Change-Id: Ic44e24057b95bb014504f02a736fb4341afc8971
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/304856
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This moves the byte block linked list structure outside of GrMemoryPool
into a new type, GrBlockAllocator. This new type is solely responsible
for managing the byte blocks, tracking where the next allocation
occurs, and creating/destroying the byte blocks. It also tries to
encapsulate all of/most alignment related math, while making it
convenient for clients to add per-allocation padding/metadata.
It has added functionality compared to the original block linked list
that was embedded in GrMemoryPool:
- Supports resetting the entire allocator
- Supports resizing the last allocation
- Is able to rewind an entire stack of allocations, instead of just the
last allocation.
- Supports multiple block growth policies instead of just a fixed size.
- Supports templated alignment, and variable alignment within a single
GrBlockAllocator
- Query the amount of available memory
- Performs as much math as possible in 32-bit ints with static asserts
that ensure overflow won't happen.
Some of this flexibility was added so that the GrBlockAllocator can be
used to implement an arena allocator similar to SkArenaAlloc, or to
replace the use of SkTArray in GrQuadBuffer. It is also likely possible
that GrAllocator can be written on top of GrBlockAllocator. I will try
to perform these consolidations in later CLs.
Change-Id: Ia6c8709642369b66b88eb1bc46264fb2aa9b62ab
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/262216
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Saves one heap allocation per DDL recorded.
Change-Id: I9393aedc3b48031cd2ea5f0160b107915077099a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/259419
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Current strategy: everything from the top
Things to look at first are the manual changes:
- added tools/rewrite_includes.py
- removed -Idirectives from BUILD.gn
- various compile.sh simplifications
- tweak tools/embed_resources.py
- update gn/find_headers.py to write paths from the top
- update gn/gn_to_bp.py SkUserConfig.h layout
so that #include "include/config/SkUserConfig.h" always
gets the header we want.
No-Presubmit: true
Change-Id: I73a4b181654e0e38d229bc456c0d0854bae3363e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209706
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Hal Canary <halcanary@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
This enables four different options in the compiler, described
below. I also added enough masks to satisfy RTCc when running
all GMs in both 8888 and gl configs.
---
/RTCc - Detects when a value is assigned to a smaller data
type and results in data loss. This happens even when casting.
Masking is required to suppress this.
/RTCs - Various stack-related checks, including uninitialized
data (by initializing locals to a non-zero value), array bounds
checking, and stack pointer corruption that can occur with a
calling convention mismatch.
/RTCu - Reports when a variable is used without having been
initialized. Mostly redundant with compile-time checks.
/guard:cf - This is more of a security option, that computes
all possible targets for indirect calls at compile time, and
verifies that those are the only targets reached at compile
time. Also generates similar logic around switch statements
that turn into jump tables.
Bug: skia:
Change-Id: I7b527af8fd67dec0b6556f38bcd0efc3fd505856
Reviewed-on: https://skia-review.googlesource.com/c/188625
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Most of this is (obviously) not necessary to do, but once
I started, I figured I'd just get it all. Tools (nanobench,
DM, skiaserve), all GMs, benches, and unit tests, plus support
code (command line parsing and config stuff).
This is almost entirely mechanical.
Bug: skia:
Change-Id: I209500f8df8c5bd43f8298ff26440d1c4d7425fb
Reviewed-on: https://skia-review.googlesource.com/131153
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Some memory allocators have very coarse size buckets, so for example on
Android (jemalloc) an attempt to allocate 32 KiB + 1 byte will end up
allocating 40 KiB, wasting 8 KiB.
GrMemoryPool ctor takes two arguments that specify prealloc / block sizes,
and then inflates them to accommodate some bookkeeping structures. Since
most places create GrMemoryPools with pow2 numbers (which have buckets in
most allocators) the inflation causes allocator to select next size bucket,
wasting memory.
