mirror of
https://github.com/darlinghq/darling-gdb.git
synced 2025-04-01 19:51:40 +00:00

the stand-alone executable. [in ChangeLog.sky:] * sky-gpuif.c (call_gs): Call properly into GPU2 library if configured --with-sim-gpu2. Use SKY_GPU2_REFRESH symbol as placeholder for future GPU2-refresh policy. [in ChangeLog:] * Makefile.in (MIPS_EXTRA_LIBS, SIM_EXTRA_LIBS): Added configurable settings for stand-alone simulator. start-sanitize-sky * configure.in: Added --with-sim-gpu2 option to specify path of sky GPU2 library. Triggers -DSKY_GPU2 for sky-gpuif.c, and links/compiles stand-alone simulator with this library. * interp.c (MEM_SIZE): Increased default sky memory size to 16MB. end-sanitize-sky * configure.in: Added X11 search, just in case. * configure: Regenerated.
This directory contains two very different simulators: o gencode (old) Gencode.c outputs a single monolithic file that is #included by interp.c o igen (new) The *.igen files are used as inputs to ../igen/igen. A number of separate, fairly modula files, are created. The new simulator has a number of advantages: o builtin support for multi-simming (single simulator image supporting a number of different instruction set architectures). o Easier maintenance. The input files are not confused by an intermixing with the generator code. gencode continues to exist so that old architectures can be emulated. *.igen should be used when adding new architectures or adding instructions to an existing ISA. Known bugs? A mips16 simulator cannot be built using igen. A custom mips16 engine.c needs to be written. In mips.igen, the semantics for many of the instructions were created using code generated by gencode. Those semantic segments could be greatly simplified. ---- Old README.Cygnus ... > README.Cygnus ------------------------------------------------------------------------------- The following are the main reasons for constructing the simulator as a generator: 1) Avoid large fixed decode source file, with lots of #ifs controlling the compilation. i.e. keep the source cleaner, smaller and easier to parse. 2) Allow optimum code to be created, without run-time checks on instruction types. Ensure that the simulator engine only includes code for the architecture being targetted. e.g. This avoids run-time checks on ISA conformance, aswell as increasing throughput. 3) Allow updates to the instruction sets to be added quickly. Having a table means that the information is together, and is easier to manipulate. Having the table generate the engine, rather than the run-time parse the table gives higher performance at simulation time. 4) Keep all the similar simulation code together. i.e. have a single place where, for example, the addition code is held. This ensures that updates to the simulation are not spread over a large flat source file maintained by the developer. ------------------------------------------------------------------------------- To keep the simulator simple (and to avoid the slight chance of mis-matched files) the manifests describing an engine, and the simulator engine itself, are held in the same source file. This means that the engine must be included twice, with the first pass controlled by the SIM_MANIFESTS definition. ------------------------------------------------------------------------------- > EOF README.Cygnus