void ReleaseView(void *view, size_t size); |
|||
|
|||
// This only finds 1 GB in 32-bit |
|||
static u8 *Find4GBBase(); |
|||
private: |
|||
|
|||
#ifdef _WIN32 |
|||
HANDLE hMemoryMapping; |
|||
#else |
|||
int fd; |
|||
#endif |
|||
}; |
|||
|
|||
enum { |
|||
MV_MIRROR_PREVIOUS = 1, |
|||
// MV_FAKE_VMEM = 2, |
|||
// MV_WII_ONLY = 4, |
|||
MV_IS_PRIMARY_RAM = 0x100, |
|||
MV_IS_EXTRA1_RAM = 0x200, |
|||
MV_IS_EXTRA2_RAM = 0x400, |
|||
}; |
|||
|
|||
struct MemoryView |
|||
{ |
|||
u8 **out_ptr_low; |
|||
u8 **out_ptr; |
|||
u32 virtual_address; |
|||
u32 size; |
|||
u32 flags; |
|||
}; |
|||
|
|||
// Uses a memory arena to set up an emulator-friendly memory map according to |
|||
// a passed-in list of MemoryView structures. |
|||
u8 *MemoryMap_Setup(const MemoryView *views, int num_views, u32 flags, MemArena *arena); |
|||
void MemoryMap_Shutdown(const MemoryView *views, int num_views, u32 flags, MemArena *arena); |
|||
Write
Preview
Loading…
Cancel
Save
Reference in new issue