Clean up round and align in PartitionAlloc

Clean up round and align in PartitionAlloc

|partitionDirectMapSize| and |partitionRoundUpToSystemPage| were duplicates of
eachother, modulo the assertion. The rounding function was abused to align
pointers in addition to rounding sizes. With this CL there is one function for
size rounding and two functions to align pointers (up and down).

Note that the reinterpret cast magic was already there. It relies on the fact
that on the platforms we build for, pointers are integers. This CL simply
moves the casts into the function to avoid repetition.

[Primiano Tucci (use gerrit), 2015-09-03 11:24:53]

primiano@chromium.org changed reviewers:
+ primiano@chromium.org

[Primiano Tucci (use gerrit), 2015-09-03 11:24:54]

some comments, I think here you just want to fix the names and the castin of the
roundUp functions and leave the directMapSize at it is.

Also, at this point please fix the commit message accordingly.

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.cpp
File Source/wtf/PartitionAlloc.cpp (right):

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.c...
Source/wtf/PartitionAlloc.cpp:514: size_t addr = reinterpret_cast<size_t>(ptr);
these should be uintptr_t, also + const (i.e. const uintptr_t)

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.c...
Source/wtf/PartitionAlloc.cpp:521: size_t addr = reinterpret_cast<size_t>(ptr);
same here

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.h
File Source/wtf/PartitionAlloc.h (left):

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.h...
Source/wtf/PartitionAlloc.h:729: ALWAYS_INLINE size_t
partitionDirectMapSize(size_t size)
I think this should stay as it is (as the assert is map-specific)

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.h...
Source/wtf/PartitionAlloc.h:752: size = partitionDirectMapSize(size);
ditto

[Ruud van Asseldonk, 2015-09-03 13:44:47]

ruuda@google.com changed reviewers:
+ haraken@chromium.org

[Ruud van Asseldonk, 2015-09-03 13:44:48]

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.cpp
File Source/wtf/PartitionAlloc.cpp (right):

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.c...
Source/wtf/PartitionAlloc.cpp:514: size_t addr = reinterpret_cast<size_t>(ptr);
On 2015/09/03 at 11:24:53, Primiano Tucci wrote:
> these should be uintptr_t, also + const (i.e. const uintptr_t)

Sure, but note that the current code uses size_t.

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.h
File Source/wtf/PartitionAlloc.h (left):

https://codereview.chromium.org/1330593002/diff/1/Source/wtf/PartitionAlloc.h...
Source/wtf/PartitionAlloc.h:729: ALWAYS_INLINE size_t
partitionDirectMapSize(size_t size)
On 2015/09/03 at 11:24:53, Primiano Tucci wrote:
> I think this should stay as it is (as the assert is map-specific)

Done.

[Ruud van Asseldonk, 2015-11-26 16:49:40]

Description was changed from

==========
Clean up round and align in PartitionAlloc

|partitionDirectMapSize| and |partitionRoundUpToSystemPage| were duplicates of
eachother, modulo the assertion. The rounding function was abused to align
pointers in addition to rounding sizes. With this CL there is one function for
size rounding and two functions to align pointers (up and down).

Note that the reinterpret cast magic was already there. It relies on the fact
that on the platforms we build for, pointers are integers. This CL simply
moves the casts into the function to avoid repetition.
==========

to

==========
Clean up round and align in PartitionAlloc

|partitionDirectMapSize| and |partitionRoundUpToSystemPage| were duplicates of
eachother, modulo the assertion. The rounding function was abused to align
pointers in addition to rounding sizes. With this CL there is one function for
size rounding and two functions to align pointers (up and down).

Note that the reinterpret cast magic was already there. It relies on the fact
that on the platforms we build for, pointers are integers. This CL simply
moves the casts into the function to avoid repetition.
==========