Hello, I'm designing several Xilinx FPGAs and CPLDs and I was wondering if there is a way to set the way ISE encodes state machines other than in project navigator? I'd like to be able to set it in the VHDL file or in the UCF. It's really a quality control problem. As the project is touched by many people and as it's upgraded to newer versions of ISE software the encoding method seems to get changed. I want to be able to guarantee that it's set to what I want it to be set to. Thanks in advance for the help. Dale
Setting FSM encoding in VHDL or in UCF for Xilinx
A developer sought a way to explicitly define Finite State Machine (FSM) encoding within VHDL source code to ensure consistency across different Xilinx ISE versions and team members. The discussion identifies specific XST synthesis attributes that allow for individual FSM control, which is more granular than global project settings.
The thread also explores a deeper design issue where changing encoding styles revealed glitches in UART control signals. Participants concluded that while attributes can force encoding, the root cause was likely combinatorial output logic rather than the encoding itself, recommending registered outputs for better stability.
- To force FSM encoding in XST, use the 'fsm_encoding' attribute applied directly to the signal rather than the type.
- Valid string values for the attribute include "one-hot", "compact", "sequential", "gray", "johnson", and "none".
- Synthesis reports should be checked to confirm that the tool correctly recognized the FSM and applied the requested encoding.
- Setting encoding to "none" (fsm_extract = "no") prevents the synthesizer from re-optimizing the state registers.
- Glitches on FSM outputs are often caused by combinatorial logic; registering these outputs is recommended for synchronous reliability.
Dale wrote:> Hello, > I'm designing several Xilinx FPGAs and CPLDs and I was wondering if > there is a way to set the way ISE encodes state machines other than in > project navigator? I'd like to be able to set it in the VHDL file or > in the UCF. It's really a quality control problem. As the project is > touched by many people and as it's upgraded to newer versions of ISE > software the encoding method seems to get changed. I want to be able > to guarantee that it's set to what I want it to be set to. > Thanks in advance for the help. > > Daletype state_type is (foo, bar, ...); attribute syn_enum_encoding of state_type: type is "onehot"; signal state: state_type; -Jeff
On Nov 7, 11:04=A0am, Jeff Cunningham <j...@sover.net> wrote:> Dale wrote: > > Hello, > > I'm designing several Xilinx FPGAs and CPLDs and I was wondering if > > there is a way to set the way ISE encodes state machines other than in > > project navigator? =A0I'd like to be able to set it in the VHDL file or > > in the UCF. =A0It's really a quality control problem. =A0As the project=is> > touched by many people and as it's upgraded to newer versions of ISE > > software the encoding method seems to get changed. =A0I want to be able > > to guarantee that it's set to what I want it to be set to. > > Thanks in advance for the help. > > > Dale > > type state_type is (foo, bar, ...); > attribute syn_enum_encoding of state_type: type is "onehot"; > signal state: state_type; > > -JeffJeff, thanks for the help. When I do that I get the following error. ERROR:HDLParsers:3312 - "C:/Projects/HAL/Hardware/FPGA_B/trunk/ ADC_Audio.vhd" Line 51. Undefined symbol 'syn_enum_encoding'. Do I need to include a library or something? Thanks.
