a³ release 11.04

New targets

Qualification Support Kits (QSKs)

  • One QSK package for all operating systems.
  • Added test for the GUI switch “Always read program headers”.
  • Changed command line interface for run_test.sh by adding the operating system as a required first parameter.

a³ toolchain

  • Reduced memory consumption and increased analysis speed. Some analyses may now be 25–50% faster.
  • Mapping only allows infinite call string length; max-length will be ignored in AIS files.

XTC

Minor extension to the XTC specification allowing the selection of the used core.

a³ GUI

  • Analysis messages now include the time and memory consumption of each analysis step.
  • Improved display of analysis messages:
    • easily configurable verbosity
    • improved highlighting of errors and warnings
    • retaining results of analyses performed earlier
    • new menu action to show the selected point in a control flow graph
  • New Stack Contribution view.
  • Added auto completion for analysis start points in Files view.
  • Added global search in all source files.
  • Improved error handling (Windows).
  • Improved usability of XML query dialogs.
  • Improved, consistent filtering and search functionality across all Information views (Symbols, Variables, Functions…).
  • Fixes for some glitches and crashes related to graph visualization.
  • Better handling of machine-setting files and improved annotation generation for machine settings.
  • Editors: selecting text and hitting Ctrl + F now enables searching in the selected text only.
  • TraceValidator: user can now visualize traces in trace validation analysis.
  • The default names for reports and AIS files are now based on the analysis ID.
  • Various fixes and enhancements related to changing or opening a project.
  • Interactive value analysis: filter out infeasible contexts.
  • Global text report file and analysis text report file may no longer have the same filename.
  • Improved tab order in GUI.
  • aiSee Quick Reference added to the Help menu.
  • Target-specific changes:
    • MPC7448/MPC7448s/MPC603e: DRTRY mode settings fixed.
    • MPC5XX: hardware configuration register SRAMMCR has been renamed to CRAMMCR_A and CRAMMCR_B.
    • MPC603e: added switch for enabling/disabling 60x bus pipelining.
    • C16x: configure external bus via register contents.
    • V850: allow to choose register mode for the GHS compiler. More core specific changes.
    • M68020, x86, H8: added MAP-file entry in the Files view.

AIS

  • Contexts: the context parameter max-length is now obsolete. The Analyzer will always use “max-length=infinite” to increase analysis precision. If complexity is too large, decrease max-unroll. Hence, no more “any-history” and “...,” in the call string outputs.
  • ResultCombinator: entry fields for expressions now allow C and C++ style comments.
  • The annotation
    area .. features ddr = 1;
    is replaced by the more general
    area .. features memory_type = "DDR"
  • PCP2: added AIS Quick Reference Guide.

Decoding

  • Fixed the ”immediately return“ annotation for instructions followed by a delay slot.
  • Improved debug information for arrays of executables using the DWARF1 format.
  • General improvements to reading DWARF1 debug information.
  • PPC:
    • more switch table patterns (GHS, DiabData)
    • more computed-call table patterns (DiabData)
    • improved switch table decoding (GHS)
    • allow ASAM file to be read via the --asam-file parameter of exec2crl, and ASAM variable names to be written to the report in addition to DWARF names.
  • SPARC: improved detection of tail calls.
  • V850: improved switch table decoding.

Stack, loop, and value analysis

  • Improved automatic loop-bound detection.
  • Improved precision of sign extensions for value analysis.
  • PowerPC: improved precision of value analysis.
  • MPC603e/MPC750/MPC755/MPC7448: improved reporting of speculative memory accesses in the XML report for accesses that occur in code locations that are found to be infeasible.

Path analysis

  • Improved ILP solver (clpsolve).
  • “Prediction file based (ILP)” now supports busy waiting loops.
  • Removed the GLPK based ILP solver for performance reasons.
  • Replaced the pathan2 toolchain with poptimize2ilp. Consolidation of tools for path analysis.
  • Better handling of program ends in the middle of the program, such as external routines.
  • More strict call-return constraints. If a routine must always return, this is now ensured. This can lead to more tight bounds and a better approximation of the WCET path in the graph.

Cache and pipeline analysis

  • Optimized memory usage of the pipeline analysis framework.
  • Correctly treating computed call targets annotated as
    instruction <program point> target is not analyzed
    as potential program ends.
  • Am486:
    • fixed internal errors for infeasible entries to multi-entry loops
    • fixed issues with cache locking for code memory areas
  • MPC5xx:
    • improved handling of memory modules governed by an external memory controller
    • improved handling of accesses to registers via the internal memory bus
  • MPC755: fixed handling of write-back penalties for data caches.
  • MPC755, MPC603e: improved cache access counting.
  • MPC603e, PPC750: improved the PCI backend. Assuming a 32-bit PCI bus width, 64-bit accesses are now supported by splitting into two separate 32-bit accesses.
  • PPC750: improved handling of tail instructions.
  • TriCore 179x: improved model for burst accesses across DMI.
  • V850:
    • The V850E2M CPU subsystem model has been re-implemented according to new information from Renesas.
    • The pipeline interlock treatment in the V850E2M pipeline model has been refined according to new information from Renesas.

Visualization and reporting

  • Matching XSDs and XSLs are bundled for the release in the share/xsd directory. Users will always have local available copies.
  • Performance improvements to WCET graph visualization.
  • Better reporting of assertion violations in value analysis.
  • The warning about missing timing behaviour now follows the format:
    No memory access timing found for [code fetch|data read access|data write access] of instruction <address>[ to <address range>][, context any-history].
  • Fixed problem with interactive pipeline mode, which led to corruption of text and XML reports.
  • C33: the annotation “area .. count accesses;” now produces output.