Thanks, Chip. Here are my final comments on the instruction set. Most suggestions are new, the others repeated for completeness. My thinking is that if one instruction must move, the set will have changed and might as well move some others and have done with it. Numbering is by ascending opcodes, not importance.
* = instruction moved or changed
1. TEST and others
Move as shown below, so that all three instructions that use S/!S, namely AND/ANDN and MOV/NOT and TEST/TESTN, have opcode bits [23:21] = 000/001, making the coding more consistent and harmonious.
Does not involve D (nor S), so out of place in with the registered D instructions. I think it could be moved in front of POLLINT, with bit 17 changed from 0 to 1 for POLLxxx and WAITxxx.
Possibly move as shown, so that SPLITx/MERGEx are split, not merged!
Before:
EEEE 1101011000 DDDDDDDDD 001100000SPLITB D
EEEE 1101011000 DDDDDDDDD 001100001MERGEB D
EEEE 1101011000 DDDDDDDDD 001100010SPLITW D
EEEE 1101011000 DDDDDDDDD 001100011MERGEW D
After:
EEEE 1101011000 DDDDDDDDD 001100000SPLITB D
EEEE 1101011000 DDDDDDDDD 001100001 * SPLITW D
EEEE 1101011000 DDDDDDDDD 001100010 * MERGEB D
EEEE 1101011000 DDDDDDDDD 001100011MERGEW D
Modified instructions.txt with above changes attached.
Whether or not any of the above are implemented, that's me finished with the instruction set!
TonyB_, think in terms of what looks most sensible to programmers and tool makers. At this point, there are no huge gains to be gotten by simplifying decoding, as things are already compartmentalized behind the scenes. How it flows in peoples' minds is more important.
But what if avoiding names is promoting them? Maybe we should "beepbeepbeep" the word "Propeller" to make it the best known word in the world? What if HP outside the (electro-) technical world is expanded in the sense of PLATO
Happy new year !
I'm using - no I'm enjoying - tachyon with P1 and one thing I'm definitely missing is flow control. Although Peter stated above that there currently is no fiddling with newline delay and character delay on the P2, I'm sure in future there will be.
From my point of view the lack of flow control would avoid unmanned builds which is inconvenient in times of robots ...
Best regards,
proplem
I also use FL fast load once I have EASYFILE loaded so that everything is buffered into a temporary or named file and loaded from the file where it is not necessary for flow control. With TAQOZ in P2 ROM this should be possible since I have FAT32 already resident, all it needs is an SD card. One advantage of downloading to a named file is that it is also an easy way to transfer source code onto the card. Maybe some of the newer USB serial chips have proper flow control built into them because as we know you can't rely upon the O/S to do it.
I think I've got the documentation all up-to-date, including all CLKSET functions and the timeout modes in the smart pins.
Chip,
Changing the encoding of the selector bits to be at LSbit end will allow for the reset and write protect functions to be encoded as single instructions.
Eg:
CLKSET D/# - Configure global circuit selected by MSBs
%DDDD_DDMM_MMMM_MMMM_PPPP_CCSS_Exxx_xx10 Set clock generator mode
%0000_0000_0000_0000_0000_0001_0000_0000 Hard reset, reboots chip
%0000_0000_0000_0000_0000_000P_1000_0000 Set write-protect of last16KB RAM to P
%0000_0000_0000_0000_RRLL_TTTT_T100_0000 Set filter R to length L and tap T
%DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDD1 Seed Xoroshiro128+ PRNG with D
I think I've got the documentation all up-to-date, including all CLKSET functions and the timeout modes in the smart pins.
Chip,
Changing the encoding of the selector bits to be at LSbit end will allow for the reset and write protect functions to be encoded as single instructions.
Eg:
CLKSET D/# - Configure global circuit selected by MSBs
%DDDD_DDMM_MMMM_MMMM_PPPP_CCSS_Exxx_xx10 Set clock generator mode
%0000_0000_0000_0000_0000_0001_0000_0000 Hard reset, reboots chip
%0000_0000_0000_0000_0000_000P_1000_0000 Set write-protect of last16KB RAM to P
%0000_0000_0000_0000_RRLL_TTTT_T100_0000 Set filter R to length L and tap T
%DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDD1 Seed Xoroshiro128+ PRNG with D
What I don't like about that is that it makes setting the clock require ## for most cases. I think setting the clock is going to be the most common use of this instruction, so it needs to be the simplest.
