Pedro, and all:
It's interesting from your point of view, that the problem happens on older, less capable machines, because of their floating-point handling, as well as because of having to reduce the polyphony to avoid excessive processing overhead.
So I guess the older machines get hit by it both ways at once. I know of the problem because I actually test my courseware on older, less capable machines.
It is actually a promotional point for my offering, that you can take a machine that is gathering dust in the basement (because it won't run modern versions of Windows), and turn it into a wonderful music workstation.
I can still demonstrate this successfully using a 450 megahertz HP Vectra machine with only 384 megabytes of RAM.
As I think more about the problem, from what you said, I wonder if the problem I noticed, and the problem you reported (that David fixed), are actually the same problem. Perhaps it was something else fixed in that later version that fixed my problem.
I observed my problem on faster, more capable machines as well. It definitely happened on my Dell Dimension 1100 machine, which has an Intel(R) Celeron(R) CPU of 2.53 gigahertz.
I also observed that applying the re-packaged (altered version-designation) PPA fix (originally for Ubuntu 11.10) solved the problem, in that the very poor quality sound (where random new notes are dropped altogether, or sound only very briefly) does not occur with the PPA version of libfluidsynth1.
The problem also happened on my Acer Aspire laptop (which has two Celeron (R) processors of 2.0 GHz, but uses 64-bit architecture. That was observed on Ubuntu 11.10 (I have not yet converted that machine to 12.04). Again, the PPA fix (which was just a later version of libfluidsynth1), made the problem's symptoms no longer appear.
In fact, none of the machines I own - even the fastest and most capable, are immune from the problem.
But that being said, I tend to purchase computer hardware from the trailing edge of the technology curve, and Linux has prolonged their usefulness - perhaps indefinitely (though with some of the later releases, I am beginning to have my doubts).
- Aere
On Mon, 2012-05-28 at 15:56 +0200, Pedro Lopez-Cabanillas wrote:
Hi,
On Wednesday 23 May 2012, David Henningsson wrote:
> FluidSynth upstream) didn't make a bug fix release either. Perhaps we
> should do that some time soon...so that the fixes will go into 12.10 at
> least...
I'm not sure if only fixing that particular bug justifies a release. It
doesn't harm to do it, of course. But we may face very similar bugs in the
future. Let me explain.
I've found the bug you are talking about some time ago, while I was working
with fluidsynth for a Nokia device having an ARM processor:
http://sourceforge.net/apps/trac/fluidsynth/ticket/100
This bug was solved by David with this patch:
http://sourceforge.net/apps/trac/fluidsynth/changeset/435
The problem was very specific to lowering the polyphony, but the cause was not
uncommon and had plagued FluidSynth for a long time. The root problem is
denormals in floating point calculations:
http://en.wikipedia.org/wiki/Denormal_number
To show a similar problem, here is a simple test scenario.
1.- Sample MIDI file: http://www.miditrax.com/Aug/salsa.mid
2.- GeneralUser GS Soundfont: http://www.schristiancollins.com/generaluser.php
3.- Build FluidSynth (SVN head, or any previous release) in debug mode with
floats instead of doubles and FPE checks:
$ cmake -DCMAKE_BUILD_TYPE=Debug \
-Denable-debug=on \
-Denable-floats=on \
-Denable-fpe-check=on
Runtime results on an AMD Athlon(tm) XP processor:
fluidsynth: warning: FPE exception (before or in voice_write vol env):
Underflow
fluidsynth: warning: FPE exception (before or in voice_write vol env):
Denormal number
fluidsynth: warning: FPE exception (before or in voice_filter): Underflow
fluidsynth: warning: FPE exception (before or in voice_write vol env):
Denormal number Underflow
fluidsynth: warning: FPE exception (before or in voice_write vol env):
Denormal number
fluidsynth: warning: FPE exception (before or in voice_filter): Denormal
number Underflow
When running FS in some old CPUs, denormals penalize CPU performance and sound
quality, while in modern CPUs these denormal numbers are automatically fixed
in hardware, eliminating the problem.
My oldest machine is the only one I can access today, which is (slightly)
affected by denormals in FluidSynth:
$ cat /proc/cpuinfo
processor : 0
vendor_id : AuthenticAMD
cpu family : 6
model : 10
model name : AMD Athlon(tm) XP 2800+
stepping : 0
cpu MHz : 2086.903
cache size : 512 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow up
bogomips : 4173.80
clflush size : 32
cache_alignment : 32
address sizes : 34 bits physical, 32 bits virtual
power management: ts
My other machines with Intel processors (Core2Duo, Core-i7) implementing SSE2
instructions are immune. However, you can tell the GCC compiler to not use the
SSE2 instructions with this flag: -mno-sse2
$ cmake ... -DCMAKE_C_FLAGS=-mno-sse2 ...
The above configuration allows some FPU exceptions to be reported on runtime,
even in a modern CPU. I would like to know if somebody is able to reproduce my
test results and help fixing them.
Regards,
Pedro
_______________________________________________
fluid-dev mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/fluid-dev