Commit Graph
210 Commits
Author SHA1 Message Date
Mathias Agopian 6d68a4a4c6 libmath benchmarks 2018-12-03 11:58:30 -08:00
Philip Rideout edda436405 Documentation for new filamesh flags, etc. 2018-12-02 16:46:31 -08:00
Philip Rideout 61992f556c Add unit test for filamesh.
The motivation behind a new regression test is the fact that we will
soon be adding compression features to filamesh.
2018-12-02 16:46:31 -08:00
Philip Rideout aa3a7085cc Factor out common filamesh / meshio code.
This creates a filamesh namespace and removes several duplicated
classes, including Box and filamesh::Header.
2018-12-02 16:46:31 -08:00
Philip Rideout af799e734a FilamentApp can now use KTX environments.
When cmgen creates ktx files, it creates a pair: one with an _ibl suffix
and one with _skybox. This convention is now honored by FilamentApp.

Note that we already have several KTX environments in the repo, so you
can clone a fresh repo and do:

  gltf_viewer --ibl={FILAMENT}/docs/webgl/pillars_2k DamagedHelmet.gltf

Note that we have several hundred files in samples/envs, we should
consider replacing those or copying the environments in docs/webgl
(which git would internally de-dup, due to content hashing).
2018-11-30 14:09:04 -08:00
Philip Rideout 7366bdb500 More KtxUtility improvements.
This makes it easier for C++ clients to use KTX files and simplifies
the JavaScript bindings.
2018-11-30 13:16:47 -08:00
Philip Rideout bcfe03c8e1 KtxUtility improvements.
This makes it easier for C++ clients to use KTX files and simplifies
the JavaScript bindings.
2018-11-29 16:39:30 -08:00
Mathias Agopian e8895fee11 Add a binary search benchmark
this one basically shows that the stl is is a good as our "branch less"
versions. this is because clang generate branchless code in the end.
2018-11-29 11:44:52 -08:00
Mathias Agopian 42a3d5ea33 use utils::Mutex instead of std::mutex for Allocators locking policy
on Android it's about 2x faster and uses much less instructions, which
is better for i-cache pressure.
2018-11-29 11:44:52 -08:00
Mathias Agopian 785ce645e2 pool allocators benchmark
this benchmark shows that, as we expected, on Android mutexes,
spinlocks and lock-free algorithms are running at similar performance.
2018-11-29 11:44:52 -08:00
Ben Doherty fe4907d0ff Remove use of iostream in filamat (#549) 2018-11-29 10:35:37 -08:00
Philip Rideout 239317ee10 Use resgen for libshaders. 2018-11-29 07:37:56 -08:00
Ben Doherty b759074c8f Deprecate set method on MaterialBuilder (#545)
* Remove the defines entry in material definitions

* Deprecate set function

* Remove some getter methods from MaterilaBuilder

* Fix tests, add to CI
2018-11-28 15:35:19 -08:00
Ben Doherty 3a677fa832 Fix matc crash (#546) 2018-11-28 15:00:45 -08:00
Mathias Agopian cf5606e940 Fix another bug in AtomicFreeList
An assert checking invariants would sometime trigger, the problem
was a logic error that would be exposed by a race when running out
of space in the list.

The root of the problem is that in one place we were not remapping the 
-1 offset to nullptr, storing a pointer violating our invariants.

Also added more asserts!!!
2018-11-28 12:10:23 -08:00
Philip Rideout 91abaf575b Add tangents utility to fix #532.
I tested this by writing a quick-and-dirty Sphere demo.
2018-11-28 11:27:37 -08:00
Mathias Agopian 05854e0d05 port the "filament_calls" benchmark to google benchmark 2018-11-27 15:07:45 -08:00
Ben Doherty a1ed3b81a3 Move static code analysis into filamat (#534)
* Handle static code analysis automatically