This CL makes GrMemoryPool to stop inflating sizes it was created with, and
allocate specified amounts exactly. Part of allocated memory is then used for
bookkeeping structures. Additionally, GrObjectMemoryPool template is provided,
which takes prealloc / block object counts (instead of sizes) and guarantees
that specified number of objects will fit in prealloc / block spaces.
BUG=651872
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2525773002
Review-Url: https://codereview.chromium.org/2525773002
Replace with std::unique_ptr.
Change-Id: I5806cfbb30515fcb20e5e66ce13fb5f3b8728176
Reviewed-on: https://skia-review.googlesource.com/4381
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
This code requires fewer macros to use it (just one), has less code in macro
definitions, and has simpler synchronization code (just atomic ints, no SkOnce,
no SkMutex, etc.)
A minor downside, we lose indentation and reverse-ordering in the final report:
Leaked SkRefCntBase: 7
Leaked SkFontMgr: 1
Leaked SkWeakRefCnt: 1
Leaked SkTypeface: 1
Leaked SkFlattenable: 3
Leaked SkXfermode: 3
Leaked SkPathRef: 1
Leaked SkPixelRef: 1
Leaked SkMallocPixelRef: 1
becomes
Leaked SkXfermode: 3
Leaked SkMallocPixelRef: 1
Leaked SkPixelRef: 1
Leaked SkPathRef: 1
Leaked SkFlattenable: 3
Leaked SkTypeface: 1
Leaked SkWeakRefCnt: 1
Leaked SkFontMgr: 1
Leaked SkRefCntBase: 7
This is motivated by wanting to land https://codereview.chromium.org/806473006/,
which makes sure all static use of SkOnce are in global scope. The current
implementation of SkInstCnt uses them in function scope, which isn't safe.
BUG=skia:
No public API changes.
TBR=reed@google.com
Review URL: https://codereview.chromium.org/841263004
Add INHERITED declarations to class declarations that prevent
compilation with the flag.
Remove SK_DEFINE_INST_COUNT from all class implementations. Instead,
use function-local static variables in the reference count helper
classes to create the global instances to store the needed info. The
accessor functions are defined inline in the helper classes, so
definitions are not needed. The initialization point of the variables
should be as well defined as previously.
Remove SK_DECLARE_INST_COUNT_TEMPLATE and use SK_DECLARE_INST_COUNT
instead. This avoids possible future compilation errors further.
For SK_ENABLE_INST_COUNT=0 compilation, add an empty static member
function to all classes that use SK_DECLARE_INST_COUNT and
SK_DECLARE_INST_COUNT_ROOT macros. The function ensures that classes
contain public INHERITED typedef. This member function seems to be
compiled away. This shouĺd ensure that part of the compilation errors
are caught earlier.
Also adds DSK_DECLARE_INST_COUNT to few SkPDFDict subclasses.
R=robertphillips@google.com, richardlin@chromium.org, bsalomon@google.com
Author: kkinnunen@nvidia.com
Review URL: https://codereview.chromium.org/98703002
git-svn-id: http://skia.googlecode.com/svn/trunk@12501 2bbb7eff-a529-9590-31e7-b0007b416f81
unused headers and fixes a couple of bugs exposed by changing the random
number generator:
First, the function SkMatrix::getMaxStretch() had an error where it was testing
the square of a number against near-zero. This led to it occasionally taking a
cheaper but imprecise path for computing the eigenvalues of the matrix. It's
been replaced with a check against the square of SK_ScalarNearlyZero.
The second case was a failure in ClipStackTest, where it hit the rare case of
a practically empty clip stack (it has a single Union) and we set a tight
bounds. The bounds rect doesn't get set by GrReducedClip::ReduceClipStack() in
this case, so when it clips the reduced stack it's clipping against garbage,
and the resulting regions don't match. The solution is to initialize the
tightBounds rect.
git-svn-id: http://skia.googlecode.com/svn/trunk@7952 2bbb7eff-a529-9590-31e7-b0007b416f81
