summaryrefslogtreecommitdiffstatshomepage
path: root/src/mame/drivers/midxunit.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/mame/drivers/midxunit.cpp')
-rw-r--r--src/mame/drivers/midxunit.cpp19
1 files changed, 12 insertions, 7 deletions
diff --git a/src/mame/drivers/midxunit.cpp b/src/mame/drivers/midxunit.cpp
index d47202dfb11..c233650567f 100644
--- a/src/mame/drivers/midxunit.cpp
+++ b/src/mame/drivers/midxunit.cpp
@@ -12,13 +12,18 @@
Known bugs:
* POST test ends with "CUSTOM CHIP U76 BAD" message,
- however this is caused by fully clipped blitter DMA request:
- DMA command 8003: (bpp=0 skip=0 xflip=0 yflip=0 preskip=0 postskip=0)
- offset=00000000 pos=(0,257) w=448 h=1 clip=(0,0)-(511,0)
- palette=0505 color=0000 lskip=00 rskip=00 xstep=0100 ystep=0100 test=0000 config=0010
- it is not clear if it's due to DMA emulation flaws (unemulated command bit 6 - clipping mode, 0=offset??? method),
- or if some protection device may modify routine code in the RAM (unlikely).
- set BP 20D31340 to see check routine (notice: 20D31550: MOVE A14,@C08000E0h,0 have no effect because of FS0 selected).
+ However, this is caused by failed blitter DMA test, it's buggy code:
+ 20D31460: MOVI 100000h,A14 ; control register value, set H-clip access
+ 20D31490: MOVE A14,@C08000E0h,1 ; 32bit write control reg
+ 20D314C0: MOVI 1FF0000h,A14 ; H-clip range value 0-511
+ 20D314F0: MOVE A14,@C08000C0h,1 ; 32bit write H-clip reg
+ 20D31520: MOVI 300000h,A14 ; control register value, set V-clip access
+ 20D31550: MOVE A14,@C08000E0h,0 ; 16bit(!!) write, upper word ignored in MAME, control reg still set to H-clip access
+ 20D31580: MOVI 1FF0000h,A14 ; V-clip range value 0-511
+ 20D315B0: MOVE A14,@C08000C0h,1 ; 32bit write, in MAME write goes to H-clip registers
+ in result of not set V-clip range in MAME the whole DMA request is clipped.
+ It is possible X-unit DMA registers block C08000xxh access is 32-bit only (/CAS0-3 lines ignored), so any write access always update whole 32bits.
+ At the moment TMS34020 incorrectly emulated as 16bit wide data bus CPU, so we can't properly handle this issue.
***************************************************************************