RE: [boost] Math constants using templated functions - looks efficient?

From: Paul A. Bristow (boost_at_[hidden])
Date: 2001-10-23 12:26:50

I have carried out a few simplistic test of the function template
//  Based on a post by Michael Kenniston:
http://groups.yahoo.com/group/boost/message/14867

as suggested by Ed below (with his suggested removal of the redundant
constructor).

It is not as easy as might appear because the compiler tends to optimise
everything away!
To avoid this I have used two naive methods, cout << pi << endl;
and global double dummy = p;

This suggests that the code produced is indeed quite efficient,
though no FP registers are used in this example.

Do the speed freaks who made much about the importance of the interface
satisfied?

Paul

Dr Paul A Bristow, hetp Chromatography
Prizet Farmhouse
Kendal, Cumbria
LA8 8AB      UK
[phone]
Mobile [phone]
mailto:pbristow_at_[hidden]

_constpi DQ	0400921fb54442d18r		; 3.14159    <<<<< pi in hex

; 52   : 	double ccp = constpi; // assignment to a local or global const
double
; 53   : 	cout << ccp << endl; // prevent optimisation.

  00000	a1 04 00 00 00	 mov	 eax, DWORD PTR _constpi+4
  00005	8b 0d 00 00 00
        00		 mov	 ecx, DWORD PTR _constpi
  0000b	56		 push	 esi
  0000c	50		 push	 eax
  0000d	51		 push	 ecx
  0000e	b9 00 00 00 00	 mov	 ecx, OFFSET
FLAT:?cout_at_std@@3V?$basic_ostream_at_DU?$char_traits_at_D@std@@@1_at_A
  00013	e8 00 00 00 00	 call
??6?$basic_ostream_at_DU?$char_traits_at_D@std@@@std@@QAEAAV01_at_N@Z ;
std::basic_ostream<char,std::char_traits<char> >::operator<<
  00018	8b f0		 mov	 esi, eax
  0001a	6a 0a		 push	 10			; 0000000aH
  0001c	8b ce		 mov	 ecx, esi
  0001e	e8 00 00 00 00	 call
?put@?$basic_ostream_at_DU?$char_traits_at_D@std@@@std@@QAEAAV12_at_D@Z ;
std::basic_ostream<char,std::char_traits<char> >::put
  00023	8b ce		 mov	 ecx, esi
  00025	e8 00 00 00 00	 call
?flush@?$basic_ostream_at_DU?$char_traits_at_D@std@@@std@@QAEAAV12_at_XZ ;
std::basic_ostream<char,std::char_traits<char> >::flush

; 54   : 	double pd = 3.141592653589793238462643383279502884197;
; 55   : 	double p = pi; // assignment using templated function constant pi
; 56   : 	cout << p << endl; // prevent optimisation.

  0002a	68 fb 21 09 40	 push	 1074340347		; 400921fbH <<<<< pi in hex
  0002f	68 18 2d 44 54	 push	 1413754136		; 54442d18H <<<<< pi in hex
  00034	b9 00 00 00 00	 mov	 ecx, OFFSET
FLAT:?cout_at_std@@3V?$basic_ostream_at_DU?$char_traits_at_D@std@@@1_at_A
  00039	e8 00 00 00 00	 call
??6?$basic_ostream_at_DU?$char_traits_at_D@std@@@std@@QAEAAV01_at_N@Z ;
std::basic_ostream<char,std::char_traits<char> >::operator<<
  0003e	8b f0		 mov	 esi, eax
  00040	6a 0a		 push	 10			; 0000000aH
  00042	8b ce		 mov	 ecx, esi
  00044	e8 00 00 00 00	 call
?put@?$basic_ostream_at_DU?$char_traits_at_D@std@@@std@@QAEAAV12_at_D@Z ;
std::basic_ostream<char,std::char_traits<char> >::put
  00049	8b ce		 mov	 ecx, esi
  0004b	e8 00 00 00 00	 call
?flush@?$basic_ostream_at_DU?$char_traits_at_D@std@@@std@@QAEAAV12_at_XZ ;
std::basic_ostream<char,std::char_traits<char> >::flush

; 60   : 	double p = pi; // assignment using function.
; 61   : 	dummy = p; // prevent optimisation - dummy is global.

  00000	c7 05 00 00 00
        00 18 2d 44 54	 mov	 DWORD PTR ?dummy@@3NA, 1413754136 ; 54442d18H <<<<< pi
in hex
  0000a	c7 05 04 00 00
        00 fb 21 09 40	 mov	 DWORD PTR ?dummy@@3NA+4, 1074340347 ; 400921fbH <<<<<
pi in hex

> -----Original Message-----
> From: Ed Brey [mailto:edbrey_at_[hidden]]
> Sent: Tuesday, October 23, 2001 1:56 PM
> To: boost_at_[hidden]
> Subject: Re: [boost] Math constants for naive and gurus? - which
> constants do you want?
 The testing I did was rather simplistic.  Rather than share my
> results, I'd rather have you independently generate your own
> results from your own tests, now that you know how get to the
> release mode assembly.  I'm certain you'll be able to think of
> more real-world test cases than I could.
>
> Assuming that inspection of the assembly output yields positive
> results, the next test I would recommend is taking some
> signficant piece of math-intenstive code, and compiling it twice,
> once with plain old constants definitions, and once with the
> template definitions.  There would be too much output to wade
> through by hand, be one should be able to do comparisons of the
> resulting assembly, object, and executable files.  I expect that
> they will be byte-for-byte identical.