Skip to content
Permalink
master
Switch branches/tags
Go to file
 
 
Cannot retrieve contributors at this time
/// Fast Memory Manager for FPC x86_64
// - this unit is a part of the freeware Synopse mORMot framework
// licensed under a MPL/GPL/LGPL three license - see LICENSE.md
unit SynFPCx64MM;
{
*****************************************************************************
A Multi-thread Friendly Memory Manager for FPC written in x86_64 assembly
- targetting Linux (and Windows) multi-threaded Services
- only for FPC on the x86_64 target - use the RTL MM on Delphi or ARM
- based on proven FastMM4 by Pierre le Riche - with tuning and enhancements
- can report detailed statistics (with threads contention and memory leaks)
- three app modes: default mono-thread friendly, FPCMM_SERVER or FPCMM_BOOST
Usage: include this unit as the very first in your FPC project uses clause
Why another Memory Manager on FPC?
- The built-in heap.inc is well written and cross-platform and cross-CPU,
but its threadvar arena for small blocks tends to consume a lot of memory
on multi-threaded servers, and has suboptimal allocation performance
- C memory managers (glibc, Intel TBB, jemalloc) have a very high RAM
consumption (especially Intel TBB) and do panic/SIGKILL on any GPF
- Pascal alternatives (FastMM4,ScaleMM2,BrainMM) are Windows+Delphi specific
- Our lockess round-robin of tiny blocks and freemem bin list are unique
algorithms among Memory Managers, and match modern CPUs and workloads
- It was so fun diving into SSE2 x86_64 assembly and Pierre's insight
- Resulting code is still easy to understand and maintain
DISCLAMER: seems stable on Linux and Win64 but feedback is welcome!
*****************************************************************************
This file is part of Synopse framework.
Synopse framework. Copyright (C) 2022 Arnaud Bouchez
Synopse Informatique - https://synopse.info
*** BEGIN LICENSE BLOCK *****
Version: MPL 1.1/GPL 2.0/LGPL 2.1
The contents of this file are subject to the Mozilla Public License Version
1.1 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.mozilla.org/MPL
Software distributed under the License is distributed on an "AS IS" basis,
WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
for the specific language governing rights and limitations under the License.
The Original Code is Synopse mORMot framework.
The Initial Developer of the Original Code is Arnaud Bouchez.
Portions created by the Initial Developer are Copyright (C) 2022
the Initial Developer. All Rights Reserved.
Contributor(s):
Alternatively, the contents of this file may be used under the terms of
either the GNU General Public License Version 2 or later (the "GPL"), or
the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
in which case the provisions of the GPL or the LGPL are applicable instead
of those above. If you wish to allow use of your version of this file only
under the terms of either the GPL or the LGPL, and not to allow others to
use your version of this file under the terms of the MPL, indicate your
decision by deleting the provisions above and replace them with the notice
and other provisions required by the GPL or the LGPL. If you do not delete
the provisions above, a recipient may use your version of this file under
the terms of any one of the MPL, the GPL or the LGPL.
***** END LICENSE BLOCK *****
}
{ ---- Ready-To-Use Scenarios for Memory Manager Tuning }
{
TL;DR:
1. default settings target LCL/console mono-threaded apps;
2. define FPCMM_SERVER for a multi-threaded service/daemon.
}
// target a multi-threaded service on a modern CPU
// - define FPCMM_DEBUG, FPCMM_ASSUMEMULTITHREAD, FPCMM_ERMS, FPCMM_LOCKLESSFREE
// - currently mormot2tests run with no sleep when FPCMM_SERVER is set :)
// - you may try to define FPCMM_BOOST for even more aggressive settings.
{.$define FPCMM_SERVER}
// increase settings for very aggressive multi-threaded process
// - try to enable it if unexpected SmallGetmemSleepCount/SmallFreememSleepCount
// and SleepCount/SleepCycles contentions are reported by CurrentHeapStatus;
// - tiny blocks will be <= 256 bytes (instead of 128 bytes);
// - FPCMM_BOOSTER will use 2x more tiny blocks arenas - likely to be wasteful;
// - will enable FPCMM_SMALLNOTWITHMEDIUM trying to reduce medium sleeps;
// - warning: depending on the workload and hardware, it may actually be slower,
// triggering more Meidum arena contention, and consuming more RAM: consider
// FPCMM_SERVER as a fair alternative.
{.$define FPCMM_BOOST}
{.$define FPCMM_BOOSTER}
{ ---- Fine Grained Memory Manager Tuning }
// includes more detailed information to WriteHeapStatus()
{.$define FPCMM_DEBUG}
// on thread contention, don't spin executing "pause" but directly call Sleep()
// - may help on a single core CPU, or for very specific workloads
{.$define FPCMM_NOPAUSE}
// let FPCMM_DEBUG include SleepCycles information from rdtsc
// and FPCMM_PAUSE call rdtsc for its spinnning loop
// - since rdtsc is emulated so unrealiable on VM, it is disabled by default
{.$define FPCMM_SLEEPTSC}
// checks leaks and write them to the console at process shutdown
// - only basic information will be included: more debugging information (e.g.
// call stack) may be gathered using heaptrc or valgrid
{.$define FPCMM_REPORTMEMORYLEAKS}
// won't check the IsMultiThread global, but assume it is true
// - multi-threaded apps (e.g. a Server Daemon instance) will be faster with it
// - mono-threaded (console/LCL) apps are faster without this conditional
{.$define FPCMM_ASSUMEMULTITHREAD}
// let Freemem multi-thread contention use a lockless algorithm
// - on contention, Freemem won't yield the thread using an OS call, but fill
// an internal Bin list which will be released when the lock becomes available
// - beneficial from our tests on high thread contention (HTTP/REST server)
{.$define FPCMM_LOCKLESSFREE}
// won't use mremap but a regular getmem/move/freemem pattern for large blocks
// - depending on the actual system (e.g. on a VM), mremap may be slower
// - will disable Linux mremap() or Windows following block VirtualQuery/Alloc
{.$define FPCMM_NOMREMAP}
// force the tiny/small blocks to be in their own arena, not with medium blocks
// - would use a little more memory, but medium pool is less likely to sleep
// - not defined for FPCMM_SERVER because no performance difference was found
{.$define FPCMM_SMALLNOTWITHMEDIUM}
// use "rep movsb/stosd" ERMS for blocks > 256 bytes instead of SSE2 "movaps"
// - ERMS is available since Ivy Bridge, and we use "movaps" for smallest blocks
// (to not slow down older CPUs), so it is safe to enable this on FPCMM_SERVER
{.$define FPCMM_ERMS}
// try "cmp" before "lock cmpxchg" for old processors with huge lock penalty
{.$define FPCMM_CMPBEFORELOCK}
// will export libc-like functions, and not replace the FPC MM
// - e.g. to use this unit as a stand-alone C memory allocator
{.$define FPCMM_STANDALONE}
// this whole unit will compile as void
// - may be defined e.g. when compiled as Design-Time Lazarus package
{.$define FPCMM_DISABLE}
interface
{$ifdef FPC}
// cut-down version of Synopse.inc to make this unit standalone
{$mode Delphi}
{$inline on}
{$R-} // disable Range checking
{$S-} // disable Stack checking
{$W-} // disable stack frame generation
{$Q-} // disable overflow checking
{$B-} // expect short circuit boolean
{$ifdef CPUX64}
{$define FPCX64MM} // this unit is for FPC + x86_64 only
{$asmmode Intel}
{$endif CPUX64}
{$ifdef FPCMM_BOOSTER}
{$define FPCMM_BOOST}
{$endif FPCMM_BOOSTER}
{$ifdef FPCMM_BOOST}
{$define FPCMM_SERVER}
{$define FPCMM_SMALLNOTWITHMEDIUM}
{$endif FPCMM_BOOST}
{$ifdef FPCMM_SERVER}
{$define FPCMM_DEBUG}
{$define FPCMM_ASSUMEMULTITHREAD}
{$define FPCMM_LOCKLESSFREE}
{$define FPCMM_ERMS}
{$endif FPCMM_SERVER}
{$ifdef FPCMM_BOOSTER}
{$undef FPCMM_DEBUG} // when performance matters more than stats
{$endif FPCMM_BOOSTER}
{$endif FPC}
{$ifdef FPCMM_DISABLE}
{$undef FPCX64MM} // e.g. when compiled as Design-Time Lazarus package
{$endif FPCMM_DISABLE}
{$ifdef FPCX64MM}
// this unit is available only for FPC + X86_64 CPU
// other targets would compile as a void unit
type
/// Arena (middle/large) heap information as returned by CurrentHeapStatus
TMMStatusArena = record
/// how many bytes are currently reserved (mmap) to the Operating System
CurrentBytes: PtrUInt;
/// how many bytes have been reserved (mmap) to the Operating System
CumulativeBytes: PtrUInt;
{$ifdef FPCMM_DEBUG}
/// maximum bytes count reserved (mmap) to the Operating System
PeakBytes: PtrUInt;
/// how many VirtualAlloc/mmap calls to the Operating System did occur
CumulativeAlloc: PtrUInt;
/// how many VirtualFree/munmap calls to the Operating System did occur
CumulativeFree: PtrUInt;
{$endif FPCMM_DEBUG}
/// how many times this Arena did wait from been unlocked by another thread
SleepCount: PtrUInt;
end;
/// heap information as returned by CurrentHeapStatus
TMMStatus = record
/// how many tiny/small memory blocks (<=2600 bytes) are currently allocated
SmallBlocks: PtrUInt;
/// how many bytes of tiny/small memory blocks are currently allocated
// - this size is included in Medium.CurrentBytes value, even if
// FPCMM_SMALLNOTWITHMEDIUM has been defined
SmallBlocksSize: PtrUInt;
/// information about blocks up to 256KB (tiny, small and medium)
// - includes also the memory needed for tiny/small blocks
Medium: TMMStatusArena;
/// information about large blocks > 256KB
// - those blocks are directly handled by the Operating System
Large: TMMStatusArena;
{$ifdef FPCMM_DEBUG}
{$ifdef FPCMM_SLEEPTSC}
/// how much rdtsc cycles were spent within SwitchToThread/NanoSleep API
// - we rdtsc since it is an indicative but very fast way of timing on
// direct hardware
// - warning: on virtual machines, the rdtsc opcode is usually emulated so
// these SleepCycles number are non indicative anymore
SleepCycles: PtrUInt;
{$endif FPCMM_SLEEPTSC}
{$ifdef FPCMM_LOCKLESSFREE}
/// how many times Freemem() did spin to acquire its lock-less bin list
SmallFreememLockLessSpin: PtrUInt;
{$endif FPCMM_LOCKLESSFREE}
{$endif FPCMM_DEBUG}
/// how many times the Operating System Sleep/NanoSleep API was called
// - in a perfect world, should be as small as possible: mormot2tests run as
// SleepCount=0 when FPCMM_LOCKLESSFREE is defined (FPCMM_SERVER default)
SleepCount: PtrUInt;
/// how many times Getmem() did block and wait for a tiny/small block
// - see also GetSmallBlockContention() for more detailed information
SmallGetmemSleepCount: PtrUInt;
/// how many times Freemem() did block and wait for a tiny/small block
// - see also GetSmallBlockContention() for more detailed information
SmallFreememSleepCount: PtrUInt;
end;
PMMStatus = ^TMMStatus;
/// allocate a new memory buffer
// - as FPC default heap, _Getmem(0) returns _Getmem(1)
function _GetMem(size: PtrUInt): pointer;
/// allocate a new zeroed memory buffer
function _AllocMem(Size: PtrUInt): pointer;
/// release a memory buffer
// - returns the allocated size of the supplied pointer (as FPC default heap)
function _FreeMem(P: pointer): PtrUInt;
/// change the size of a memory buffer
// - won't move any data if in-place reallocation is possible
// - as FPC default heap, _ReallocMem(P=nil,Size) maps P := _getmem(Size) and
// _ReallocMem(P,0) maps _Freemem(P)
function _ReallocMem(var P: pointer; Size: PtrUInt): pointer;
/// retrieve the maximum size (i.e. the allocated size) of a memory buffer
function _MemSize(P: pointer): PtrUInt; inline;
/// retrieve high-level statistics about the current memory manager state
// - see also GetSmallBlockContention for detailed small blocks information
// - standard GetHeapStatus and GetFPCHeapStatus gives less accurate information
// (only CurrHeapSize and MaxHeapSize are set), since we don't track "free" heap
// bytes: I can't figure how "free" memory is relevant nowadays - on 21th century
// Operating Systems, memory is virtual, and reserved/mapped by the OS but
// physically hosted in the HW RAM chips only when written the first time -
// GetHeapStatus information made sense on MSDOS with fixed 640KB of RAM
// - note that FPC GetHeapStatus and GetFPCHeapStatus is only about the
// current thread (irrelevant for sure) whereas CurrentHeapStatus is global
function CurrentHeapStatus: TMMStatus;
{$ifdef FPCMM_STANDALONE}
/// should be called before using any memory function
procedure InitializeMemoryManager;
/// should be called to finalize this memory manager process and release all RAM
procedure FreeAllMemory;
{$undef FPCMM_DEBUG} // excluded FPC-specific debugging
/// IsMultiThread global variable is not correct outside of the FPC RTL
{$define FPCMM_ASSUMEMULTITHREAD}
/// not supported to reduce dependencies and console writing
{$undef FPCMM_REPORTMEMORYLEAKS}
{$else}
type
/// one GetSmallBlockContention info about unexpected multi-thread waiting
// - a single GetmemBlockSize or FreememBlockSize non 0 field is set
TSmallBlockContention = packed record
/// how many times a small block getmem/freemem has been waiting for unlock
SleepCount: cardinal;
/// the small block size on which Getmem() has been blocked - or 0
GetmemBlockSize: cardinal;
/// the small block size on which Freemem() has been blocked - or 0
FreememBlockSize: cardinal;
end;
/// small blocks detailed information as returned GetSmallBlockContention
TSmallBlockContentionDynArray = array of TSmallBlockContention;
/// one GetSmallBlockStatus information
TSmallBlockStatus = packed record
/// how many times a memory block of this size has been allocated
Total: cardinal;
/// how many memory blocks of this size are currently allocated
Current: cardinal;
/// the standard size of the small memory block
BlockSize: cardinal;
end;
/// small blocks detailed information as returned GetSmallBlockStatus
TSmallBlockStatusDynArray = array of TSmallBlockStatus;
/// sort order of detailed information as returned GetSmallBlockStatus
TSmallBlockOrderBy = (obTotal, obCurrent, obBlockSize);
/// retrieve the use counts of allocated small blocks
// - returns maxcount biggest results, sorted by "orderby" field occurence
function GetSmallBlockStatus(maxcount: integer = 10;
orderby: TSmallBlockOrderBy = obTotal; count: PPtrUInt = nil; bytes: PPtrUInt = nil;
small: PCardinal = nil; tiny: PCardinal = nil): TSmallBlockStatusDynArray;
/// retrieve all small blocks which suffered from blocking during multi-thread
// - returns maxcount biggest results, sorted by SleepCount occurence
function GetSmallBlockContention(
maxcount: integer = 10): TSmallBlockContentionDynArray;
/// convenient debugging function into the console
// - if smallblockcontentioncount > 0, includes GetSmallBlockContention() info
// up to the smallblockcontentioncount biggest occurences
procedure WriteHeapStatus(const context: shortstring = '';
smallblockstatuscount: integer = 8; smallblockcontentioncount: integer = 8;
compilationflags: boolean = false);
/// convenient debugging function into a string
// - if smallblockcontentioncount > 0, includes GetSmallBlockContention() info
// up to the smallblockcontentioncount biggest occurences
// - warning: this function is not thread-safe
function GetHeapStatus(const context: shortstring; smallblockstatuscount,
smallblockcontentioncount: integer; compilationflags, onsameline: boolean): string;
const
/// human readable information about how our MM was built
// - similar to WriteHeapStatus(compilationflags=true) output
FPCMM_FLAGS = ' '
{$ifdef FPCMM_BOOSTER} + 'BOOSTER ' {$else}
{$ifdef FPCMM_BOOST} + 'BOOST ' {$else}
{$ifdef FPCMM_SERVER} + 'SERVER ' {$endif}
{$endif}
{$endif}
{$ifdef FPCMM_ASSUMEMULTITHREAD} + ' assumulthrd' {$endif}
{$ifdef FPCMM_LOCKLESSFREE} + ' lockless' {$endif}
{$ifdef FPCMM_PAUSE} + ' pause' {$endif}
{$ifdef FPCMM_SLEEPTSC} + ' rdtsc' {$endif}
{$ifndef BSD}
{$ifdef FPCMM_NOMREMAP} + ' nomremap' {$endif}
{$endif BSD}
{$ifdef FPCMM_SMALLNOTWITHMEDIUM}+ ' smallpool' {$endif}
{$ifdef FPCMM_ERMS} + ' erms' {$endif}
{$ifdef FPCMM_DEBUG} + ' debug' {$endif}
{$ifdef FPCMM_REPORTMEMORYLEAKS} + ' repmemleak' {$endif};
{$endif FPCMM_STANDALONE}
{$endif FPCX64MM}
implementation
{
High-level Allocation Strategy Description
--------------------------------------------
The allocator handles the following families of memory blocks:
- TINY <= 128 B (<= 256 B for FPCMM_BOOST)
Round-robin distribution into several arenas, fed from shared tiny/small pool
(fair scaling from multi-threaded calls, with no threadvar nor GC involved)
- SMALL <= 2600 B
One arena per block size, fed from shared tiny/small pool
- MEDIUM <= 256 KB
Separated pool of bitmap-marked chunks, fed from 1MB of OS mmap/virtualalloc
- LARGE > 256 KB
Directly fed from OS mmap/virtualalloc with mremap when growing
The original FastMM4 was enhanced as such, especially in FPCMM_SERVER mode:
- FPC compatibility, even on POSIX/Linux, also for FPC specific API behavior;
- x86_64 code was refactored and tuned in regard to 2020's hardware;
- Inlined SSE2 movaps loop or ERMS are more efficient that subfunction(s);
- New round-robin thread-friendly arenas of tiny blocks;
- Tiny and small blocks can fed from their own pool, not the medium pool;
- Additional bin list to reduce small/tiny Freemem() thread contention;
- Memory leaks and thread sleep tracked with almost no performance loss;
- Large blocks logic has been rewritten, especially realloc;
- On Linux, mremap is used for efficient realloc of large blocks.
