VTR Change Log¶

Note that changes from release candidates (e.g. v8.0.0-rc1, v8.0.0-rc2) are included/repeated in the final release (e.g. v8.0.0) change log.

Unreleased¶

The following are changes which have been implemented in the VTR master branch but have not yet been included in an official release.

Added¶

Changed¶

Fixed¶

Deprecated¶

Removed¶

v9.0.0 - 2024-12-23¶

Added¶

Support for Advanced Architectures:
- 3D FPGA and RAD architectures.
- Architectures with hard Networks-on-Chip (NoCs).
- Distinct horizontal and vertical channel widths and types.
- Diagonal routing wires and other complex wire shapes (L-shaped, T-shaped, ….).
New Benchmark Suites:
- Koios: A deep-learning-focused benchmark suite with various design sizes.
- Hermes: Benchmarks utilizing hard NoCs.
- TitanNew: Large benchmarks targeting the Stratix 10 architecture.
Commercial FPGAs Architecture Captures:
- Intel’s Stratix 10 FPGA architecture.
- AMD’s 7-series FPGA architecture.
Parmys Logic Synthesis Flow:
- Better Verilog language coverage
- More efficient hard block mapping
VPR Graphics Visualizations:
- New interface for improved usability and underlying graphics rewritten using EZGL/GTK to allow more UI widgets.
- Algorithm breakpoint visualizations for placement and routing algorithm debugging.
- User-guided (manual) placement optimization features.
- Enabled a live connection for client graphical application to VTR engines through sockets (server mode).
- Interactive timing path analysis (IPA) client using server mode.
Performance Enhancements:
- Parallel router for faster inter-cluster routing or flat routing.
Re-clustering API to modify packing decisions during the flow.
Support for floorplanning and placement constraints.
Unified intra- and inter-cluster (flat) routing.
Comprehensive web-based VTR utilities and API documentation.

Changed¶

The default values of many command line options (e.g. inner_num is 0.5 instead of 1.0)
Changes to placement engine
- Smart centroid initial placement algorithm.
- Multiple smart placement directed moves.
- Reinforcement learning-based placement algorithm.
Changes to routing engine
- Faster lookahead creation.
- More accurate lookahead for large blocks.
- More efficient heap and pruning strategies.
- max pres_fac capped to avoid possible numeric issues.

Fixed¶

Many algorithmic and coding bugs are fixed in this release

Removed¶

Breadth-first (non-timing-driven) router.
Non-linear congestion placement cost.

v8.0.0 - 2020-03-24¶

Added¶

Support for arbitrary FPGA device grids/floorplans
Support for clustered blocks with width > 1
Customizable connection-block and switch-blocks patterns (controlled from FPGA architecture file)
Fan-out dependent routing mux delays
VPR can generate/load a routing architecture (routing resource graph) in XML format
VPR can load routing from a .route file
VPR can performing analysis (STA/Power/Area) independently from optimization (via vpr --analysis)
VPR supports netlist primitives with multiple clocks
VPR can perform hold-time (minimum delay) timing analysis
Minimum delays can be annotated in the FPGA architecture file
Flow supports formal verification of circuit implementation against input netlist
Support for generating FASM to drive bitstream generators
Routing predictor which predicts and aborts impossible routings early (saves significant run-time during minimum channel width search)
Support for minimum routable channel width ‘hints’ (reduces minimum channel width search run-time if accurate)
Improved VPR debugging/verbosity controls
VPR can perform basic netlist cleaning (e.g. sweeping dangling logic)
VPR graphics visualizations:
- Critical path during placement/routing
- Cluster pin utilization heatmap
- Routing utilization heatmap
- Routing resource cost heatmaps
- Placement macros
VPR can route constant nets
VPR can route clock nets
VPR can load netlists in extended BLIF (eBLIF) format
Support for generating post-placement timing reports
Improved router ‘map’ lookahead which adapts to routing architecture structure
Script to upgrade legacy architecture files (vtr_flow/scripts/upgrade_arch.py)
Support for Fc overrides which depend on both pin and target wire segment type
Support for non-configurable switches (shorts, inline-buffers) used to model structures like clock-trees and non-linear wires (e.g. ‘L’ or ‘T’ shapes)
Various other features since VTR 7

