This commit is contained in:
Araq
2017-01-30 19:01:03 +01:00
parent 6050102b76
commit b26e6e3589
3 changed files with 68 additions and 44 deletions

View File

@@ -100,7 +100,7 @@ path="$lib/pure"
@if windows:
#gcc.path = r"$nim\dist\mingw\bin"
@if gcc:
tlsEmulation:on
#tlsEmulation:on
@end
@end
@@ -110,7 +110,13 @@ path="$lib/pure"
gcc.options.always = "-w"
gcc.cpp.options.always = "-w -fpermissive"
@else:
gcc.options.always = "-w"
@if cpu32:
gcc.options.always = "-w -m32"
gcc.options.linker = "-m32"
@else:
gcc.options.always = "-w -m64"
gcc.options.linker = "-m64"
@end
gcc.cpp.options.always = "-w -fpermissive"
@end

View File

@@ -54,7 +54,7 @@ type
BaseChunk {.pure, inheritable.} = object
prevSize: int # size of previous chunk; for coalescing
size: int # if < PageSize it is a small chunk
used: bool # later will be optimized into prevSize...
origSize: int # 0th bit == 1 if 'used'
SmallChunk = object of BaseChunk
next, prev: PSmallChunk # chunks of the same size
@@ -65,11 +65,11 @@ type
BigChunk = object of BaseChunk # not necessarily > PageSize!
next, prev: PBigChunk # chunks of the same (or bigger) size
align: int
heapLink: PBigChunk # linked list of all chunks for bulk 'deallocPages'
data: AlignType # start of usable memory
template smallChunkOverhead(): expr = sizeof(SmallChunk)-sizeof(AlignType)
template bigChunkOverhead(): expr = sizeof(BigChunk)-sizeof(AlignType)
template smallChunkOverhead(): untyped = sizeof(SmallChunk)-sizeof(AlignType)
template bigChunkOverhead(): untyped = sizeof(BigChunk)-sizeof(AlignType)
# ------------- chunk table ---------------------------------------------------
# We use a PtrSet of chunk starts and a table[Page, chunksize] for chunk
@@ -98,6 +98,7 @@ type
currMem, maxMem, freeMem: int # memory sizes (allocated from OS)
lastSize: int # needed for the case that OS gives us pages linearly
freeChunksList: PBigChunk # XXX make this a datastructure with O(1) access
heapLink: PBigChunk # used to link every chunk for bulk 'deallocPages'
chunkStarts: IntSet
root, deleted, last, freeAvlNodes: PAvlNode
locked, blockChunkSizeIncrease: bool # if locked, we cannot free pages.
@@ -241,7 +242,7 @@ proc isSmallChunk(c: PChunk): bool {.inline.} =
return c.size <= SmallChunkSize-smallChunkOverhead()
proc chunkUnused(c: PChunk): bool {.inline.} =
result = not c.used
result = (c.origSize and 1) == 0
iterator allObjects(m: var MemRegion): pointer {.inline.} =
m.locked = true
@@ -310,12 +311,14 @@ proc requestOsChunks(a: var MemRegion, size: int): PBigChunk =
incCurrMem(a, size)
inc(a.freeMem, size)
result.heapLink = a.heapLink
result.origSize = size
a.heapLink = result
sysAssert((cast[ByteAddress](result) and PageMask) == 0, "requestOsChunks 1")
#zeroMem(result, size)
result.next = nil
result.prev = nil
result.used = false
result.size = size
# update next.prevSize:
var nxt = cast[ByteAddress](result) +% size
@@ -336,19 +339,21 @@ proc requestOsChunks(a: var MemRegion, size: int): PBigChunk =
result.prevSize = 0 # unknown
a.lastSize = size # for next request
proc freeOsChunks(a: var MemRegion, p: pointer, size: int) =
# update next.prevSize:
var c = cast[PChunk](p)
var nxt = cast[ByteAddress](p) +% c.size
sysAssert((nxt and PageMask) == 0, "freeOsChunks")
var next = cast[PChunk](nxt)
if pageIndex(next) in a.chunkStarts:
next.prevSize = 0 # XXX used
excl(a.chunkStarts, pageIndex(p))
osDeallocPages(p, size)
decCurrMem(a, size)
dec(a.freeMem, size)
#c_fprintf(stdout, "[Alloc] back to OS: %ld\n", size)
when false:
# with the new linked list design this is not possible anymore:
proc freeOsChunks(a: var MemRegion, p: pointer, size: int) =
# update next.prevSize:
var c = cast[PChunk](p)
var nxt = cast[ByteAddress](p) +% c.size
sysAssert((nxt and PageMask) == 0, "freeOsChunks")
var next = cast[PChunk](nxt)
if pageIndex(next) in a.chunkStarts:
next.prevSize = 0 # XXX used