About locking:
- Tiny and Small blocks have their own per-size lock;
- Tiny and Small blocks have one giant lock when fedding from their pool;
- Medium blocks have one giant lock over their own pool;
- Large blocks have one giant lock over mmap/virtualalloc system calls;
- SwitchToThread/FpNanoSleep OS call is done after initial spinning;
- FPCMM_LOCKLESSFREE reduces Freemem() thread contention;
- FPCMM_DEBUG / WriteHeapStatus helps identifying the lock contention(s).
}
{$ifdef FPCX64MM}
// this unit is available only for FPC + X86_64 CPU
{$ifndef FPCMM_NOPAUSE}
// on contention problem, execute "pause" opcode and spin retrying the lock
// - defined by default to follow Intel recommendatations from
// https://software.intel.com/content/www/us/en/develop/articles/benefitting-power-and-performance-sleep-loops.html
// - spinning loop is either using constants or rdtsc (if FPCMM_SLEEPTSC is set)
// - on SkylakeX (Intel 7th gen), "pause" opcode went from 10-20 to 140 cycles
// so our constants below will favor those latest CPUs with a longer pause
{$define FPCMM_PAUSE}
{$endif FPCMM_NOPAUSE}
{ ********* Operating System Specific API Calls }
{$ifdef MSWINDOWS}
// Win64: any assembler function with sub-calls should have a stack frame
// -> nostackframe is defined only on Linux or for functions with no nested call
{$undef NOSFRAME}
const
kernel32 = 'kernel32.dll';
MEM_COMMIT = $1000;
MEM_RESERVE = $2000;
MEM_RELEASE = $8000;
MEM_FREE = $10000;
MEM_TOP_DOWN = $100000;
PAGE_READWRITE = 4;
PAGE_GUARD = $0100;
PAGE_VALID = $00e6; // PAGE_READONLY or PAGE_READWRITE or PAGE_EXECUTE or
// PAGE_EXECUTE_READ or PAGE_EXECUTE_READWRITE or PAGE_EXECUTE_WRITECOPY
type
// VirtualQuery() API result structure
TMemInfo = record
BaseAddress, AllocationBase: pointer;
AllocationProtect: cardinal;
PartitionId: word;
RegionSize: PtrUInt;
State, Protect, MemType: cardinal;
end;
function VirtualAlloc(lpAddress: pointer;
dwSize: PtrUInt; flAllocationType, flProtect: cardinal): pointer;
stdcall; external kernel32 name 'VirtualAlloc';
function VirtualFree(lpAddress: pointer; dwSize: PtrUInt;
dwFreeType: cardinal): LongBool;
stdcall; external kernel32 name 'VirtualFree';
procedure SwitchToThread;
stdcall; external kernel32 name 'SwitchToThread';
function VirtualQuery(lpAddress, lpMemInfo: pointer; dwLength: PtrUInt): PtrUInt;
stdcall; external kernel32 name 'VirtualQuery';
function AllocMedium(Size: PtrInt): pointer; inline;
begin
// bottom-up allocation to reduce fragmentation
result := VirtualAlloc(nil, Size, MEM_COMMIT, PAGE_READWRITE);
end;
function AllocLarge(Size: PtrInt): pointer; inline;
begin
// top-down allocation of large blocks to reduce fragmentation
// (MEM_TOP_DOWN is not available on POSIX, but seems less needed)
result := VirtualAlloc(nil, Size, MEM_COMMIT or MEM_TOP_DOWN, PAGE_READWRITE);
end;
procedure FreeMediumLarge(ptr: pointer; Size: PtrInt); inline;
begin
VirtualFree(ptr, 0, MEM_RELEASE);
end;
{$ifndef FPCMM_NOMREMAP}
function RemapLarge(addr: pointer; old_len, new_len: size_t): pointer;
var
meminfo: TMemInfo;
next: pointer;
nextsize: PtrUInt;
begin
// old_len and new_len have 64KB granularity, so match Windows page size
nextsize := new_len - old_len;
if PtrInt(nextsize) > 0 then
begin
// try to allocate the memory just after the existing one
FillChar(meminfo, SizeOf(meminfo), 0);
next := addr + old_len;
if (VirtualQuery(next, @meminfo, SizeOf(meminfo)) = SizeOf(meminfo)) and
(meminfo.State = MEM_FREE) and
(meminfo.RegionSize >= nextsize) and // enough space?
// reserve the address space in two steps for thread safety
(VirtualAlloc(next, nextsize, MEM_RESERVE, PAGE_READWRITE) <> nil) and
(VirtualAlloc(next, nextsize, MEM_COMMIT, PAGE_READWRITE) <> nil) then
begin
result := addr; // in-place realloc: no need to move memory :)
exit;
end;
end;
// we need to use the slower but safe Alloc/Move/Free pattern
result := AllocLarge(new_len);
if new_len > old_len then
new_len := old_len; // handle size up or down
Move(addr^, result^, new_len); // RTL non-volatile asm or our AVX MoveFast()
FreeMediumLarge(addr, old_len);
end;
{$endif FPCMM_NOMREMAP}
// experimental VirtualQuery detection of object class - use at your own risk
{$define FPCMM_REPORTMEMORYLEAKS_EXPERIMENTAL}
{$else}
uses
{$ifndef DARWIN}
syscall,
{$endif DARWIN}
BaseUnix;
// in practice, SYSV ABI seems to not require a stack frame, as Win64 does, for
// our use case of nested calls with no local stack storage and direct kernel
// syscalls - but since it is clearly undocumented, we set it on LINUX only
// -> appears to work with no problem from our tests: feedback is welcome!
// -> see FPCMM_NOSFRAME conditional to disable it on LINUX
{$ifdef LINUX}
{$define NOSFRAME}
{$endif LINUX}
// we directly call the OS Kernel, so this unit doesn't require any libc
function AllocMedium(Size: PtrInt): pointer; inline;
begin
result := fpmmap(nil, Size, PROT_READ or PROT_WRITE,
MAP_PRIVATE or MAP_ANONYMOUS, -1, 0);
end;
function AllocLarge(Size: PtrInt): pointer; inline;
begin
result := AllocMedium(size); // same API (no MEM_TOP_DOWN option)
end;
procedure FreeMediumLarge(ptr: pointer; Size: PtrInt); inline;
begin
fpmunmap(ptr, Size);
end;
{$ifdef LINUX}
{$ifndef FPCMM_NOMREMAP}
const
syscall_nr_mremap = 25; // valid on x86_64 Linux and Android
MREMAP_MAYMOVE = 1;
function RemapLarge(addr: pointer; old_len, new_len: size_t): pointer; inline;
begin
// let the Linux Kernel mremap() the memory using its TLB magic
result := pointer(do_syscall(syscall_nr_mremap, TSysParam(addr),
TSysParam(old_len), TSysParam(new_len), TSysParam(MREMAP_MAYMOVE)));
end;
{$endif FPCMM_NOMREMAP}
// experimental detection of object class - use at your own risk
{$define FPCMM_REPORTMEMORYLEAKS_EXPERIMENTAL}
// (untested on BSD/DARWIN)
{$else}
{$define FPCMM_NOMREMAP} // mremap is a Linux-specific syscall
{$endif LINUX}
procedure SwitchToThread;
var
t: Ttimespec;
begin
// note: nanosleep() adds a few dozen of microsecs for context switching
t.tv_sec := 0;
t.tv_nsec := 10; // empirically identified on a recent Linux Kernel
fpnanosleep(@t, nil);
end;
{$endif MSWINDOWS}
// fallback to safe and simple Alloc/Move/Free pattern
{$ifdef FPCMM_NOMREMAP}
function RemapLarge(addr: pointer; old_len, new_len: size_t): pointer;
begin
result := AllocLarge(new_len);
if new_len > old_len then
new_len := old_len; // resize down
Move(addr^, result^, new_len); // RTL non-volatile asm or our AVX MoveFast()
FreeMediumLarge(addr, old_len);
end;
{$endif FPCMM_NOMREMAP}
{ ********* Some Assembly Helpers }
// low-level conditional to disable nostackframe code on Linux
{$ifdef FPCMM_NOSFRAME}
{$undef NOSFRAME}
{$endif FPCMM_NOSFRAME}
var
HeapStatus: TMMStatus;
{$ifdef FPCMM_DEBUG}
procedure ReleaseCore;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
asm
{$ifdef FPCMM_SLEEPTSC}
rdtsc // returns the TSC in EDX:EAX
shl rdx, 32
or rax, rdx
push rax
call SwitchToThread
pop rcx
rdtsc
shl rdx, 32
or rax, rdx
lea rdx, [rip + HeapStatus]
sub rax, rcx
lock add qword ptr [rdx + TMMStatus.SleepCycles], rax
{$else}
call SwitchToThread
lea rdx, [rip + HeapStatus]
{$endif FPCMM_SLEEPTSC}
lock inc qword ptr [rdx + TMMStatus.SleepCount]
end;
{$else}
procedure ReleaseCore;
begin
SwitchToThread;
inc(HeapStatus.SleepCount); // indicative counter
end;
{$endif FPCMM_DEBUG}
procedure NotifyArenaAlloc(var Arena: TMMStatusArena; Size: PtrUInt);
nostackframe; assembler;
asm
{$ifdef FPCMM_DEBUG}
lock add qword ptr [Arena].TMMStatusArena.CurrentBytes, Size
lock add qword ptr [Arena].TMMStatusArena.CumulativeBytes, Size
lock inc qword ptr [Arena].TMMStatusArena.CumulativeAlloc
mov rax, qword ptr [Arena].TMMStatusArena.CurrentBytes
cmp rax, qword ptr [Arena].TMMStatusArena.PeakBytes
jbe @s
mov qword ptr [Arena].TMMStatusArena.PeakBytes, rax
@s: {$else}
add qword ptr [Arena].TMMStatusArena.CurrentBytes, Size
add qword ptr [Arena].TMMStatusArena.CumulativeBytes, Size
{$endif FPCMM_DEBUG}
end;
procedure NotifyMediumLargeFree(var Arena: TMMStatusArena; Size: PtrUInt);
nostackframe; assembler;
asm
neg Size
{$ifdef FPCMM_DEBUG}
lock add qword ptr [Arena].TMMStatusArena.CurrentBytes, Size
lock inc qword ptr [Arena].TMMStatusArena.CumulativeFree
{$else}
add qword ptr [Arena].TMMStatusArena.CurrentBytes, Size
{$endif FPCMM_DEBUG}
end;
{ ********* Constants and Data Structures Definitions }
const
// (sometimes) the more arenas, the better multi-threadable
{$ifdef FPCMM_BOOSTER}
NumTinyBlockTypesPO2 = 4;
NumTinyBlockArenasPO2 = 4; // will probably end up with Medium lock contention
{$else}
{$ifdef FPCMM_BOOST}
NumTinyBlockTypesPO2 = 4; // tiny are <= 256 bytes
NumTinyBlockArenasPO2 = 3; // 8 arenas
{$else}
// default (or FPCMM_SERVER) settings
NumTinyBlockTypesPO2 = 3; // multiple arenas for tiny blocks <= 128 bytes
NumTinyBlockArenasPO2 = 3; // 8 round-robin arenas (including main) by default
{$endif FPCMM_BOOST}
{$endif FPCMM_BOOSTER}
NumSmallBlockTypes = 46;
MaximumSmallBlockSize = 2608;
SmallBlockSizes: array[0..NumSmallBlockTypes - 1] of word = (
16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224, 240, 256,
272, 288, 304, 320, 352, 384, 416, 448, 480, 528, 576, 624, 672, 736, 800,
880, 960, 1056, 1152, 1264, 1376, 1504, 1648, 1808, 1984, 2176, 2384,
MaximumSmallBlockSize, MaximumSmallBlockSize, MaximumSmallBlockSize);
NumTinyBlockTypes = 1 shl NumTinyBlockTypesPO2;
NumTinyBlockArenas = (1 shl NumTinyBlockArenasPO2) - 1; // -1 = main Small[]
NumSmallInfoBlock = NumSmallBlockTypes + NumTinyBlockArenas * NumTinyBlockTypes;
SmallBlockGranularity = 16;
TargetSmallBlocksPerPool = 48;
MinimumSmallBlocksPerPool = 12;
SmallBlockDownsizeCheckAdder = 64;
SmallBlockUpsizeAdder = 32;
{$ifdef FPCMM_LOCKLESSFREE}
