Karl Wiberg d2c6967f1d Optimize RoboCaller::AddReceiver() for code size
Essentially, instead of having the inlined UntypedFunction::Create(f)
return an UntypedFunction which is then passed as an argument to
non-inlined RoboCallerReceivers::AddReceiverImpl(), we let
UntypedFunction::PrepareArgs(f) return a few different kinds of
trivial structs (depending on what sort of type f has) which are
passed as arguments to non-inlined RoboCallerReceivers::AddReceiver()
(which then converts them to UntypedFunction by calling
UntypedFunction::Create()). These structs are smaller than
UntypedFunction and optimized for argument passing, so many fewer
instructions are needed.

Example code:

  struct Foo {
    void Receive(int, float, int, float);
    void TestAddLambdaReceiver();
    webrtc::RoboCaller<int, float, int, float> rc;
  };

  void Foo::TestAddLambdaReceiver() {
    rc.AddReceiver([this](int a, float b, int c, float d){
        Receive(a, b, c, d);});
  }

On arm32, we get before this CL:

  Foo::TestAddLambdaReceiver():
        push    {r11, lr}
        mov     r11, sp
        sub     sp, sp, #24
        ldr     r1, .LCPI0_0
        mov     r2, #0
        stm     sp, {r0, r2}
        add     r1, pc, r1
        str     r2, [sp, #20]
        str     r1, [sp, #16]
        mov     r1, sp
        bl      RoboCallerReceivers::AddReceiverImpl
        mov     sp, r11
        pop     {r11, pc}
  .LCPI0_0:
        .long   CallInlineStorage<Foo::TestAddLambdaReceiver()::$_0>
  CallInlineStorage<Foo::TestAddLambdaReceiver()::$_0>:
        ldr     r0, [r0]
        b       Foo::Receive(int, float, int, float)

After this CL:

  Foo::TestAddLambdaReceiver():
        ldr     r3, .LCPI0_0
        mov     r2, r0
        add     r3, pc, r3
        b       RoboCallerReceivers::AddReceiver<1u>
  .LCPI0_0:
        .long   CallInlineStorage<Foo::TestAddLambdaReceiver()::$_0>
  CallInlineStorage<Foo::TestAddLambdaReceiver()::$_0>:
        ldr     r0, [r0]
        b       Foo::Receive(int, float, int, float)

(Symbol names abbreviated so that they'll fit on one line.)

So a reduction from 64 to 28 bytes. The improvements on arm64 and
x86_64 are similar.

Bug: webrtc:11943
Change-Id: I93fbba083be0235051c3279d3e3f6852a4a9fdad
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/185960
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32244}
2020-09-30 06:59:44 +00:00
2018-10-05 14:40:21 +00:00
2020-09-29 20:24:07 +00:00
2020-09-29 13:51:59 +00:00
2020-09-30 06:14:57 +00:00
2020-09-29 20:24:07 +00:00
2019-10-28 12:27:50 +00:00
2020-09-09 14:36:03 +00:00
2020-03-30 12:15:56 +00:00
2020-09-07 08:34:44 +00:00
2020-07-13 11:42:07 +00:00
2020-04-16 11:08:43 +00:00
2020-09-06 10:13:23 +00:00

WebRTC is a free, open software project that provides browsers and mobile applications with Real-Time Communications (RTC) capabilities via simple APIs. The WebRTC components have been optimized to best serve this purpose.

Our mission: To enable rich, high-quality RTC applications to be developed for the browser, mobile platforms, and IoT devices, and allow them all to communicate via a common set of protocols.

The WebRTC initiative is a project supported by Google, Mozilla and Opera, amongst others.

Development

See here for instructions on how to get started developing with the native code.

Authoritative list of directories that contain the native API header files.

More info

Description
The idea is to make CMake build for WebRTC m130 version - for audio processing module
Readme BSD-3-Clause 446 MiB
Languages
C++ 90.3%
Java 2.9%
C 2.2%
Objective-C++ 2%
Python 1.3%
Other 1%