This refactors the invalidation control logic into an explicit state
machine. The top-level icache_invalid_o signal is also removed.
Replaced with an explicit scramble key request instead.
This has all been done to better deal with corner cases around a new
invalidation being requested whilst another is still going on.
Previously there was a bug wher an invalidation request in the final
cycle of an ongoing invalidation didn't restart the invalidation but did
rotate the scrambling key producing an ECC failure and an alert.
This allows us to actually use Icache in our tests beecause
before this commit key_valid was tied to 0 which means everytime
we receive a FENCE.I instruction, we wouldn't be able to successfully
flag inval_done. Which means we weren't probably using ICache correctly.
Signed-off-by: Canberk Topal <ctopal@lowrisc.org>
This commit adds ibex_icache_ram_if to connect between DUT and tag /
data RAMs.
This interface injects 1 or 2 bit error on rdata if enable_ecc_errors
bit is set. It also checks that ecc_err_o pin is asserted by DUT
whenever an ecc error is injected.
ibex_icache_ecc_vseq and ibex_icache_base_vseq have been modified to
inject ecc errors through the ram interface.
This commit removes extra hierarchy of ic_top inside icache TB and moves
the scrambling request generation logic and instantiation of data and
tag RAMs to tb.
This commit adds a new scrambling agent to drive scrambling key and
valid to the data and tag memory interfaces.
Update lowrisc_ip to lowRISC/opentitan@7c4f8b3fd
Update code from upstream repository
https://github.com/lowRISC/opentitan to revision
7c4f8b3fde4bb625ac3330ff52d3f66507190fe5
Signed-off-by: Prajwala Puttappa <prajwalaputtappa@lowrisc.org>
Earlier the design supported single clock cycle error responses from PMP
block whenever a read was done from blocked memory. Now there is at
least one clock cycle delay after the request has been granted for the
error to be asserted. Therefore, this commit removes the support for
single clock cycle PMP error response.
Update code from upstream repository
https://github.com/lowRISC/opentitan to revision
7c4f8b3fde4bb625ac3330ff52d3f66507190fe5
Please note that we're adding push_pull_agent for the first time in this
commit.
Signed-off-by: Prajwala Puttappa <prajwalaputtappa@lowrisc.org>
There was issue with rtespect to calculating number of instructions per
word and this commit fixes that issue.
Number of instructions per word = 1/4*1 + 3/4(1/4*3/2 + 3/4*2) = 53/32.
Earlier th5s was calculated as 7/4.
Ideal window length needed to calculate fetch ratio percentage is
calculated as 53/32*C*2 = 848. Earlier it was calculated to be 300.
This commit includes switching to a scrambling RAM primitive for
ICache data and tag RAMs. Also introduces minor changes to ICache
to handle scrambling key valid signal.
It also includes a minor bug fix regarding not initializing
`fill_way_q` signal without ResetAll parameter. When the parameter
is not set and we have our first hit right after ICache enables,
the signal hangs.
Signed-off-by: Canberk Topal <ctopal@lowrisc.org>
- Instruction addresses are now checked in the IF stage, after the cache
and after the prefetch buffer.
- To deal with unaligned instructions, the PMP logic checks the current
address and the next in parallel.
- The spec_branch timing hack has been removed as it's no longer
relevant with the PMP logic moved.
- Various updates made to the icache testbench to account for the
changes.
- Relates to #1471
Signed-off-by: Tom Roberts <tomroberts@lowrisc.org>
This commit creates a new top level wrapping the core, register file and
icache RAMs. The tracing top level is also renamed to ibex_top_tracing
to match. This new top level is intended to enable a dual core lockstep
implementation of Ibex.
There are no functional changes in this commit, only wiring.
Signed-off-by: Tom Roberts <tomroberts@lowrisc.org>
No functional change. These parameters are effectively fixed. Moving
them to the pkg eases top-level wiring of RAM signals.
Signed-off-by: Tom Roberts <tomroberts@lowrisc.org>
Update code from upstream repository
https://github.com/lowRISC/opentitan to revision
f29a0f7a7115e03fba734b1c00691c253aceb07e. The list of OpenTitan
changes that are merged in appears at the bottom of the commit.
There are some manual changes needed to adapt the code to work with
these changes.
- The ICache monitors need some extra types to adapt to the (rather
odd) data model that the OpenTitan dv_lib code now uses, where a
monitor needs to know an agent's associated sequence type.
- Verilator simulations now use MemArea slightly differently
OpenTitan changes:
* [dv] Allow monitor items to have different types from sequence items
(Rupert Swarbrick)
* [dvsim] Fix primary_cfg handling (Srikrishna Iyer)
* [dvsim] Deal with non unicode chars in log files (Srikrishna Iyer)
* [dvsim] Added common build fail patterns (Srikrishna Iyer)
* [dvsim] Minot cleanup to the lint flow (Srikrishna Iyer)
* [dvsim] Minor cleanups to to formal flow (Srikrishna Iyer)
* [dvsim] Fixes to UNR and cov analysis flows (Srikrishna Iyer)
* [dvsim] Very minor cleanup of Deploy class (Srikrishna Iyer)
* [dvsim] LsfLauncher report early errors as F (Srikrishna Iyer)
* [dvsim] Minor fix in clean_odirs function (Srikrishna Iyer)
* [chip dv] Set +sw_images as comma-separated list (Srikrishna Iyer)
* [flash_ctrl] Split tl intefaces for flash_ctrl and prim_flash_cfg
(Timothy Chen)
* [keymgr] Fix input value checks (Timothy Chen)
* [formal/script] Update generic formal flow naming from `fpv` to
`formal` (Cindy Chen)
* [top, prim] Address wmask and data width mismatch issue (Timothy
Chen)
* [dvsim] Add GUI mode for running simulations (Srikrishna Iyer)
* [dv] Fix reg backdoor (Weicai Yang)
* [dpi] Make an "ECC32" flavour of MemArea (Rupert Swarbrick)
* [uvmdvgen] Fix has_interrupts in env_cfg (Cindy Chen)
* [dvsim] Keep dependencies list (Srikrishna Iyer)
* [prim_prince] Reverse the k0||k1 mapping to match with the paper
(Michael Schaffner)
* [dvsim] Fix printing of last 10 lines (Srikrishna Iyer)
* [primgen] Minor fix to enable types with underscores (Michael
Schaffner)
* [dvsim] Prevent command echo suppression (Srikrishna Iyer)
* [dvsim] Spot fixes for LSF and internal launcher (Srikrishna Iyer)
* [sva] csr assertion dependency update (Cindy Chen)
* [memutil] Change DpiMemUtil so that it no longer owns MemAreas
(Rupert Swarbrick)
* [memutil] Factor out MemArea as a class (Rupert Swarbrick)
* [prim] Split out PRESENT and PRINCE support from prim:all (Rupert
Swarbrick)
* [fpv/otp_ctrl] Disable assertions due to lc_esc_en (Cindy Chen)
* [prim_prince] Annotate some arrays to avoid UNOPTFLAT warnings
(Rupert Swarbrick)
* [top] Hook up latest ast ports and complete a few other integration
(Timothy Chen)
* Eliminate `#pragma once` in favor of include guards (Chris Frantz)
* [sw,dv] Update headers to pass fix_include_guards.py (Alex Bradbury)
* [xbar/dv] Fix assertion error due to short reset (Weicai Yang)
* [sram] Add memory initialization (Timothy Chen)
* [uvmdvgen] Update links in checklist template (Philipp Wagner)
* [dv/uvmdvgen] Add comment for testplan (Cindy Chen)
* [dv/top_earlgrey] chip csr_aliasing timeout (Cindy Chen)
* [dvsim] Cosmetic updates to launcher methods (Srikrishna Iyer)
* [dv] Update csr_wr to support field write (Weicai Yang)
* [dv/common] Fix regression warnings (Cindy Chen)
* [dv] Get blocks with multiple device interfaces working with chip DV
(Rupert Swarbrick)
* [doc] Use relative links in Hjson-related shortcodes (Philipp
Wagner)
* [dvsim] minor enhancement to clean_odir (Srikrishna Iyer)
* [dvsim] Statically display jobs' status (Srikrishna Iyer)
* [dvsim] Do weighted scheduling of jobs (Srikrishna Iyer)
* [dvsim] Schedule jobs by dependency (Srikrishna Iyer)
* [dv] Xcelium UNR typo (Srikrishna Iyer)
* [dvsim] Implement LsfLauncher (Srikrishna Iyer)
