On the plasma demo I noticed when running with the -hw option (the option to
put the surface in video memory) it actually ran slower than using the regular
system memory surface. I was running as root and in fullscreen both times.
I thought that accessing the surface in video memory would be faster? Im
thinking maybe it has something to do with locking the surface?
Nope, raw video memory accesses really are slower than accelerated blits
using the X server. Try building SDL on a 2.2 kernel (enabling MTRR support),
and see if it’s faster.
-Sam Lantinga (slouken at devolution.com)
Lead Programmer, Loki Entertainment Software–
“Any sufficiently advanced bug is indistinguishable from a feature”
– Rich Kulawiec
On the plasma demo I noticed when running with the -hw option (the option to
put the surface in video memory) it actually ran slower than using the regular
system memory surface. I was running as root and in fullscreen both times.
I thought that accessing the surface in video memory would be faster? Im
thinking maybe it has something to do with locking the surface?
-Garrett, WPI student majoring in Computer Science.
“He who joyfully marches in rank and file has already earned
my contempt. He has been given a large brain by mistake, since
for him the spinal cord would suffice.” -Albert Einstein
ummm wow. I just compiled MTRR support into my kernel and my framerate for
the fullscreen hardware plasma demo went from ~60fps to ~170fps on average. Its
about the same as running it in system memory now.
-Garrett, WPI student majoring in Computer Science.
“He who joyfully marches in rank and file has already earned
my contempt. He has been given a large brain by mistake, since
for him the spinal cord would suffice.” -Albert EinsteinOn Fri, 09 Jul 1999, you wrote:
On the plasma demo I noticed when running with the -hw option (the option to
put the surface in video memory) it actually ran slower than using the regular
system memory surface. I was running as root and in fullscreen both times.
I thought that accessing the surface in video memory would be faster? Im
thinking maybe it has something to do with locking the surface?
Nope, raw video memory accesses really are slower than accelerated blits
using the X server. Try building SDL on a 2.2 kernel (enabling MTRR support),
and see if it’s faster.
Garrett wrote:
On the plasma demo I noticed when running with the -hw option (the option to
put the surface in video memory) it actually ran slower than using the regular
system memory surface. I was running as root and in fullscreen both times.
I thought that accessing the surface in video memory would be faster? Im
thinking maybe it has something to do with locking the surface?
Nope, raw video memory accesses really are slower than accelerated blits
using the X server. Try building SDL on a 2.2 kernel (enabling MTRR support),
and see if it’s faster.
ummm wow. I just compiled MTRR support into my kernel and my framerate for
the fullscreen hardware plasma demo went from ~60fps to ~170fps on average. Its
about the same as running it in system memory now.
I am already getting about the same framerate for hardware as I am for software, does
that mean I already have mtrr support built in?
Also the framerate I am getting is from ~29fps to ~40fps. This is running on a 400mhz
98meg of ram CPU. What kind of computer do you have to yield a 170fps framerate? Are
there any tricks to compiling SDL for speed that I should know about? I tried it with
and without NASM installed and it didn’t make a difference with the cvs version.> On Fri, 09 Jul 1999, you wrote:
-Garrett, WPI student majoring in Computer Science.
“He who joyfully marches in rank and file has already earned
my contempt. He has been given a large brain by mistake, since
for him the spinal cord would suffice.” -Albert Einstein
ummm wow. I just compiled MTRR support into my kernel and my framerate for
the fullscreen hardware plasma demo went from ~60fps to ~170fps on average. Its
about the same as running it in system memory now.
I am already getting about the same framerate for hardware as I am for software, does
that mean I already have mtrr support built in?
Best way to check is check your kernel configuration.
Also the framerate I am getting is from ~29fps to ~40fps. This is running on a 400mhz
98meg of ram CPU. What kind of computer do you have to yield a 170fps framerate? Are
there any tricks to compiling SDL for speed that I should know about? I tried it with
and without NASM installed and it didn’t make a difference with the cvs version.
Im just did a regular compile and had nasm. Im also on a pii400 with 128 megs
of ram and a Millenium II vid card. I think its more important what vid card
you have. I think the MTRR is only going to make a difference on lots of small
operations to the video memory, because it groups up all the operations and
sends them across the PCI bus in one big group. If your sending large blts
across the bus it probably won’t make any difference.
-Garrett, WPI student majoring in Computer Science.
“He who joyfully marches in rank and file has already earned
my contempt. He has been given a large brain by mistake, since
for him the spinal cord would suffice.” -Albert EinsteinOn Sat, 10 Jul 1999, you wrote: