(no commit message)

diff --git a/meetings/sync_up/sync_up_2024-05-07.mdwn b/meetings/sync_up/sync_up_2024-05-07.mdwn
new file mode 100644 (file)
index 0000000..667e0f1
--- /dev/null
+++ b/meetings/sync_up/sync_up_2024-05-07.mdwn
@@ -0,0 +1,33 @@
+# Tuesday 7th May 16:00 UTC
+
+* Previous notes: [[/meetings/sync_up/sync_up_2024-04-30]]
+* Next week's notes: [[/meetings/sync_up/sync_up_2024-05-14]]
+
+# Main Agenda
+
+Meeting notes:
+
+# Cesar
+
+Attended. Did a bit of work on pin definitions for Nexys Video ([bug #1004](https://bugs.libre-soc.org/show_bug.cgi?id=1004)). Added I2C and DisplayPort.
+Updated pull nmigen-boards pull request
+([link](https://gitlab.com/nmigen/nmigen-boards/-/merge_requests/6)).
+See the current state of the work
+[here](https://gitlab.com/nmigen/nmigen-boards/-/blob/3819bf24e4710d9696cca0f78bc89e8775b79b25/nmigen_boards/nexys_video.py).
+Next: SDRAM, PMODs, Ethernet, QSPI Flash, SD card, etc.
+
+# Jacob
+
+Absent. No updates.
+
+# Luke
+
+Attended.
+
+# Dmitry
+
+Absent.
+
+# Rita
+
+Absent.

diff --git a/meetings/sync_up/sync_up_2024-05-14.mdwn b/meetings/sync_up/sync_up_2024-05-14.mdwn
new file mode 100644 (file)
index 0000000..01d1c06
--- /dev/null
+++ b/meetings/sync_up/sync_up_2024-05-14.mdwn
@@ -0,0 +1,32 @@
+# Tuesday 14th May 16:00 UTC
+
+* Previous notes: [[/meetings/sync_up/sync_up_2024-05-07]]
+* Next week's notes: [[/meetings/sync_up/sync_up_2024-05-21]]
+
+# Main Agenda
+
+Meeting notes:
+
+# Cesar
+
+Attended. 
+
+Figured out criteria and bug list for Jacob to do RFPs.
+See thread starting at
+[[https://lists.libre-soc.org/pipermail/libre-soc-dev/2024-May/006209.html]].
+
+# Jacob
+
+Absent. Worked on getting ready for and submitting an RFP.
+
+# Luke
+
+Absent.
+
+# Dmitry
+
+Absent.
+
+# Rita
+
+Present. Had research interviews with people and with Luke.

diff --git a/meetings/sync_up/sync_up_2024-05-21.mdwn b/meetings/sync_up/sync_up_2024-05-21.mdwn
new file mode 100644 (file)
index 0000000..d3c9a6c
--- /dev/null
+++ b/meetings/sync_up/sync_up_2024-05-21.mdwn
@@ -0,0 +1,29 @@
+# Tuesday 21th May 16:00 UTC
+
+* Previous notes: [[/meetings/sync_up/sync_up_2024-05-14]]
+* Next week's notes: [[/meetings/sync_up/sync_up_2024-05-28]]
+
+# Main Agenda
+
+Meeting notes:
+
+# Cesar
+
+Attended.
+Worked on pin definitions for Nexys Video ([bug #1004](https://bugs.libre-soc.org/show_bug.cgi?id=1004)). Added audio codec and Ethernet.
+
+# Jacob
+
+Absent. Planning on submitting the RFP for the gfb* instructions Tuesday night.
+
+# Luke
+
+Absent.
+
+# Dmitry
+
+Absent.
+
+# Rita
+
+Present. Working on my data and article.

diff --git a/openpower/sv/cookbook/ed25519.mdwn b/openpower/sv/cookbook/ed25519.mdwn
new file mode 100644 (file)
index 0000000..1301258
--- /dev/null
+++ b/openpower/sv/cookbook/ed25519.mdwn
@@ -0,0 +1,60 @@
+# Elliptic Curve ed25519
+
+Links:
+
+* bugrepoet: <https://bugs.libre-soc.org/show_bug.cgi?id=1166>
+* [ed25519_mul.py](https://git.libre-soc.org/?p=openpower-isa.git;a=blob_plain;f=src/openpower/decoder/isa/ed25519/curve25519_mul.py;hb=de10b86f7b3)
+* [donna-ed25519](https://bugs.libre-soc.org/show_bug.cgi?id=773#c1)
+* [Triangular REMAP discussion](https://lists.libre-soc.org/pipermail/libre-soc-dev/2022-May/004881.html)
+
+ed25519 is strategically important as its implementation was highly optimised
+during its design, for high security.
+[Edwards-curve Digital Signature Algorithm (EdDSA)](https://en.wikipedia.org/wiki/EdDSA#Ed25519)
+was also designed to be fast.
+
+In the donna-ed25519 implementation, key functions such as ed25519_mul
+are laid out explicitly by loop-unrolling:
+
+```
+t[0]  =  r0 * s0
+t[1]  =  r0 * s1 + r1 * s0;
+t[2]  =  r0 * s2 + r1 * s1 + r2 * s0;
+t[3]  =  r0 * s3 + r1 * s2 + r2 * s1 + r3 * s0;
+t[4]  =  r0 * s4 + r1 * s3 + r2 * s2 + r3 * s1 + r4 * s0;
+```
+
+Note the very obvious patterns here which are triangular in nature.
+With the very existence of Simple-V's REMAP subsystem it is quite
+natural to see if triangular remapping can be added and used.
+It turns out to be quite easy, and there are two possible techniques:
+Vertical-First and Horizontal-First.
+
+With Vertical-First, the multiply is done first as a scalar item,
+into a temporary register, followed by an addition of the scalar
+into the actual target (t0 thru t4)
+
+```
+sv.mul temp, *r, *s   # temporary target scalar register
+sv.add *t,*t,temp     # add temporary scalar onto target vector
+```
+
+With Horizontal-First it is extremely simple: use `madd` -
+integer multiply-and-accumulate:
+
+```
+sv.madd *t, *r, *s
+```
+
+In both cases, all three target registers are set up with the
+same REMAP Schedules.  Additionally in both cases, t0-t4 must
+be pre-initialised to zeros.
+
+As always with Simple-V, the power of simplicity comes primarily
+from the REMAP subsystem. However in a secure environment,
+reduced instruction count is also critical not just for power
+consumption but to get the size of the binary down small enough
+that it could fit easily into a few lines of L1 Cache.
+If a huge number of loop-unrolled instructions (the normal
+way of handing these algorithms) are reduced down to a
+bare handful, with the looping covered in hardware, then it
+is easy to understand how valuable Simple-V and REMAP is.