- Apr 12, 2018
-
-
David Hancock authored
-
- Apr 11, 2018
-
-
David Hancock authored
Add mininet script argument for automating mn commands
-
- Dec 04, 2017
-
-
David Hancock authored
Action handling now involves checking whether meta_primitive_state.primitive[+Y+] == 0 between primitives, which indicates no further primitives remain, but that value was used to indicate a modify_field primitive, so the value for modify_field has been changed.
-
- Dec 01, 2017
-
-
David Hancock authored
-
David Hancock authored
Requires update to compiler in hp4-ctrl repository
-
David Hancock authored
-
- Nov 20, 2017
-
-
David Hancock authored
-
David Hancock authored
This folder housed old tests that have long since been rendered obsolete by 1) changes to HyPer4; and 2) the controller repository (hp4-ctrl: git@gitlab.flux.utah.edu:hp4/hp4-ctrl.git). Saved two analysis scripts that may yet be useful in hp4/analysis.
-
David Hancock authored
All files were obsolete, though hp4_design.md may still have some interest (but probably only to me). This file was nevertheless removed from the repo and moved to a local folder for archived design analysis.
-
David Hancock authored
-
David Hancock authored
-
David Hancock authored
-
David Hancock authored
Files not meant for the public have been moved to a local directory that is gitignored
-
- Nov 07, 2017
-
-
David Hancock authored
-
David Hancock authored
Makes virtual network compositions need to use explicit forwarding instead of copying ingress port to egress_spec, but in the meantime, this commit better supports vdev chains, esp. vdevs like the arp proxy.
-
- Nov 05, 2017
-
-
David Hancock authored
Also, comments added to help identify conditionals, supporting debugging via the event logger. The next_table field was 4 bits, but init_program_state was assigning a value of '99' corresponding to matchless table type. When a modify_field tries to assign a value larger than the width of the target, bmv2 silently fails. Over in hp4-ctrl, t06 now works in its most essential test - a chain is created with simple_router -> firewall, and the router routes, and the firewall blocks/doesn't block correctly!
-
- Nov 03, 2017
-
-
David Hancock authored
-
- Oct 31, 2017
-
-
David Hancock authored
-
- Oct 30, 2017
-
-
David Hancock authored
Conflicts: hp4/hp4commands.txt hp4/p4src/includes/parser.p4 hp4/p4src/includes/setup.p4 hp4/template/includes/parser.p4t hp4/template/includes/setup.p4t
-
David Hancock authored
p4c-hp4 belongs to the hp4-ctrl repo. Script for nano_msg client updated. Misc files deleted.
-
- Oct 27, 2017
-
-
David Hancock authored
-
- Oct 26, 2017
-
-
David Hancock authored
Now, packets drop when we tell them to.
-
- Oct 25, 2017
-
-
David Hancock authored
Thought avoiding resubmits would be key to significant increase in throughput, but this commit casts doubt on the idea that resubmits are the main contributor to the throughput penalty. The next obvious thing is the number of actions that have dependencies among their primitives, which would induce stalls in the pipeline.
-
David Hancock authored
-
David Hancock authored
-
- Oct 24, 2017
-
-
David Hancock authored
This reverts commit b729fcc0. Turns out that made performance worse, significantly.
-
David Hancock authored
Minimum Ethernet frame is 64 B for good reason - regardless of link layer protocol, I doubt we'll ever see a packet smaller than 64 B. This change should avoid most of the resubmits previously incurred by HyPer4's programmable parser.
-
- Oct 13, 2017
-
-
David Hancock authored
-
David Hancock authored
-
David Hancock authored
-
- Oct 12, 2017
-
-
David Hancock authored
run.sh works the same way in the p4-projects repo; it invokes run_demo.sh after setting a path variable. Use run.sh instead of run_demo.sh from now on.
-
- Oct 05, 2017
-
-
David Hancock authored
-
David Hancock authored
A better word for 'extracted' would be 'parsed representation', but alas
-
David Hancock authored
The previous commit overlooked a few things...
-
David Hancock authored
-
- Sep 26, 2017
-
-
David Hancock authored
Revise/improve support for arp proxy scenario; remove assumption that commands.txt file exists and should be fed to P4 switch by default This commit corresponds to a commit in the p4-projects repo
-
- Sep 24, 2017
-
-
David Hancock authored
P4-projects repo is where 'native' functions are housed and tested; now hp4 sims are kicked off the same way, and topo.py and run_demo.sh and run.sh are identical (env.sh can be different to accomodate different paths for each setting)
-
- Sep 22, 2017
-
-
David Hancock authored
-
- Sep 21, 2017
-
-
David Hancock authored
This commit brings the repo up to date but is not meant to convey pursuit of full INT demo at this time.
-
- Sep 17, 2017
-
-
David Hancock authored
Having the match field simplifies controller tasks related to composition management. It has to do with handles associated with table entries. Without the vdev_ID match field, essentially we have a tough time coordinating between vdevs that employ the drop primitive in the same stage/action (esp: two vdevs created from the same .p4). Also in this commit: hp4commands adds the startup rules that are required for proper functioning but independent of any vdev.
-