SmallBlockTypePO2 = 8; // SizeOf(TSmallBlockType)=256 with Bin list
SmallBlockBinCount = (((1 shl SmallBlockTypePO2) - 64) div 8) - 1; // =23
{$else}
SmallBlockTypePO2 = 6; // SizeOf(TSmallBlockType)=64
{$endif FPCMM_LOCKLESSFREE}
MediumBlockPoolSizeMem = 20 * 64 * 1024;
MediumBlockPoolSize = MediumBlockPoolSizeMem - 16;
MediumBlockSizeOffset = 48;
MinimumMediumBlockSize = 11 * 256 + MediumBlockSizeOffset;
MediumBlockBinsPerGroup = 32;
MediumBlockBinGroupCount = 32;
MediumBlockBinCount = MediumBlockBinGroupCount * MediumBlockBinsPerGroup;
MediumBlockGranularity = 256;
MaximumMediumBlockSize =
MinimumMediumBlockSize + (MediumBlockBinCount - 1) * MediumBlockGranularity;
OptimalSmallBlockPoolSizeLowerLimit =
29 * 1024 - MediumBlockGranularity + MediumBlockSizeOffset;
OptimalSmallBlockPoolSizeUpperLimit =
64 * 1024 - MediumBlockGranularity + MediumBlockSizeOffset;
MaximumSmallBlockPoolSize =
OptimalSmallBlockPoolSizeUpperLimit + MinimumMediumBlockSize;
MediumInPlaceDownsizeLimit = MinimumMediumBlockSize div 4;
IsFreeBlockFlag = 1;
IsMediumBlockFlag = 2;
IsSmallBlockPoolInUseFlag = 4;
IsLargeBlockFlag = 4;
PreviousMediumBlockIsFreeFlag = 8;
LargeBlockIsSegmented = 8;
DropSmallFlagsMask = -8;
ExtractSmallFlagsMask = 7;
DropMediumAndLargeFlagsMask = -16;
ExtractMediumAndLargeFlagsMask = 15;
{$ifdef FPCMM_SLEEPTSC}
// pause using rdtsc (30 cycles latency on hardware but emulated on VM)
SpinMediumLockTSC = 10000;
SpinLargeLockTSC = 10000;
{$ifdef FPCMM_PAUSE}
SpinSmallGetmemLockTSC = 1000;
SpinSmallFreememLockTSC = 1000; // _freemem has more collisions
{$ifdef FPCMM_LOCKLESSFREE}
SpinSmallFreememBinTSC = 2000;
{$endif FPCMM_LOCKLESSFREE}
{$endif FPCMM_PAUSE}
{$else}
// pause with constant spinning counts (empirical values from fastmm4-avx)
SpinMediumLockCount = 2500;
SpinLargeLockCount = 5000;
{$ifdef FPCMM_PAUSE}
SpinSmallGetmemLockCount = 500;
SpinSmallFreememLockCount = 500;
{$ifdef FPCMM_LOCKLESSFREE}
SpinFreememBinCount = 500;
{$endif FPCMM_LOCKLESSFREE}
{$endif FPCMM_PAUSE}
{$endif FPCMM_SLEEPTSC}
{$ifdef FPCMM_ERMS}
// pre-ERMS expects at least 256 bytes, IvyBridge+ with ERMS is good from 64
// see https://stackoverflow.com/a/43837564/458259 for explanations and timing
// -> "movaps" loop is used up to 256 bytes of data: good on all CPUs
// -> "movnt" Move/MoveFast is used for large blocks: always faster than ERMS
ErmsMinSize = 256;
{$endif FPCMM_ERMS}
type
PSmallBlockPoolHeader = ^TSmallBlockPoolHeader;
// information for each small block size - 64/256 bytes long >= CPU cache line
TSmallBlockType = record
Locked: boolean;
AllowedGroupsForBlockPoolBitmap: byte;
BlockSize: Word;
MinimumBlockPoolSize: Word;
OptimalBlockPoolSize: Word;
NextPartiallyFreePool: PSmallBlockPoolHeader;
PreviousPartiallyFreePool: PSmallBlockPoolHeader;
NextSequentialFeedBlockAddress: pointer;
MaxSequentialFeedBlockAddress: pointer;
CurrentSequentialFeedPool: PSmallBlockPoolHeader;
GetmemCount: cardinal;
FreememCount: cardinal;
GetmemSleepCount: cardinal;
FreememSleepCount: cardinal;
{$ifdef FPCMM_LOCKLESSFREE} // 192 optional bytes for FreeMem Bin (= 13KB)
BinLocked: boolean; // dedicated lock for less contention
BinCount: byte;
BinSpinCount: cardinal;
BinInstance: array[0.. SmallBlockBinCount - 1] of pointer;
{$endif FPCMM_LOCKLESSFREE}
end;
PSmallBlockType = ^TSmallBlockType;
TSmallBlockTypes = array[0..NumSmallBlockTypes - 1] of TSmallBlockType;
TTinyBlockTypes = array[0..NumTinyBlockTypes - 1] of TSmallBlockType;
TSmallBlockInfo = record
Small: TSmallBlockTypes;
Tiny: array[0..NumTinyBlockArenas - 1] of TTinyBlockTypes;
GetmemLookup: array[0..
(MaximumSmallBlockSize - 1) div SmallBlockGranularity] of byte;
{$ifndef FPCMM_ASSUMEMULTITHREAD}
IsMultiThreadPtr: PBoolean; // safe access to IsMultiThread global variable
{$endif FPCMM_ASSUMEMULTITHREAD}
TinyCurrentArena: integer;
end;
TSmallBlockPoolHeader = record
BlockType: PSmallBlockType;
{$ifdef CPU32}
Padding32Bits: cardinal;
{$endif CPU32}
NextPartiallyFreePool: PSmallBlockPoolHeader;
PreviousPartiallyFreePool: PSmallBlockPoolHeader;
FirstFreeBlock: pointer;
BlocksInUse: cardinal;
SmallBlockPoolSignature: cardinal;
FirstBlockPoolPointerAndFlags: PtrUInt;
end;
PMediumBlockPoolHeader = ^TMediumBlockPoolHeader;
TMediumBlockPoolHeader = record
PreviousMediumBlockPoolHeader: PMediumBlockPoolHeader;
NextMediumBlockPoolHeader: PMediumBlockPoolHeader;
Reserved1: PtrUInt;
FirstMediumBlockSizeAndFlags: PtrUInt;
end;
PMediumFreeBlock = ^TMediumFreeBlock;
TMediumFreeBlock = record
PreviousFreeBlock: PMediumFreeBlock;
NextFreeBlock: PMediumFreeBlock;
end;
// medium locks occurs at getmem: freemem bin list got MaxCount=0 :(
{.$define FPCMM_LOCKLESSFREEMEDIUM}
{$ifdef FPCMM_LOCKLESSFREEMEDIUM}
const
MediumBlockLocklessBinCount = 255;
type
// used by TMediumBlockInfo to reduce thread pressure
TMediumLocklessBin = record
Locked: boolean; // dedicated lock for less contention
Count: byte;
MaxCount: byte;
Instance: array[0 .. MediumBlockLocklessBinCount - 1] of pointer;
end;
{$endif FPCMM_LOCKLESSFREEMEDIUM}
TMediumBlockInfo = record
Locked: boolean;
PoolsCircularList: TMediumBlockPoolHeader;
LastSequentiallyFed: pointer;
SequentialFeedBytesLeft: cardinal;
BinGroupBitmap: cardinal;
{$ifndef FPCMM_ASSUMEMULTITHREAD}
IsMultiThreadPtr: PBoolean; // safe access to IsMultiThread global variable
{$endif FPCMM_ASSUMEMULTITHREAD}
BinBitmaps: array[0..MediumBlockBinGroupCount - 1] of cardinal;
Bins: array[0..MediumBlockBinCount - 1] of TMediumFreeBlock;
{$ifdef FPCMM_LOCKLESSFREEMEDIUM}
LocklessBin: TMediumLocklessBin;
{$endif FPCMM_LOCKLESSFREEMEDIUM}
end;
PLargeBlockHeader = ^TLargeBlockHeader;
TLargeBlockHeader = record
PreviousLargeBlockHeader: PLargeBlockHeader;
NextLargeBlockHeader: PLargeBlockHeader;
Reserved: PtrUInt;
BlockSizeAndFlags: PtrUInt;
end;
const
BlockHeaderSize = SizeOf(pointer);
SmallBlockPoolHeaderSize = SizeOf(TSmallBlockPoolHeader);
MediumBlockPoolHeaderSize = SizeOf(TMediumBlockPoolHeader);
LargeBlockHeaderSize = SizeOf(TLargeBlockHeader);
LargeBlockGranularity = 65536;
var
SmallBlockInfo: TSmallBlockInfo;
MediumBlockInfo: TMediumBlockInfo;
SmallMediumBlockInfo: TMediumBlockInfo
{$ifdef FPCMM_SMALLNOTWITHMEDIUM} ;
{$else} absolute MediumBlockInfo;
{$endif FPCMM_SMALLNOTWITHMEDIUM}
LargeBlocksLocked: boolean;
LargeBlocksCircularList: TLargeBlockHeader;
{ ********* Shared Routines }
{$define FPCMM_CMPBEFORELOCK_SPIN}
// during spinning, there is clearly thread contention: in this case, plain
// "cmp" before "lock cmpxchg" is mandatory to leverage the CPU cores
procedure LockMediumBlocks;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
// on input/output: r10=TMediumBlockInfo
asm
{$ifdef FPCMM_SLEEPTSC}
@s: rdtsc // tsc in edx:eax
shl rdx, 32
lea r9, [rax + rdx + SpinMediumLockTSC] // r9 = endtsc
@sp: pause
rdtsc
shl rdx, 32
or rax, rdx
cmp rax, r9
ja @rc // timeout
{$else}
@s: mov edx, SpinMediumLockCount
@sp: pause
dec edx
jz @rc //timeout
{$endif FPCMM_SLEEPTSC}
mov rcx, r10
mov eax, $100
{$ifdef FPCMM_CMPBEFORELOCK_SPIN}
cmp byte ptr [r10].TMediumBlockInfo.Locked, true
je @sp
{$endif FPCMM_CMPBEFORELOCK_SPIN}
lock cmpxchg byte ptr [rcx].TMediumBlockInfo.Locked, ah
je @ok
jmp @sp
@rc: push rsi // preserve POSIX and Win64 ABI registers
push rdi
push r10
push r11
call ReleaseCore
pop r11
pop r10
pop rdi
pop rsi
lea rax, [rip + HeapStatus]
{$ifdef FPCMM_DEBUG} lock {$endif}
inc qword ptr [rax].TMMStatus.Medium.SleepCount
jmp @s
@ok:
end;
procedure InsertMediumBlockIntoBin; nostackframe; assembler;
// rcx=P edx=blocksize r10=TMediumBlockInfo - even on POSIX
asm
mov rax, rcx
// Get the bin number for this block size
sub edx, MinimumMediumBlockSize
shr edx, 8
// Validate the bin number
sub edx, MediumBlockBinCount - 1
sbb ecx, ecx
and edx, ecx
add edx, MediumBlockBinCount - 1
mov r9, rdx
// Get the bin address in rcx
shl edx, 4
lea rcx, [r10 + rdx + TMediumBlockInfo.Bins]
// Bins are LIFO, se we insert this block as the first free block in the bin
mov rdx, TMediumFreeBlock[rcx].NextFreeBlock
mov TMediumFreeBlock[rax].PreviousFreeBlock, rcx
mov TMediumFreeBlock[rax].NextFreeBlock, rdx
mov TMediumFreeBlock[rdx].PreviousFreeBlock, rax
mov TMediumFreeBlock[rcx].NextFreeBlock, rax
// Was this bin empty?
cmp rdx, rcx
jne @Done
// Get ecx=bin number, edx=group number
mov rcx, r9
mov rdx, r9
shr edx, 5
// Flag this bin as not empty
mov eax, 1
shl eax, cl
or dword ptr [r10 + TMediumBlockInfo.BinBitmaps + rdx * 4], eax
// Flag the group as not empty
mov eax, 1
mov ecx, edx
shl eax, cl
or [r10 + TMediumBlockInfo.BinGroupBitmap], eax
@Done:
end;
procedure RemoveMediumFreeBlock; nostackframe; assembler;
asm
// rcx=MediumFreeBlock r10=TMediumBlockInfo - even on POSIX
// Get the current previous and next blocks
mov rdx, TMediumFreeBlock[rcx].PreviousFreeBlock
mov rcx, TMediumFreeBlock[rcx].NextFreeBlock
// Remove this block from the linked list
mov TMediumFreeBlock[rcx].PreviousFreeBlock, rdx
mov TMediumFreeBlock[rdx].NextFreeBlock, rcx
// Is this bin now empty? If the previous and next free block pointers are
// equal, they must point to the bin
cmp rcx, rdx
jne @Done
// Get ecx=bin number, edx=group number
lea r8, [r10 + TMediumBlockInfo.Bins]
sub rcx, r8
mov edx, ecx
shr ecx, 4
shr edx, 9
// Flag this bin as empty
mov eax, -2
rol eax, cl
and dword ptr [r10 + TMediumBlockInfo.BinBitmaps + rdx * 4], eax
jnz @Done
// Flag this group as empty
mov eax, -2
mov ecx, edx
rol eax, cl
and [r10 + TMediumBlockInfo.BinGroupBitmap], eax
@Done:
end;
procedure BinMediumSequentialFeedRemainder(
var Info: TMediumBlockInfo); nostackframe; assembler;
asm
mov r10, Info
mov eax, [Info + TMediumBlockInfo.SequentialFeedBytesLeft]
test eax, eax
jz @Done
// Is the last fed sequentially block free?
