Re: [Sdcc-user] STM8 - optimizations

[Ben S. <pow...@16...>]

georg,
1. could please show a piece of c code which can be compiled to the redundant asm ? as simple & short as possible.

2. you are welcome to create patch tickets on sourceforge with any stm8 optimized functions.

ben



--
发自我的网易邮箱手机智能版


在 2014-10-02 22:57:08，"Georg Ottinger" <ge...@ot...> 写道：
>ok - after figuring out that A is 8Bit wide (what i should have expected
>on an 8Bit arch) - i have two proposals for Peephole optimizations.
>
>1.)
>
>ld a,(x)
>ld xl,a
>inc (direct)
>ld a,xl
>ret
>
>to
>
>ld a,(x)
>inc (direct)
>ret
>
>2.)
>
>ld xl, a
>ld a, xh
>
>to
>
>rlwa (rotate word left through accumulator)
>
>(under the condition that the current content of xh is not used elsewhere)
>
>the same can be achived the other way round with RRWA
>
>
>Am 2014-10-02 um 16:29 schrieb Georg Ottinger:
>> hmm ...
>>
>> is ld a, (x) on stm8 a byte wise operation???
>>
>>
>>
>> Am 2014-10-02 um 16:10 schrieb Georg Ottinger:
>>> I just figured out that
>>> s 
>>> ld a,(x)
>>> ld xl, a
>>> ld a, xl
>>>
>>> acutally is usefull code - to zero the Upper Byte of A ....
>>>
>>>
>>>
>>> Am 2014-10-02 um 16:03 schrieb Georg Ottinger:
>>>> congrats to Ben Shi - his lastest patch - just shrunk my firmware from
>>>> 4210 Bytes to 4039 Bytes thanks to _mullong_stm8
>>>>
>>>> I had a look at the asm code of my project and there more points where
>>>> optimization could be achived. - maybe I can convince someone to give it
>>>> a try - as I am not an expert to do this ;)
>>>>
>>>> for example an optimized _divulong routine - which should be
>>>> interessting as the STM8 has a hardware divder instruction.
>>>>
>>>> I also make use of _strncmp which could be optimzed but maybe that is
>>>> not so promising.
>>>>
>>>> I also noted some parts which could be optimized via peeps.def:
>>>>
>>>>
>>>> 000052 F6               [ 1]  126     ld    a, (x)
>>>> 000053 97               [ 1]  127     ld    xl, a
>>>> 128 ;    rx_ringbuffer.c: 65: rx_tail++;
>>>> 000054 72 5Cu01u01      [ 1]  129     inc    _rx_tail+0
>>>> 130 ;    rx_ringbuffer.c: 67: return temp;
>>>> 000058 9F               [ 1]  131     ld    a, xl
>>>> 000059 81               [ 4]  132     ret
>>>>
>>>>
>>>> => I think that the "131 ld a,xl" instruction could be omited - but I
>>>> have no Idea how to craft a peephole definition for this case.
>>>>
>>>>
>>>> best wishes, Georg
>>>>
>>
>> ------------------------------------------------------------------------------
>> Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
>> Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports
>> Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper
>> Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer
>> http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk
>> _______________________________________________
>> Sdcc-user mailing list
>> Sdc...@li...
>> https://lists.sourceforge.net/lists/listinfo/sdcc-user
>
>
>------------------------------------------------------------------------------
>Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
>Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports
>Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper
>Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer
>http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk
>_______________________________________________
>Sdcc-user mailing list
>Sdc...@li...
>https://lists.sourceforge.net/lists/listinfo/sdcc-user