Possible bug with AUGS followed by ALTS

Cluso99

Here is the suspect code...

                alts	y, ##_TABLE_OPC0		'\ set next S to offset+table addrs
                mov	p, #0-0				'/

_TABLE_OPC0 is a table located in hub ram
y is an index into the table

When y + _TABLE_OPC0 causes an overflow from >$1FF, the overflow is lost.

Here is a sample output where the first hex is the value of "y", the second hex is the hub location of "_TABLE_OPC0", and the third hex is the resultant (sum) "p" value.

=== Cluso's P2v10a Debugger v.105x ===
000000C0+00001508=000015C8  MIN
000000C8+00001508=000015D0  MAX
000000D0+00001508=000015D8  MINS
000000D8+00001508=000015E0  MAXS
000000E0+00001508=000015E8  SUMC
000000E8+00001508=000015F0  SUMNC
000000F0+00001508=000015F8  SUMZ
000000F8+00001508=00001400
00000100+00001508=00001408
00000108+00001508=00001410
00000110+00001508=00001418

Note that when $F8 is added to $1508 the result should be $1600 but is shown as $1400 (ie the $200 carry is lost). It seems that the addition is limiting the size of the addition to 9 bits.

The attached file uses my LMM Serial Debugger code for display purposes on the serial P30/31 port at 115200 baud.

ozpropdev

Cluso99 wrote: »

                alts	y, ##_TABLE_OPC0		'\ set next S to offset+table addrs
                mov	p, #0-0				'/

The AUGS instruction only applies to the ALTS instruction and not the following MOV instruction.
The MOV instruction only has a single '#' immediate symbol so will always be limited to 9 bits unless it also has its own AUGS preceeding it.

ozpropdev

The ALTS instruction should cancel the AUGS mechanism after it's use.
What seems to be happening is the AUGS is also being applied to the MOV instruction too.

cgracey

ALTS is used to indirectly access cog registers, not hub RAM.

I actually added a section to the doc file on this. I'm going to go add in the ALTI description now.

Cluso99

Thanks Chip.

I can use this sequence instead...

                alts	lmm_w, ##_TABLE_CCCC		'\ set next S to offset+table addrs
                mov	lmm_p, #0-0			'/
replace with...
		mov	lmm_p, ##_TABLE_CCCC
		add	lmm_p, lmm_w

Here is where I use the sequence to decode the condition bits to an "if_xxx" string

                and     lmm_w, #$0F                     ' extract 'cccc'
		mul	lmm_w, #12			' cccc offset *12

'                alts	lmm_w, ##_TABLE_CCCC		'\ set next S to offset+table addrs
'                mov	lmm_p, #0-0			'/
		mov	lmm_p, ##_TABLE_CCCC
		add	lmm_p, lmm_w

                mov     lmm_c, #12                      ' set count of chars for the "if" part
.ifcccc         rdbyte  lmm_x, lmm_p                    ' get the byte from table
                add     lmm_p, #1                       ' point to next byte
              CALL   	#_HubTx                         '               	< call: transmit char(s) >
		djnz	lmm_c, #.ifcccc

I have tables in hub for the various conditions, opcode sets, etc, for decoding instructions in my Lmm Serial Debugger.