Cesar Strauss [Fri, 1 Jan 2021 15:06:48 +0000 (12:06 -0300)]
Make output write enables independent of valid_o
Just combinatiorally decode the operation.
This is because MultiCompUnit depends on *_ok being kept valid.
Cesar Strauss [Fri, 1 Jan 2021 14:11:24 +0000 (11:11 -0300)]
Move NOP test case earlier
Better way to see alu_o_ok being disabled during the instruction.
Cesar Strauss [Fri, 1 Jan 2021 12:59:49 +0000 (09:59 -0300)]
Disable data value output on NOP
Cesar Strauss [Fri, 1 Jan 2021 12:54:45 +0000 (09:54 -0300)]
Add condition register (CR) output
Cesar Strauss [Thu, 31 Dec 2020 21:41:18 +0000 (18:41 -0300)]
Implement and test NOP in the test ALU
Change the output port from Signal to Data, to allow for the output to be
masked-out.
Specify a masked-out output in the NOP test case.
Cesar Strauss [Thu, 31 Dec 2020 20:43:34 +0000 (17:43 -0300)]
Don't use OP_NOP for zero-delay subtraction
We are going to implement an actual NOP
Cesar Strauss [Thu, 31 Dec 2020 20:31:41 +0000 (17:31 -0300)]
Test first input port being masked out
Cesar Strauss [Thu, 31 Dec 2020 20:28:05 +0000 (17:28 -0300)]
Sign extend the second input port
Cesar Strauss [Thu, 31 Dec 2020 20:06:56 +0000 (17:06 -0300)]
Test masked-out second input port
Sign extend uses only the first port.
Cesar Strauss [Thu, 31 Dec 2020 19:51:03 +0000 (16:51 -0300)]
Add sign extend to the Test ALU
Cesar Strauss [Thu, 31 Dec 2020 13:36:58 +0000 (10:36 -0300)]
Show rdmaskn and wrmask in GTKWave
Cesar Strauss [Thu, 31 Dec 2020 13:08:41 +0000 (10:08 -0300)]
Use the increment operator
Cesar Strauss [Thu, 31 Dec 2020 13:04:05 +0000 (10:04 -0300)]
Add support for masked write operations
Note that the test ALU currently does not have any masked writes.
Cesar Strauss [Thu, 31 Dec 2020 12:20:13 +0000 (12:20 +0000)]
Clarify reason for holding rdmaskn valid during the entire cycle
Cesar Strauss [Thu, 31 Dec 2020 10:42:27 +0000 (10:42 +0000)]
Remove previous version of the CompUnit parallel unit test
It was too detailed, modeling properties which are better checked by
formal verification.
Other than that, the new version has reached feature parity, so the old
code can finally be removed.
Cesar Strauss [Thu, 31 Dec 2020 10:22:04 +0000 (10:22 +0000)]
Only hold the decoder signals for one cycle, along with issue_i
The exception is rdmaskn, which is not latched, and must be held valid
for the entire instrucion cycle.
Cesar Strauss [Wed, 30 Dec 2020 21:00:38 +0000 (21:00 +0000)]
Test the rdmaskn control signal
The operation issuer now can drive the rdmaskn signals, and check
that the operand fetch on any masked ports is supressed.
Cesar Strauss [Tue, 29 Dec 2020 11:26:29 +0000 (11:26 +0000)]
Remove left-over comments.
The debug signals were removed in a previous commit, but the comment
lines remained.
Luke Kenneth Casson Leighton [Mon, 28 Dec 2020 20:30:34 +0000 (20:30 +0000)]
add CR1 to power_enums
Cesar Strauss [Sun, 20 Dec 2020 14:18:34 +0000 (11:18 -0300)]
Add support for CXXSim simulation
Cesar Strauss [Sun, 13 Dec 2020 18:16:29 +0000 (15:16 -0300)]
Ignore formal verification output in the source directory
This is similarly done in other sister directories.
Cesar Strauss [Sun, 13 Dec 2020 18:03:59 +0000 (15:03 -0300)]
Allow more test cases to be run with CXXSim
Luke Kenneth Casson Leighton [Sat, 12 Dec 2020 16:37:49 +0000 (16:37 +0000)]
skip madd, not implemented
Luke Kenneth Casson Leighton [Wed, 9 Dec 2020 18:13:28 +0000 (18:13 +0000)]
update submodules
Luke Kenneth Casson Leighton [Wed, 9 Dec 2020 18:10:48 +0000 (18:10 +0000)]
update submodules
Cesar Strauss [Mon, 7 Dec 2020 21:44:40 +0000 (18:44 -0300)]
Display the instruction type as a vector on cxxsim
It doesn't support enums traces yet.