Here is how I see the the use cases in order of descending frequency:
1) setting the clock (most common, as apps need to save power)
2) write-protecting the last 16KB of RAM (used once/twice? by OS-type software)
3) setting the filters (probably a one-time thing)
4) resetting the chip (this has got to be rare)
5) seeding the PRNG (done by boot ROM, no need after)
It would be nice to be able to set 160MHz PLL operation from a 20MHz crystal. That would require something like this:
%P_PPPD_DDDD_DMMM_MMMM_MMME_CCSS
CLKSET #%1111_1_10_11
That seems kind of messy, though, as it reorders the PLL config fields in a way that doesn't flow logically from left to right.
The way it is now is maybe fine, considering the alternatives.
It would be nice to be able to set 160MHz PLL operation from a 20MHz crystal. That would require something like this:
%P_PPPD_DDDD_DMMM_MMMM_MMME_CCSS
CLKSET #%1111_1_10_11
That seems kind of messy, though, as it reorders the PLL config fields in a way that doesn't flow logically from left to right.
The way it is now is maybe fine, considering the alternatives.
I've noticed E field can be derived from CC field. How about splitting up the PLL config from the clock selection:
CLKSET D/# - Configure global circuit selected by MSBs
%0000_0000_PPPP_DDDD_DDMM_MMMM_MMMM_CC10 Config clock PLL generator
%0000_0000_0000_0000_0000_0000_SS10_0000 Select clock source
%0000_0000_0000_0000_0000_0001_0000_0000 Hard reset, reboots chip
%0000_0000_0000_0000_0000_000P_1000_0000 Set write-protect of last16KB RAM to P
%0000_0000_0000_0000_RRLL_TTTT_T100_0000 Set filter R to length L and tap T
%DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDD1 Seed Xoroshiro128+ PRNG with D
It would be nice to be able to set 160MHz PLL operation from a 20MHz crystal. That would require something like this:
%P_PPPD_DDDD_DMMM_MMMM_MMME_CCSS
CLKSET #%1111_1_10_11
That seems kind of messy, though, as it reorders the PLL config fields in a way that doesn't flow logically from left to right.
The way it is now is maybe fine, considering the alternatives.
I've noticed E field can be derived from CC field. How about splitting up the PLL config from the clock selection:
CLKSET D/# - Configure global circuit selected by MSBs
%0000_0000_PPPP_DDDD_DDMM_MMMM_MMMM_CC10 Config clock PLL generator
%0000_0000_0000_0000_0000_0000_SS10_0000 Select clock source
%0000_0000_0000_0000_0000_0001_0000_0000 Hard reset, reboots chip
%0000_0000_0000_0000_0000_000P_1000_0000 Set write-protect of last16KB RAM to P
%0000_0000_0000_0000_RRLL_TTTT_T100_0000 Set filter R to length L and tap T
%DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDD1 Seed Xoroshiro128+ PRNG with D
But, what if you wanted the crystal oscillator running, but not the PLL, too? It takes extra power to run the PLL.
Hmm ... hadn't thought of that option. The E field can be put where you suggested at bit4. I can't say I saw any flow issues.
I'm going to keep it how I've currently got it. I'm part way through compiling FPGA images now, and I don't want to start over. It's of marginal worth, either way. I think what we have now is simplest.
But, what if you wanted the crystal oscillator running, but not the PLL, too? It takes extra power to run the PLL.
If you wanted a CMOS clock source, and no Xtal amplifier, that seems to be missing ?
Select of %CC=%01 is close, but has a signal on XO that is not strictly needed.
The closest to 'Halt and wait for reset' looks to be 20kHz Osc select ?
But, what if you wanted the crystal oscillator running, but not the PLL, too? It takes extra power to run the PLL.
If you wanted a CMOS clock source, and no Xtal amplifier, that seems to be missing ?
Select of %CC=%01 is close, but has a signal on XO that is not strictly needed.
The closest to 'Halt and wait for reset' looks to be 20kHz Osc select ?
If you drive a signal into XI, XO will output the opposite with CC=%01. There will be no loading caps. I figured the clock-input mode %01 could follow on XO to allow some custom C-loading for a lower-C crystal.
Yes. 20KHz would draw way less power than the quiescent leakage. That is the power-down mode.