mov rax, [Info + TMediumBlockInfo.LastSequentiallyFed]
test byte ptr [rax - BlockHeaderSize], IsFreeBlockFlag
jnz @LastBlockFedIsFree
// Set the "previous block is free" flag in the last block fed
or qword ptr [rax - BlockHeaderSize], PreviousMediumBlockIsFreeFlag
// Get edx=remainder size, rax=remainder start
mov edx, [r10 + TMediumBlockInfo.SequentialFeedBytesLeft]
sub rax, rdx
@BinTheRemainder:
// Store the size of the block as well as the flags
lea rcx, [rdx + IsMediumBlockFlag + IsFreeBlockFlag]
mov [rax - BlockHeaderSize], rcx
// Store the trailing size marker
mov [rax + rdx - 16], rdx
// Bin this medium block
cmp edx, MinimumMediumBlockSize
jb @Done
mov rcx, rax
jmp InsertMediumBlockIntoBin // rcx=APMediumFreeBlock, edx=AMediumBlockSize
@Done: ret
@LastBlockFedIsFree:
// Drop the flags
mov rdx, DropMediumAndLargeFlagsMask
and rdx, [rax - BlockHeaderSize]
// Free the last block fed
cmp edx, MinimumMediumBlockSize
jb @DontRemoveLastFed
// Last fed block is free - remove it from its size bin
mov rcx, rax
call RemoveMediumFreeBlock // rcx = APMediumFreeBlock
// Re-read rax and rdx
mov rax, [r10 + TMediumBlockInfo.LastSequentiallyFed]
mov rdx, DropMediumAndLargeFlagsMask
and rdx, [rax - BlockHeaderSize]
@DontRemoveLastFed:
// Get the number of bytes left in ecx
mov ecx, [r10 + TMediumBlockInfo.SequentialFeedBytesLeft]
// rax = remainder start, rdx = remainder size
sub rax, rcx
add edx, ecx
jmp @BinTheRemainder
end;
procedure FreeMedium(ptr: PMediumBlockPoolHeader);
begin
FreeMediumLarge(ptr, MediumBlockPoolSizeMem);
NotifyMediumLargeFree(HeapStatus.Medium, MediumBlockPoolSizeMem);
end;
function AllocNewSequentialFeedMediumPool(BlockSize: cardinal;
var Info: TMediumBlockInfo): pointer;
var
old: PMediumBlockPoolHeader;
new: pointer;
begin
BinMediumSequentialFeedRemainder(Info);
new := AllocMedium(MediumBlockPoolSizeMem);
if new <> nil then
begin
old := Info.PoolsCircularList.NextMediumBlockPoolHeader;
PMediumBlockPoolHeader(new).PreviousMediumBlockPoolHeader := @Info.PoolsCircularList;
Info.PoolsCircularList.NextMediumBlockPoolHeader := new;
PMediumBlockPoolHeader(new).NextMediumBlockPoolHeader := old;
old.PreviousMediumBlockPoolHeader := new;
PPtrUInt(PByte(new) + MediumBlockPoolSize - BlockHeaderSize)^ := IsMediumBlockFlag;
Info.SequentialFeedBytesLeft :=
(MediumBlockPoolSize - MediumBlockPoolHeaderSize) - BlockSize;
result := pointer(PByte(new) + MediumBlockPoolSize - BlockSize);
Info.LastSequentiallyFed := result;
PPtrUInt(PByte(result) - BlockHeaderSize)^ := BlockSize or IsMediumBlockFlag;
NotifyArenaAlloc(HeapStatus.Medium, MediumBlockPoolSizeMem);
end
else
begin
Info.SequentialFeedBytesLeft := 0;
result := nil;
end;
end;
procedure LockLargeBlocks;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
asm
@s: mov eax, $100
lea rcx, [rip + LargeBlocksLocked]
lock cmpxchg byte ptr [rcx], ah
je @ok
{$ifdef FPCMM_SLEEPTSC}
rdtsc
shl rdx, 32
lea r9, [rax + rdx + SpinLargeLockTSC] // r9 = endtsc
@sp: pause
rdtsc
shl rdx, 32
or rax, rdx
cmp rax, r9
ja @rc // timeout
{$else}
mov edx, SpinLargeLockCount
@sp: pause
dec edx
jz @rc // timeout
{$endif FPCMM_SLEEPTSC}
mov eax, $100
{$ifdef FPCMM_CMPBEFORELOCK_SPIN}
cmp byte ptr [rcx], true
je @sp
{$endif FPCMM_CMPBEFORELOCK_SPIN}
lock cmpxchg byte ptr [rcx], ah
je @ok
jmp @sp
@rc: call ReleaseCore
lea rax, [rip + HeapStatus]
{$ifdef FPCMM_DEBUG} lock {$endif}
inc qword ptr [rax].TMMStatus.Large.SleepCount
jmp @s
@ok: // reset the stack frame before ret
end;
function AllocateLargeBlockFrom(size: PtrUInt;
existing: pointer; oldsize: PtrUInt): pointer;
var
blocksize: PtrUInt;
header, old: PLargeBlockHeader;
begin
blocksize := (size + LargeBlockHeaderSize +
LargeBlockGranularity - 1 + BlockHeaderSize) and -LargeBlockGranularity;
if existing = nil then
header := AllocLarge(blocksize)
else
header := RemapLarge(existing, oldsize, blocksize);
if header <> nil then
begin
NotifyArenaAlloc(HeapStatus.Large, blocksize);
if existing <> nil then
NotifyMediumLargeFree(HeapStatus.Large, oldsize);
header.BlockSizeAndFlags := blocksize or IsLargeBlockFlag;
LockLargeBlocks;
old := LargeBlocksCircularList.NextLargeBlockHeader;
header.PreviousLargeBlockHeader := @LargeBlocksCircularList;
LargeBlocksCircularList.NextLargeBlockHeader := header;
header.NextLargeBlockHeader := old;
old.PreviousLargeBlockHeader := header;
LargeBlocksLocked := False;
inc(header);
end;
result := header;
end;
function AllocateLargeBlock(size: PtrUInt): pointer;
begin
result := AllocateLargeBlockFrom(size, nil, 0);
end;
procedure FreeLarge(ptr: PLargeBlockHeader; size: PtrUInt);
begin
NotifyMediumLargeFree(HeapStatus.Large, size);
FreeMediumLarge(ptr, size);
end;
function FreeLargeBlock(p: pointer): PtrInt;
var
header, prev, next: PLargeBlockHeader;
begin
header := pointer(PByte(p) - LargeBlockHeaderSize);
if header.BlockSizeAndFlags and IsFreeBlockFlag <> 0 then
begin
// try to release the same pointer twice
result := 0;
exit;
end;
LockLargeBlocks;
prev := header.PreviousLargeBlockHeader;
next := header.NextLargeBlockHeader;
next.PreviousLargeBlockHeader := prev;
prev.NextLargeBlockHeader := next;
LargeBlocksLocked := False;
result := DropMediumAndLargeFlagsMask and header.BlockSizeAndFlags;
FreeLarge(header, result);
end;
function ReallocateLargeBlock(p: pointer; size: PtrUInt): pointer;
var
oldavail, minup, new: PtrUInt;
prev, next, header: PLargeBlockHeader;
begin
header := pointer(PByte(p) - LargeBlockHeaderSize);
oldavail := (DropMediumAndLargeFlagsMask and header^.BlockSizeAndFlags) -
(LargeBlockHeaderSize + BlockHeaderSize);
new := size;
if size > oldavail then
begin
// size-up with 1/8 or 1/4 overhead for any future growing realloc
if oldavail > 128 shl 20 then
minup := oldavail + oldavail shr 3
else
minup := oldavail + oldavail shr 2;
if size < minup then
new := minup;
end
else
begin
result := p;
oldavail := oldavail shr 1;
if size >= oldavail then
// small size-up within current buffer -> no reallocate
exit
else
// size-down and move just the trailing data
oldavail := size;
end;
if new < MaximumMediumBlockSize then
begin
// size was reduced to a small/medium block: use GetMem/Move/FreeMem
result := _GetMem(new);
if result <> nil then
Move(p^, result^, oldavail); // RTL non-volatile asm or our AVX MoveFast()
_FreeMem(p);
end
else
begin
// remove large block from current chain list
LockLargeBlocks;
prev := header^.PreviousLargeBlockHeader;
next := header^.NextLargeBlockHeader;
next.PreviousLargeBlockHeader := prev;
prev.NextLargeBlockHeader := next;
LargeBlocksLocked := False;
size := DropMediumAndLargeFlagsMask and header^.BlockSizeAndFlags;
// on Linux, call Kernel mremap() and its TLB magic
// on Windows, try to reserve the memory block just after the existing
// otherwise, use Alloc/Move/Free pattern, with asm/AVX move
result := AllocateLargeBlockFrom(new, header, size);
end;
end;
{ ********* Main Memory Manager Functions }
function _GetMem(size: PtrUInt): pointer;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
asm
{$ifndef MSWINDOWS}
mov rcx, size
{$else}
push rsi
push rdi
{$endif MSWINDOWS}
push rbx
// Since most allocations are for small blocks, determine small block type
lea rbx, [rip + SmallBlockInfo]
@VoidSizeToSomething:
lea rdx, [size + BlockHeaderSize - 1]
shr rdx, 4 // div SmallBlockGranularity
// Is it a tiny/small block?
cmp size, (MaximumSmallBlockSize - BlockHeaderSize)
ja @NotTinySmallBlock
test size, size
jz @VoidSize
{$ifndef FPCMM_ASSUMEMULTITHREAD}
mov rax, qword ptr [rbx].TSmallBlockInfo.IsMultiThreadPtr
{$endif FPCMM_ASSUMEMULTITHREAD}
// Get the tiny/small TSmallBlockType[] offset in rcx
movzx ecx, byte ptr [rbx + rdx].TSmallBlockInfo.GetmemLookup
mov r8, rbx
shl ecx, SmallBlockTypePO2
// ---------- Acquire block type lock ----------
{$ifndef FPCMM_ASSUMEMULTITHREAD}
cmp byte ptr [rax], false
je @GotLockOnSmallBlock // no lock if IsMultiThread=false
{$endif FPCMM_ASSUMEMULTITHREAD}
// Can use one of the several arenas reserved for tiny blocks?
cmp ecx, SizeOf(TTinyBlockTypes)
jae @NotTinyBlockType
// ---------- TINY (size<=128B) block lock ----------
@LockTinyBlockTypeLoop:
// Round-Robin attempt to lock next SmallBlockInfo.Tiny[]
// -> fair distribution among calls to reduce thread contention
mov dl, NumTinyBlockArenas + 1 // 8/16 arenas (including Small[])
@TinyBlockArenaLoop:
mov eax, SizeOf(TTinyBlockTypes)
// note: "lock xadd" decreases the loop iterations but is slower
xadd dword ptr [r8 + TSmallBlockInfo.TinyCurrentArena], eax
lea rbx, [r8 + rcx]
and eax, ((NumTinyBlockArenas + 1) * SizeOf(TTinyBlockTypes)) - 1
jz @TinySmall // Arena 0 = TSmallBlockInfo.Small[]
lea rbx, [rax + rbx + TSmallBlockInfo.Tiny - SizeOf(TTinyBlockTypes)]
@TinySmall:
mov eax, $100
{$ifdef FPCMM_CMPBEFORELOCK}
cmp byte ptr [rbx].TSmallBlockType.Locked, false
jnz @NextTinyBlockArena1
{$endif FPCMM_CMPBEFORELOCK}
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @GotLockOnSmallBlockType
@NextTinyBlockArena1:
dec dl
jnz @TinyBlockArenaLoop
// Fallback to SmallBlockInfo.Small[] next 2 small sizes - never occurs
lea rbx, [r8 + rcx + TSmallBlockInfo.Small + SizeOf(TSmallBlockType)]
mov eax, $100
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @GotLockOnSmallBlockType
add rbx, SizeOf(TSmallBlockType) // next two small sizes
mov eax, $100
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @GotLockOnSmallBlockType
// Thread Contention (_Freemem is more likely)
{$ifdef FPCMM_DEBUG} lock {$endif}
inc dword ptr [rbx].TSmallBlockType.GetmemSleepCount
push r8
push rcx
call ReleaseCore
pop rcx
pop r8
jmp @LockTinyBlockTypeLoop
// ---------- SMALL (size<2600) block lock ----------
@NotTinyBlockType:
lea rbx, [r8 + rcx].TSmallBlockInfo.Small
@LockBlockTypeLoopRetry:
{$ifdef FPCMM_PAUSE}
{$ifdef FPCMM_SLEEPTSC}
rdtsc
shl rdx, 32
lea r9, [rax + rdx + SpinSmallGetmemLockTSC] // r9 = endtsc
{$else}
mov edx, SpinSmallGetmemLockCount
{$endif FPCMM_SLEEPTSC}
{$endif FPCMM_PAUSE}
@LockBlockTypeLoop:
// Grab the default block type
mov eax, $100
{$ifdef FPCMM_CMPBEFORELOCK}
cmp byte ptr [rbx].TSmallBlockType.Locked, false
jnz @NextLockBlockType1
{$endif FPCMM_CMPBEFORELOCK}
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @GotLockOnSmallBlockType
// Try up to two next sizes
mov eax, $100
@NextLockBlockType1:
add rbx, SizeOf(TSmallBlockType)
{$ifdef FPCMM_CMPBEFORELOCK}
cmp byte ptr [rbx].TSmallBlockType.Locked, al
jnz @NextLockBlockType2
{$endif FPCMM_CMPBEFORELOCK}
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @GotLockOnSmallBlockType
mov eax, $100
@NextLockBlockType2:
add rbx, SizeOf(TSmallBlockType)
pause
{$ifdef FPCMM_CMPBEFORELOCK}
cmp byte ptr [rbx].TSmallBlockType.Locked, al
jnz @NextLockBlockType3
{$endif FPCMM_CMPBEFORELOCK}
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @GotLockOnSmallBlockType
@NextLockBlockType3:
sub rbx, 2 * SizeOf(TSmallBlockType)
{$ifdef FPCMM_PAUSE}
pause
{$ifdef FPCMM_SLEEPTSC}
rdtsc
shl rdx, 32
or rax, rdx
cmp rax, r9
jb @LockBlockTypeLoop // continue spinning until timeout
{$else}
dec edx
jnz @LockBlockTypeLoop // continue until spin count reached
{$endif FPCMM_SLEEPTSC}
{$endif FPCMM_PAUSE}
// Block type and two sizes larger are all locked - give up and sleep
{$ifdef FPCMM_DEBUG} lock {$endif}
inc dword ptr [rbx].TSmallBlockType.GetmemSleepCount
call ReleaseCore
jmp @LockBlockTypeLoopRetry
// ---------- TINY/SMALL block registration ----------
{$ifndef FPCMM_ASSUMEMULTITHREAD}
@GotLockOnSmallBlock:
add rbx, rcx
{$endif FPCMM_ASSUMEMULTITHREAD}
@GotLockOnSmallBlockType:
// set rdx=NextPartiallyFreePool rax=FirstFreeBlock rcx=DropSmallFlagsMask
mov rdx, [rbx].TSmallBlockType.NextPartiallyFreePool
add [rbx].TSmallBlockType.GetmemCount, 1
mov rax, [rdx].TSmallBlockPoolHeader.FirstFreeBlock
mov rcx, DropSmallFlagsMask
// Is there a pool with free blocks?
cmp rdx, rbx
je @TrySmallSequentialFeed
add [rdx].TSmallBlockPoolHeader.BlocksInUse, 1
// Set the new first free block and the block header
and rcx, [rax - BlockHeaderSize]
mov [rdx].TSmallBlockPoolHeader.FirstFreeBlock, rcx
mov [rax - BlockHeaderSize], rdx
// Is the chunk now full?
jz @RemoveSmallPool
// Unlock the block type and leave
mov byte ptr [rbx].TSmallBlockType.Locked, false
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@VoidSize:
inc size // "we always need to allocate something" (see RTL heap.inc)
jmp @VoidSizeToSomething
@TrySmallSequentialFeed:
// Feed a small block sequentially
movzx ecx, [rbx].TSmallBlockType.BlockSize
mov rdx, [rbx].TSmallBlockType.CurrentSequentialFeedPool
add rcx, rax
// Can another block fit?
cmp rax, [rbx].TSmallBlockType.MaxSequentialFeedBlockAddress
ja @AllocateSmallBlockPool
// Adjust number of used blocks and sequential feed pool
mov [rbx].TSmallBlockType.NextSequentialFeedBlockAddress, rcx
add [rdx].TSmallBlockPoolHeader.BlocksInUse, 1
// Unlock the block type, set the block header and leave
mov byte ptr [rbx].TSmallBlockType.Locked, false
mov [rax - BlockHeaderSize], rdx
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@RemoveSmallPool:
// Pool is full - remove it from the partially free list
mov rcx, [rdx].TSmallBlockPoolHeader.NextPartiallyFreePool
mov [rcx].TSmallBlockPoolHeader.PreviousPartiallyFreePool, rbx
mov [rbx].TSmallBlockType.NextPartiallyFreePool, rcx
// Unlock the block type and leave
mov byte ptr [rbx].TSmallBlockType.Locked, false
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@AllocateSmallBlockPool:
// Access shared information about Medium blocks storage
lea rcx, [rip + SmallMediumBlockInfo]
mov r10, rcx
{$ifndef FPCMM_ASSUMEMULTITHREAD}
mov rax, [rcx + TMediumBlockinfo.IsMultiThreadPtr]
cmp byte ptr [rax], false
je @MediumLocked1 // no lock if IsMultiThread=false
{$endif FPCMM_ASSUMEMULTITHREAD}
mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumBlockInfo.Locked, ah
je @MediumLocked1
call LockMediumBlocks
@MediumLocked1:
// Are there any available blocks of a suitable size?
