performance idea for old computers

-for older computers..

Firefox should Have...
MMX builds optimized.

- MMX is in many CPUs...

- MMX should be in the compiler parameters...

- MMX was introduced in 1997

http://en.wikipedia.org/wiki/MMX_(instruction_set)
_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

hi - what parts of the firefox code do you think would generate SIMD
operations and benefit from this, and on what compilers?

imho the  best way to engage folks in a serious discussion of the pros
and cons would be with some benchmarks showing some area improving and
an analysis of why - or alternatively a specific proposal on what could
be coded differently to leverage things like mmx/sse.

-Patrick


On Sun, 2008-09-14 at 14:42 -0700, dlhdavidlh@... wrote:

> -for older computers..
>
> Firefox should Have...
> MMX builds optimized.
>
> - MMX is in many CPUs...
>
> - MMX should be in the compiler parameters...
>
> - MMX was introduced in 1997
>
> http://en.wikipedia.org/wiki/MMX_(instruction_set)
> _______________________________________________
> dev-performance mailing list
> dev-performance@...
> https://lists.mozilla.org/listinfo/dev-performance

_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

  - this what i have found...
- i have tests on
 http://dlhdragon.110mb.com/webbrowsertests.html

-------------------------------------------------------------------

  C:\Programme\Microsoft Visual Studio 8\VC>cl -? > C:\cl.txt
    Microsoft (R) 32-bit C/C++ Optimizing Compiler Version
14.00.50727.42 for 80x86
    Copyright (C) Microsoft Corporation.  All rights reserved.

                             C/C++ COMPILER OPTIONS


                                  -OPTIMIZATION-

    /O1 minimize space                      /O2 maximize speed
    /Ob<n> inline expansion (default n=0)   /Od disable optimizations
(default)
    /Og enable global optimization          /Oi[-] enable intrinsic
functions
    /Os favor code space                    /Ot favor code speed
    /Ox maximum optimizations               /Oy[-] enable frame
pointer omission

                                 -CODE GENERATION-

    /GF enable read-only string pooling     /Gm[-] enable minimal
rebuild
    /Gy[-] separate functions for linker    /GS[-] enable security
checks
    /GR[-] enable C++ RTTI                  /GX[-] enable C++ EH (same
as /EHsc)
    /EHs enable C++ EH (no SEH exceptions)  /EHa enable C++ EH (w/ SEH
exceptions)
    /EHc extern "C" defaults to nothrow
    /fp:<except[-]|fast|precise|strict> choose floating-point model:
        except[-] - consider floating-point exceptions when generating
code
        fast - "fast" floating-point model; results are less
predictable
        precise - "precise" floating-point model; results are
predictable
        strict - "strict" floating-point model (implies /fp:except)
    /GL[-] enable link-time code generation /GA optimize for Windows
Application
    /Ge force stack checking for all funcs  /Gs[num] control stack
checking calls
    /Gh enable _penter function call        /GH enable _pexit function
call
    /GT generate fiber-safe TLS accesses    /RTC1 Enable fast checks (/
RTCsu)
    /RTCc Convert to smaller type checks    /RTCs Stack Frame runtime
checking
    /RTCu Uninitialized local usage checks
    /clr[:option] compile for common language runtime, where option
is:
        pure - produce IL-only output file (no native executable code)
        safe - produce IL-only verifiable output file
        oldSyntax - accept the Managed Extensions syntax from Visual C+
+ 2002/2003
        initialAppDomain - enable initial AppDomain behavior of Visual
C++ 2002
        noAssembly - do not produce an assembly
    /Gd __cdecl calling convention          /Gr __fastcall calling
convention
    /Gz __stdcall calling convention        /GZ Enable stack checks (/
RTCs)
    /QIfist[-] use FIST instead of ftol()
    /hotpatch ensure function padding for hotpatchable images
    /arch:<SSE|SSE2> minimum CPU architecture requirements, one of:
        SSE - enable use of instructions available with SSE enabled
CPUs
        SSE2 - enable use of instructions available with SSE2 enabled
CPUs

                                  -OUTPUT FILES-

    /Fa[file] name assembly listing file    /FA[scu] configure
assembly listing
    /Fd[file] name .PDB file                /Fe<file> name executable
file
    /Fm[file] name map file                 /Fo<file> name object file
    /Fp<file> name precompiled header file  /Fr[file] name source
browser file
    /FR[file] name extended .SBR file
    /doc[file] process XML documentation comments and optionally name
the .xdc file

                                  -PREPROCESSOR-

    /AI<dir> add to assembly search path    /FU<file> forced using
assembly/module
    /C don't strip comments                 /D<name>{=|#}<text> define
macro
    /E preprocess to stdout                 /EP preprocess to stdout,
no #line
    /P preprocess to file                   /Fx merge injected code to
file
    /FI<file> name forced include file      /U<name> remove predefined
macro
    /u remove all predefined macros         /I<dir> add to include
search path
    /X ignore "standard places"

