David Black

I went and looked at the website where this stuff was supposed to be and found problems. I then dug around and found an archived copy, did a quick sanity check, and I've now published it on GitHub under github.com/dcblack/SCFTGU_BOOK As noted, I am working on a modern version, but time is limited.

Perhaps you would be willing to share your work as an Apache 2.0 licensed github project with the minimal code needed and a screen snapshot of the results.

Reset is accomplished by throwing an exception. You cannot perform cleanup *before*, but you can perform cleanup on the way out in two manners. First, automatically created objects will of course run their destructors as guaranteed by C++. Second, you can catch exceptions by strategic placement of try-catch blocks and looking for the appropriate sc_exception. It is required by SystemC that you rethrow after catching; otherwise, you will corrupt the SystemC kernel.

sc_spawn creates same kind of processes as sC_METHOD, SC_THREAD, etc, but is not limited to the elaboration phase. Please either read the official documentation for IEEE-1666 (available as pdf free via Accellera) or get a good book on SystemC or take a class.

A derived class could inherit this.

I am able to reproduce the problem and will attempt a fix. Unless you are using async_request_update() in your code, you can safely ignore this problem for now. CORRECTION: While there is a bug with the following deprecated feature issue, this does not solve the problem. Stay tuned for a real fix. There is a bug in the implementation of SystemC due to Apple removing support for POSIX sem_init, which is a non-required API by the POSIX standard. See <https://stackoverflow.com/questions/1413785/sem-init-on-os-x/24617282> for details. [Pure speculation: I suspect the reason for removing support was that Apple has recently moved to an all 64-bit coding model. Potentially because they are positioning themselves to be able to port quickly to Arm v8A lacking Aarch32 on certain hardware.] When building on OSX using Cmake I noticed a clue: I use the adage, "A warning is usually a potential bug leading to a real error." Never ignore warnings from compilations. Yes, I know there is a lot of code out there with superfluous warnings. Shame on them for leaving them in. So if you see a warning, track it down. If it is truly a don't care (rarely), then it can be overridden with a #pragma. Almost all warnings can be fixed with proper coding. I am going to attempt a fix to sc_host_semaphore.h, but if you're in a hurry go to Linux.

Yes, SystemC is doing exactly what is supposed to do. Perfect behavior and standard reaction when you don't understand event driven simulators. In Verilog/SystemVerilog, this behavior is called non-blocking assignment (NBA). In VHDL, this is the behavior of a signal.

Here is a short list of topics in no particular order you need to be comfortable with in order to be have an easier time learning SystemC: [Note: Others might chime in with variations on this list (add/subtract), and this is not necessarily a complete list, but I am fairly certain if you are able to comfortably use the topics I list below, you will have very little trouble syntactically with learning SystemC. In addition to C++, it helps if you have some familiarity with event driven simulation (e.g. SystemVerilog or VHDL). Also, if you have deep knowledge in another OO language (e.g. Java or SystemVerilog), you might have an easier time learning the C++ part.] Difference between declaration and definition Pass by value vs pass by reference Use of const (5 distinct cases) Casting C++ style (4 types) Implicit vs explicit conversions Use of function overloading and how to deal with ambiguity issues Use of std::string Use of streaming I/O How to declare, define and use classes Definition of default constructor Purpose and syntax of copy constructor How to declare and use namespaces Operator overloading as member functions and global functions. The difference between overloading and overriding. Relationship between class and struct How to extend classes and multiple inheritance Purpose of public and private Storage types and lifetimes: static, automatic, dynamic How to properly use new and delete Use of pointers and understanding of issues with pointer arithmetic Use of arrays and issues Advantages and use of std::vector<> Use of try-catch and throw Use of initializer list in constructor and a proper understanding of the order of construction Polymorphism and RTTI RAII Rule of 4 (6 if using C++11 or later) How and where to define templates/generic programming (does not need to be deep knowledge - just the basics) Use of templates and nested templates. Definition of full and partial template specialization. Different types of constructors and destructors Use of virtual inheritance (hint: it's not polymorphism) Extra topics: More STL including at least std::map<>, std::set<> Boost Modern C++ users (2011 onward) should know about: nullptr Uniform initialization Use of auto Use of ranged for Lambda definition, binding and use constexpr std::unique_ptr<>, std::shared_ptr<>

First, the title of this post is completely nonsensical. SystemC is a library written in C++ and everything about it is C++. It is much more likely that you simply are not very proficient with C++. One aspect of SystemC I always warn newbies about is that SystemC uses pretty much all of the features of C++. You really cannot get by with just thinking C++ is merely a better C. You have to really know about polymorphism, function/operator overloading, templates, exception handling, STL libraries, and all things C++. If you do not, I suggest you get some deeper learning on C++. Your problem is that sc_module disallows copy construction. Prior to C++11, this was accomplished by declaring the copy constructor private, which is why you are getting the error. sc_vector really is a better way of creating an array of modules, and it looks as if somebody has given you this hint.

