TEMU 3.0.4 addresses two issues related to IRQ remapping on the GR740
In addition the following features have been added:
cpu0.stacktrace
in the command line interface.TEMU 3.0.3 addresses the following bugs:
experimental
namespace for the SoC ID and version properties in the AhbCtrl
added in TEMU 3.0.2 has been dropped. Properties should now be named without the experimental
tag.TEMU 3.0.2 has been released it addresses the following issues:
GRGPREG
model. This model is a dummy, but provides the correct PNP info and user configurable registers.GRGPRBANK
model. This model is a dummy, but provides the correct PNP info and user configurable registers.AhbCtrl
model to expose undocumented SoC ID and version IDs.TEMU 3.0.2 also adds the following enhancements:
ExecutionProfiler
model that can monitor performance metrics over time.Sampler
model that can sample arbitrary properties over time.ScriptSequencer
model that can sequence TEMU scripts on specific times.TEMU 3.0.1 has been released, it addresses the following issues:
TEMU 3.0.0 has now been released. TEMU 3.0.0 is a major milestone in the TEMU product evolution.
The major focus for TEMU 3 has been on the performance and we add two major changes to system to increase performance.
The first change is the introduction of a pre-decoded dispatch interpreter, which converts target architecture instructions to an internal intermediate representation format. TEMU 2 used to fetch and decode instructions every time they were executed, but with TEMU 3 we only decode them once. This speeds up the interpreter by around 50% in measurements we have performed.
The second change for performance is the introduction of a dynamic binary translator engine. This engine can yield an additional 100% performance increase (exact numbers depend on the software running, and they can be higher or lower).
TEMU 3 also introduce the ability to define custom idle patterns, which although the API is currently unstable, could be used to define custom idle loop matchers (e.g. memory scrubbers etc). This is possible since the interpreter (and binary translator) decodes blocks of instructions now.
TEMU 3 also switch to semantic versioning for this and future releases, and although TEMU 2 almost use this system, it was not formally semver compliant. This will add clarity to the question of API and ABI stability and the compatibility of different versions.
New Models
Removals