I'm going to keep it how I've currently got it. I'm part way through compiling FPGA images now, and I don't want to start over. It's of marginal worth, either way. I think what we have now is simplest.
If you do revisit it then turning it into four separate instructions also seems sensible. One for the random seed, one for xtal/pll config, one for pin filters, and one for clock-select and reset and write-protect.
That would provide maximal immediate addressing and, being single operand, they don't consume any valuable encoding space.
* Streamer pins now wrap around for 16-and 32-bit data paths on upper 8-pin offsets
* AND/ANDN, TEST/TESTN, and SETPIV/SETPIX have swapped encodings
* SCLU/SCL have been renamed to SCA/SCAS
* CLKSET has been renamed to HUBSET
* New MODC/MODZ aliases for MODCZ
This is the latest (and I think darn-near final) Verilog. This is in OnSemi's hands now and they are developing their compilation scripts with it.
I also use FL fast load once I have EASYFILE loaded so that everything is buffered into a temporary or named file and loaded from the file where it is not necessary for flow control. With TAQOZ in P2 ROM this should be possible since I have FAT32 already resident, all it needs is an SD card. One advantage of downloading to a named file is that it is also an easy way to transfer source code onto the card. Maybe some of the newer USB serial chips have proper flow control built into them because as we know you can't rely upon the O/S to do it.
@Peter,
is this FL something like redirecting serial in into a temporary file on sdcard and then loading from that file so that compilarion/execution is delayed but save?
Maybe you can answer in the tachyon thread to keep others informed about this. Thanks!
I also use FL fast load once I have EASYFILE loaded so that everything is buffered into a temporary or named file and loaded from the file where it is not necessary for flow control. With TAQOZ in P2 ROM this should be possible since I have FAT32 already resident, all it needs is an SD card. One advantage of downloading to a named file is that it is also an easy way to transfer source code onto the card. Maybe some of the newer USB serial chips have proper flow control built into them because as we know you can't rely upon the O/S to do it.
@Peter,
is this FL something like redirecting serial in into a temporary file on sdcard and then loading from that file so that compilarion/execution is delayed but save?
Maybe you can answer in the tachyon thread to keep others informed about this. Thanks!
that is what Peter is saying above in other words.
Re new MODC/MODZ aliases, shouldn't flag writing be shown in the spreadsheet as compulsory, as with GETRND without D?
GETRNDWC/WZ/WCZMODC c WCMODZ z WZ
Also, how does the assembler know which of the two instructions called CALLD to assemble?
Flag writing is not compulsory, though one wouldn't use the instruction without WC/WZ. For self-modifying code, though, we should not make flag writing compulsory.
The assembler does some sleuthing to know which CALLD to use.
I haven't heard much since I put up v31. Is anyone using it?
I loaded it onto my Prop123a9 board and ran the blink demo. I also started reading the documentation for LLVM. It looks to be quite a steep learning curve.
Comments
Next, I'll put the instruction set to bed for, hopefully, the last time by pushing out the latest internal Verilog into v31.
After that, I have a lot of communicating to do with OnSemi.
And after that, I want to get back to Spin2.
That's a great idea. Do you think we could make it shorter, though? That's way better than MulticoreEnhancedRealtimeKelvinEscalatingLaptop.
https://docs.google.com/spreadsheets/d/12GWGrEmp72MUIBvoXWadjWwqngrgQRqh9eYq6XhpEYQ/edit?usp=sharing
Thanks, Chip. Here are my final comments on the instruction set. Most suggestions are new, the others repeated for completeness. My thinking is that if one instruction must move, the set will have changed and might as well move some others and have done with it. Numbering is by ascending opcodes, not importance.
* = instruction moved or changed
1. TEST and others
Move as shown below, so that all three instructions that use S/!S, namely AND/ANDN and MOV/NOT and TEST/TESTN, have opcode bits [23:21] = 000/001, making the coding more consistent and harmonious.
