shouse: sound RAM2 error
jotego opened this issue · comments
Entering test mode displays a "SOUND RAM2 ERROR" message. This might be the reason why music/FX gets weird sometimes in rompers or the missing fire sound in bakutotu.
RAM2 Error seems to refer to the Tri RAM. It currently has a hack around xwe, but removing it conditionally does not make the test pass:
wire xwe_aux = debug_bus[0] ? xwe : xwe && (xaddr !=0 || xdout == 8'ha6);
/* verilator tracing_off */
jtframe_dual_ram #(.AW(11)) u_ram(
// Port 0: main and sub CPUs
.clk0 ( clk ),
.data0 ( bdout ),
.addr0 ( baddr ),
.we0 ( bwe ),
.q0 ( p_bdin ),
// Port 1
.clk1 ( clk ),
.data1 ( xdout ),
.addr1 ( xaddr ),
.we1 ( xwe_aux ), // see https://github.com/jotego/jtcores/issues/410
.q1 ( xdin )
);This might be the reason why dangseed cannot enter service mode and stays trapped in a loop.
This:
diff --git a/cores/shouse/hdl/jtshouse_game.v b/cores/shouse/hdl/jtshouse_game.v
index f1338832..03b8bbbb 100644
--- a/cores/shouse/hdl/jtshouse_game.v
+++ b/cores/shouse/hdl/jtshouse_game.v
@@ -241,7 +241,7 @@ jtshouse_triram u_triram(
.snd_cen ( cen_snd ),
.mcu_cen ( cen_mcu ),
- .snd_sel ( snd_sel ),
+ .snd_sel ( cen_sndq ),
.baddr ( baddr[10:0] ),
.mcu_addr ( eerom_addr ),
diff --git a/cores/shouse/hdl/jtshouse_triram.v b/cores/shouse/hdl/jtshouse_triram.v
index b2350135..253a63ba 100644
--- a/cores/shouse/hdl/jtshouse_triram.v
+++ b/cores/shouse/hdl/jtshouse_triram.v
@@ -123,7 +123,7 @@ jtframe_dual_ram #(.AW(11)) u_ram(
.clk1 ( clk ),
.data1 ( xdout ),
.addr1 ( xaddr ),
- .we1 ( xwe && (xaddr !=0 || xdout == 8'ha6) ), // see https://github.com/jotego/jtcores/issues/410
+ .we1 ( xwe && (xaddr !=0 || xdout == 8'ha6 || xsel) ), // see https://github.com/jotego/jtcores/issues/410
.q1 ( xdin )
);
fixes Rompers' service mode, but not dangseed's. I don't think the two issues have the same cause, the sound cpu doesn't even select tri-ram.
Not related, but I noticed that JTFRAME_LF_SDRAM_BUFFER is defined for the SiDi128 target. It's only needed for Outrun, isn't it?
Dangseed's reset loop is due to the watchdog.
But the watchdog kicks in because the CPU goes out of track.
If you keep pressing reset, eventually it'll enter service mode.
Not related, but I noticed that JTFRAME_LF_SDRAM_BUFFER is defined for the SiDi128 target. It's only needed for Outrun, isn't it?
Currently outrun is the only core using it, yes.
Not related, but I noticed that JTFRAME_LF_SDRAM_BUFFER is defined for the SiDi128 target. It's only needed for Outrun, isn't it?
Currently outrun is the only core using it, yes.
But then it was wrong to move it to the global target.def, because now all cores has 'no driver for SDRAM2_XXX'.
PR #658 fixed the RAM2 error message. Thank you!
I thought this would fix #641 too, but it didn't. Bad luck! At least dangseed service mode works.
I have tested the service mode on a few more romsets. Some of them produce errors. Not sure whether these errors were there before the PR or not. This is not a comprehensive list, some games are not here:
dangseed ok
splatter ok
quester ROM NO. 5 ERROR
mmaze TEST PROGRAM INITIALIZE ERROR
wldcourt ROM NO. 6 ERROR
ws89 ROM NO. 0 ERROR
bakutotu ok
rompers ok
pacmaniao ok
pistoldm ok
shland ok
soukobdx ok
tankfrcej ok
Not related, but I noticed that JTFRAME_LF_SDRAM_BUFFER is defined for the SiDi128 target. It's only needed for Outrun, isn't it?
Currently outrun is the only core using it, yes.
But then it was wrong to move it to the global target.def, because now all cores has 'no driver for SDRAM2_XXX'.
Let's take that discussion to a new issue
PR #658 fixed the RAM2 error message. Thank you! I thought this would fix #641 too, but it didn't. Bad luck! At least dangseed service mode works.
I have tested the service mode on a few more romsets. Some of them produce errors. Not sure whether these errors were there before the PR or not. This is not a comprehensive list, some games are not here:
dangseed ok splatter ok quester ROM NO. 5 ERROR mmaze TEST PROGRAM INITIALIZE ERROR wldcourt ROM NO. 6 ERROR ws89 ROM NO. 0 ERROR bakutotu ok rompers ok pacmaniao ok pistoldm ok shland ok soukobdx ok tankfrcej ok
At least not RAM errors :)
In mmaze and rompers, I hear music-like tunes, probably filters or MiSTer related bug?
I cannot reproduce the mmaze failure (it worked previously, it still works for me).
For ws, mirroring the ROMs fixes the ROM error. In terms of the MRA file, it should look like:
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg0.bin" crc="b0234298"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg1.bin" crc="dfd72bed"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part name="ws1_prg2.bin" crc="bb09fa9b"/>
<part repeat="0x200000"> FF</part>
<part name="ws1_prg7.bin" crc="28712eba"/>
<part repeat="0x10000"> FF</part>
ws1_prg7.bin does not need the mirroring to pass the ROM check, though, although mirroring it once does not break either the ROM check or the MRA file.
I am going to check if this mirroring can be done by jtframe mra
The ROM error problem was fixed by duplicating the files. A small change was introduced in jtframe mra to allow more files repetition, although for this case it was enough with the 2x already built it.
I cannot see the mmaze error now either, it might be related to the NVRAM.
I think we can close this issue. @jtmiki marked it as done. Thanks @gyurco