 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
You have to link with "cc -xarch=v9" not with "cc" only.
--
#macro N(D,I)#if(I<6)cylinder{M()#local D[I]=div(D[I],104);M().5,2pigment{
rgb M()}}N(D,(D[I]>99?I:I+1))#end#end#macro M()<mod(D[I],13)-6,mod(div(D[I
],13),8)-3,10>#end blob{N(array[6]{11117333955,
7382340,3358,3900569407,970,4254934330},0)}// - Warp -
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Warp wrote:
>
> You have to link with "cc -xarch=v9" not with "cc" only.
*SMACK* = the sound made when ones open hand hits ones forehead soundly.
OK .. I hope no one at work sees this.
Thanks.
Next issue, what is the difference between the sparc v9 and v9a processor? The
UltraSparc I is referred to as the v8plusa yes? Then what is the v9a?
UltraSPARC II I guess. I'm going to compile for that architecture as well and
then test with your sample file until I get the fastest time. The sample file I
am using is :
----------8<----------8<----------8<----------8<----------8<----------
camera { location <0,2,-2> look_at y*.2 angle 35 }
light_source { <10,10,-5>, 1 }
#declare WPRadius = 1;
#declare WPCurvRadius = 1.5;
intersection
{ #declare IPH = sqrt(pow(WPCurvRadius,2)-pow(WPRadius,2));
torus { WPRadius, WPCurvRadius translate -y*IPH}
cylinder { 0, y*(WPCurvRadius-IPH), WPRadius }
pigment
{ spiral1 1 color_map { [0 rgb <.5,.75,1>][1 rgb <0,.5,1>] }
turbulence .1
rotate x*90
scale <.1,1,.1>
}
normal
{ spiral1 1 .15 slope_map
{ [0 <0,0>][.25 <.5,1>][.5 <1,0>][.75 <.5,-1>][1 <0,0>] }
rotate x*90
scale <.2,1,.2>
}
finish { specular .5 reflection .3 }
}
----------8<----------8<----------8<----------8<----------8<----------
which comes from one of your earlier posts. The system that I'm testing
with is a Sun Ultra 2 with two 300MHz processors. I believe that the CPUs have
2Mb of L2 cache each. The WorkShop compiler will allow me to specify a target
system and that will, in theory, tailor a few other options to maximize
performance on that system. I really don't know how much I believe that but any
binary built for the Ultra 2 will probably show poor performance on a SunBlade
100 with only 256Kb of cache.
I'll post my results. At the moment I'm trying something a tad extreme.
CFLAGS=-dalign -fsimple=0 -xO4 -ftrap=%none -libmil
-s -temp=/tmp/compiler -v -Xa -xarch=v9a
Dennis Clarke
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Results :
Binary:Sparc v9a ELF 64-bit MSB executable SPARCV9 Version 1,
UltraSPARC1 Extensions Required, dynamically linked, not stripped
:command : povray_sparc_v9a -I./test.pov -f -d -x -p -w640 -h480 +a.1
:Total Time: 0 hours 2 minutes 33.0 seconds (153 seconds)
Binary:Sparc v9 ELF 64-bit MSB executable SPARCV9 Version 1,
dynamically linked, not stripped
:command : povray_sparc_v9 -I./test.pov -f -d -x -p -w640 -h480 +a.1
:Total Time: 0 hours 2 minutes 36.0 seconds (156 seconds)
Binary:Sparc v8plusa ELF 32-bit MSB executable SPARC32PLUS Version 1,
V8+ Required, UltraSPARC1 Extensions Required,
dynamically linked, notstripped
:command : povray_sparc_v8 -I./test.pov -f -d -x -p -w640 -h480 +a.1
:Total Time: 0 hours 2 minutes 17.0 seconds (137 seconds)
Binary:Sparc v1 ELF 32-bit MSB executable SPARC Version 1,
dynamically linked, not stripped
:command : povray_sparc_v1 -I./test.pov -f -d -x -p -w640 -h480 +a.1
:Total Time: 0 hours 2 minutes 37.0 seconds (157 seconds)
Binary:Sparc Generic ELF 32-bit MSB executable SPARC Version 1,
dynamically linked, not stripped
:command : povray_sparc_generic -I./test.pov -f -d -x -p -w640 -h480 +a.1
:Total Time: 0 hours 4 minutes 28.0 seconds (268 seconds)
Clearly the Sparc v8plusa is the winner here. It was compiled with options
-fast -xarch=v8plusa -Xa. I recompiled the binary with the following options
just to be sure
-dalign -fsimple=0 -xO4 -ftrap=%none -libmil -s -temp=/tmp/compiler -v -Xa
-xarch=v8plusa and that resulted in Total Time: 0 hours 2 minutes 32.0
seconds (152 seconds)
Dennis Clarke
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Dennis Clarke <dcl### [at] blastwave com> wrote:
: Clearly the Sparc v8plusa is the winner here. It was compiled with options
: -fast -xarch=v8plusa -Xa. I recompiled the binary with the following options
: just to be sure
: -dalign -fsimple=0 -xO4 -ftrap=%none -libmil -s -temp=/tmp/compiler -v -Xa
: -xarch=v8plusa and that resulted in Total Time: 0 hours 2 minutes 32.0
: seconds (152 seconds)
AFAIK the option -fast turns on the optimal flags (or what it thinks are
the optimal flags) for the target architecture (besides using -xO5). Probably
the flags you used in that latter case are not the optimal for the CPU and
that explains the 15 seconds difference.