* [dv/chip] solve same_csr_outstanding_timeout (Cindy Chen)
* [dv] make dv_base_agent work for high-level agent (Weicai Yang)
* [tools/dv] added UNR flow for xcelium (Rasmus Madsen)
* [prim] Split prim:subreg out of prim:all (Rupert Swarbrick)
* [prim] Split prim_alert_* out of prim:all (Rupert Swarbrick)
* [prim] Split out fifos into a prim_fifo core (Rupert Swarbrick)
* [prim] Split out arbiters into a prim_arbiter core (Rupert
Swarbrick)
* [prim] Make prim:flop_2sync depend on prim:flop (Rupert Swarbrick)
Signed-off-by: Rupert Swarbrick <rswarbrick@lowrisc.org>
These changes correspond to similar changes in the prefetch buffer to
support branch prediction. A registered version of fill_ext_done was
required to prevent a combinational loop from branch_i in to valid_o
out.
Multiplexing priorities for fifo_addr have been swapped to match
fetch_addr_d in the same module and all similar multiplexing in the
icache (prioritize incoming branch_i over branch_mispredict_i). Note
however that it is not expected that these conditions will actually
occur together, and an assertion has been added to check that.
Signed-off-by: Tom Roberts <tomroberts@lowrisc.org>
Amusingly, there was also actually a bug in the combo sequence logic:
it took seq_idx to be the number of sequences that we'd run so far,
but it is actually an index that chooses which sequence we're about to
run.
dvsim.py runs make as a subprocess, which gets rather confused if
MAKEFLAGS appears in its environment. The proper fix is to clear them
from the environment in the dvsim command: we'll do that on the
OpenTitan side[1] and can revert this patch once that change is vendored
in.
[1] https://github.com/lowRISC/opentitan/pull/4325
This commit amends some paths in the vendoring hjson file (and updates
config files to use things at the new paths). Finally it re-runs the
vendoring tool:
Update code from upstream repository
https://github.com/lowRISC/opentitan to revision
92e9242424c72c59008e267dd3779e2af5ec8e83
which just ends up with a load of file renames.
Signed-off-by: Rupert Swarbrick <rswarbrick@lowrisc.org>
Some sequences run in "no invalidate" mode. Unfortunately, sometimes
we *have* to invalidate, either because we have a stale seed, or
because we're at the start of a sequence.
It turns out that "no invalidate" is a soft requirement, so we can
weaken it here. This patch also renames the flag to
"avoid_invalidation" to make it clear that this is allowed to be
violated occasionally.
Closes#1101.
The shortlog from the vendor tool's automated patch is reproduced at
the bottom of this commit message.
The automated commit is squashed with one to update how we depend on
bus parameters. Before, we had to provide an "Ibex top package". This
behaved the same as OpenTitan's "lowrisc:constants:top_pkg", but
avoided having to vendor in that file.
On the OpenTitan side, this has been tidied up with commit
d266c68 ("[dv] Update dv_utils sources to use bus_params_pkg"). This
changes the dependency of dv_utils to
"lowrisc:opentitan:bus_params_pkg". We still have to provide our
own (now called "lowrisc:ibex:bus_params_pkg") and need to patch the
dv_utils dependency, but this is a bit cleaner because dv_utils is
less likely to accidentally include dependencies on OpenTitan
internals.
On our side, we have to update the vendoring patch for dv_utils (and
change its name). We also need an equivalent patch for dv_lib. Then we
rename our hacky "Ibex top package" to "bus_params_pkg". The ICache DV
environment also needs patching to use the bus parameters properly.