Before:
EEEE 0111000 CZI DDDDDDDDD SSSSSSSSS INCMOD D,S/# {WC/WZ/WCZ} EEEE 0111001 CZI DDDDDDDDD SSSSSSSSS DECMOD D,S/# {WC/WZ/WCZ} EEEE 0111010 CZI DDDDDDDDD SSSSSSSSS ZEROX D,S/# {WC/WZ/WCZ} EEEE 0111011 CZI DDDDDDDDD SSSSSSSSS SIGNX D,S/# {WC/WZ/WCZ} EEEE 0111100 CZI DDDDDDDDD SSSSSSSSS ENCOD D,S/# {WC/WZ/WCZ} EEEE 0111101 CZI DDDDDDDDD SSSSSSSSS ONES D,S/# {WC/WZ/WCZ} EEEE 0111110 CZI DDDDDDDDD SSSSSSSSS TEST D,S/# {WC/WZ/WCZ} EEEE 0111111 CZI DDDDDDDDD SSSSSSSSS TESTN D,S/# {WC/WZ/WCZ}After:EEEE 0111000 CZI DDDDDDDDD SSSSSSSSS * TEST D,S/# {WC/WZ/WCZ} EEEE 0111001 CZI DDDDDDDDD SSSSSSSSS * TESTN D,S/# {WC/WZ/WCZ} EEEE 0111010 CZI DDDDDDDDD SSSSSSSSS * ENCOD D,S/# {WC/WZ/WCZ} EEEE 0111011 CZI DDDDDDDDD SSSSSSSSS * ONES D,S/# {WC/WZ/WCZ} EEEE 0111100 CZI DDDDDDDDD SSSSSSSSS * INCMOD D,S/# {WC/WZ/WCZ} EEEE 0111101 CZI DDDDDDDDD SSSSSSSSS * DECMOD D,S/# {WC/WZ/WCZ} EEEE 0111110 CZI DDDDDDDDD SSSSSSSSS * ZEROX D,S/# {WC/WZ/WCZ} EEEE 0111111 CZI DDDDDDDDD SSSSSSSSS * SIGNX D,S/# {WC/WZ/WCZ}2. RDFAST and others
Move as shown, so that RDFAST/WRFAST opcodes differ by only one bit.
Before:
EEEE 1100011 1LI DDDDDDDDD SSSSSSSSS RDFAST D/#,S/# EEEE 1100100 0LI DDDDDDDDD SSSSSSSSS WRFAST D/#,S/# EEEE 1100100 1LI DDDDDDDDD SSSSSSSSS FBLOCK D/#,S/#After:EEEE 1100011 1LI DDDDDDDDD SSSSSSSSS * FBLOCK D/#,S/# EEEE 1100100 0LI DDDDDDDDD SSSSSSSSS * RDFAST D/#,S/# EEEE 1100100 1LI DDDDDDDDD SSSSSSSSS * WRFAST D/#,S/#3. RFBYTE and others
Move as shown, so that RFVAR/RFVARS, RFBYTE/WFBYTE, RFWORD/WFWORD and RFLONG/WFLONG differ by the same single bit.
Before:
EEEE 1101011 CZ0 DDDDDDDDD 000010000 RFBYTE D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010001 RFWORD D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010010 RFLONG D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010011 RFVAR D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010100 RFVARS D {WC/WZ/WCZ} EEEE 1101011 00L DDDDDDDDD 000010101 WFBYTE D/# EEEE 1101011 00L DDDDDDDDD 000010110 WFWORD D/# EEEE 1101011 00L DDDDDDDDD 000010111 WFLONG D/#After:EEEE 1101011 CZ0 DDDDDDDDD 000010000 * RFVAR D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010001 * RFBYTE D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010010 * RFWORD D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010011 * RFLONG D {WC/WZ/WCZ} EEEE 1101011 CZ0 DDDDDDDDD 000010100 RFVARS D {WC/WZ/WCZ} EEEE 1101011 00L DDDDDDDDD 000010101 WFBYTE D/# EEEE 1101011 00L DDDDDDDDD 000010110 WFWORD D/# EEEE 1101011 00L DDDDDDDDD 000010111 WFLONG D/#4. MODCZ and others
Does not involve D (nor S), so out of place in with the registered D instructions. I think it could be moved in front of POLLINT, with bit 17 changed from 0 to 1 for POLLxxx and WAITxxx.