And by the way. You should never restrict your benchmarks to just one
POV-Ray file, but you should always test with several files which use
different aspects of POV-Ray. Different features the optimized differently
with the several options and CPU targets.
I recommend you to test at least with POV files like:
- A scene which is slow to parse but very fast to render.
- A scene which doesn't take any considerable memory, but takes time to
render.
- A scene which takes huge amounts of memory (eg. several tens of megabytes)
and render time.
- Perhaps scenes which test individual features of POV-Ray, like lighting,
several types of primitives, meshes, complex textures and so on.
Judging based on just one simple scene may give wrong results in some
cases.
--
#macro N(D,I)#if(I<6)cylinder{M()#local D[I]=div(D[I],104);M().5,2pigment{
rgb M()}}N(D,(D[I]>99?I:I+1))#end#end#macro M()<mod(D[I],13)-6,mod(div(D[I
],13),8)-3,10>#end blob{N(array[6]{11117333955,
7382340,3358,3900569407,970,4254934330},0)}// - Warp -
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Dennis Clarke <dcl### [at] blastwavecom> wrote:
: Next issue, what is the difference between the sparc v9 and v9a processor? The
: UltraSparc I is referred to as the v8plusa yes? Then what is the v9a?
AFAIK v8plusa compiles 32-bit code for the UltraSPARC. The man page for CC
tells the following:
v8plusa Compile for the V8plusa version of the
SPARC-V9 ISA.
By definition, V8plusa means the V8plus
architecture, plus the Visual Instruc-
tion Set (VIS) version 1.0, and with
UltraSPARC extensions. This option
enables the compiler to generate code
for good performance on the UltraSPARC
architecture, but limited to the 32-bit
subset defined by the V8plus specifica-
tion. The resulting object code is in
SPARC-V8+ ELF32 format and only executes
in a Solaris UltraSPARC environment --
it does not run on a V7 or V8 processor.
v9 compiles 64-bit code for the same architecture:
v9 Compile for the SPARC-V9 ISA.
Enables the compiler to generate code
for good performance on the V9 SPARC
architecture. The resulting .o object
files are in ELF64 format and can only
be linked with other SPARC-V9 object
files in the same format. The resulting
executable can only be run on an
UltraSPARC processor running a 64-bit
enabled Solaris operating environment
with the 64-bit kernel.
-xarch=v9 is only available when compil-
ing in a 64-bit enabled Solaris environ-
ment.
v9a is the same as v9, but it uses some UltraSPARC extensions:
v9a Compile for the SPARC-V9 ISA with
UltraSPARC extensions.
Adds to the SPARC-V9 ISA the Visual
Instruction Set (VIS) and extensions
specific to UltraSPARC processors, and
enables the compiler to generate code
for good performance on the V9 SPARC
architecture. The resulting .o object
files are in ELF64 format and can only
be linked with other SPARC-V9 object
files in the same format. The resulting
executable can only be run on an
UltraSPARC processor running a 64-bit
enabled Solaris operating environment
with the 64-bit kernel.
-xarch=v9a is only available when com-
piling in a 64-bit enabled Solaris
operating environment.
There exists also a v9b, but it works only if you have an UltraSPARC-III.
--
#macro N(D,I)#if(I<6)cylinder{M()#local D[I]=div(D[I],104);M().5,2pigment{
rgb M()}}N(D,(D[I]>99?I:I+1))#end#end#macro M()<mod(D[I],13)-6,mod(div(D[I
],13),8)-3,10>#end blob{N(array[6]{11117333955,
7382340,3358,3900569407,970,4254934330},0)}// - Warp -
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
