- 07 Mar, 2012 40 commits
-
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[new mpm]: debugged input and output (first version which does not loose pages during the basic tests !!!)
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[new mpm]: added FSM to manage Linked List writes, it enables for some performance optimizatio,prevents STALLing, loosing pages
-
Maciej Lipinski authored
swcore[new mpm]: modified linked list to enable clearing pckstart_addr with the last page of the previous pck
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[new mpm]: bugfix in asynchronous MPM (the LL address mux-ed in p_ll_mux_addr was not synch-ed with delayed grand and data registering in mpm_rpath_io_block
-
Maciej Lipinski authored
swcore[new mpm]: freeing module adapted to new usecnt handling (the usecnt of pck stored only with the first page). This is the first swcore version working fully with the new MPM, however, it is still buggy and needs tests
-
Maciej Lipinski authored
swcore[new mpm]: testbenches and wrappers changed to include new (for new mpm, i/o blocks) parameters
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[new mpm]: page allocator (both old and new) modified to enable a new way of page usecnt handling: we store the usecnt only in the first page of the pck, the non-first pages have usecnt=1. When pck is sent on the output port, only the usecnt of the first page is decremented unless the usecnt==1, then we free all the pages allocated to the pck, so the page allocator needs to indicate if the usecnt is 1
-
Maciej Lipinski authored
swcore[new mpm]: bugfix in asynchronous MPM (dreq=LOW on the one but last word caused readout problem and dlast to be HIGH when dvalid=LOW while not yet the last word)
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[new_mpm + cut-through]: modified LinkedList for the needs of the new_mpm (direct access with core_clock) and enabled cut-through (validation of page entry, and waiting on non-validated entry. Additionally, the write access has been speeded up (two writes, one for valid entry, second for the next non-valid entry), and the Round Robin arbitration has been speeded up
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[new mpm]: modified (thoroughly) input_block (+ other necessary changes) to work with the new mpm (asynch by Tom)
-
Maciej Lipinski authored
swcore[generic-ing]: generic-azed hard-coded values (connected with rr_arbiter usage), identified limitation (in prio_encoder,TODO), added some description, tested with different port number values
-
Maciej Lipinski authored
swcore[generic-azing]: generic simulation (for any number of ports) works, cleaned up, added README to testbenches, changed names to add clarity to the naming
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
swcore[generic-azing]: no more global constants !!! -> fully generic-azed, removed global constants from the swc_swcore_pkg.vhd (not tested yet for different generic values, there is a known problem inherited from global consants - does not work when some configs are changed
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-
Maciej Lipinski authored
-