                                    -LANGUAGE-

    /Zi enable debugging information        /Z7 enable old-style debug
info
    /Zp[n] pack structs on n-byte boundary  /Za disable extensions
    /Ze enable extensions (default)         /Zl omit default library
name in .OBJ
    /Zg generate function prototypes        /Zs syntax check only
    /vd{0|1|2} disable/enable vtordisp      /vm<x> type of pointers to
members
    /Zc:arg1[,arg2] C++ language conformance, where arguments can be:
        forScope[-] - enforce Standard C++ for scoping rules
        wchar_t[-] - wchar_t is the native type, not a typedef
    /ZI enable Edit and Continue debug info
    /openmp enable OpenMP 2.0 language extensions

                                  -MISCELLANEOUS-

    @<file> options response file           /?, /help print this help
message
    /bigobj generate extended object format /c compile only, no link
    /errorReport:option Report internal compiler errors to Microsoft
        none - do not send report
        prompt - prompt to immediately send report
        queue - at next admin logon, prompt to send report (default)
        send - send report automatically
    /FC use full pathnames in diagnostics   /H<num> max external name
length
    /J default char type is unsigned        /nologo suppress copyright
message
    /showIncludes show include file names   /Tc<source file> compile
file as .c
    /Tp<source file> compile file as .cpp   /TC compile all files
as .c
    /TP compile all files as .cpp           /V<string> set version
string
    /w disable all warnings                 /wd<n> disable warning n
    /we<n> treat warning n as an error      /wo<n> issue warning n
once
    /w<l><n> set warning level 1-4 for n    /W<n> set warning level
(default n=1)
    /Wall enable all warnings               /WL enable one line
diagnostics
    /WX treat warnings as errors            /Yc[file] create .PCH file
    /Yd put debug info in every .OBJ        /Yl[sym] inject .PCH ref
for debug lib
    /Yu[file] use .PCH file                 /Y- disable all PCH
options
    /Zm<n> max memory alloc (% of default)  /Wp64 enable 64 bit
porting warnings

                                    -LINKING-

    /LD Create .DLL                         /LDd Create .DLL debug
library
    /LN Create a .netmodule                 /F<num> set stack size
    /link [linker options and libraries]    /MD link with MSVCRT.LIB
    /MT link with LIBCMT.LIB                /MDd link with MSVCRTD.LIB
debug lib
    /MTd link with LIBCMTD.LIB debug lib

_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

this is in a Firefox build with SSE...

about:buildconfig

Build platform
target
i686-pc-mingw32

Build tools
Compiler Version Compiler flags
cl 14.00.50727.762 -GL -wd4624 -wd4952 -TC -nologo -W3 -Gy -Fd$
(PDBFILE)
cl 14.00.50727.762 -GR- -GL -wd4624 -wd4952 -TP -nologo -Zc:wchar_t-
-W3 -Gy -Fd$(PDBFILE)

Configure arguments
--enable-application=browser '--enable-optimize=-Ox -GAFs -GL -Gy -
arch:SSE -fp:fast' --enable-update-channel=release --enable-update-
packaging --disable-debug --disable-tests --enable-strip --enable-
strip-libs --enable-jemalloc
_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

Firefox should Have...

SSE builds optimized.

SSE is in many CPUs...


this site test Firefox with SSE and without
Written by: Shivanand Sharma
http://www.binaryturf.com/a-firefox-faster-than-firefox/

_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

[Boris Zbarsky]

dlhdavidlh@... wrote:
> this site test Firefox with SSE and without
> Written by: Shivanand Sharma
> http://www.binaryturf.com/a-firefox-faster-than-firefox/

That doesn't seem to show much of a difference; certainly not worth the
hassle of maintaining multiple different builds, if that's what would be
involved.

I realize it's just one benchmark, by the way.

-Boris
_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

[Boris Zbarsky]

dlhdavidlh@... wrote:
>   - this what i have found...
> - i have tests on
>  http://dlhdragon.110mb.com/webbrowsertests.html

Out of curiosity, was the "Firefox 3.0.1" build your own build or the
Mozilla.org one?

-Boris
_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

On Sep 19, 9:38 pm, Boris Zbarsky <bzbar...@...> wrote:
> dlhdavi...@... wrote:
> >   - this what i have found...
> > - i have tests on
> >  http://dlhdragon.110mb.com/webbrowsertests.html
>
> Out of curiosity, was the "Firefox 3.0.1" build your own build or the
> Mozilla.org one?
>
> -Boris

two are from Mozilla (Firefox 3.0.1 and Firefox 3.1b1)

the others are from (Firefox 3.0.1 w/SSE)
 http://binaryturf.com/software/blazing-fast-firefox-optimized-distributions/
_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance

On Sep 19, 9:36 pm, Boris Zbarsky <bzbar...@...> wrote:

> dlhdavi...@... wrote:
> > this site test Firefox with SSE and without
> > Written by: Shivanand Sharma
> >http://www.binaryturf.com/a-firefox-faster-than-firefox/
>
> That doesn't seem to show much of a difference; certainly not worth the
> hassle of maintaining multiple different builds, if that's what would be
> involved.
>
> I realize it's just one benchmark, by the way.
>
> -Boris

i am still testing and still need to add more...

also..

www.binaryturf.com has some benchmarks / tests..(Links)

here is the link
http://www.binaryturf.com/a-firefox-faster-than-firefox/
_______________________________________________
dev-performance mailing list
dev-performance@...
https://lists.mozilla.org/listinfo/dev-performance