Doesn't make as much sense to those who use Windows and Visual Studio. Much more understandable if you use Linux. Tip: You can access those two arguments from anywhere in your SystemC design using the two functions: sc_argc() and sc_argv(), which are in the sc_core namespace. Keep in mind that they only provide access to c_str values. If you want to interpret a numeric value from the command-line, you will need to convert it yourself.

SystemC behaves in almost identical manner to other event driven simulators (e.g. Verilog and VHDL). The simulator provides a co-operative multitasking model to simplify the programming paradigm. SC_THREAD processes yield by calling wait() and SC_METHOD processes yield by returning. There are 3 semantics for event notification: notify() immediate notify(SC_ZERO_TIME) at the end of the current delta cycle notify(NON_ZERO) scheduled to occur at a scheduled time in the future Furthermore, you need to be aware that SystemC is not inherently thread safe.

Adding ignorable extensions and using the generic payload means that your TLM models would be reusable and interoperable with other TLM components that conform to the standard payload. By deriving a new class, you would make your model less usable in alternate systems.

[I assume that when you say "TLM", you mean SystemC TLM 2.0.] You need to understand the difference between modeling styles. TLM is precisely about not modeling at the level of RTL. The SystemC TLM 2.0 also has two different modeling styles: Loosely Timed (LT) and Approximately Timed (AT). Let's look at each using a specific case. Suppose you are modeling two UARTs operating at 9600 baud (bits per second) with 8-bits, no parity, and 1 stop bit to transfer the message "Hello World\n". This configuration results in 960 characters per second (1.042 ms/char), which is quite slow, so probably you would be transmitting/receiving characters slowly enough that most systems would either process them one at a time or provide a FIFO (e.g. 16 bytes) and only process empty/full events. There is one more question to answer though. Consider the diagram below. The connections between sender to UART and UART to receiver are clearly memory mapped for most systems. So there is no question of modeling. The connection UART to UART is not memory mapped, which means you need to create a custom protocol. Furthermore, for TLM, it actually requires to connections since communication can be invoked bi-directionally (for a full UART). You need to decide what is important to model. For a high level model and efficiency, I would either transfer as much data as I could. It might even make sense to use TLM 1.0 rather than TLM 2.0. Do you have the requirement to inject errors? For my example, you would configure the transmitter, and then transfer a burst of 12 characters into the transmit FIFO on one end of the transfer and generate an empty FIFO interrupt at 12.5 ms later. The receiver side would be similar. What about the UART/UART transaction? An efficient approach might be as follows: Create a required extension that carries the transmit configuration information (baud rate, bits, parity, etc.) Use TLM_WRITE_COMMAND because all transactions over this socket pair are initiated from the sender. The second pair in the opposite direction would do the same thing. Check and insist that the address always be 0 and the streaming width is 1. Byte enables would be illegal. Check that the configuration matches before accepting data. Place all received data into an unbounded queue and then indicate the size allowed by the hardware model. Send interrupts using the sc_signal when the received queue goes non-empty. Consider the error situation when the timing indicates characters would be lost due to FIFO full and timing of characters. You will have to decide how to deal with interrupts received in your thread process. Notice that I do not model at the bit level. If you wish to add bit-level error injection, then inject errors at the point of transmission.

Where is this code executing (constructor, SC_THREAD, start_of_simulation)?

Remember that the SystemC OS thread is completely asynchronous to external threads and processes. Async_request_update() simply says to call update() once at the end of the delta cycle for a single object. So you can get multiple notifications within the same delta and only get one call-back. To keep from losing these you should setup a threadsafe mailbox/queue. Rather than just notify the event, you should also put an entry into the mailbox on each activation. Then your threadsafe update() method can see the multiple requests accurately.

These questions have little to do with SystemC per se, and are really about C++. Templates are all about compile-time elaboration and template arguments must be compile-time computable. If you use C++11 or later, then various forms of constexpr functions may be available, but they are still compile-time issues. You could of course use sc_bv_base and its constructors, but keep in mind that modules, ports, and other "hardware" constructs are not allowed to be modified after end_of_elaboration. KEY POINT: To be an effective SystemC designer, you MUST be proficient at C++. Minimal C++ is NOT enough. Knowledge of C (even expert knowledge) is totally inadequate and in some cases downright harmful. Furthermore, really good SystemC often requires excellent C++ skills. Therefore, before you even consider learning much in SystemC, you really should invest in a solid C++ course. Expert SystemC practitioners take time to continually update their C++ skills. If this does not sound like fun to you, then I would advise choosing a different discipline.

To be slightly more precise, the back-slash character (\) has two contexts. In this context, it is an escape character for CPP, the C Pre-Processor. It should only be used when using the pre-processor #define directive. Without \, the #define must be kept entirely on a single line. The other context for which \ is used, is inside literal strings and and character constants. Examples: "Hello world\r\n" and "\n". In these cases, it serves to provide an escape to introduce non-printable ASCII characters.