Before:
EEEE 1101011 CZ0 000000000 000100100 POLLINT {WC/WZ/WCZ} EEEE 1101011 CZ0 000000001 000100100 POLLCT1 {WC/WZ/WCZ} EEEE 1101011 CZ0 000000010 000100100 POLLCT2 {WC/WZ/WCZ} EEEE 1101011 CZ0 000000011 000100100 POLLCT3 {WC/WZ/WCZ} EEEE 1101011 CZ0 000000100 000100100 POLLSE1 {WC/WZ/WCZ} EEEE 1101011 CZ0 000000101 000100100 POLLSE2 {WC/WZ/WCZ} EEEE 1101011 CZ0 000000110 000100100 POLLSE3 {WC/WZ/WCZ} EEEE 1101011 CZ0 000000111 000100100 POLLSE4 {WC/WZ/WCZ} EEEE 1101011 CZ0 000001000 000100100 POLLPAT {WC/WZ/WCZ} EEEE 1101011 CZ0 000001001 000100100 POLLFBW {WC/WZ/WCZ} EEEE 1101011 CZ0 000001010 000100100 POLLXMT {WC/WZ/WCZ} EEEE 1101011 CZ0 000001011 000100100 POLLXFI {WC/WZ/WCZ} EEEE 1101011 CZ0 000001100 000100100 POLLXRO {WC/WZ/WCZ} EEEE 1101011 CZ0 000001101 000100100 POLLXRL {WC/WZ/WCZ} EEEE 1101011 CZ0 000001110 000100100 POLLATN {WC/WZ/WCZ} EEEE 1101011 CZ0 000001111 000100100 POLLQMT {WC/WZ/WCZ} EEEE 1101011 CZ0 000010000 000100100 WAITINT {WC/WZ/WCZ} EEEE 1101011 CZ0 000010001 000100100 WAITCT1 {WC/WZ/WCZ} EEEE 1101011 CZ0 000010010 000100100 WAITCT2 {WC/WZ/WCZ} EEEE 1101011 CZ0 000010011 000100100 WAITCT3 {WC/WZ/WCZ} EEEE 1101011 CZ0 000010100 000100100 WAITSE1 {WC/WZ/WCZ} EEEE 1101011 CZ0 000010101 000100100 WAITSE2 {WC/WZ/WCZ} EEEE 1101011 CZ0 000010110 000100100 WAITSE3 {WC/WZ/WCZ} EEEE 1101011 CZ0 000010111 000100100 WAITSE4 {WC/WZ/WCZ} EEEE 1101011 CZ0 000011000 000100100 WAITPAT {WC/WZ/WCZ} EEEE 1101011 CZ0 000011001 000100100 WAITFBW {WC/WZ/WCZ} EEEE 1101011 CZ0 000011010 000100100 WAITXMT {WC/WZ/WCZ} EEEE 1101011 CZ0 000011011 000100100 WAITXFI {WC/WZ/WCZ} EEEE 1101011 CZ0 000011100 000100100 WAITXRO {WC/WZ/WCZ} EEEE 1101011 CZ0 000011101 000100100 WAITXRL {WC/WZ/WCZ} EEEE 1101011 CZ0 000011110 000100100 WAITATN {WC/WZ/WCZ} EEEE 1101011 CZ1 0cccczzzz 001101111 MODCZ c,z {WC/WZ/WCZ}After:EEEE 1101011 CZ0 0cccczzzz 000100100 * MODCZ c,z {WC/WZ/WCZ} EEEE 1101011 CZ0 100000000 000100100 * POLLINT {WC/WZ/WCZ} EEEE 1101011 CZ0 100000001 000100100 * POLLCT1 {WC/WZ/WCZ} EEEE 1101011 CZ0 100000010 000100100 * POLLCT2 {WC/WZ/WCZ} EEEE 1101011 CZ0 100000011 000100100 * POLLCT3 {WC/WZ/WCZ} EEEE 1101011 CZ0 100000100 000100100 * POLLSE1 {WC/WZ/WCZ} EEEE 1101011 CZ0 100000101 000100100 * POLLSE2 {WC/WZ/WCZ} EEEE 1101011 CZ0 100000110 000100100 * POLLSE3 {WC/WZ/WCZ} EEEE 1101011 CZ0 100000111 000100100 * POLLSE4 {WC/WZ/WCZ} EEEE 1101011 CZ0 100001000 000100100 * POLLPAT {WC/WZ/WCZ} EEEE 1101011 CZ0 100001001 000100100 * POLLFBW {WC/WZ/WCZ} EEEE 1101011 CZ0 100001010 000100100 * POLLXMT {WC/WZ/WCZ} EEEE 1101011 CZ0 100001011 000100100 * POLLXFI {WC/WZ/WCZ} EEEE 1101011 CZ0 100001100 000100100 * POLLXRO {WC/WZ/WCZ} EEEE 1101011 CZ0 100001101 000100100 * POLLXRL {WC/WZ/WCZ} EEEE 1101011 CZ0 100001110 000100100 * POLLATN {WC/WZ/WCZ} EEEE 1101011 CZ0 100001111 000100100 * POLLQMT {WC/WZ/WCZ} EEEE 1101011 CZ0 100010000 000100100 * WAITINT {WC/WZ/WCZ} EEEE 1101011 CZ0 100010001 000100100 * WAITCT1 {WC/WZ/WCZ} EEEE 1101011 CZ0 100010010 000100100 * WAITCT2 {WC/WZ/WCZ} EEEE 1101011 CZ0 100010011 000100100 * WAITCT3 {WC/WZ/WCZ} EEEE 1101011 CZ0 100010100 000100100 * WAITSE1 {WC/WZ/WCZ} EEEE 1101011 CZ0 100010101 000100100 * WAITSE2 {WC/WZ/WCZ} EEEE 1101011 CZ0 100010110 000100100 * WAITSE3 {WC/WZ/WCZ} EEEE 1101011 CZ0 100010111 000100100 * WAITSE4 {WC/WZ/WCZ} EEEE 1101011 CZ0 100011000 000100100 * WAITPAT {WC/WZ/WCZ} EEEE 1101011 CZ0 100011001 000100100 * WAITFBW {WC/WZ/WCZ} EEEE 1101011 CZ0 100011010 000100100 * WAITXMT {WC/WZ/WCZ} EEEE 1101011 CZ0 100011011 000100100 * WAITXFI {WC/WZ/WCZ} EEEE 1101011 CZ0 100011100 000100100 * WAITXRO {WC/WZ/WCZ} EEEE 1101011 CZ0 100011101 000100100 * WAITXRL {WC/WZ/WCZ} EEEE 1101011 CZ0 100011110 000100100 * WAITATN {WC/WZ/WCZ}5. STALLI and others
Move as shown so that TRGINTx/NIXINTx differ by same bit. Also, bit 17 changed from 0 to 1 due to MODCZ move.
Before:
EEEE 1101011 000 000100000 000100100 ALLOWI EEEE 1101011 000 000100001 000100100 STALLI EEEE 1101011 000 000100010 000100100 TRGINT1 EEEE 1101011 000 000100011 000100100 TRGINT2 EEEE 1101011 000 000100100 000100100 TRGINT3 EEEE 1101011 000 000100101 000100100 NIXINT1 EEEE 1101011 000 000100110 000100100 NIXINT2 EEEE 1101011 000 000100111 000100100 NIXINT3After:EEEE 1101011 000 100100000 000100100 * ALLOWI EEEE 1101011 000 100100001 000100100 * TRGINT1 EEEE 1101011 000 100100010 000100100 * TRGINT2 EEEE 1101011 000 100100011 000100100 * TRGINT3 EEEE 1101011 000 100100100 000100100 * STALLI EEEE 1101011 000 100100101 000100100 * NIXINT1 EEEE 1101011 000 100100110 000100100 * NIXINT2 EEEE 1101011 000 100100111 000100100 * NIXINT36. MERGEB and another
Possibly move as shown, so that SPLITx/MERGEx are split, not merged!
Before:
EEEE 1101011 000 DDDDDDDDD 001100000 SPLITB D EEEE 1101011 000 DDDDDDDDD 001100001 MERGEB D EEEE 1101011 000 DDDDDDDDD 001100010 SPLITW D EEEE 1101011 000 DDDDDDDDD 001100011 MERGEW DAfter:EEEE 1101011 000 DDDDDDDDD 001100000 SPLITB D EEEE 1101011 000 DDDDDDDDD 001100001 * SPLITW D EEEE 1101011 000 DDDDDDDDD 001100010 * MERGEB D EEEE 1101011 000 DDDDDDDDD 001100011 MERGEW DModified instructions.txt with above changes attached.
Whether or not any of the above are implemented, that's me finished with the instruction set!
I also use FL fast load once I have EASYFILE loaded so that everything is buffered into a temporary or named file and loaded from the file where it is not necessary for flow control. With TAQOZ in P2 ROM this should be possible since I have FAT32 already resident, all it needs is an SD card. One advantage of downloading to a named file is that it is also an easy way to transfer source code onto the card. Maybe some of the newer USB serial chips have proper flow control built into them because as we know you can't rely upon the O/S to do it.
