aboutsummaryrefslogtreecommitdiff
path: root/src/test/scala/firrtlTests/ReplSeqMemTests.scala
AgeCommit message (Collapse)Author
2019-08-01Followup to PR #1142chick
Fixes a threading bug in where lazy reading of file caused a problem for multithreaded access to the that was read. Changes all uses of io.Source to use new API getText getLines getTextResource getLinesResouce Make style to only import FileUtils and not its methods So code is more explicit as e.g. FileUtils.getText()
2019-03-19Designs with no SeqMems should produce empty MemConf strings, and this ↵John Wright
should be parsable without excepting (#1060)
2019-03-07Add a data structure for memory conf reading and writing (#1041)John Wright
* Copy MemConf.scala from ucb-bar/barstools#35 into memlib. This provides a data structure wrapper around the existing memory conf format which contains both reading and writing methods, making it easier to write code that needs to read the format. * Add MemConf tests and use a Map[MemPort, Int] for port lists instead of a Seq[MemPort] which is a bit less fragile.
2018-08-23Fix NoDedupMem to be cognizant of Module scope (#876)Jack Koenig
Previously, mems marked no dedup would prevent mems with the same instance name in other modules from deduping
2018-07-03Improve code generation for smem wmode and [w]mask ports (#834)Andrew Waterman
[skip formal checks] LEC passes with Formality * Improve code generation for smem RW-port wmode port A common case for these port-enables is wen = valid & write ren = valid & !write which the RW-port transform currently turns into en = (valid & write) | (valid & !write) wmode = valid & write because it proved `wen` and `ren` are mutually exclusive via `write`. Synthesis tools can trivially optimize `en` to `valid`, so that's not a problem, but the wmode field can't be optimized if going into a black box. This PR instead sets `wmode` to whatever node was used to prove mutual exclusion, which is always a simpler expression. In this case: en = (valid & write) | (valid & !write) wmode = write * In RemoveCHIRRTL, infer mask relative to port definition Previously, it was inferred relative to the memory definition causing the mask condition to be redundantly conjoined with the enable signal. Also enable ReplSeqMems to ignore all ValidIfs (not just on Clocks) to improve QoR.
2018-02-27Refactor Annotations (#721)Jack Koenig
- Old Annotation renamed to deprecated LegacyAnnotation - Annotation is now a trait that can be extended - New JsonProtocol for Annotation [de]serialization - Replace AnnotationMap with AnnotationSeq - Deprecate Transform.getMyAnnotations - Update Transforms - Turn on deprecation warnings - Remove deprecated Driver.compile - Make AnnotationTests abstract with Legacy and Json subclasses - Add functionality to convert LegacyAnnotations of built-in annos This will give a noisy warning and is more of a best effort than a robust solution. Fixes #475 Closes #609
2017-06-28Promote ConstProp to a transformJack Koenig
2017-05-11Improved Global Dead Code Elimination (#549)Jack Koenig
Performs DCE by constructing a global dependency graph starting with top-level outputs, external module ports, and simulation constructs as circuit sinks. External modules can optionally be eligible for DCE via the OptimizableExtModuleAnnotation. Dead code is eliminated across module boundaries. Wires, ports, registers, and memories are all eligible for removal. Components marked with a DontTouchAnnotation will be treated as a circuit sink and thus anything that drives such a marked component will NOT be removed. This transform preserves deduplication. All instances of a given DefModule are treated as the same individual module. Thus, while certain instances may have dead code due to the circumstances of their instantiation in their parent module, they will still not be removed. To remove such modules, use the NoDedupAnnotation to prevent deduplication.
2017-05-10Update rename2 (#478)Adam Izraelevitz
* Added pass name to debug logger * Addresses #459. Rewords transform annotations API. Now, any annotation not propagated by a transform is considered deleted. A new DeletedAnnotation is added in place of it. * Added more stylized debugging style * WIP: make pass transform * WIP: All tests pass, need to pull master * Cleaned up PR * Added rename updates to all core transforms * Added more rename tests, and bugfixes * Renaming tracks non-leaf subfields E.g. given: wire x: {a: UInt<1>, b: UInt<1>[2]} Annotating x.b will eventually annotate x_b_0 and x_b_1 * Bugfix instance rename lowering broken * Address review comments * Remove check for seqTransform, UnknownForm too restrictive check
2017-03-23Pass now subclasses Transform (#477)Adam Izraelevitz
2017-03-06Fix mistake when rebasingAdam Izraelevitz
2017-03-06Added more stylized debugging styleAdam Izraelevitz
2017-03-06Addresses #459. Rewords transform annotations API.Adam Izraelevitz
Now, any annotation not propagated by a transform is considered deleted. A new DeletedAnnotation is added in place of it.