Don't change the default entry point and don't define main. Main is already pre-compiled inside the library!

You are specifically referring to static sensitivity. Perhaps you should consider using dynamic sensitivity instead. You can use sc_event_and_list and sc_event_or_list to build up appropriate sensitivities.

This happens because sc_port overloads operator[]. You need to use sc_vector tlm::tlm_initiator_socket<32,tlm::tlm_base_protocol_types,0,SC_ZERO_OR_MORE_BOUND> Out; SC_MODULE(Target) { sc_core::sc_vector< tlm::tlm_target_socket<32,tlm::tlm_base_protocol_types,1,SC_ZERO_OR_MORE_BOUND> > In; SC_CTOR(Target) : In("In", 2) { ... } //< Number of elemements selected via sc_vector constructor } initiator_inst->Out.bind(target_inst->In); Note: Above is pseudo-code (i.e. not in all the right places), and I have not taken time to verify the detail, but the principle of using sc_vector is correct. You may need to consult the literature.

The problem with using SC_METHOD is that it forces a style of coding that is harder to design/debug and slows down the design process. Schedule-wise SystemC is supposed to be written quickly. If you take too much time writing/debugging it, then it can become a counterproductive effort and you might as well write RTL alone. I usually characterize it as follows: Loosely-Timed models should be written in a matter of days (not weeks or months). Approximately-Timed models should be written in terms of a few weeks (not months). Anything longer means that either you are adding too much detail (which also slows down model execution) or you are not properly reusing other people's work (purchasing/acquiring pre-written models). So yes, I would use SC_METHOD when it's use is obvious/intuitive, but I would rather express ideas in the most natural manner to speed up development time, which often means using SC_THREAD.

Using your own report handler is a bit extreme; although, instructive. I suggest using the sc_report_handler::set_action() method. Please read the standards document (downloadable from IEEE via link on Accellera). It's quite readable.

Ankur, I'm not sure where you got the advice that SC_THREADs are bad or cause context switching, because that is mostly just an old wives-tale. I did a study years ago that showed the difference between SC_METHODs and SC_THREADs is only 3%. Furthermore, it is a simulator implementation issue as the PoC implementation is only one of several. I know at least one implementation of the SystemC kernel actually had the advantage reversed (and it was faster than the PoC implementation). The real problem is context switching, which cannot be totally avoided in any case. Context switches occur everytime your SC_METHODs return or your SC_THREADs call sc_core::wait() even if indirectly (e.g. sc_fifo::read()). If you are having performance issues, you need to be profiling your code to identify the real problems. You can get rid of a lot of context switching if you simply elminate sc_clock's entirely from your design. Most engineers (esp. hardware focused) have a difficult time with this concept, but sc_clock causes two context switches per cycle and in many designs is entirely not needed. I did a presentation "Look Ma! No clocks!" (available somewhere under www.nascug.org) and demonstrated how a timer could be modeled accurately without clocks. A couple of other areas often missed are I/O and compiler switches. I/O is much bigger than context switching in terms of slowing down simulations, and yet still I often see engineers putting SC_REPORT_INFO (or worse: std::cout and printf) in their code to dump all the activity to aid debug. Except for phases before SC_SIMULATION, you should really use SC_REPORT_INFO_VERB with SC_DEBUG or equivalent. As to compile-time switches, -O3 is really quite decent once you have most bugs out, and you can leave debug -g on without worry.

Re. the error SC_METHOD is called repeatedly (every time an element of the static sensitivity list triggers); however, SC_METHOD processes must execute in single delta cycle (zero time). Blocking methods such as wait() is illegal there. Methods such as sc_fifo::write() call wait() and are therefore also illegal. To wait 10 ns, you can do something like this: #include <systemc> #include <iostream> struct Ex1 : sc_core::sc_module { Ex1( void ) { SC_HAS_PROCESS(Ex1); SC_METHOD(Method10ns); } void Method10ns( void ) { next_trigger( 10, SC_NS ); std::cout << "Now " << sc_core::sc_time_stamp() << sdt::endl; } }; SC_THREAD is called only one, and therefore normally contains an infinite loop to properly model hardware. Thus the above is coded more simply as: #include <systemc> #include <iostream> struct Ex1 : sc_core::sc_module { Ex1( void ) { SC_HAS_PROCESS(Ex1); SC_THREAD(Thread10ns); } void Thread10ns( void ) { for(;;) { std::cout << "Now " << sc_core::sc_time_stamp() << sdt::endl; wait( 10, SC_NS ); } } };

Re. SC_HAS_PROCESS That would appear to be an error in the text. The best place to put SC_HAS_PROCESS is inside the constructor as the first line in the body of the constructor. It is syntactically legal there and the only things using it are the SC_METHOD, SC_THREAD, and SC_CTHREAD macros.