Improvements and fix to the dineroIV plugin #3
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Misaligned accesses are currently ignored as it requires more processing for cost computation with dinero. Work around costs greater than the cach hierarchy depth as returned by the library. It is unclear why we get costs > 2 for instance on a 3 level cache hierarchy.
Added computation of cycles estimate by using the DINAROIV_LATENCIES env var (default: 2,40). These latencies are used to compute an estimated cost of data cache/memory reads. This variable defines the memory latency at each cache level (2 cycles for fetch from L1, and 40 cycles for fetch from RAM as the default). For instance on a 2 level cache + RAM confighuration: DINAROIV_LATENCIES=2,5,40 means 2 cycles for L1, 5 cycles for L2 and 40 cycles for RAM read accesses.
Add -O2 to compile flags for dineroIV library objects.
Change initialisation of memory levels to be fully dynamic, and removed the previous limit of three mem levels.
Reimplement caches cost computation from actual number of fetches at each cache level. Add DINEROIV_OUTPUTS envvar for selecting outputs kinds.
Fix computation of address on 32 bits targets. Remove output of ld/st value which is not useful for cache statistics.and requires more caution as the register value may be killed after the operation.
…s depth." This reverts commit b1ff9c9. Change no more required with new cost computation.
…mary on exit." This reverts commit 93fc533. Revert change to dineroIV, no more needed with new cost computation.
Fix compilation after rebase on top of QEMU 1.7.0. Fix installation path for plugins.
Compute instruction count in helper directly. Remove inlining of instruction count computation at translation time. This is unecessary as the overhead is the same as soon as the helper is called after each instruction.
Do not expect unecessary parameters (cmd, latencies) when not required. Compute cycles and stats only when requests. In particular when trace mode is requested, dinero is not used at all, nor initialised.
The dyncount plugin counts dynamic instructions per mnemonics and per groups as available in capstone library. The plugin support for now x86_64 and i386 guests and relied on libcapstone >= 3.0. Add --enable/--disable-casptone option at configure time. Defaults to disabled when --enable-tcg-plugin is not set. When --enable-tcg-plugin is set and the capstone library is not found or of incompatible version, the dyncount plugin is a no-op and it emits aa warning at runtime.
The dyntrace plugin outputs a disassembled execution trace. The current implementation requires capstone >= 3.0.
When confiured with --debug, some plugins were not compiling correctly due to some TCGV args mismatches.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
A set of changes covering: