Slow compiling with OS4.1FE

[xenic]

Since I switched to OS4.1FE it seemed like compiles were taking forever. I decided to run some tests to compare OS4.1FE compile times with OS4.1u6 compile times. I tested compiling several large multifile programs with the timer command from OS4Depot and discovered that compiling takes almost twice as long with OS4.1FE as with OS4.1u6! Here are the timing comparisons for compiling the Dopus5 Program only (no library or modules):

With OS4.1 Update6:
Compile times: 225.2183 seconds & 222.6382 seconds
CleanAll times: 0.2826 seconds & 0.3621 seconds

With OS4.1FE:
Compile times: 541.3016 seconds & 546.668 seconds
CleanAll times: 1.4830 seconds & 1.3842 seconds

The above timings are for the program only. Compiling the entire Dopus5 application (Program, Library & modules) takes a huge amount of time with FE. Turnaround compile time for debugging is almost unbearable.

I also tested compiling a large single-file program with OS4.1FE and OS4.1u6 but the difference in compile times were insignificant. I could be wrong but my conclusion is that something has gone wrong with loading source files, commands and/or Sobjs in OS4.1FE.

I've had to revert to OS4.1 Update6 for compiling and debugging some open-source programs that I work on.

[tonyw]

Intriguing results.

What is the difference between the two (boot) systems? I mean, how are the volumes arranged that hold the sources, the SDK installation, the system, etc?

Are they all on the same device (HD) ? What filesystems do they all use?

[xenic]

Intriguing results.

What is the difference between the two (boot) systems? I mean, how are the volumes arranged that hold the sources, the SDK installation, the system, etc?

Are they all on the same device (HD) ? What filesystems do they all use?

Both boot partitions and the SDK are all on the same hard-disk. The FE partition is an SFS partition as recommended. The Update6 partition is an FFS partition. I switch boot partitions by changing the boot priorities with Media Toolbox and cold rebooting. I check that I'm using the correct kickstart by checking the version of dos.library. The SDK is located on an SFS2 partition. The sources being compiled are on another SFS2 partition.

The results should be easy to confirm with any sources that have multiple source files and a makefile. Just D/L timer from OS4Depot if you don't already have it, enter a source directory and enter "timer make".

[Slayer]

That is peculiar and needs investigating!

[zzd10h]

X1000
2 SFS/2 boot partitions (4.1.6 & FE)
SDK on the same SFS2 partition

Compiling little program (SysMon) makes no difference

4.1.6
SysMon> timer make
Command duration: 12.2669 sec.

FE
SysMon> timer make
Command duration: 12.3504 sec.


Compiling bigger program (latest SMTube svn revision) seems to be slower
but only 33% slower for me not >100% like for DOpus. SMTube is not as big as DOpus.

4.1.6
SMTube > timer gmake
Command duration: 157.0790 sec.

FE
SMTube > timer gmake
Command duration: 195.6692 sec.

[xenic]

@zzd10h
Yes. The larger the compile (number of source and header files) the bigger the time difference between Update6 and FE. Here are my compile times for the entire Dopus5 sources (program, library, 20 modules, 3 commands):

Compile time with OS4.1 Update6: 506.3468 seconds (8.44 minutes)
Compile time with OS4.1FE: 1375.7705 seconds (22.93 minutes)

That's a huge difference for anyone compiling a major application for OS4.

Here are the "clean" times for the entire Dopus5 sources:

Clean time with OS4.1 Update6: 4.3809 seconds
Clean time with OS4.1FE: 105.6004 seconds

That's 20 times slower!! Even if there are no object files to delete (clean already done) the times are about the same. It's rediculous. When the sources have already been cleaned, the clean operation is just loading sub-makes and scanning for object files.

[xenic]

That is peculiar and needs investigating!

I investigated some more and discovered that the problem is that SDK:C/make doesn't work as it should. I had to revert back to the "make" command from SDK v51.22 to get a normal compile that is just as fast as the current "make" running with OS4.1 Update6.

[xenic]

I did some further investigation and found that the make command works at normal speeds (as fast as with Update6) if I boot from an FFS partition. I also discovered that if I edit the startup-sequence on an SFS boot partition and change the PYTHON: assignment to the Python directory on an FFS partition, the make command will operate at normal speed when I boot from the SFS partition with the changed PYTHON: assignment. Strange.

[Raziel]

It would ne nice to know what exactly make has to do with Python then

It would make more sense if it's a problem in Python as this part has been updated for FE, if i recall right?

[zzd10h]

@Xenic
I tried the gmake that you sent me.

No difference under FE for the same SMTube source


gmake -v => GNU Make 3.81rc2-4
timer gmake => 189.0125 sec.

previous test with original gmake under FE took 195.6692 sec.