excl(a.chunkStarts, pageIndex(p))
osDeallocPages(p, size)
decCurrMem(a, size)
dec(a.freeMem, size)
#c_fprintf(stdout, "[Alloc] back to OS: %ld\n", size)
proc isAccessible(a: MemRegion, p: pointer): bool {.inline.} =
result = contains(a.chunkStarts, pageIndex(p))
@@ -414,19 +419,20 @@ proc freeBigChunk(a: var MemRegion, c: PBigChunk) =
excl(a.chunkStarts, pageIndex(c))
c = cast[PBigChunk](le)
if c.size < ChunkOsReturn or doNotUnmap or a.locked:
incl(a, a.chunkStarts, pageIndex(c))
updatePrevSize(a, c, c.size)
listAdd(a.freeChunksList, c)
c.used = false
else:
freeOsChunks(a, c, c.size)
#if c.size < ChunkOsReturn or doNotUnmap or a.locked:
incl(a, a.chunkStarts, pageIndex(c))
updatePrevSize(a, c, c.size)
listAdd(a.freeChunksList, c)
# set 'used' to false:
c.origSize = c.origSize and not 1
#else:
# freeOsChunks(a, c, c.size)
proc splitChunk(a: var MemRegion, c: PBigChunk, size: int) =
var rest = cast[PBigChunk](cast[ByteAddress](c) +% size)
sysAssert(rest notin a.freeChunksList, "splitChunk")
rest.size = c.size - size
rest.used = false
rest.origSize = 0 # not used and size irrelevant
rest.next = nil
rest.prev = nil
rest.prevSize = size
@@ -461,7 +467,8 @@ proc getBigChunk(a: var MemRegion, size: int): PBigChunk =
if result.size > size:
splitChunk(a, result, size)
result.prevSize = 0 # XXX why is this needed?
result.used = true
# set 'used' to to true:
result.origSize = result.origSize or 1
incl(a, a.chunkStarts, pageIndex(result))
dec(a.freeMem, size)
@@ -704,18 +711,26 @@ proc realloc(allocator: var MemRegion, p: pointer, newsize: Natural): pointer =
proc deallocOsPages(a: var MemRegion) =
# we free every 'ordinarily' allocated page by iterating over the page bits:
for p in elements(a.chunkStarts):
var page = cast[PChunk](p shl PageShift)
when not doNotUnmap:
var size = if page.size < PageSize: PageSize else: page.size
osDeallocPages(page, size)
else:
# Linux on PowerPC for example frees MORE than asked if 'munmap'
# receives the start of an originally mmap'ed memory block. This is not
# too bad, but we must not access 'page.size' then as that could trigger
# a segfault. But we don't need to access 'page.size' here anyway,
# because calling munmap with PageSize suffices:
osDeallocPages(page, PageSize)
var it = a.heapLink
while it != nil:
let next = it.heapLink
sysAssert it.origSize >= PageSize, "origSize too small"
# note:
osDeallocPages(it, it.origSize and not 1)
it = next
when false:
for p in elements(a.chunkStarts):
var page = cast[PChunk](p shl PageShift)
when not doNotUnmap:
var size = if page.size < PageSize: PageSize else: page.size
osDeallocPages(page, size)
else:
# Linux on PowerPC for example frees MORE than asked if 'munmap'
# receives the start of an originally mmap'ed memory block. This is not
# too bad, but we must not access 'page.size' then as that could trigger
# a segfault. But we don't need to access 'page.size' here anyway,
# because calling munmap with PageSize suffices:
osDeallocPages(page, PageSize)
# And then we free the pages that are in use for the page bits:
llDeallocAll(a)

View File

@@ -132,7 +132,7 @@ elif defined(windows):
header: "<windows.h>", stdcall, importc: "VirtualAlloc".}
proc virtualFree(lpAddress: pointer, dwSize: int,
dwFreeType: int32) {.header: "<windows.h>", stdcall,
dwFreeType: int32): cint {.header: "<windows.h>", stdcall,
importc: "VirtualFree".}
proc osAllocPages(size: int): pointer {.inline.} =
@@ -151,7 +151,10 @@ elif defined(windows):
# Windows :-(. We have to live with MEM_DECOMMIT instead.
# Well that used to be the case but MEM_DECOMMIT fragments the address
# space heavily, so we now treat Windows as a strange unmap target.
when reallyOsDealloc: virtualFree(p, 0, MEM_RELEASE)
when reallyOsDealloc:
if virtualFree(p, 0, MEM_RELEASE) == 0:
cprintf "yes, failing!"
quit 1
#VirtualFree(p, size, MEM_DECOMMIT)
elif hostOS == "standalone":