z88dk forums

It's an interesting one that shows the benefit of code analysis and a couple of compiler techniques that sccz80 doesn't use: first of all knowing what value is held in registers and secondly hoisting up a common sub-expression and keeping that value live in a register. You can get a benefit for both...

There’s a few: https://github.com/z88dk/z88dk/wiki/Cla ... -Libraries

Best bet is to stick —math32 onto compilation line.

I'm not certain to be honest - I've got a memory of a comment somewhere indicating that when a __asm block is hit, the registers are reloaded afterwards if required.

But, in this case, what not use use vdp_vwrite() from <video/tms9918.h> ?

Thanks for the report, I've pushed a fix + tests so will hopefully work now.

You can have as much space as you like really. The default is that code_graphics starts at 50000: https://github.com/z88dk/z88dk/blob/mas ... pm.cfg#L59

I think it's failing/failling through the gaps because it's being held as constant by the codegenerator. I'll fix it up. I have to admit I normally use the syntax: extern char arr[] @ 64000; so I wouldn't have come across this bug.

I've just gone on a little Excalibur64 memory refresh. It's one of those machines where the screen is paged in low in memory, so we put anything that accesses VRAM into the code_graphics section which gets relocated to high memory on startup. It looks there are bunch of configurations for the the lo...

I use the following to debug disc spec errors: 1. In disc_create() change the #if 0 to #if 1 - this will fill the disc with track/sector/head 2. In cpm_write_file() comment out this memcpy: memcpy(h->image + offset, data, len); This means in mame you can then stick a breakpoint at 0x100, dump the me...

By the way I'm not sure the current appmake implementation is able to generate a a valid IMD image for the MBC-1200 series, as you can see I tricked CPMTOOLS by suggesting to work on a single side disk image. The skew handling in appmake is fairly "funky" - is that the problem or is it so...

After the work on VDP library last week, it's probably not a surprise that I've been workng on V9938 support. I've done an initial sweep and made the new modes compatible with the standard libraries, that is generic console and "monochrome" graphics are running if appropriate. I've not imp...

VG5000 is EF9345

Have you got a link to a page with details of the hardware? It sounds like an interesting target with modern features (as you get in RC2014 style machines) on an older design.

It's been a while since I looked at that feature, but I seem to remember that the initial benchmarks gained about 2-3% on pure C compiled code with sdcc. If you're calling some assembler code, odds are that you've already gained more than that by writing it assembler rather than compiling it in the ...

There's a BINARY directive available to include a raw binary file into a .asm file.

This is a good point: are we on a path to actually deprecate Newlib? I find it somehow inconsistent that we are suppossedly comming back to Classic but then the Arkos sources had to go under libsrc/_DEVELOPMENT and the Classic makefiles adjusted for that.. Also, do SP1 sources really need to go in ...

Over the past week or so, I've merged a series of improvements to the VDP library. I've not spotted any regressions, in fact there's been several improvements. In summary: - In the past all the VDP access routines were accessed via <msx.h> which although it made sense when only the MSX was supported...

That makes sense, I guess the di/ei we've got in the keyboard reading meant that any interaction wasn't observed.

That's great. The binaries should be signed (by me). I probably ought to add in the notarise step which I hope will mean that the binaries will just be able to run. I think the fact that you get a "can't be verified" and not "unknown developer" means that the signing actually wor...

Some targets supported by z88dk may be headless and the console is presented using a serial connection (either physical or bridged to network). On these machines, you may still want to display graphics, well now you can with the z88dk libraries! - z88dk now provides monochrome graphic support for Te...

DarkSchneider wrote: ↑Fri Mar 17, 2023 11:18 am To use it I have to update all the z88dk to new version or downloading the new stddef.h is enough?

It was a change in sccz80, so a new build of that is needed. The new version is in the 202030317 nightly build.

I've committed a fix that works with offsetof() in stddef.h #include <stddef.h> struct x { long table1[20]; double tablen[10]; }; typedef struct x xt; static struct x data; long offs_x = offsetof(struct x, tablen); long offs_x2 = offsetof(x, tablen); int offs_xt = offsetof(xt, tablen); int offs_data...

It looks like that sort of operation is supported by sdcc, but not by sccz80 (so I''ll add it in).

If you include <stddef.h> you'll get the usual offsetof() macro which maps to a __builtin function

The addresses are documented here: https://github.com/z88dk/z88dk/wiki/Classic-TMS9918 - they basically match the default MSX configuration. The VDP code is located here: https://github.com/z88dk/z88dk/tree/master/libsrc/video/tms9918 There's (assembler) variables named: __tms9918_pattern_name and _...

I've got a memory of something like this coming in the past and I thought it was great idea and started work on it.

However, I can't find any trace of it unfortunately.

Having noted a few people tripping over adding a new target to z88dk and the +z80 target only being available for newlib I thought I'd spend a Sunday afternoon documenting things and hopefully making life easier. There's been a fair bit of chatter about the newlib vs classic over the past year or so...