On Nov 7, 9:14 am, Dale <dale.prat...@gmail.com> wrote:> On Nov 7, 11:04 am, Jeff Cunningham <j...@sover.net> wrote: > > > > > Dale wrote: > > > Hello, > > > I'm designing several Xilinx FPGAs and CPLDs and I was wondering if > > > there is a way to set the way ISE encodes state machines other than in > > > project navigator? I'd like to be able to set it in the VHDL file or > > > in the UCF. It's really a quality control problem. As the project is > > > touched by many people and as it's upgraded to newer versions of ISE > > > software the encoding method seems to get changed. I want to be able > > > to guarantee that it's set to what I want it to be set to. > > > Thanks in advance for the help. > > > > Dale > > > type state_type is (foo, bar, ...); > > attribute syn_enum_encoding of state_type: type is "onehot"; > > signal state: state_type; > > > -Jeff > > Jeff, thanks for the help. When I do that I get the following error. > > ERROR:HDLParsers:3312 - "C:/Projects/HAL/Hardware/FPGA_B/trunk/ > ADC_Audio.vhd" Line 51. Undefined symbol 'syn_enum_encoding'. > > Do I need to include a library or something? > > Thanks.Typo; try this: attribute SYN_ENCODING of state_type: type is "onehot"; -or- attribute type_encoding_style of state_type: type is "onehot"; One of those should work, Alex
Thanks for all your input. I'm still not sure if I am where I need to be. I have the following and it synthesizes. type State_Machine is (S0,S1,S2,S3,S4,S5,S6,S7); attribute syn_enum_encoding: string; attribute syn_enum_encoding of State_Machine: type is "onehot"; signal FSM_0 : State_Machine; Then I changed the "onehot" to "jibberish" for an experiment. It sill synthesizes. So, what effect does the above syntax really have? Does it do anything?
Dale wrote:> I'm designing several Xilinx FPGAs and CPLDs and I was wondering if > there is a way to set the way ISE encodes state machines other than in > project navigator? I'd like to be able to set it in the VHDL file or > in the UCF. It's really a quality control problem.It is possible to write synthesis code such that the encoding selection does not affect the fpga operation. -- Mike Treseler
On Fri, 7 Nov 2008 08:45:59 -0800 (PST) Dale <dale.prather@gmail.com> wrote:> Thanks for all your input. I'm still not sure if I am where I need to > be. > > I have the following and it synthesizes. > > type State_Machine is (S0,S1,S2,S3,S4,S5,S6,S7); > attribute syn_enum_encoding: string; > attribute syn_enum_encoding of State_Machine: type is > "onehot"; signal FSM_0 : State_Machine; > > Then I changed the "onehot" to "jibberish" for an experiment. It sill > synthesizes. So, what effect does the above syntax really have? Does > it do anything?Then that might not be the right attribute. Thing to remember about passing attributes to the synthesizer is that (far too) many of them are synthesizer specific. I didn't see you mention what synthesis tool your're using, but if you're using XST, then you want: type State_Machine is (S0,S1,S2,S3,S4,S5,S6,S7); signal FSM_0 : State_Machine; attribute fsm_encoding : string; attribute fsm_encoding of FSM_0 : signal is "one-hot"; Note that not only is the attribute name different, but for some unfathomable reason it's applied to the signal, rather than to the type. -- Rob Gaddi, Highland Technology Email address is currently out of order
Yes, I'm using XST. I'm using a Spartan 3. I took your suggestion, Rob and again replaced "onehot" with garbage and it still synthesized. Any chance that attribute is invalid with a spartan 3? I "think" I read that somewhere. Thanks.
On Fri, 7 Nov 2008 10:00:36 -0800 (PST) Dale <dale.prather@gmail.com> wrote:> Yes, I'm using XST. I'm using a Spartan 3. I took your suggestion, > Rob and again replaced "onehot" with garbage and it still synthesized. > Any chance that attribute is invalid with a spartan 3? I "think" I > read that somewhere. > Thanks.Not at all. I use Spartan 3s, and have successfully used that attribute to force the encoding. One thing is that "onehot" is just as much garbage as "asdfhalksdjfh"; XST wants "one-hot". In the 10.1 documentation, FSM_ENCODING is covered in xst.pdf, page 370, with a full list of what options it supports. I'm not 100% sure that, if you specify an invalid encoding type, that it will throw an error rather than just assume "auto" and keep right on going. What you should try is asking it for "one-hot" and for one of the other defined types, then looking through the synthesis report file and seeing how it got encoded. There'll be a block of text beginning with the phrase Found finite state machine down in the HDL synthesis section, and it'll tell you the encoding style that it used. -- Rob Gaddi, Highland Technology Email address is currently out of order
" Found finite state machine down in the HDL synthesis section, and it'll tell you the encoding style that it used. " Great idea. That's solid proof. I'm on it.