Changed¶

VPR will exit with code 1 on errors (something went wrong), and code 2 when unable to implement a circuit (e.g. unroutable)
VPR now gives more complete help about command-line options (vpr -h)
Improved a wide variety of error messages
Improved STA timing reports (more details, clearer format)
VPR now uses Tatum as its STA engine
VPR now detects missmatched architecture (.xml) and implementation (.net/.place/.route) files more robustly
Improved router run-time and quality through incremental re-routing and improved handling of high-fanout nets
The timing edges within each netlist primitive must now be specified in the section of the architecture file
All interconnect tags must have unique names in the architecture file
Connection block input pin switch must now be specified in section of the architecture file
Renamed switch types buffered/pass_trans to more descriptive tristate/pass_gate in architecture file
Require longline segment types to have no switchblock/connectionblock specification
Improve naming (true/false -> none/full/instance) and give more control over block pin equivalnce specifications
VPR will produce a .route file even if the routing is illegal (aids debugging), however analysis results will not be produced unless vpr --analsysis is specified
VPR long arguments are now always prefixed by two dashes (e.g. --route) while short single-letter arguments are prefixed by a single dash (e.g. -h)
Improved logic optimization through using a recent 2018 version of ABC and new synthesis script
Significantly improved implementation quality (~14% smaller minimum routable channel widths, 32-42% reduced wirelength, 7-10% lower critical path delay)
Significantly reduced run-time (~5.5-6.3x faster) and memory usage (~3.3-5x lower)
Support for non-contiguous track numbers in externally loaded RR graphs
Improved placer quality (reduced cost round-off)
Various other changes since VTR 7

Fixed¶

FPGA Architecture file tags can be in arbitary orders
SDC command arguments can be in arbitary orders
Numerous other fixes since VTR 7

Removed¶

Classic VPR timing analyzer
IO channel distribution section of architecture file

Deprecated¶

VPR’s breadth-first router (use the timing-driven router, which provides supperiour QoR and Run-time)

Docker Image¶

A docker image is available for VTR 8.0 release on mohamedelgammal/vtr8:latest. You can run it using the following commands:

$ sudo docker pull mohamedelgammal/vtr8:latest
$ sudo docker run -it mohamedelgammal/vtr8:latest

v8.0.0-rc2 - 2019-08-01¶

Changed¶

Support for non-contiguous track numbers in externally loaded RR graphs
Improved placer quality (reduced cost round-off)

v8.0.0-rc1 - 2019-06-13¶

Added¶

Support for arbitrary FPGA device grids/floorplans
Support for clustered blocks with width > 1
Customizable connection-block and switch-blocks patterns (controlled from FPGA architecture file)
Fan-out dependent routing mux delays
VPR can generate/load a routing architecture (routing resource graph) in XML format
VPR can load routing from a .route file
VPR can performing analysis (STA/Power/Area) independently from optimization (via vpr --analysis)
VPR supports netlist primitives with multiple clocks
VPR can perform hold-time (minimum delay) timing analysis
Minimum delays can be annotated in the FPGA architecture file
Flow supports formal verification of circuit implementation against input netlist
Support for generating FASM to drive bitstream generators
Routing predictor which predicts and aborts impossible routings early (saves significant run-time during minimum channel width search)
Support for minimum routable channel width ‘hints’ (reduces minimum channel width search run-time if accurate)
Improved VPR debugging/verbosity controls
VPR can perform basic netlist cleaning (e.g. sweeping dangling logic)
VPR graphics visualizations:
- Critical path during placement/routing
- Cluster pin utilization heatmap
- Routing utilization heatmap
- Routing resource cost heatmaps
- Placement macros
VPR can route constant nets
VPR can route clock nets
VPR can load netlists in extended BLIF (eBLIF) format
Support for generating post-placement timing reports
Improved router ‘map’ lookahead which adapts to routing architecture structure
Script to upgrade legacy architecture files (vtr_flow/scripts/upgrade_arch.py)
Support for Fc overrides which depend on both pin and target wire segment type
Support for non-configurable switches (shorts, inline-buffers) used to model structures like clock-trees and non-linear wires (e.g. ‘L’ or ‘T’ shapes)
Various other features since VTR 7

Changed¶

VPR will exit with code 1 on errors (something went wrong), and code 2 when unable to implement a circuit (e.g. unroutable)
VPR now gives more complete help about command-line options (vpr -h)
Improved a wide variety of error messages
Improved STA timing reports (more details, clearer format)
VPR now uses Tatum as its STA engine
VPR now detects missmatched architecture (.xml) and implementation (.net/.place/.route) files more robustly
Improved router run-time and quality through incremental re-routing and improved handling of high-fanout nets
The timing edges within each netlist primitive must now be specified in the section of the architecture file
All interconnect tags must have unique names in the architecture file
Connection block input pin switch must now be specified in section of the architecture file
Renamed switch types buffered/pass_trans to more descriptive tristate/pass_gate in architecture file
Require longline segment types to have no switchblock/connectionblock specification
Improve naming (true/false -> none/full/instance) and give more control over block pin equivalnce specifications
VPR will produce a .route file even if the routing is illegal (aids debugging), however analysis results will not be produced unless vpr --analsysis is specified
VPR long arguments are now always prefixed by two dashes (e.g. --route) while short single-letter arguments are prefixed by a single dash (e.g. -h)
Improved logic optimization through using a recent 2018 version of ABC and new synthesis script
Significantly improved implementation quality (~14% smaller minimum routable channel widths, 32-42% reduced wirelength, 7-10% lower critical path delay)
Significantly reduced run-time (~5.5-6.3x faster) and memory usage (~3.3-5x lower)
Various other changes since VTR 7