Phew!
* [dv] Update prim_present cov opt (Srikrishna Iyer)
* [dv] Align VCS and Xcelium cov var names (Srikrishna Iyer)
* [dv] Split coverage for functional and auto tests (Srikrishna Iyer)
* [dvsim] Do builds smartly (Srikrishna Iyer)
* [syn] Carry over synthesis flow updates from bronze (Michael
Schaffner)
* [dvsim] Lint cleanup (Srikrishna Iyer)
* [dvsim] Allow testplan to be omitted (Srikrishna Iyer)
* [dvsim] Address lowRISC/opentitan#3071 comments (Srikrishna Iyer)
* [dvsim] lint cleanup (Srikrishna Iyer)
* [dvsim] Add support for second-level indirection (Srikrishna Iyer)
* [dvsim] Change cores-root to avoid conflicts with autogen'd core
files (Michael Schaffner)
* [dvsim] Update `tests` key behavior in regressions (Srikrishna Iyer)
* [lint] Minor update of ERROR patterns in parser script (Michael
Schaffner)
* [packer] Revise the implementation (Eunchan Kim)
* [flow] Remove lint makefile (Timothy Chen)
* [flows] Various updates to tools and documents to suppose top/ip
select (Timothy Chen)
* [dv/shadow_reg] shadow_reg update error (Cindy Chen)
* [rtl/alert] change the naming from _en_i to _req_i (Cindy Chen)
* [dvsim] Tidy up config file loading in FlowCfg.py (Rupert Swarbrick)
* [dvsim] Make it simpler to derive from FlowCfg (Rupert Swarbrick)
* [lint] Update warning/error exclusions in parser scripts (Michael
Schaffner)
* [dvsim] Fix for `--tool` override (Srikrishna Iyer)
* [dvsim] Bug fix in LintCfg.py (Srikrishna Iyer)
* [prim/dv] Integrate LFSR TB with dvsim (Udi Jonnalagadda)
* [uvmdvgen] Update template to reflect bind reorg (Srikrishna Iyer)
* [dv] remove prim_lfsr_bind (Srikrishna Iyer)
* [dv] Cleanup lint warnings in csr_utils_pkg (Srikrishna Iyer)
* [dv] Cleanup lint warnings in clk_rst_if (Srikrishna Iyer)
* [dvsim] Fix coverage dashboard link (Srikrishna Iyer)
* [prim] Rename prim_util_memload.sv to svh (Philipp Wagner)
* [lint/doc] Update linting readme to reflect recent updates (Michael
Schaffner)
* [lint] Remove legacy Makefile flow for linting tools (Michael
Schaffner)
* [dvsim/lint] Enable Verilator lint in Dvsim (Michael Schaffner)
* [prim_arbiter_fixed/fpv] Add generated FPV testbench (Michael
Schaffner)
* [prim_arbiter_fixed] This adds a fixed priority arbiter (Michael
Schaffner)
* [prim] Domain-Oriented Masking AND logic (Eunchan Kim)
* [dv] Update dv_utils sources to use bus_params_pkg (Srikrishna Iyer)
* [dv] Update mem_model to use bus-params_pkg (Srikrishna Iyer)
* [dv] Update dv_lib sources to use bus_params_pkg (Srikrishna Iyer)
* [uvmdvgen] Support for setting vendor name in VLNV (Srikrishna Iyer)
This fixes a test failure that I was seeing when following a "many
errors" test by something different. To reproduce,
make -C dv/uvm/icache/dv \
SEED=1465832714 \
TESTS=ibex_icache_stress_all_with_reset
There are actually two different ways this can come unstuck:
(1) Memory request goes out and gets put into the response queue.
req_i goes low. Sequence changes. req_i goes high and we get the
response from the previous request (but mem_err_shift has changed
in the meantime).
To fix this, we pair up the memory seed and its associated
mem_err_shift in the scoreboard queue, rather than retrieving
mem_err_shift from the config object when the response comes in.
(2) Memory request goes out. Sequence changes. Memory request is
handled (with new mem_err_shift). Scoreboard sees the result. New
sequence generates its first item.
