Android application to manipulate sd image from the site
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
Chris Johnson (125) 825 posts |
ROOL in their latest DDE release deprecate the use of !Make. For a long time I have used amu directly, and the makefile is a simple text file (filetype text) which I produce manually. I actually have a skeleton default makefile which I modify as necessary for different projects. The makefile can be made compatible with both AMU and the GCC Make utility, although the reverse is not necessarily true – you can do a lot more in GCC Make. |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaaa |
Ronald May (387) 407 posts |
Compressing to sqar and it looks OK, but decompressing the sqar and I get why -lz?I’ve stopped linking in zlib, but I guess it was sitting idle in the binary anyway. I compiled and run on linux machine. A couple of warnings The binary compress’s a tar but when decompressing, the output is like a quick repeat of the compressing stage.Maybe .sqar is not being checked for properly? |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
Ronald May (387) 407 posts |
I see, On one occasion I got an output for the top directory that looked the right number of characters but wrong values. |
WPB (1391) 352 posts |
ADFS doesn’t mandate an encoding scheme, so the OS will interpret the filename strings using whatever alphabet is set for the system. Most of the time that’s going to be Latin1, in which 0xA0 is defined as a non-breaking space. You will need to assume 8 bit “characters”, not 7 bit, definitely. If you want to convert to UTF-8, it’s a simple conversion from Latin1 to UTF-8. |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
h0bby1 (2567) 480 posts |
aaaa |
Rick Murray (539) 13850 posts |
The C library sets up a big jump table in order to avoid the overheads in taking a SWI call. Don’t pay to much attention to the quoted time given, that was probably correct for an ARM2 ;-) Just know that it is the difference between a SWI exception (a lot of code) and a BL (to jump table) followed by a B (into CLib). |
Ronald May (387) 407 posts |
I have compiled both stdc and ro versions with with !GCC and both are working on my Iyonix.I will upload them if you would like them to compare to your own versions. In your gcc makefile Converting names and directories is not necessary for compiling in RISC OS, The standard !GCC works with norcroft c and h dirs, maybe you are making it so you can use norcroft source on other platforms?
|
h0bby1 (2567) 480 posts |
aaaaa |
h0bby1 (2567) 480 posts |
aaaaa |
Ronald May (387) 407 posts |
But well yes in the absolute, the goal is to have a single source tree for visual studio and gcc/linux and riscos. I think it would be better to leave the makefile alone and make a script for each platform to convert norcroft to plain.A RISC OS obey file with would do it for RISC OS, other platforms would have their own scripting.
Which directory/archive did you use for time difference testing? |
h0bby1 (2567) 480 posts |
aaaa |
Ronald May (387) 407 posts |
I use the riscos source tree archive from this site for testing :) OK compressing the BCM2835Dev on Iyonix Edit: with the -O2 switch no improvement for STDC, but RO now compresses in 168 secs. |
h0bby1 (2567) 480 posts |
aaaa |
Rick Murray (539) 13850 posts |
Using CLib (as a module) sets up the jump table. It doesn’t issue SWIs for each call. The alternative is calling SWIs. Let’s see… The processor takes the SWI vector. The OS needs to load the instruction (as data, polluting data cache) to read the SWI number. The OS then needs to go through all the logic of working out what module to actually call. Then the module will interpret the SWI number via its own jump table. Then the function you wanted can be executed. Phew! |
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15