Circular buffer should use conditional statement instead of modulo

[deepikabhavnani]

Description

1. Resolves: Circular buffer should use if statement instead of % to handle index overruns #7701
2. Moved Circular buffer test to platform folder

Pull request type

[X] Fix
[X] Refactor
[ ] Target update
[ ] Functionality change
[ ] Breaking change

[0xc0170]

AStyle CI review (fix needed)

AStyle fix done

[deepikabhavnani]

@SenRamakri @geky - Please review

[kegilbert]

LGTM!

[cmonr]

...huh. Kinda surprised to find that we're using modulo, since I would expect masking arythmetic to be used for this kind of rollover. Still, an if statement is a bit of an improvement.

/morph build

[mbed-ci]

Build : SUCCESS

Build number : 3158
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/7890/

Triggering tests

/morph test
/morph export-build
/morph mbed2-build

[mbed-ci]

Exporter Build : FAILURE

Build number : 2762
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/7890/

[mbed-ci]

Test : SUCCESS

Build number : 2969
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/7890/2969

[geky]

...huh. Kinda surprised to find that we're using modulo, since I would expect masking arythmetic to be used for this kind of rollover. Still, an if statement is a bit of an improvement.

In this case BufferSize is a compile-time constant (template). So the compiler can trivially turn it into a mask when it's a power of 2.

[cmonr]

06:46:19 FATAL: java.io.IOException: Unexpected termination of the channel

Wow. It's been a while since we last saw something like this.

/morph export

[cmonr]

In this case BufferSize is a compile-time constant (template). So the compiler can trivially turn it into a mask when it's a power of 2.

Gotcha. Is the compiler still able to do the optimization with this PR change?

[geky]

Nope: https://godbolt.org/z/aYf-MO

Though note how much worse it gets when you change SIZE to a non-power of two.

Note that is we really wanted to, we could have a template specialization for powers of two, and a template specialization for non-powers of two. But this may take much much more effort than it's really worth.

[pan-]

@cmonr No the compiler doesn't apply the optimisation: https://godbolt.org/z/yEKolO .
However the proposed change is much better when the buffer size is not a power of 2: https://godbolt.org/z/uDriqd

Edit: As @geky pointed out there's performance salvation with partial specialisation however it'll be maintenance hell for uninitiated programmers.

[0xc0170]

/morph export-build

[mbed-ci]

Exporter Build : SUCCESS

Build number : 2776
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/7890/