* Fix matc test cases

* Only call InitializeProcess once per process
2018-11-27 14:11:27 -08:00
Sebastian Hartte 6b030c9694 More documentation of the free list 2018-11-27 11:44:32 -08:00
Sebastian Hartte df4a2bb86e Fixes #542 by moving the assert to after checking for a possible race condition 2018-11-27 11:44:32 -08:00
Ben Doherty 0548b28535 Update glslang (#526)
* Update glslang

* Update builtinResource

* Add glslang update README
2018-11-27 09:52:03 -08:00
Ben Doherty 594f1abbbc Rename GlslChunk to TextChunk (#527)
* Rename Glsl -> Text

* Make MaterialTextChunk and DictionaryTextChunk more generic
2018-11-26 14:02:04 -08:00
Pixelflinger cf4410f656 Add some libutils benchmarks 2018-11-26 11:51:17 -08:00
Pixelflinger bc0f2f26f2 Tweak Profiler API so it works in multithread
Instead of using a singleton, Profiler has to be instantiated in the
scope/thread it is used.
2018-11-26 11:51:17 -08:00
Ben Doherty 6d158bbc80 Remove codeGenTargetApi (#511) 2018-11-26 10:51:34 -08:00
shartte 9ab9333fa5 Fixes various MSVC issues in public headers (#523) 2018-11-26 09:17:19 -08:00
Mathias Agopian d1b19efa55 a fast approximation of pow() and pow(x, 2.2f)
fast::pow(x, 2.2f) is about 100x faster than std::pow.
std::pow(x, 2.2f) takes roughly 1000 cycles.
2018-11-21 13:46:56 -08:00
Mathias Agopian 9e39004b1f Add wait_for and wait_until to our Condition implementation
We're using timed condition variable in one place, but the STL version
pulls in a lot of code because it does clock calculations in 
"long double" (!!!!). Since we already had an implementation
of condition_variable, we just add the timed version.
This saves several KiB of code.

Also don't use unique_lock() lock/unlock because it can throw exceptions.
2018-11-21 13:46:07 -08:00
Mathias Agopian 1212728e45 get rid of std::string in OpenGLProgram
This was generating tons of code.
2018-11-21 13:46:07 -08:00
Mathias Agopian 815fb36b98 code size optimizations
- prevent unrolling of some loops. e.g. one loop generated more than
  1 KB of code!
2018-11-21 13:46:07 -08:00
Mathias Agopian 5bcc4cfff1 Improve the slow-path of creating a program
This removes a lot of heap allocations/deallocations, and reduces
code size. Most improvements come from using CString and
StaticString instead of std::string and better using move
semantics.
2018-11-21 13:46:07 -08:00
Mathias Agopian c1a7164084 Improve usage of CString
This cuts down 7 calls to free() to 1, when
instantiating a new program. There are still many low-hanging fruits
there.
2018-11-21 13:46:07 -08:00
prideout 776284b2b5 Raise sampler limit to 16.
In practice this limits materials to 10 samplers since Filament uses 5
for lights and skips 1 slot for the post-process sampler. This can
probably be optimized.
2018-11-20 16:05:16 -08:00
prideout cfa9895b96 Improve error reporting for bindings overflow.
We now check for sampler overflow during material compilation rather
than waiting for run-time checks. This allows for Kokoro-based
validation of sample materials, and allows developers to catch this
issue in their asset pipeline.

We should also probably raise the upper limit (Qualcomm allows up to 16
samplers in their Vulkan implementation) but that can be a separate PR.

Fixes #507
2018-11-20 15:28:02 -08:00
Ben Doherty b5fa822e86 Move GLSLPostProcessor and dependencies into filamat (#500)
* Move GLSLPostProcessor into filamat
2018-11-20 12:30:26 -08:00
Philip Rideout aa0090a85b Fix build warnings. 2018-11-19 15:59:31 -08:00
Philip Rideout 0d476b9f6d Fix debug builds. 2018-11-19 15:45:40 -08:00
prideout 694eab703c Remove usage of std::string in VulkanDriver. 2018-11-19 14:18:39 -08:00
Philip Rideout edb545919a Fix regression with non-Vulkan backends.
This fixes the "shader is not ASCII" error seen with GL backend,
introduced when we changed blob ownership semantics to accommodate
shader compression. It was due to missing null terminators.