Changing the encoding of the selector bits to be at LSbit end will allow for the reset and write protect functions to be encoded as single instructions.
Eg:
CLKSET D/# - Configure global circuit selected by MSBs %DDDD_DDMM_MMMM_MMMM_PPPP_CCSS_Exxx_xx10 Set clock generator mode %0000_0000_0000_0000_0000_0001_0000_0000 Hard reset, reboots chip %0000_0000_0000_0000_0000_000P_1000_0000 Set write-protect of last 16KB RAM to P %0000_0000_0000_0000_RRLL_TTTT_T100_0000 Set filter R to length L and tap T %DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDD1 Seed Xoroshiro128+ PRNG with DWhat I don't like about that is that it makes setting the clock require ## for most cases. I think setting the clock is going to be the most common use of this instruction, so it needs to be the simplest.
Here is how I see the the use cases in order of descending frequency:
1) setting the clock (most common, as apps need to save power)
2) write-protecting the last 16KB of RAM (used once/twice? by OS-type software)
3) setting the filters (probably a one-time thing)
4) resetting the chip (this has got to be rare)
5) seeding the PRNG (done by boot ROM, no need after)
It would be nice to be able to set 160MHz PLL operation from a 20MHz crystal. That would require something like this:
%P_PPPD_DDDD_DMMM_MMMM_MMME_CCSS
CLKSET #%1111_1_10_11
That seems kind of messy, though, as it reorders the PLL config fields in a way that doesn't flow logically from left to right.
The way it is now is maybe fine, considering the alternatives.
CLKSET D/# - Configure global circuit selected by MSBs %0000_0000_PPPP_DDDD_DDMM_MMMM_MMMM_CC10 Config clock PLL generator %0000_0000_0000_0000_0000_0000_SS10_0000 Select clock source %0000_0000_0000_0000_0000_0001_0000_0000 Hard reset, reboots chip %0000_0000_0000_0000_0000_000P_1000_0000 Set write-protect of last 16KB RAM to P %0000_0000_0000_0000_RRLL_TTTT_T100_0000 Set filter R to length L and tap T %DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDDD_DDD1 Seed Xoroshiro128+ PRNG with DBut, what if you wanted the crystal oscillator running, but not the PLL, too? It takes extra power to run the PLL.
I'm going to keep it how I've currently got it. I'm part way through compiling FPGA images now, and I don't want to start over. It's of marginal worth, either way. I think what we have now is simplest.
If you wanted a CMOS clock source, and no Xtal amplifier, that seems to be missing ?
Select of %CC=%01 is close, but has a signal on XO that is not strictly needed.
The closest to 'Halt and wait for reset' looks to be 20kHz Osc select ?
If you drive a signal into XI, XO will output the opposite with CC=%01. There will be no loading caps. I figured the clock-input mode %01 could follow on XO to allow some custom C-loading for a lower-C crystal.
Yes. 20KHz would draw way less power than the quiescent leakage. That is the power-down mode.
That would provide maximal immediate addressing and, being single operand, they don't consume any valuable encoding space.
WHAT's NEW:
* Streamer pins now wrap around for 16-and 32-bit data paths on upper 8-pin offsets
* AND/ANDN, TEST/TESTN, and SETPIV/SETPIX have swapped encodings
* SCLU/SCL have been renamed to SCA/SCAS
* CLKSET has been renamed to HUBSET
* New MODC/MODZ aliases for MODCZ
This is the latest (and I think darn-near final) Verilog. This is in OnSemi's hands now and they are developing their compilation scripts with it.
Please try it out, if you can.
Thanks!
Cheers
Now that sounds a lot like a milestone !!
is this FL something like redirecting serial in into a temporary file on sdcard and then loading from that file so that compilarion/execution is delayed but save?
Maybe you can answer in the tachyon thread to keep others informed about this. Thanks!
that is what Peter is saying above in other words.
GETRND WC/WZ/WCZ MODC c WC MODZ z WZAlso, how does the assembler know which of the two instructions called CALLD to assemble?
Flag writing is not compulsory, though one wouldn't use the instruction without WC/WZ. For self-modifying code, though, we should not make flag writing compulsory.
The assembler does some sleuthing to know which CALLD to use.
Super! Glad you're on board.