movsx esi, [rbx].TSmallBlockType.AllowedGroupsForBlockPoolBitmap
and esi, [r10 + TMediumBlockInfo.BinGroupBitmap]
jz @NoSuitableMediumBlocks
// Compute rax = bin group number with free blocks, rcx = bin number
bsf eax, esi
lea r9, [rax * 4]
mov ecx, [r10 + TMediumBlockInfo.BinBitmaps + r9]
bsf ecx, ecx
lea rcx, [rcx + r9 * 8]
// Set rdi = @bin, rsi = free block
lea rsi, [rcx * 8] // SizeOf(TMediumBlockBin) = 16
lea rdi, [r10 + TMediumBlockInfo.Bins + rsi * 2]
mov rsi, TMediumFreeBlock[rdi].NextFreeBlock
// Remove the first block from the linked list (LIFO)
mov rdx, TMediumFreeBlock[rsi].NextFreeBlock
mov TMediumFreeBlock[rdi].NextFreeBlock, rdx
mov TMediumFreeBlock[rdx].PreviousFreeBlock, rdi
// Is this bin now empty?
cmp rdi, rdx
jne @MediumBinNotEmpty
// rbx = block type, rax = bin group number,
// r9 = bin group number * 4, rcx = bin number, rdi = @bin, rsi = free block
// Flag this bin (and the group if needed) as empty
mov edx, - 2
mov r11d, [r10 + TMediumBlockInfo.BinGroupBitmap]
rol edx, cl
btr r11d, eax // btr reg,reg is faster than btr [mem],reg
and [r10 + TMediumBlockInfo.BinBitmaps + r9], edx
jnz @MediumBinNotEmpty
mov [r10 + TMediumBlockInfo.BinGroupBitmap], r11d
@MediumBinNotEmpty:
// rsi = free block, rbx = block type
// Get the size of the available medium block in edi
mov rdi, DropMediumAndLargeFlagsMask
and rdi, [rsi - BlockHeaderSize]
cmp edi, MaximumSmallBlockPoolSize
jb @UseWholeBlock
// Split the block: new block size is the optimal size
mov edx, edi
movzx edi, [rbx].TSmallBlockType.OptimalBlockPoolSize
sub edx, edi
lea rcx, [rsi + rdi]
lea rax, [rdx + IsMediumBlockFlag + IsFreeBlockFlag]
mov [rcx - BlockHeaderSize], rax
// Store the size of the second split as the second last pointer
mov [rcx + rdx - 16], rdx
// Put the remainder in a bin (it will be big enough)
call InsertMediumBlockIntoBin // rcx=APMediumFreeBlock, edx=AMediumBlockSize
jmp @GotMediumBlock
@NoSuitableMediumBlocks:
// Check the sequential feed medium block pool for space
movzx ecx, [rbx].TSmallBlockType.MinimumBlockPoolSize
mov edi, [r10 + TMediumBlockInfo.SequentialFeedBytesLeft]
cmp edi, ecx
jb @AllocateNewSequentialFeed
// Get the address of the last block that was fed
mov rsi, [r10 + TMediumBlockInfo.LastSequentiallyFed]
// Enough sequential feed space: Will the remainder be usable?
movzx ecx, [rbx].TSmallBlockType.OptimalBlockPoolSize
lea rdx, [rcx + MinimumMediumBlockSize]
cmp edi, edx
cmovae edi, ecx
sub rsi, rdi
// Update the sequential feed parameters
sub [r10 + TMediumBlockInfo.SequentialFeedBytesLeft], edi
mov [r10 + TMediumBlockInfo.LastSequentiallyFed], rsi
jmp @GotMediumBlock
@AllocateNewSequentialFeed:
// Use the optimal size for allocating this small block pool
{$ifdef MSWINDOWS}
movzx ecx, word ptr [rbx].TSmallBlockType.OptimalBlockPoolSize
lea rdx, [rip + SmallMediumBlockInfo]
push rcx
push rdx
{$else}
movzx edi, word ptr [rbx].TSmallBlockType.OptimalBlockPoolSize
lea rsi, [rip + SmallMediumBlockInfo]
push rdi
push rsi
{$endif MSWINDOWS}
call AllocNewSequentialFeedMediumPool
pop r10
pop rdi // restore edi=blocksize and r10=TMediumBlockInfo
mov rsi, rax
test rax, rax
jnz @GotMediumBlock // rsi=freeblock rbx=blocktype edi=blocksize
mov [r10 + TMediumBlockInfo.Locked], al
mov [rbx].TSmallBlockType.Locked, al
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@UseWholeBlock:
// rsi = free block, rbx = block type, edi = block size
// Mark this block as used in the block following it
and byte ptr [rsi + rdi - BlockHeaderSize], NOT PreviousMediumBlockIsFreeFlag
@GotMediumBlock:
// rsi = free block, rbx = block type, edi = block size
// Set the size and flags for this block
lea rcx, [rdi + IsMediumBlockFlag + IsSmallBlockPoolInUseFlag]
mov [rsi - BlockHeaderSize], rcx
// Unlock medium blocks and setup the block pool
xor eax, eax
mov [r10 + TMediumBlockInfo.Locked], al
mov TSmallBlockPoolHeader[rsi].BlockType, rbx
mov TSmallBlockPoolHeader[rsi].FirstFreeBlock, rax
mov TSmallBlockPoolHeader[rsi].BlocksInUse, 1
mov [rbx].TSmallBlockType.CurrentSequentialFeedPool, rsi
// Return the pointer to the first block, compute next/last block addresses
lea rax, [rsi + SmallBlockPoolHeaderSize]
movzx ecx, [rbx].TSmallBlockType.BlockSize
lea rdx, [rax + rcx]
mov [rbx].TSmallBlockType.NextSequentialFeedBlockAddress, rdx
add rdi, rsi
sub rdi, rcx
mov [rbx].TSmallBlockType.MaxSequentialFeedBlockAddress, rdi
// Unlock the small block type, set header and leave
mov byte ptr [rbx].TSmallBlockType.Locked, false
mov [rax - BlockHeaderSize], rsi
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
// ---------- MEDIUM block allocation ----------
@NotTinySmallBlock:
// Do we need a Large block?
lea r10, [rip + MediumBlockInfo]
cmp rcx, MaximumMediumBlockSize - BlockHeaderSize
ja @IsALargeBlockRequest
// Get the bin size for this block size (rounded up to the next bin size)
lea rbx, [rcx + MediumBlockGranularity - 1 + BlockHeaderSize - MediumBlockSizeOffset]
mov rcx, r10
and ebx, - MediumBlockGranularity
add ebx, MediumBlockSizeOffset
{$ifndef FPCMM_ASSUMEMULTITHREAD}
mov rax, [r10 + TMediumBlockinfo.IsMultiThreadPtr]
cmp byte ptr [rax], false
je @MediumLocked2 // no lock if IsMultiThread=false
{$endif FPCMM_ASSUMEMULTITHREAD}
mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumBlockInfo.Locked, ah
je @MediumLocked2
call LockMediumBlocks
@MediumLocked2:
// Compute ecx = bin number in ecx and edx = group number
lea rdx, [rbx - MinimumMediumBlockSize]
mov ecx, edx
shr edx, 8 + 5
shr ecx, 8
mov eax, -1
shl eax, cl
and eax, [r10 + TMediumBlockInfo.BinBitmaps + rdx * 4]
jz @GroupIsEmpty
and ecx, - 32
bsf eax, eax
or ecx, eax
jmp @GotBinAndGroup
@GroupIsEmpty:
// Try all groups greater than this group
mov eax, - 2
mov ecx, edx
shl eax, cl
and eax, [r10 + TMediumBlockInfo.BinGroupBitmap]
jz @TrySequentialFeedMedium
// There is a suitable group with enough space
bsf edx, eax
mov eax, [r10 + TMediumBlockInfo.BinBitmaps + rdx * 4]
bsf ecx, eax
mov eax, edx
shl eax, 5
or ecx, eax
jmp @GotBinAndGroup
@TrySequentialFeedMedium:
mov ecx, [r10 + TMediumBlockInfo.SequentialFeedBytesLeft]
// Can block be fed sequentially?
sub ecx, ebx
jc @AllocateNewSequentialFeedForMedium
// Get the block address, store remaining bytes, set the flags and unlock
mov rax, [r10 + TMediumBlockInfo.LastSequentiallyFed]
sub rax, rbx
mov [r10 + TMediumBlockInfo.LastSequentiallyFed], rax
mov [r10 + TMediumBlockInfo.SequentialFeedBytesLeft], ecx
or rbx, IsMediumBlockFlag
mov [rax - BlockHeaderSize], rbx
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@AllocateNewSequentialFeedForMedium:
{$ifdef MSWINDOWS}
mov ecx, ebx
lea rdx, [rip + MediumBlockInfo]
{$else}
mov edi, ebx
lea rsi, [rip + MediumBlockInfo]
{$endif MSWINDOWS}
call AllocNewSequentialFeedMediumPool
mov byte ptr [rip + MediumBlockInfo.Locked], false
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@GotBinAndGroup:
// ebx = block size, ecx = bin number, edx = group number
// Compute rdi = @bin, rsi = free block
lea rax, [rcx + rcx]
lea rdi, [r10 + TMediumBlockInfo.Bins + rax * 8]
mov rsi, TMediumFreeBlock[rdi].NextFreeBlock
// Remove the first block from the linked list (LIFO)
mov rax, TMediumFreeBlock[rsi].NextFreeBlock
mov TMediumFreeBlock[rdi].NextFreeBlock, rax
mov TMediumFreeBlock[rax].PreviousFreeBlock, rdi
// Is this bin now empty?
cmp rdi, rax
jne @MediumBinNotEmptyForMedium
// edx=bingroupnumber, ecx=binnumber, rdi=@bin, rsi=freeblock, ebx=blocksize
// Flag this bin and group as empty
mov eax, - 2
mov r11d, [r10 + TMediumBlockInfo.BinGroupBitmap]
rol eax, cl
btr r11d, edx // btr reg,reg is faster than btr [mem],reg
and [r10 + TMediumBlockInfo.BinBitmaps + rdx * 4], eax
jnz @MediumBinNotEmptyForMedium
mov [r10 + TMediumBlockInfo.BinGroupBitmap], r11d
@MediumBinNotEmptyForMedium:
// rsi = free block, ebx = block size
// Get rdi = size of the available medium block, rdx = second split size
mov rdi, DropMediumAndLargeFlagsMask
and rdi, [rsi - BlockHeaderSize]
mov edx, edi
sub edx, ebx
jz @UseWholeBlockForMedium
// Split the block in two
lea rcx, [rsi + rbx]
lea rax, [rdx + IsMediumBlockFlag + IsFreeBlockFlag]
mov [rcx - BlockHeaderSize], rax
// Store the size of the second split as the second last pointer
mov [rcx + rdx - 16], rdx
// Put the remainder in a bin
cmp edx, MinimumMediumBlockSize
jb @GotMediumBlockForMedium
call InsertMediumBlockIntoBin // rcx=APMediumFreeBlock, edx=AMediumBlockSize
jmp @GotMediumBlockForMedium
@UseWholeBlockForMedium:
// Mark this block as used in the block following it
and byte ptr [rsi + rdi - BlockHeaderSize], NOT PreviousMediumBlockIsFreeFlag
@GotMediumBlockForMedium:
// Set the size and flags for this block
lea rcx, [rbx + IsMediumBlockFlag]
mov [rsi - BlockHeaderSize], rcx
// Unlock medium blocks and leave
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
mov rax, rsi
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
// ---------- LARGE block allocation ----------
@IsALargeBlockRequest:
xor rax, rax
test rcx, rcx
js @Done
// Note: size is still in the rcx/rdi first param register
call AllocateLargeBlock
@Done: // restore registers and the stack frame before ret
pop rbx
{$ifdef MSWINDOWS}
pop rdi
pop rsi
{$endif MSWINDOWS}
end;
function FreeMediumBlock(arg1: pointer): PtrUInt;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
// rcx=P rdx=[P-BlockHeaderSize] r10=TMediumBlockInfo
// (arg1 is used only for proper call of pascal functions below on all ABI)
asm
// Drop the flags, and set r11=P rbx=blocksize
and rdx, DropMediumAndLargeFlagsMask
push rbx
push rdx // save blocksize
mov rbx, rdx
mov r11, rcx
// Lock the Medium blocks
mov rcx, r10
{$ifndef FPCMM_ASSUMEMULTITHREAD}
mov rax, [r10 + TMediumBlockinfo.IsMultiThreadPtr]
cmp byte ptr [rax], false
je @MediumBlocksLocked // no lock if IsMultiThread=false
{$endif FPCMM_ASSUMEMULTITHREAD}
mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumBlockInfo.Locked, ah
je @MediumBlocksLocked
{$ifdef FPCMM_LOCKLESSFREEMEDIUM}
// locked: try to put r11=P in TMediumBlockInfo.LocklessBin.Instance[]
lea rcx, [rcx].TMediumBlockInfo.LocklessBin
cmp byte ptr [rcx].TMediumLocklessBin.Count, MediumBlockLocklessBinCount
je @DoLock // all slots are filled
mov r9d, SpinFreememBinCount
@BinSp: mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumLocklessBin.Locked, ah
je @BinOk
@BinNo: pause
dec r9d
jnz @BinSp
jmp @DoLock
@BinOk: // we acquired TMediumLocklessBin.Locked
movzx eax, byte ptr [rcx].TMediumLocklessBin.Count
cmp al, MediumBlockLocklessBinCount
je @DoLoc2
add byte ptr [rcx].TMediumLocklessBin.Count, 1
cmp al, byte ptr [rcx].TMediumLocklessBin.MaxCount
jb @max
mov byte ptr [rcx].TMediumLocklessBin.MaxCount, al
@max:
mov [rcx + TMediumLocklessBin.Instance + rax * 8], r11
mov byte ptr [rcx].TMediumLocklessBin.Locked, false
jmp @Quit
@DoLoc2:mov byte ptr [rcx].TMediumLocklessBin.Locked, false
@DoLock:{$endif FPCMM_LOCKLESSFREEMEDIUM}
call LockMediumBlocks
@MediumBlocksLocked:
// Get rcx = next block size and flags
mov rcx, [r11 + rbx - BlockHeaderSize]
// Can we combine this block with the next free block?