Luke Kenneth Casson Leighton [Sun, 6 Dec 2020 19:35:57 +0000 (19:35 +0000)]
attempt to split into two separate GPIO banks due to a coriolis2 compile error
Cesar Strauss [Sun, 6 Dec 2020 12:31:31 +0000 (09:31 -0300)]
Whitespace
Cesar Strauss [Sun, 6 Dec 2020 11:34:35 +0000 (08:34 -0300)]
Update GTKWave documents to work with latest cxxsim
* Hierarchy begins at "top", just like pysim
* Avoid intermediate signals, that work differently on both
* Use the new "submodule" style in write_gtkw
Cesar Strauss [Sat, 5 Dec 2020 12:40:20 +0000 (09:40 -0300)]
Write a GTKWave document to investigate why the proof fails
Cesar Strauss [Sat, 5 Dec 2020 12:37:18 +0000 (09:37 -0300)]
Use the DummyALU regspec and its corresponding OpSubset
Luke Kenneth Casson Leighton [Thu, 3 Dec 2020 16:52:09 +0000 (16:52 +0000)]
put ls180 litex bus width back to 32 bit temporarily
Luke Kenneth Casson Leighton [Thu, 3 Dec 2020 15:51:56 +0000 (15:51 +0000)]
argh issue with yosys ABC
Luke Kenneth Casson Leighton [Thu, 3 Dec 2020 15:34:33 +0000 (15:34 +0000)]
add 3 more 4k SRAMs, change WB bus width to 64 in ls180 litex
Cesar Strauss [Sat, 28 Nov 2020 17:59:30 +0000 (14:59 -0300)]
Fix signal names: go/rel -> go_i/rel_o
Cesar Strauss [Tue, 24 Nov 2020 11:06:30 +0000 (08:06 -0300)]
Fix some typos and whitespace
Cesar Strauss [Tue, 24 Nov 2020 10:53:14 +0000 (07:53 -0300)]
Port the DummyALU test case to the new parallel issuer
In the process, fix its OpSubset to be consistent with the one which is
really used by this ALU.
This required adapting the issuer, to cope with the absence of some fields
in the OpSubset.
Cesar Strauss [Mon, 23 Nov 2020 10:59:42 +0000 (07:59 -0300)]
Results are now a list, so "expected" should follow suit
Cesar Strauss [Mon, 23 Nov 2020 10:40:04 +0000 (07:40 -0300)]
Parameterize the issuer on the number of operands and results
This allows reuse for the DummyALU, which has three operands, but is
otherwise similar to the ALU in terms of operations.
Also, the issuer now creates the producers and consumers.
Cesar Strauss [Sun, 22 Nov 2020 22:07:32 +0000 (19:07 -0300)]
Refactor the ALU operation issuer into a class
This allows sharing its code with other similar test cases.
Cesar Strauss [Sun, 22 Nov 2020 19:05:15 +0000 (16:05 -0300)]
Port the ALU test case to the new parallel test style
Mostly copy & paste from the Shifter, but using the operation spec
for the ALU.
The producers transaction count will fall behind on zero_a and imm_ok
executions, so these counters were added to the invariant check.
Cesar Strauss [Sun, 22 Nov 2020 17:12:19 +0000 (14:12 -0300)]
Add a GTKWave document to the ALU test case
Luke Kenneth Casson Leighton [Sat, 14 Nov 2020 18:41:29 +0000 (18:41 +0000)]
simplify litex-core wishbone interfaces
Cesar Strauss [Thu, 19 Nov 2020 10:52:05 +0000 (07:52 -0300)]
Separate input and output ports by color
Cesar Strauss [Thu, 19 Nov 2020 10:38:11 +0000 (07:38 -0300)]
Explain the test cases
Cesar Strauss [Wed, 18 Nov 2020 10:59:15 +0000 (07:59 -0300)]
Separate individual traces for each rel_o/go_i port
Use the new "bit" attribute to select individual bits from the
wide rel_o/go_i signals.
Tobias Platen [Tue, 17 Nov 2020 19:20:16 +0000 (20:20 +0100)]
testcase for dcbz
Cesar Strauss [Mon, 16 Nov 2020 22:17:19 +0000 (19:17 -0300)]
Add a transaction counter to producers and consumers
By comparing counts, we assure no data is duplicated or dropped.
Tobias Platen [Mon, 16 Nov 2020 19:02:05 +0000 (20:02 +0100)]
add class LoadStore1(PortInterfaceBase)
Cesar Strauss [Sun, 15 Nov 2020 20:20:19 +0000 (17:20 -0300)]
Implement ResultConsumer and port the Shifter unit tests to it.