We actually never bothered including the trailing null in blob length,
which was wrong but happened to work because the BlobDictionary held a
weak reference to chunk data. SInce it now holds an actual copy, the
lack of null caused our strings to contain garbage memory.
2018-11-16 15:43:41 -08:00
Philip Rideout b2a2da532a Fix Android-with-Vulkan build. 2018-11-16 12:07:23 -08:00
Philip Rideout b83c82ea9b Enable SMOLV compression. 2018-11-16 12:07:23 -08:00
Philip Rideout e0fd196530 Exclude SMOL from Vulkan-free builds. 2018-11-15 16:55:40 -08:00
Philip Rideout 55eaad9450 Add smol-v to third_party.
Since we use this for decoding, this adds a dependency to the core
filament renderer which in practice is only used for Vulkan. However
this is a tiny library, so it's simplest just to always include it.

The next PR will add the actual compression / decompression code to
filaflat and filamat. Here are the preliminary results.

   102K =>  29K  aiDefaultMat.filamat
   102K =>  29K  aiDefaultTrans.filamat
    21K => 5.0K  bakedColor.filamat
    22K => 5.2K  bakedTexture.filamat
    21K => 5.0K  depthVisualizer.filamat
    40K =>  10K  groundShadow.filamat
   102K =>  29K  sandboxCloth.filamat
   125K =>  36K  sandboxLit.filamat
   126K =>  36K  sandboxLitFade.filamat
   126K =>  36K  sandboxLitTransparent.filamat
   109K =>  31K  sandboxSubsurface.filamat
    21K => 4.9K  sandboxUnlit.filamat
    21K => 4.9K  transparentColor.filamat

Both the "before" and "after" numbers are excluding non-Vulkan targets,
and I also changed our CMakeLists to build filamat instead of inc.
2018-11-15 16:55:40 -08:00
Mathias Agopian f011a212bd Minor clean-ups of Program 2018-11-13 16:02:44 -08:00
Mathias Agopian bc3aee118a code size and performance optimizations (#472)
* Clean-up EntityManager a bit

- use tsl::robin_set instead of std::set (which should have been unordered::set
  anyways).

- getListeners() now returns a vector which avoids to traverse a set twice.
  Turns out that copying the set wasn't as efficient as I thought.

* Improve jobsystem a bit

We recently added a job reference counting mechanism, but we were a bit
too aggressive about taking/release references.

Also make the API more complete by adding explicit retain/release,
which is needed to allow several threads to wait on the same job.

Also improve futex code by inlining it.
2018-11-12 14:11:31 -08:00
Mathias Agopian f8f24f492f minor typo / spelling / clang-tidy fixes (#468)
* Minor clean-ups

- fix a couple usage of std::function
- fix a couple usage of std::string
- remove ALIGN_LOOP, which didn't work
- fix a couple explicit/noexcept
- virtual -> override

* Fix spelling typos and other minor clang-tidy
2018-11-08 11:50:00 -08:00
Philip Rideout ee5b0268fd Introduce KtxUtility to make it easier to translate enum values. 2018-11-07 16:12:55 -08:00
Mathias Agopian 591c9f2454 Fix a crasher when allocating the last Job
Instead of returning nullptr, the pool allocator returned an
invalid pointer.
2018-11-06 13:44:35 -08:00
Mathias Agopian e580bfed56 Fix allocator test when --gtest_repeat is used 2018-11-06 13:21:33 -08:00
Mathias Agopian e2896f4632 Set cache size to 64 bytes on all platforms
We used to assume 32-bytes cache lines when running on ARM 32-bit mode,
however, that was wrong because 32-bit mode doesn't change the cpu's
cache line. On all modern platforms we support, the cache line
is 64 bytes.

Set Job storage to 48 bytes on all platforms.
2018-11-06 13:21:09 -08:00