test qword ptr [r11 + rbx - BlockHeaderSize], IsFreeBlockFlag
jnz @NextBlockIsFree
// Set the "PreviousIsFree" flag in the next block
or rcx, PreviousMediumBlockIsFreeFlag
mov [r11 + rbx - BlockHeaderSize], rcx
@NextBlockChecked:
// Re-read the flags and try to combine with previous free block
test byte ptr [r11 - BlockHeaderSize], PreviousMediumBlockIsFreeFlag
jnz @PreviousBlockIsFree
@PreviousBlockChecked:
// Check if entire medium block pool is free
cmp ebx, (MediumBlockPoolSize - MediumBlockPoolHeaderSize)
je @EntireMediumPoolFree
@Bin: // Store size of the block, flags and trailing size marker and insert into bin
lea rax, [rbx + IsMediumBlockFlag + IsFreeBlockFlag]
mov [r11 - BlockHeaderSize], rax
mov [r11 + rbx - 16], rbx
mov rcx, r11
mov rdx, rbx
call InsertMediumBlockIntoBin // rcx=P, edx=blocksize
{$ifdef FPCMM_LOCKLESSFREEMEDIUM}
// recycle any pending TMediumLocklessBin.Instance[] pointer
lea rcx, [r10].TMediumBlockInfo.LocklessBin
cmp byte ptr [rcx].TMediumLocklessBin.Count, 0
je @Done
mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumLocklessBin.Locked, ah // just try once
jne @Done
// compute r11=P and rbx=blocksize from pending pointer
movzx eax, byte ptr [rcx].TMediumLocklessBin.Count
dec byte ptr [rcx].TMediumLocklessBin.Count
mov r11, [rcx + TMediumLocklessBin.Instance - 8 + rax * 8]
mov byte ptr [rcx].TMediumLocklessBin.Locked, false
mov rbx, qword ptr [r11 - BlockHeaderSize]
and rbx, DropMediumAndLargeFlagsMask
jmp @MediumBlocksLocked
@Done: {$endif FPCMM_LOCKLESSFREEMEDIUM}
// Unlock medium blocks and leave
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
jmp @Quit
@NextBlockIsFree:
// Get rax = next block address, rbx = end of the block
lea rax, [r11 + rbx]
and rcx, DropMediumAndLargeFlagsMask
add rbx, rcx
// Was the block binned?
cmp rcx, MinimumMediumBlockSize
jb @NextBlockChecked
mov rcx, rax
call RemoveMediumFreeBlock // rcx = APMediumFreeBlock
jmp @NextBlockChecked
@PreviousBlockIsFree:
// Get rcx = size/point of the previous free block, rbx = new block end
mov rcx, [r11 - 16]
sub r11, rcx
add rbx, rcx
// Remove the previous block from the linked list
cmp ecx, MinimumMediumBlockSize
jb @PreviousBlockChecked
mov rcx, r11
call RemoveMediumFreeBlock // rcx = APMediumFreeBlock
jmp @PreviousBlockChecked
@EntireMediumPoolFree:
// Ensure current sequential feed pool is free
cmp dword ptr [r10 + TMediumBlockInfo.SequentialFeedBytesLeft], MediumBlockPoolSize - MediumBlockPoolHeaderSize
jne @MakeEmptyMediumPoolSequentialFeed
// Remove this medium block pool from the linked list stored in its header
sub r11, MediumBlockPoolHeaderSize
mov rax, TMediumBlockPoolHeader[r11].PreviousMediumBlockPoolHeader
mov rdx, TMediumBlockPoolHeader[r11].NextMediumBlockPoolHeader
mov TMediumBlockPoolHeader[rax].NextMediumBlockPoolHeader, rdx
mov TMediumBlockPoolHeader[rdx].PreviousMediumBlockPoolHeader, rax
// Unlock medium blocks and free the block pool
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
mov arg1, r11
call FreeMedium
jmp @Quit
@MakeEmptyMediumPoolSequentialFeed:
// Get rbx = end-marker block, and recycle the current sequential feed pool
lea rbx, [r11 + MediumBlockPoolSize - MediumBlockPoolHeaderSize]
mov arg1, r10
call BinMediumSequentialFeedRemainder
// Set this medium pool up as the new sequential feed pool, unlock and leave
mov qword ptr [rbx - BlockHeaderSize], IsMediumBlockFlag
mov dword ptr [r10 + TMediumBlockInfo.SequentialFeedBytesLeft], MediumBlockPoolSize - MediumBlockPoolHeaderSize
mov [r10 + TMediumBlockInfo.LastSequentiallyFed], rbx
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
@Quit: // restore registers and the stack frame
pop rax // medium block size
pop rbx
end;
{$ifdef FPCMM_REPORTMEMORYLEAKS}
const
/// mark freed blocks with 00000000 BLODLESS marker to track incorrect usage
REPORTMEMORYLEAK_FREEDHEXSPEAK = $B10D1E55;
{$endif FPCMM_REPORTMEMORYLEAKS}
function _FreeMem(P: pointer): PtrUInt;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
asm
{$ifdef FPCMM_REPORTMEMORYLEAKS}
mov eax, REPORTMEMORYLEAK_FREEDHEXSPEAK // 00000000 BLODLESS marker
{$endif FPCMM_REPORTMEMORYLEAKS}
{$ifndef MSWINDOWS}
mov rcx, P
{$endif MSWINDOWS}
test P, P
jz @Quit // void pointer
{$ifdef FPCMM_REPORTMEMORYLEAKS}
mov qword ptr [P], rax // over TObject VMT or string/dynarray header
{$endif FPCMM_REPORTMEMORYLEAKS}
mov rdx, qword ptr [P - BlockHeaderSize]
{$ifdef FPCMM_ASSUMEMULTITHREAD}
mov eax, $100
{$else}
mov rax, qword ptr [rip + SmallBlockInfo].TSmallBlockInfo.IsMultiThreadPtr
{$endif FPCMM_ASSUMEMULTITHREAD}
// Is it a small block in use?
test dl, IsFreeBlockFlag + IsMediumBlockFlag + IsLargeBlockFlag
jnz @NotSmallBlockInUse
// Get the small block type in rbx and try to grab it
push rbx
mov rbx, [rdx].TSmallBlockPoolHeader.BlockType
{$ifdef FPCMM_ASSUMEMULTITHREAD}
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
jne @CheckTinySmallLock
{$else}
cmp byte ptr [rax], false
jne @TinySmallLockLoop // lock if IsMultiThread=true
{$endif FPCMM_ASSUMEMULTITHREAD}
@FreeAndUnlock:
// rbx=TSmallBlockType rcx=P rdx=TSmallBlockPoolHeader
// Adjust number of blocks in use, set rax = old first free block
add [rbx].TSmallBlockType.FreememCount, 1
mov rax, [rdx].TSmallBlockPoolHeader.FirstFreeBlock
sub [rdx].TSmallBlockPoolHeader.BlocksInUse, 1
jz @PoolIsNowEmpty
// Store this as the new first free block
mov [rdx].TSmallBlockPoolHeader.FirstFreeBlock, rcx
// Store the previous first free block as the block header
lea r9, [rax + IsFreeBlockFlag]
mov [rcx - BlockHeaderSize], r9
// Was the pool full?
test rax, rax
jnz @SmallPoolWasNotFull
// Insert the pool back into the linked list if it was full
mov rcx, [rbx].TSmallBlockType.NextPartiallyFreePool
mov [rdx].TSmallBlockPoolHeader.PreviousPartiallyFreePool, rbx
mov [rdx].TSmallBlockPoolHeader.NextPartiallyFreePool, rcx
mov [rcx].TSmallBlockPoolHeader.PreviousPartiallyFreePool, rdx
mov [rbx].TSmallBlockType.NextPartiallyFreePool, rdx
@SmallPoolWasNotFull:
{$ifdef FPCMM_LOCKLESSFREE}
// Try to release all pending bin from this block while we have the lock
cmp byte ptr [rbx].TSmallBlockType.BinCount, 0
jne @ProcessPendingBin
{$endif FPCMM_LOCKLESSFREE}
mov byte ptr [rbx].TSmallBlockType.Locked, false
movzx eax, word ptr [rbx].TSmallBlockType.BlockSize
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@PoolIsNowEmpty:
// FirstFreeBlock=nil means it is the sequential feed pool with a single block
test rax, rax
jz @IsSequentialFeedPool
// Pool is now empty: Remove it from the linked list and free it
mov rax, [rdx].TSmallBlockPoolHeader.PreviousPartiallyFreePool
mov rcx, [rdx].TSmallBlockPoolHeader.NextPartiallyFreePool
mov TSmallBlockPoolHeader[rax].NextPartiallyFreePool, rcx
mov [rcx].TSmallBlockPoolHeader.PreviousPartiallyFreePool, rax
// Is this the sequential feed pool? If so, stop sequential feeding
xor eax, eax
cmp [rbx].TSmallBlockType.CurrentSequentialFeedPool, rdx
jne @NotSequentialFeedPool
@IsSequentialFeedPool:
mov [rbx].TSmallBlockType.MaxSequentialFeedBlockAddress, rax
@NotSequentialFeedPool:
// Unlock blocktype and release this pool
mov byte ptr [rbx].TSmallBlockType.Locked, false
mov rcx, rdx
mov rdx, qword ptr [rdx - BlockHeaderSize]
lea r10, [rip + SmallMediumBlockInfo]
call FreeMediumBlock // no call nor BinLocked to avoid race condition
movzx eax, word ptr [rbx].TSmallBlockType.BlockSize
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
{$ifdef FPCMM_LOCKLESSFREE}
@ProcessPendingBin:
// Try once to acquire BinLocked (spinning may induce race condition)
{$ifdef FPCMM_CMPBEFORELOCK}
cmp byte ptr [rbx].TSmallBlockType.BinLocked, true
je @BinAlreadyLocked
{$endif FPCMM_CMPBEFORELOCK}
mov eax, $100
lock cmpxchg byte ptr [rbx].TSmallBlockType.BinLocked, ah
jne @BinAlreadyLocked
movzx eax, byte ptr [rbx].TSmallBlockType.BinCount
test al, al
jz @NoBin
// free last pointer in TSmallBlockType.BinInstance[]
mov rcx, qword ptr [rbx + TSmallBlockType.BinInstance - 8 + rax * 8]
dec byte ptr [rbx].TSmallBlockType.BinCount
mov byte ptr [rbx].TSmallBlockType.BinLocked, false
mov rdx, [rcx - BlockHeaderSize]
jmp @FreeAndUnlock // loop until BinCount=0 or BinLocked
@NoBin: mov byte ptr [rbx].TSmallBlockType.BinLocked, false
@BinAlreadyLocked:
mov byte ptr [rbx].TSmallBlockType.Locked, false
{$endif FPCMM_LOCKLESSFREE}
movzx eax, word ptr [rbx].TSmallBlockType.BlockSize
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@NotSmallBlockInUse:
lea r10, [rip + MediumBlockInfo]
test dl, IsFreeBlockFlag + IsLargeBlockFlag
jz @DoFreeMedium
call FreeLargeBlock // P is still in rcx/rdi first param register
jmp @Quit
@DoFreeMedium:
call FreeMediumBlock
jmp @Quit
@TinySmallLockLoop:
mov eax, $100
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
je @FreeAndUnlock
@CheckTinySmallLock:
{$ifdef FPCMM_LOCKLESSFREE}
// Try to put rcx=P in TSmallBlockType.BinInstance[]
cmp byte ptr [rbx].TSmallBlockType.BinCount, SmallBlockBinCount
je @LockBlockTypeSleep // wait if all slots are filled
mov eax, $100
lock cmpxchg byte ptr [rbx].TSmallBlockType.BinLocked, ah
je @BinLocked
{$ifdef FPCMM_PAUSE}
{$ifdef FPCMM_SLEEPTSC}
push rdx
rdtsc
shl rdx, 32
lea r9, [rax + rdx + SpinSmallFreememBinTSC] // r9 = endtsc
{$else}
mov r9d, SpinFreememBinCount
{$endif FPCMM_SLEEPTSC}
@SpinBinLock:
pause
{$ifdef FPCMM_SLEEPTSC}
rdtsc
shl rdx, 32
or rax, rdx
cmp rax, r9
ja @SpinTimeout
{$else}
dec r9
jz @SpinTimeout
{$endif FPCMM_SLEEPTSC}
{$ifdef FPCMM_CMPBEFORELOCK_SPIN}
cmp byte ptr [rbx].TSmallBlockType.BinLocked, true
je @SpinBinLock
{$endif FPCMM_CMPBEFORELOCK_SPIN}
mov eax, $100
lock cmpxchg byte ptr [rbx].TSmallBlockType.BinLocked, ah
jne @SpinBinLock
{$ifdef FPCMM_SLEEPTSC}
pop rdx
{$endif FPCMM_SLEEPTSC}
jmp @BinLocked
@SpinTimeout:
{$ifdef FPCMM_SLEEPTSC}
pop rdx
{$endif FPCMM_SLEEPTSC}
{$endif FPCMM_PAUSE}
{$ifdef FPCMM_DEBUG} // no lock (informative only)
inc dword ptr [rbx].TSmallBlockType.BinSpinCount
{$endif FPCMM_DEBUG}
jmp @LockBlockTypeSleep
@BinLocked:
movzx eax, byte ptr [rbx].TSmallBlockType.BinCount
cmp al, SmallBlockBinCount
je @LockBlockType
add byte ptr [rbx].TSmallBlockType.BinCount, 1
mov [rbx + TSmallBlockType.BinInstance + rax * 8], rcx
mov byte ptr [rbx].TSmallBlockType.BinLocked, false
movzx eax, word ptr [rbx].TSmallBlockType.BlockSize
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
@LockBlockType:
// Fallback to main block lock if TSmallBlockType.BinInstance[] is full
mov byte ptr [rbx].TSmallBlockType.BinLocked, false
@LockBlockTypeSleep:
{$endif FPCMM_LOCKLESSFREE}
{$ifdef FPCMM_PAUSE}
// Spin to grab the block type (don't try too long due to contention)
{$ifdef FPCMM_SLEEPTSC}
push rdx
rdtsc
shl rdx, 32
lea r9, [rax + rdx + SpinSmallFreememLockTSC] // r9 = endtsc
@SpinLockBlockType:
pause
rdtsc
shl rdx, 32
or rax, rdx
cmp rax, r9
ja @LockBlockTypeReleaseCore
{$else}
mov r8d, SpinSmallFreememLockCount
@SpinLockBlockType:
pause
dec r8d
jz @LockBlockTypeReleaseCore
{$endif FPCMM_SLEEPTSC}
mov eax, $100
{$ifdef FPCMM_CMPBEFORELOCK_SPIN}
cmp byte ptr [rbx].TSmallBlockType.Locked, true
je @SpinLockBlockType
{$endif FPCMM_CMPBEFORELOCK_SPIN}
lock cmpxchg byte ptr [rbx].TSmallBlockType.Locked, ah
jne @SpinLockBlockType
{$ifdef FPCMM_SLEEPTSC}
pop rdx
{$endif FPCMM_SLEEPTSC}
jmp @FreeAndUnlock
@LockBlockTypeReleaseCore:
{$ifdef FPCMM_SLEEPTSC}
pop rdx
{$endif FPCMM_SLEEPTSC}
{$endif FPCMM_PAUSE}
// Couldn't grab the block type - sleep and try again
{$ifdef FPCMM_DEBUG} lock {$endif}
inc dword ptr [rbx].TSmallBlockType.FreeMemSleepCount
push rdx
push rcx
call ReleaseCore
pop rcx
pop rdx
jmp @TinySmallLockLoop
@Done: // restore rbx and the stack frame before ret
pop rbx
@Quit:
end;
// warning: FPC signature is not the same than Delphi: requires "var P"
function _ReallocMem(var P: pointer; Size: PtrUInt): pointer;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
asm
{$ifdef MSWINDOWS}
push rdi
push rsi
{$else}
mov rdx, Size
{$endif MSWINDOWS}
push rbx
push r14
push P // for assignement in @Done
mov r14, qword ptr [P]
test rdx, rdx
jz @VoidSize // ReallocMem(P,0)=FreeMem(P)
test r14, r14
jz @GetMemMoveFreeMem // ReallocMem(nil,Size)=GetMem(Size)
mov rcx, [r14 - BlockHeaderSize]
test cl, IsFreeBlockFlag + IsMediumBlockFlag + IsLargeBlockFlag
jnz @NotASmallBlock
// -------------- TINY/SMALL block -------------
// Get rbx=blocktype, rcx=available size, rax=inplaceresize
mov rbx, [rcx].TSmallBlockPoolHeader.BlockType
lea rax, [rdx * 4 + SmallBlockDownsizeCheckAdder]
movzx ecx, [rbx].TSmallBlockType.BlockSize
sub ecx, BlockHeaderSize
cmp rcx, rdx
jb @SmallUpsize
// Downsize or small growup with enough space: reallocate only if need
cmp eax, ecx
jb @GetMemMoveFreeMem // r14=P rdx=size
@NoResize:
// branchless execution if current block is good enough for this size
mov rax, r14 // keep original pointer
pop rcx
{$ifdef NOSFRAME}
pop r14
pop rbx
ret
{$else}
jmp @Quit // on Win64, a stack frame is required
{$endif NOSFRAME}
@VoidSize:
push rdx // to set P=nil
jmp @DoFree // ReallocMem(P,0)=FreeMem(P)
@SmallUpsize:
// State: r14=pointer, rdx=NewSize, rcx=CurrentBlockSize, rbx=CurrentBlockType
// Small blocks always grow with at least 100% + SmallBlockUpsizeAdder bytes
lea P, qword ptr [rcx * 2 + SmallBlockUpsizeAdder]
movzx ebx, [rbx].TSmallBlockType.BlockSize
sub ebx, BlockHeaderSize + 8
// r14=pointer, P=NextUpBlockSize, rdx=NewSize, rbx=OldSize-8
@AdjustGetMemMoveFreeMem:
// New allocated size is max(requestedsize, minimumupsize)
cmp rdx, P
cmova P, rdx
push rdx
call _GetMem
pop rdx
test rax, rax
jz @Done
jmp @MoveFreeMem // rax=New r14=P rbx=size-8
@GetMemMoveFreeMem:
// reallocate copy and free: r14=P rdx=size
mov rbx, rdx
mov P, rdx // P is the proper first argument register
call _GetMem
test rax, rax
jz @Done
test r14, r14 // ReallocMem(nil,Size)=GetMem(Size)
jz @Done
sub rbx, 8
@MoveFreeMem:
// copy and free: rax=New r14=P rbx=size-8
push rax
{$ifdef FPCMM_ERMS}
cmp rbx, ErmsMinSize // startup cost of 0..255 bytes
jae @erms
{$endif FPCMM_ERMS}
lea rcx, [r14 + rbx]
lea rdx, [rax + rbx]
neg rbx
jns @Last8
align 16
@By16: movaps xmm0, oword ptr [rcx + rbx]
movaps oword ptr [rdx + rbx], xmm0
add rbx, 16
js @By16
@Last8: mov rax, qword ptr [rcx + rbx]
mov qword ptr [rdx + rbx], rax
@DoFree:mov P, r14
call _FreeMem
pop rax
jmp @Done
{$ifdef FPCMM_ERMS}
@erms: cld
mov rsi, r14
mov rdi, rax
lea rcx, [rbx + 8]
rep movsb
jmp @DoFree
{$endif FPCMM_ERMS}
@NotASmallBlock:
// Is this a medium block or a large block?