Cesar Strauss [Sat, 14 Nov 2020 22:29:05 +0000 (19:29 -0300)]
Move the DUT driver to within the test case process
This reduces verbosity, as parameters are replaced by local variables in
the external scope.
Another way would be to save the parameters in a class, and transform the
function into a method.
Cesar Strauss [Sat, 14 Nov 2020 18:15:44 +0000 (15:15 -0300)]
Fix and enable the regspec test for the Shifter
1) use correct names for the Shifter ports in the regspec
2) migrate to the new OperandProducer
3) add the test on __main__
Luke Kenneth Casson Leighton [Sat, 14 Nov 2020 14:29:26 +0000 (14:29 +0000)]
sigh, direction wrong in IOtypes litex core
Luke Kenneth Casson Leighton [Fri, 13 Nov 2020 17:51:13 +0000 (17:51 +0000)]
reduce number of nc in ls180 to 24
Luke Kenneth Casson Leighton [Fri, 13 Nov 2020 17:47:46 +0000 (17:47 +0000)]
reduce clkcsel ls180 width (2 pins), rename pll_18 signal
Luke Kenneth Casson Leighton [Fri, 13 Nov 2020 16:10:42 +0000 (16:10 +0000)]
rename and add pll lock signal to ls180
Luke Kenneth Casson Leighton [Fri, 13 Nov 2020 16:04:15 +0000 (16:04 +0000)]
rename ls180 litex pll_48 output to pll_18
Luke Kenneth Casson Leighton [Fri, 13 Nov 2020 15:48:21 +0000 (15:48 +0000)]
add enable/disable arguments (not ideal but it works) to issuer_verilog.py
Luke Kenneth Casson Leighton [Fri, 13 Nov 2020 15:45:51 +0000 (15:45 +0000)]
remove io_in/out now it is not needed for niolib
Tobias Platen [Wed, 11 Nov 2020 18:51:41 +0000 (19:51 +0100)]
dcbz and tlbie first test, still incomplete
Tobias Platen [Wed, 11 Nov 2020 18:09:52 +0000 (19:09 +0100)]
fu/mmu/test/test_pipe_caller.py test case for mfspr
Luke Kenneth Casson Leighton [Tue, 10 Nov 2020 19:44:09 +0000 (19:44 +0000)]
add build commands to Makefile for versa ecp5
Luke Kenneth Casson Leighton [Tue, 10 Nov 2020 19:37:49 +0000 (19:37 +0000)]
submodule update
Luke Kenneth Casson Leighton [Tue, 10 Nov 2020 16:40:33 +0000 (16:40 +0000)]
remove ClockSelect module, use DummyPLL
Luke Kenneth Casson Leighton [Tue, 10 Nov 2020 15:49:56 +0000 (15:49 +0000)]
add separate DummyPLL module, according to API discussed at
https://bugs.libre-soc.org/show_bug.cgi?id=155#c21
;
Tobias Platen [Sun, 8 Nov 2020 12:05:36 +0000 (13:05 +0100)]
mmu fsm testcase: add check_fsm_outputs based on function from soc/fu/div/test/helper.py
Tobias Platen [Sun, 8 Nov 2020 09:31:11 +0000 (10:31 +0100)]
Merge branch 'master' of ssh://git.libre-riscv.org:922/soc
Tobias Platen [Sun, 8 Nov 2020 09:30:08 +0000 (10:30 +0100)]
mmu/fsm: test case for mtspr
Luke Kenneth Casson Leighton [Sat, 7 Nov 2020 22:27:38 +0000 (22:27 +0000)]
update submodule
Tobias Platen [Sat, 7 Nov 2020 14:43:07 +0000 (15:43 +0100)]
fixed a bug in src/soc/fu/mmu/fsm.py
Luke Kenneth Casson Leighton [Fri, 6 Nov 2020 11:48:01 +0000 (11:48 +0000)]
sigh sorting out litex pin-connections to sdram
Luke Kenneth Casson Leighton [Wed, 4 Nov 2020 19:10:28 +0000 (19:10 +0000)]
move back to 3.3v on X3 VERSA ECP5 connector
Tobias Platen [Wed, 4 Nov 2020 17:49:31 +0000 (18:49 +0100)]
MMU: begin test case for 'dcbz'
Tobias Platen [Tue, 3 Nov 2020 18:48:08 +0000 (19:48 +0100)]
fix broken unittest after installing power-instruction-analyzer
Luke Kenneth Casson Leighton [Tue, 3 Nov 2020 13:53:41 +0000 (13:53 +0000)]
swap jtag pinorder to match ulx3s
Luke Kenneth Casson Leighton [Tue, 3 Nov 2020 13:40:28 +0000 (13:40 +0000)]
change LVCMOS level on versa ecp5 jtag to 2.5v
Cesar Strauss [Sun, 1 Nov 2020 17:02:52 +0000 (14:02 -0300)]
Add a check for liveness.