2017-03-06Add ability to emit 1 file per module (#443)Jack Koenig
Changes Emitters to also be Transforms and use Annotations for both telling an emitter to do emission as well as getting the emitted result. Helper functions ease the use of the new interface. Also adds a FirrtlExecutionOptions field as well as a command-line option. Use of Writers in Compilers and Emitters is now deprecated.
2017-02-21Implementation of nodedupe mem (#447)Colin Schmidt
This allows the replseqmem transform to not deduplicate some memories, based on their name.
2017-01-27Fix signed types (#422)Angie Wang
* type conversions between sint/fixed and uint added at memory interfaces for replseqmem * turns out asFixed requires bp as constant in PrimOps (really should be documented) * fixed legalizeconnects to handle FixedPt * added tests for replseqmem failure with signed types
2016-12-08Clk2clock - rename the implicit "clk" module input "clock" (#387)Jim Lawson
* Rename implict module "clk" input to "clock". This doesn't rename all the "self-contained" test instances. nor the memory "clk" enables, nor the implict module "clk"s in the regress .fir files. * Consistency: rename implict module "clk" input to "clock" in "self-contained" test instances. This doesn't rename the memory "clk" enables, nor the implict module "clk"s in the regress .fir files.
2016-11-23Stringified annotations (#367)Adam Izraelevitz
Restricts annotations to be string-based (and thus less typesafe) Makes annotations more easily serializable and interact with Chisel
2016-11-04Cleanup license at top of every file (#364)Jack Koenig
Replace with more sensible comment to see LICENSE rather than including the whole license in every file
2016-11-04Refactor Compilers and Transformsjackkoenig
* Transform Ids now handled by Class[_ <: Transform] instead of magic numbers * Transforms define inputForm and outputForm * Custom transforms can be inserted at runtime into compiler or the Driver * Current "built-in" custom transforms handled via above mechanism * Verilog-specific passes moved to the Verilog emitter
2016-10-30Keep package name + directory structure consistent (#354)Colin Schmidt
* Keep package name + directory structure consistent This annoyed me so heres a PR * fix InferReadWrite references * delete .ConvertFixedToSInt.scala.swo
2016-10-17Reorganized memory blackboxing (#336)Adam Izraelevitz
* Reorganized memory blackboxing Moved to new package memlib Added comments Moved utility functions around Removed unused AnnotateValidMemConfigs.scala * Fixed tests to pass * Use DefAnnotatedMemory instead of AppendableInfo * Broke passes up into simpler passes AnnotateMemMacros -> (ToMemIR, ResolveMaskGranularity) UpdateDuplicateMemMacros -> (RenameAnnotatedMemoryPorts, ResolveMemoryReference) * Fixed to make tests run * Minor changes from code review * Removed vim comments and renamed ReplSeqMem
2016-09-21refactor AnnotateMemMacrosDonggyu Kim
2016-09-14Fixed infinite loop for finding connect origin in ReplSeqMem (#300)Angie Wang
* Addressed the fact that a node can be connected to itself (updating reg)
2016-09-12Bug fix -- remove all empty expressions after ReplSeqMem passes (#294)Angie Wang
* Bug fix -- remove all empty expressions after ReplSeqMem passes * Added test to make sure ReplSeqMem can handle BundleType SMem (EmptyExpression leakage)
2016-09-08Remove brittle ReplSeqMemTestjackkoenig
This test is breaks with any minor change to code generation. It should be replaced with a more robust test.
2016-09-08clean up ExpandWhensDonggyu Kim
2016-09-06Address style feedback and add tests for getConnectOrigin utilityAngie
2016-09-06Expanded annotations for valid memory sizesAngie
2016-09-06Added simple unit test for ReplSeqMemAngie