test cl, IsFreeBlockFlag + IsLargeBlockFlag
jnz @PossibleLargeBlock
// -------------- MEDIUM block -------------
// rcx=CurrentSize+Flags, r14=P, rdx=RequestedSize, r10=TMediumBlockInfo
lea rsi, [rdx + rdx]
lea r10, [rip + MediumBlockInfo]
mov rbx, rcx
and ecx, DropMediumAndLargeFlagsMask
lea rdi, [r14 + rcx]
sub ecx, BlockHeaderSize
and ebx, ExtractMediumAndLargeFlagsMask
// Is it an upsize or a downsize?
cmp rdx, rcx
ja @MediumBlockUpsize
// rcx=CurrentBlockSize-BlockHeaderSize, rbx=CurrentBlockFlags,
// rdi=@NextBlock, r14=P, rdx=RequestedSize
// Downsize reallocate and move data only if less than half the current size
cmp rsi, rcx
jae @NoResize
// In-place downsize? Ensure not smaller than MinimumMediumBlockSize
cmp edx, MinimumMediumBlockSize - BlockHeaderSize
jae @MediumBlockInPlaceDownsize
// Need to move to another Medium block pool, or into a Small block?
cmp edx, MediumInPlaceDownsizeLimit
jb @GetMemMoveFreeMem
// No need to realloc: resize in-place (if not already at the minimum size)
mov edx, MinimumMediumBlockSize - BlockHeaderSize
cmp ecx, MinimumMediumBlockSize - BlockHeaderSize
jna @NoResize
@MediumBlockInPlaceDownsize:
// Round up to the next medium block size
lea rsi, [rdx + BlockHeaderSize + MediumBlockGranularity - 1 - MediumBlockSizeOffset]
and rsi, - MediumBlockGranularity
add rsi, MediumBlockSizeOffset
// Get the size of the second split
add ecx, BlockHeaderSize
sub ecx, esi
mov ebx, ecx
// Lock the medium blocks
mov rcx, r10
{$ifndef FPCMM_ASSUMEMULTITHREAD}
mov rax, [r10 + TMediumBlockinfo.IsMultiThreadPtr]
cmp byte ptr [rax], false
je @MediumBlocksLocked1 // no lock if IsMultiThread=false
{$endif FPCMM_ASSUMEMULTITHREAD}
mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumBlockInfo.Locked, ah
je @MediumBlocksLocked1
call LockMediumBlocks
@MediumBlocksLocked1:
mov ecx, ebx
// Reread the flags - may have changed before medium blocks could be locked
mov rbx, ExtractMediumAndLargeFlagsMask
and rbx, [r14 - BlockHeaderSize]
@DoMediumInPlaceDownsize:
// Set the new size in header, and get rbx = second split size
or rbx, rsi
mov [r14 - BlockHeaderSize], rbx
mov ebx, ecx
// If the next block is used, flag its previous block as free
mov rdx, [rdi - BlockHeaderSize]
test dl, IsFreeBlockFlag
jnz @MediumDownsizeNextBlockFree
or rdx, PreviousMediumBlockIsFreeFlag
mov [rdi - BlockHeaderSize], rdx
jmp @MediumDownsizeDoSplit
@MediumDownsizeNextBlockFree:
// If the next block is free, combine both
mov rcx, rdi
and rdx, DropMediumAndLargeFlagsMask
add rbx, rdx
add rdi, rdx
cmp edx, MinimumMediumBlockSize
jb @MediumDownsizeDoSplit
call RemoveMediumFreeBlock // rcx=APMediumFreeBlock
@MediumDownsizeDoSplit:
// Store the trailing size field and free part header
mov [rdi - 16], rbx
lea rcx, [rbx + IsMediumBlockFlag + IsFreeBlockFlag];
mov [r14 + rsi - BlockHeaderSize], rcx
// Bin this free block (if worth it)
cmp rbx, MinimumMediumBlockSize
jb @MediumBlockDownsizeDone
lea rcx, [r14 + rsi]
mov rdx, rbx
call InsertMediumBlockIntoBin // rcx=APMediumFreeBlock, edx=AMediumBlockSize
@MediumBlockDownsizeDone:
// Unlock the medium blocks, and leave with the new pointer
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
mov rax, r14
jmp @Done
@MediumBlockUpsize:
// ecx = Current Block Size - BlockHeaderSize, bl = Current Block Flags,
// rdi = @Next Block, r14 = P, rdx = Requested Size
// Try to make in-place upsize
mov rax, [rdi - BlockHeaderSize]
test al, IsFreeBlockFlag
jz @CannotUpsizeMediumBlockInPlace
// Get rax = available size, rsi = available size with the next block
and rax, DropMediumAndLargeFlagsMask
lea rsi, [rax + rcx]
cmp rdx, rsi
ja @CannotUpsizeMediumBlockInPlace
// Grow into the next block
mov rbx, rcx
mov rcx, r10
{$ifndef FPCMM_ASSUMEMULTITHREAD}
mov rax, [r10 + TMediumBlockinfo.IsMultiThreadPtr]
cmp byte ptr [rax], false
je @MediumBlocksLocked2 // no lock if IsMultiThread=false
{$endif FPCMM_ASSUMEMULTITHREAD}
mov eax, $100
lock cmpxchg byte ptr [rcx].TMediumBlockInfo.Locked, ah
je @MediumBlocksLocked2
mov rsi, rdx
call LockMediumBlocks
mov rdx, rsi
@MediumBlocksLocked2:
// Re-read info once locked, and ensure next block is still free
mov rcx, rbx
mov rbx, ExtractMediumAndLargeFlagsMask
and rbx, [r14 - BlockHeaderSize]
mov rax, [rdi - BlockHeaderSize]
test al, IsFreeBlockFlag
jz @NextMediumBlockChanged
and eax, DropMediumAndLargeFlagsMask
lea rsi, [rax + rcx]
cmp rdx, rsi
ja @NextMediumBlockChanged
@DoMediumInPlaceUpsize:
// Bin next free block (if worth it)
cmp eax, MinimumMediumBlockSize
jb @MediumInPlaceNoNextRemove
push rcx
push rdx
mov rcx, rdi
call RemoveMediumFreeBlock // rcx=APMediumFreeBlock
pop rdx
pop rcx
@MediumInPlaceNoNextRemove:
// Medium blocks grow a minimum of 25% in in-place upsizes
mov eax, ecx
shr eax, 2
add eax, ecx
// Get the maximum of the requested size and the minimum growth size
xor edi, edi
sub eax, edx
adc edi, -1
and eax, edi
// Round up to the nearest block size granularity
lea rax, [rax + rdx + BlockHeaderSize + MediumBlockGranularity - 1 - MediumBlockSizeOffset]
and eax, -MediumBlockGranularity
add eax, MediumBlockSizeOffset
// Calculate the size of the second split and check if it fits
lea rdx, [rsi + BlockHeaderSize]
sub edx, eax
ja @MediumInPlaceUpsizeSplit
// Grab the whole block: Mark it as used in the next block, and adjust size
and qword ptr [r14 + rsi], NOT PreviousMediumBlockIsFreeFlag
add rsi, BlockHeaderSize
jmp @MediumUpsizeInPlaceDone
@MediumInPlaceUpsizeSplit:
// Store the size of the second split as the second last pointer
mov [r14 + rsi - BlockHeaderSize], rdx
// Set the second split header
lea rdi, [rdx + IsMediumBlockFlag + IsFreeBlockFlag]
mov [r14 + rax - BlockHeaderSize], rdi
mov rsi, rax
cmp edx, MinimumMediumBlockSize
jb @MediumUpsizeInPlaceDone
lea rcx, [r14 + rax]
call InsertMediumBlockIntoBin // rcx=APMediumFreeBlock, edx=AMediumBlockSize
@MediumUpsizeInPlaceDone:
// No need to move data at upsize: set the size and flags for this block
or rsi, rbx
mov [r14 - BlockHeaderSize], rsi
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
mov rax, r14
jmp @Done
@NextMediumBlockChanged:
// The next block changed during lock: reallocate and move data
mov byte ptr [r10 + TMediumBlockInfo.Locked], false
@CannotUpsizeMediumBlockInPlace:
// rcx=OldSize-8, rdx=NewSize
mov rbx, rcx
mov eax, ecx
shr eax, 2
lea P, qword ptr [rcx + rax] // NextUpBlockSize = OldSize+25%
jmp @AdjustGetMemMoveFreeMem // P=BlockSize, rdx=NewSize, rbx=OldSize-8
@PossibleLargeBlock:
// -------------- LARGE block -------------
test cl, IsFreeBlockFlag + IsMediumBlockFlag
jnz @Error
{$ifdef MSWINDOWS}
mov rcx, r14
{$else}
mov rdi, r14
mov rsi, rdx
{$endif MSWINDOWS}
call ReallocateLargeBlock // with restored proper registers
jmp @Done
@Error: xor eax, eax
@Done: // restore registers and the stack frame before ret
pop rcx
mov qword ptr [rcx], rax // store new pointer in var P
@Quit: pop r14
pop rbx
{$ifdef MSWINDOWS}
pop rsi
pop rdi
{$endif MSWINDOWS}
end;
function _AllocMem(Size: PtrUInt): pointer;
{$ifdef NOSFRAME} nostackframe; {$endif} assembler;
asm
push rbx
// Compute rbx = size rounded down to the last pointer
lea rbx, [Size - 1]
and rbx, - 8
// Perform the memory allocation
call _GetMem
// Could a block be allocated? rcx = 0 if yes, -1 if no
cmp rax, 1
sbb rcx, rcx
// Point rdx to the last pointer
lea rdx, [rax + rbx]
// Compute Size (1..8 doesn't need to enter the SSE2 loop)
or rbx, rcx
jz @LastQ
// Large blocks from mmap/VirtualAlloc are already zero filled
cmp rbx, MaximumMediumBlockSize - BlockHeaderSize
jae @Done
{$ifdef FPCMM_ERMS}
cmp rbx, ErmsMinSize // startup cost of 0..255 bytes
jae @erms
{$endif FPCMM_ERMS}
neg rbx
pxor xmm0, xmm0
align 16
@FillLoop: // non-temporal movntdq not needed with small/medium size
movaps oword ptr [rdx + rbx], xmm0
add rbx, 16
js @FillLoop
// fill the last pointer
@LastQ: xor rcx, rcx
mov qword ptr [rdx], rcx
{$ifdef FPCMM_ERMS}
{$ifdef NOSFRAME}
pop rbx
ret
{$else}
jmp @Done // on Win64, a stack frame is required
{$endif NOSFRAME}
// ERMS has a startup cost, but "rep stosd" is fast enough on all CPUs
@erms: mov rcx, rbx
push rax
{$ifdef MSWINDOWS}
push rdi
{$endif MSWINDOWS}
cld
mov rdi, rdx
xor eax, eax
sub rdi, rbx
shr ecx, 2
mov qword ptr [rdx], rax
rep stosd
{$ifdef MSWINDOWS}
pop rdi
{$endif MSWINDOWS}
pop rax
{$endif FPCMM_ERMS}
@Done: // restore rbx register and the stack frame before ret
pop rbx
end;
function _MemSize(P: pointer): PtrUInt;
begin
// AFAIK used only by fpc_AnsiStr_SetLength() in FPC RTL
// also used by our static SQLite3 for its xSize() callback
P := PPointer(PByte(P) - BlockHeaderSize)^;
if (PtrUInt(P) and (IsMediumBlockFlag or IsLargeBlockFlag)) = 0 then
result := PSmallBlockPoolHeader(PtrUInt(P) and DropSmallFlagsMask).