There was already a check for the correctness of the results, but there
was no guarantee that any result would be produced at all.
Cole Poirier [Fri, 30 Oct 2020 21:33:00 +0000 (14:33 -0700)]
versa_ecp5.py add 4 arbitrarily assigned gpio pins to be used by
Libre-SOC JTAG interface on ulx3s
Cesar Strauss [Sat, 31 Oct 2020 18:29:37 +0000 (15:29 -0300)]
Check that the read and write counters differ at most by one
This will assure there are no dropped work items.
Cesar Strauss [Sat, 31 Oct 2020 13:46:00 +0000 (10:46 -0300)]
Remove stray comment
It was part of a code block that was removed.
Luke Kenneth Casson Leighton [Fri, 30 Oct 2020 18:47:59 +0000 (18:47 +0000)]
add JTAG extension to versa_ecp5 then we can use it
Cesar Strauss [Wed, 28 Oct 2020 22:57:55 +0000 (19:57 -0300)]
Implement an operand producer that talks the rel_o/go_i handshake
It can be instantiated once for each operand port, working in parallel
with the main test-bench process.
Luke Kenneth Casson Leighton [Sat, 24 Oct 2020 22:15:01 +0000 (23:15 +0100)]
submodule update
Cesar Strauss [Sat, 24 Oct 2020 17:47:16 +0000 (14:47 -0300)]
Create a GTKWave document for the test ALU unit tests
Luke Kenneth Casson Leighton [Thu, 22 Oct 2020 15:59:02 +0000 (16:59 +0100)]
add query about cross-domain on the JTAG enable of WB
https://bugs.libre-soc.org/show_bug.cgi?id=520
Luke Kenneth Casson Leighton [Thu, 22 Oct 2020 15:50:37 +0000 (16:50 +0100)]
add detection and disable of Instruction Wishbone based on JTAG command
Luke Kenneth Casson Leighton [Thu, 22 Oct 2020 15:49:11 +0000 (16:49 +0100)]
add detection and disable of LoadStore Wishbone based on JTAG command
Luke Kenneth Casson Leighton [Thu, 22 Oct 2020 15:47:23 +0000 (16:47 +0100)]
add JTAG enable/disable of wishbone to TestIssuer
Luke Kenneth Casson Leighton [Thu, 22 Oct 2020 13:13:11 +0000 (14:13 +0100)]
add means to JTAG interface to enable/disable "stuff" currently just WB
Cole Poirier [Wed, 21 Oct 2020 21:26:10 +0000 (14:26 -0700)]
versa_ecp5 adds ability to build and load for ulx3s85f, fixes testgpio
feature
Luke Kenneth Casson Leighton [Wed, 21 Oct 2020 16:47:15 +0000 (17:47 +0100)]
fix up asserts (check correct pads/cores)
https://bugs.libre-soc.org/show_bug.cgi?id=511#c12
Tobias Platen [Tue, 20 Oct 2020 17:42:50 +0000 (19:42 +0200)]
s/alu/fsm/g
Tobias Platen [Tue, 20 Oct 2020 16:41:39 +0000 (18:41 +0200)]
test case for FSMMMUStage
Cole Poirier [Sun, 18 Oct 2020 00:39:22 +0000 (17:39 -0700)]
use random.seed to generate repro cases of the two different failure
modes of test_icache()
Luke Kenneth Casson Leighton [Fri, 16 Oct 2020 19:51:09 +0000 (20:51 +0100)]
experiment swapping dummy trap stage over to input
Luke Kenneth Casson Leighton [Fri, 16 Oct 2020 18:29:16 +0000 (19:29 +0100)]
re-enable tests
Luke Kenneth Casson Leighton [Fri, 16 Oct 2020 18:28:36 +0000 (19:28 +0100)]
manually run coresync clock for test issuer
Luke Kenneth Casson Leighton [Fri, 16 Oct 2020 18:26:05 +0000 (19:26 +0100)]
set defaults in pspec
Luke Kenneth Casson Leighton [Fri, 16 Oct 2020 18:14:50 +0000 (19:14 +0100)]
update submodule
Luke Kenneth Casson Leighton [Fri, 16 Oct 2020 18:09:40 +0000 (19:09 +0100)]
add extra (test dummy stage in trap to see if combinatorial latency is reduced