From: lkcl <lkcl@web>
Date: Sun, 25 Sep 2022 20:42:05 +0000 (+0100)
Subject: (no commit message)
X-Git-Tag: opf_rfc_ls005_v1~281
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=a6bb0222cf6da7f806f5b05d761ab84b588a1bd4;p=libreriscv.git

---

diff --git a/openpower/sv/biginteger/analysis.mdwn b/openpower/sv/biginteger/analysis.mdwn
index 3d8bdbe05..5e03e1f2b 100644
--- a/openpower/sv/biginteger/analysis.mdwn
+++ b/openpower/sv/biginteger/analysis.mdwn
@@ -156,6 +156,21 @@ carry-propagation through multiple elements does not occur.
 With Scalar shift and rotate operations in the Power ISA already being
 complex and very comprehensive, it is hard to justify creating complex
 3-in 2-out variants when a sequence of 3 simple instructions will suffice.
+However it is reasonably justifiable to have a 3-in 1-out instruction
+with an implicit source, based around the inner operation:
+
+```
+    # r[i] = (un[i] >> s) | (un[i + 1] << (64 - s));
+    t <- ROT128(RA || RA1, RB[58:63])
+    RT <- t[64:127]
+``
+
+RA1 is implicitly (or explicitly, RC) greater than RA by one
+scalar register number, and like the other operations below,
+a 128/64 shift is performed, truncating to take the lower
+64 bits.  By taking a Vector source RA and assuming lower-numbered
+registers are lower-significant digits in the biginteger operation
+the entire biginteger source may be shifted by a scalar.
 
 For larger shift amounts beyond an element bitwidth standard register move
 operations may be used, or, if the shift amount is static,