BlockType.BlockSize - BlockHeaderSize
else
begin
result := (PtrUInt(P) and DropMediumAndLargeFlagsMask) - BlockHeaderSize;
if (PtrUInt(P) and IsMediumBlockFlag) = 0 then
dec(result, LargeBlockHeaderSize);
end;
end;
function _FreeMemSize(P: pointer; size: PtrUInt): PtrInt;
begin
// should return the chunk size - only used by heaptrc AFAIK
if (P <> nil) and
(size <> 0) then
result := _FreeMem(P)
else
result := 0;
end;
{ ********* Information Gathering }
{$ifdef FPCMM_STANDALONE}
procedure Assert(flag: boolean);
begin
end;
{$else}
function _GetFPCHeapStatus: TFPCHeapStatus;
var
mm: PMMStatus;
begin
mm := @HeapStatus;
{$ifdef FPCMM_DEBUG}
result.MaxHeapSize := mm^.Medium.PeakBytes + mm^.Large.PeakBytes;
{$else}
result.MaxHeapSize := 0;
{$endif FPCMM_DEBUG}
result.MaxHeapUsed := result.MaxHeapSize;
result.CurrHeapSize := mm^.Medium.CurrentBytes + mm^.Large.CurrentBytes;
result.CurrHeapUsed := result.CurrHeapSize;
result.CurrHeapFree := 0;
end;
function _GetHeapStatus: THeapStatus;
begin
FillChar(result, sizeof(result), 0);
with HeapStatus do
result.TotalAllocated := Medium.CurrentBytes + Large.CurrentBytes;
result.TotalAddrSpace := result.TotalAllocated;
end;
type
// match both TSmallBlockStatus and TSmallBlockContention
TRes = array[0..2] of cardinal;
// details are allocated on the stack, not the heap
TResArray = array[0..(NumSmallInfoBlock * 2) - 1] of TRes;
procedure QuickSortRes(var Res: TResArray; L, R, Level: PtrInt);
var
I, J, P: PtrInt;
pivot: cardinal;
tmp: TRes;
begin
if L < R then
repeat
I := L;
J := R;
P := (L + R) shr 1;
repeat
pivot := Res[P, Level];
while Res[I, Level] > pivot do
inc(I);
while Res[J, Level] < pivot do
dec(J);
if I <= J then
begin
tmp := Res[J];
Res[J] := Res[I];
Res[I] := tmp;
if P = I then
P := J
else if P = J then
P := I;
inc(I);
dec(J);
end;
until I > J;
if J - L < R - I then
begin
// use recursion only for smaller range
if L < J then
QuickSortRes(Res, L, J, Level);
L := I;
end
else
begin
if I < R then
QuickSortRes(Res, I, R, Level);
R := J;
end;
until L >= R;
end;
procedure SetSmallBlockStatus(var res: TResArray; out small, tiny: cardinal);
var
i, a: integer;
p: PSmallBlockType;
d: ^TSmallBlockStatus;
begin
small := 0;
tiny := 0;
d := @res;
p := @SmallBlockInfo;
for i := 1 to NumSmallBlockTypes do
begin
inc(small, ord(p^.GetmemCount <> 0));
d^.Total := p^.GetmemCount;
d^.Current := p^.GetmemCount - p^.FreememCount;
d^.BlockSize := p^.BlockSize;
inc(d);
inc(p);
end;
for a := 1 to NumTinyBlockArenas do
begin
d := @res; // aggregate counters
for i := 1 to NumTinyBlockTypes do
begin
inc(tiny, ord(p^.GetmemCount <> 0));
inc(d^.Total, p^.GetmemCount);
inc(d^.Current, p^.GetmemCount - p^.FreememCount);
inc(d);
inc(p);
end;
end;
assert(p = @SmallBlockInfo.GetmemLookup);
end;
function SortSmallBlockStatus(var res: TResArray; maxcount, orderby: PtrInt;
count, bytes: PPtrUInt): PtrInt;
var
i: PtrInt;
begin
QuickSortRes(res, 0, NumSmallBlockTypes - 1, orderby);
if count <> nil then
begin
count^ := 0;
for i := 0 to NumSmallBlockTypes - 1 do
inc(count^, res[i, orderby]);
end;
if bytes <> nil then
begin
bytes^ := 0;
for i := 0 to NumSmallBlockTypes - 1 do
inc(bytes^, res[i, orderby] * res[i, ord(obBlockSize)]);
end;
result := maxcount;
if result > NumSmallBlockTypes then
result := NumSmallBlockTypes;
while (result > 0) and
(res[result - 1, orderby] = 0) do
dec(result);
end;
function SetSmallBlockContention(var res: TResArray; maxcount: integer): integer;
var
i: integer;
p: PSmallBlockType;
d: ^TSmallBlockContention;
begin
result := 0;
d := @res;
p := @SmallBlockInfo;
for i := 1 to NumSmallInfoBlock do
begin
if p^.GetmemSleepCount <> 0 then
begin
d^.SleepCount := p^.GetmemSleepCount;
d^.GetmemBlockSize := p^.BlockSize;
d^.FreememBlockSize := 0;
inc(d);
inc(result);
end;
if p^.FreememSleepCount <> 0 then
begin
d^.SleepCount := p^.FreememSleepCount;
d^.GetmemBlockSize := 0;
d^.FreememBlockSize := p^.BlockSize;
inc(d);
inc(result);
end;
inc(p);
end;
if result = 0 then
exit;
QuickSortRes(res, 0, result - 1, 0);
if result > maxcount then
result := maxcount;
end;
const
K_: array[0..4] of string[1] = (
'P', 'T', 'G', 'M', 'K');
function K(i: PtrUInt): ShortString;
var
j, n: PtrUInt;
tmp: PShortString;
begin
tmp := nil;
n := 1 shl 50;
for j := 0 to high(K_) do
if i >= n then
begin
i := i div n;
tmp := @K_[j];
break;
end
else
n := n shr 10;
str(i, result);
if tmp <> nil then
result := result + tmp^;
end;
function S(i: PtrUInt): ShortString;
begin
str(i, result);
end;
type
// allow to write into a temp string or the console
TGetHeapStatusWrite =
procedure(const V: array of ShortString; CRLF: boolean = true);
procedure WriteHeapStatusDetail(const arena: TMMStatusArena;
const name: ShortString; Wr: TGetHeapStatusWrite);
begin
Wr([name, K(arena.CurrentBytes),
'B/', K(arena.CumulativeBytes), 'B '], {crlf=}false);
{$ifdef FPCMM_DEBUG}
Wr([' peak=', K(arena.PeakBytes),
'B current=', K(arena.CumulativeAlloc - arena.CumulativeFree),
' alloc=', K(arena.CumulativeAlloc),
' free=', K(arena.CumulativeFree)], false);
{$endif FPCMM_DEBUG}
Wr([' sleep=', K(arena.SleepCount)]);
end;
procedure ComputeHeapStatus(const context: ShortString; smallblockstatuscount,
smallblockcontentioncount: integer; compilationflags: boolean;
Wr: TGetHeapStatusWrite);
var
res: TResArray; // no heap allocation involved
i, n, smallcount: PtrInt;
t, b: PtrUInt;
small, tiny: cardinal;
begin
if context[0] <> #0 then
Wr([context]);
if compilationflags then
Wr([' Flags:' + FPCMM_FLAGS]);
with CurrentHeapStatus do
begin
Wr([' Small: blocks=', K(SmallBlocks),
' size=', K(SmallBlocksSize), 'B (part of Medium arena)']);
WriteHeapStatusDetail(Medium, ' Medium: ', Wr);
WriteHeapStatusDetail(Large, ' Large: ', Wr);
if SleepCount <> 0 then
Wr([' Total Sleep: count=', K(SleepCount)
{$ifdef FPCMM_SLEEPTSC} , ' rdtsc=', K(SleepCycles) {$endif}]);
smallcount := SmallGetmemSleepCount + SmallFreememSleepCount
{$ifdef FPCMM_LOCKLESSFREE} {$ifdef FPCMM_DEBUG}
+ SmallFreememLockLessSpin {$endif} {$endif};
if smallcount <> 0 then
Wr([' Small Sleep: getmem=', K(SmallGetmemSleepCount),
' freemem=', K(SmallFreememSleepCount)
{$ifdef FPCMM_LOCKLESSFREE} {$ifdef FPCMM_DEBUG} ,
' locklessspin=', K(SmallFreememLockLessSpin) {$endif} {$endif}]);
end;
if (smallblockcontentioncount > 0) and
(smallcount <> 0) then
begin
n := SetSmallBlockContention(res, smallblockcontentioncount);
for i := 0 to n - 1 do
with TSmallBlockContention(res[i]) do
begin
if GetmemBlockSize <> 0 then
Wr([' getmem(', S(GetmemBlockSize)], {crlf=}false)
else
Wr([' freemem(', S(FreememBlockSize)], false);
Wr([')=' , K(SleepCount)], false);
if (i and 3 = 3) or
(i = n - 1) then
Wr([]);
end;
end;
if smallblockstatuscount > 0 then
begin
SetSmallBlockStatus(res, small, tiny);
n := SortSmallBlockStatus(res, smallblockstatuscount, ord(obTotal), @t, @b) - 1;
Wr([' Small Blocks since beginning: ', K(t), '/', K(b),
'B (as small=', K(small), '/', S(NumSmallBlockTypes),
' tiny=', K(tiny), '/', S(NumTinyBlockArenas * NumTinyBlockTypes), ')']);
for i := 0 to n do
with TSmallBlockStatus(res[i]) do
begin
Wr([' ', S(BlockSize), '=', K(Total)], false);
if (i and 7 = 7) or
(i = n) then
Wr([]);
end;
n := SortSmallBlockStatus(res, smallblockstatuscount, ord(obCurrent), @t, @b) - 1;
Wr([' Small Blocks current: ', K(t), '/', K(b), 'B']);
for i := 0 to n do
with TSmallBlockStatus(res[i]) do
begin
Wr([' ', S(BlockSize), '=', K(Current)], false);
if (i and 7 = 7) or
(i = n) then
Wr([]);
end;
end;
end;
var
WrStrTemp: string; // we don't require thread safety here
WrStrOnSameLine: boolean;
procedure WrStr(const V: array of ShortString; CRLF: boolean);
var
i: PtrInt;
begin // we don't have format() nor formatutf8() -> this is good enough
for i := 0 to high(V) do
WrStrTemp := WrStrTemp + string(V[i]); // fast enough
if CRLF and
not WrStrOnSameLine then
WrStrTemp := WrStrTemp + #13#10;
end;
function GetHeapStatus(const context: ShortString; smallblockstatuscount,
smallblockcontentioncount: integer; compilationflags, onsameline: boolean): string;
begin
WrStrOnSameLine := onsameline;
ComputeHeapStatus(context, smallblockstatuscount, smallblockcontentioncount,
compilationflags, WrStr);
result := WrStrTemp;
WrStrTemp := '';
end;
procedure WrConsole(const V: array of ShortString; CRLF: boolean);
var
i: PtrInt;
begin // direct write to the console with no memory heap allocation
{$I-}
for i := 0 to high(V) do
write(V[i]);
if CRLF then
writeln;
ioresult;
{$I+}
end;
procedure WriteHeapStatus(const context: ShortString; smallblockstatuscount,
smallblockcontentioncount: integer; compilationflags: boolean);
begin
ComputeHeapStatus(context, smallblockstatuscount, smallblockcontentioncount,
compilationflags, WrConsole);
end;
function GetSmallBlockStatus(maxcount: integer; orderby: TSmallBlockOrderBy;
count, bytes: PPtrUInt; small, tiny: PCardinal): TSmallBlockStatusDynArray;
var
res: TResArray;
sm, ti: cardinal;
begin
assert(SizeOf(TRes) = SizeOf(TSmallBlockStatus));
result := nil;
if maxcount <= 0 then
exit;
SetSmallBlockStatus(res, sm, ti);
if small <> nil then
small^ := sm;
if tiny <> nil then
tiny^ := ti;
maxcount := SortSmallBlockStatus(res, maxcount, ord(orderby), count, bytes);
if maxcount = 0 then
exit;
SetLength(result, maxcount);
Move(res[0], result[0], maxcount * SizeOf(res[0]));
end;
function GetSmallBlockContention(maxcount: integer): TSmallBlockContentionDynArray;
var
n: integer;
res: TResArray;
begin
result := nil;
if maxcount <= 0 then
exit;
n := SetSmallBlockContention(res, maxcount);
if n = 0 then
exit;
SetLength(result, n);
Move(res[0], result[0], n * SizeOf(res[0]));
end;
{$endif FPCMM_STANDALONE}
function CurrentHeapStatus: TMMStatus;
var
i: integer;
small: PtrUInt;
p: PSmallBlockType;
begin
result := HeapStatus;
p := @SmallBlockInfo;
for i := 1 to NumSmallInfoBlock do
begin
inc(result.SmallGetmemSleepCount, p^.GetmemSleepCount);
inc(result.SmallFreememSleepCount, p^.FreememSleepCount);
small := p^.GetmemCount - p^.FreememCount;
if small <> 0 then
begin
inc(result.SmallBlocks, small);
inc(result.SmallBlocksSize, small * p^.BlockSize);
end;
{$ifdef FPCMM_LOCKLESSFREE}
{$ifdef FPCMM_DEBUG}
inc(result.SmallFreememLockLessSpin, p^.BinSpinCount);
{$endif FPCMM_LOCKLESSFREE}
{$endif FPCMM_DEBUG}
inc(p);
end;
end;
{ ********* Initialization and Finalization }
procedure InitializeMediumPool(var Info: TMediumBlockInfo);
var
i: PtrInt;
medium: PMediumFreeBlock;
begin
{$ifndef FPCMM_ASSUMEMULTITHREAD}
Info.IsMultiThreadPtr := @IsMultiThread;
{$endif FPCMM_ASSUMEMULTITHREAD}
Info.PoolsCircularList.PreviousMediumBlockPoolHeader := @Info.PoolsCircularList;
Info.PoolsCircularList.NextMediumBlockPoolHeader := @Info.PoolsCircularList;
for i := 0 to MediumBlockBinCount -1 do
begin
medium := @Info.Bins[i];
medium.PreviousFreeBlock := medium;
medium.NextFreeBlock := medium;
end;
end;
procedure InitializeMemoryManager;
var
small: PSmallBlockType;
a, i, min, poolsize, num, perpool, size, start, next: PtrInt;
begin
small := @SmallBlockInfo;
assert(SizeOf(small^) = 1 shl SmallBlockTypePO2);
for a := 0 to NumTinyBlockArenas do
for i := 0 to NumSmallBlockTypes - 1 do
begin
if (i = NumTinyBlockTypes) and
(a > 0) then
break;
size := SmallBlockSizes[i];
assert(size and 15 = 0);
small^.BlockSize := size;
small^.PreviousPartiallyFreePool := pointer(small);
small^.NextPartiallyFreePool := pointer(small);
small^.MaxSequentialFeedBlockAddress := pointer(0);
small^.NextSequentialFeedBlockAddress := pointer(1);
min := ((size * MinimumSmallBlocksPerPool +
(SmallBlockPoolHeaderSize + MediumBlockGranularity - 1 - MediumBlockSizeOffset))
and -MediumBlockGranularity) + MediumBlockSizeOffset;
if min < MinimumMediumBlockSize then
min := MinimumMediumBlockSize;
num := (min + (- MinimumMediumBlockSize +
MediumBlockBinsPerGroup * MediumBlockGranularity div 2)) div
(MediumBlockBinsPerGroup * MediumBlockGranularity);
if num > 7 then
num := 7;
small^.AllowedGroupsForBlockPoolBitmap := byte(byte(-1) shl num);
small^.MinimumBlockPoolSize := MinimumMediumBlockSize +
num * (MediumBlockBinsPerGroup * MediumBlockGranularity);
poolsize := ((size * TargetSmallBlocksPerPool +
(SmallBlockPoolHeaderSize + MediumBlockGranularity - 1 - MediumBlockSizeOffset))