In this case, the scoreboard will expect the old mem_err_shift and
see the new one. To fix this, we add an extra entry to the list of
valid states in the scoreboard if needed so that we also check the
mem_err_shift currently in the config object.
You might worry about what happens if we have two back-to-back
sequence changes that change mem_err_shift without ever changing seed:
what happens if we have a situation like (1), but for the "middle"
sequence. To avoid this problem, we actually add the extra entry in
the fix for (2), so it will look like a new seed arrived as part of
the middle sequence, so long as we have read at least one
result (always true in the core sequence).
This now allows you to specify how many seeds to run. Sadly, you can't
say "give me 10 seeds, starting at 1234" because dvsim doesn't support
that at the moment. But it does at least avoid quite such long command
lines.
Instead of:
../../../../vendor/lowrisc_ip/dvsim/dvsim.py \
ibex_icache_sim_cfg.hjson \
--scratch-root ../../../../build \
--reseed 5 \
-c
you can run:
make RESEED=5 COVERAGE=1
The idea is that this can supply top_pkg.sv, a top-level thing in
OpenTitan, for DV code we vendor from there. It's probably better to
do this than to directly vendor in OpenTitan's top_pkg, because the
latter has information about e.g. flash memory layout, which doesn't
really have any meaning for Ibex.
This is like the stress_all test, picking other sequences at random
and running them back-to-back. The difference is in the reset
behaviour, where we randomly pull the reset line at unexpected times
to try to trigger any strange glitches this might cause.
This requires slight changes to the core and memory drivers, which
need to learn to stop and return early from the current item when they
see a reset.
When we chain sequences together, we are careful to pass seeds between
neighbouring sequences. However, I didn't think to check
mem_err_shift. Before this patch, you see problems if you have a
"caching" sequence followed by a "many_errors" sequence with no reset
and no change of seed and they both happen to pick the same address
range.
The problem is that if the data at address A is cached in the first
sequence, the icache will merrily return it when address A comes up in
the second. However, the change to mem_err_shift might mean that this
would cause a memory error if it hadn't been cached, causing the
scoreboard to get upset.
This patch ensures that we always start a sequence with an
invalidation if there was a previous sequence with a different value
of mem_err_shift.
To do this cleanly, the patch also moves some of the "grab the guts of
the old sequence and put it in the new one" logic from
ibex_icache_combo_vseq and into the underlying sequence classes. The
trick is that a sequence now has a handle to the previous sequence (if
there was one), and can use that to extract whatever information it
needs.
This fixes several problems. Firstly, the window_reset function was
switching off tracking until it next saw busy_o go low, which is
correct at the start of time, but not what we want after we've
started. This patch splits that behaviour into a new tracking_reset
function (which calls window_reset). This is called on reset or
invalidate.
Secondly, this check was occasionally failing where we'd have an ECC
sequence (which should disable the check) immediately followed by a
caching sequence with similar addresses. If the window ended in the
caching sequence, we'd see a high fetch ratio and conclude that
something had gone wrong.
Now we clear the window completely whenever we fetch an instruction
when the check is disabled, which should avoid the problem (at worst,
you might get 1 instruction overlap, which is unlikely to matter).
Finally, we move the call to tracking_reset up to the end of the reset
sequence. It doesn't usually matter, but if there's a pending item
from the core monitor with busy = 0, we need to make sure that item
comes in before we set not_invalidating = 1. Otherwise, the scoreboard
incorrectly thinks it's seen the end of the invalidation
sequence (before it's even started) and starts tracking fetch ratios
too early.
This runs sequences back-to-back, occasionally resetting between
sequences.
Because our virtual sequences are composed of several smaller
sequences, we have to stop them when the core sequence finishes (see
the calls to kill() in ibex_icache_base_vseq). We also have to make
sure that we don't drop items in the memory sequence, which can be
pre-empted as part of sending a response (see the peek/get code
there).
Finally, the memory sequence also has a current seed and a list of
pending grants: this patch has to copy those across between sequences
to make everything work correctly.
