(no commit message)

diff --git a/openpower/sv/svp64/appendix.mdwn b/openpower/sv/svp64/appendix.mdwn
index 3d6a4ad0e58511f91f1317e284261c3e6628926f..1a5fd656b5a7fbcbbb0954f84dd3cffe41034a6d 100644 (file)
--- a/openpower/sv/svp64/appendix.mdwn
+++ b/openpower/sv/svp64/appendix.mdwn
@@ -108,11 +108,30 @@ followed by
 
 # LOAD/STORE Elwidths <a name="ldst"></a>
 
+Loads and Stores are almost unique in that the OpenPOWER Scalar ISA provides a width for the operation (lb, lh, lw, ld).  There are therefore three widths involved:
+
+* operation width (lb=8, lh=16, lw=32, ld=64)
+* source width override
+* destination element override
+
+The reason for all three is because Saturation (and other transformations) may occur in between, which rely on the source and destination width, and have nothing to do (per se) with the operation width.
+
+Below is the pseudocode for Unit-Strided LD.  Note the following:
+
+* `scalar identity behaviour` SV Context parameter conditions turn this
+  into a straight absolute fully-compliant Scalar v3.0B LD operation
+* brev selects whether the operation is the byte-reversed variant (ldbrx
+  rather than ld)
+* op_width specifies the operation width (lb, lh, lw, ld)
+* imm_offs specifies the immediate offset `ld r3, imm_offs(r5)`
+* svctx specifies the SV Context and includes VL as well as source and
+  destination elwidth overrides.
+
 
     # LD not VLD! (ldbrx if brev=True)
     # this covers unit stride mode
     function op_ld(rd, rs, brev, op_width, imm_offs, svctx)
-      for (int i = 0, int j = 0; i < VL && j < VL;):
+      for (int i = 0, int j = 0; i < svctx.VL && j < svctx.VL;):
 
         # unit stride mode, compute the address
         srcbase = ireg[rs] + i * op_width;
@@ -127,10 +146,13 @@ followed by
         if (bytereverse):
             memread = byteswap(memread, op_width)
 
-        # now truncate to source over-ridden width
+        # now truncate to source over-ridden width.
         if (svctx.src_elwidth != default)
             memread = adjust_wid(memread, op_width, svctx.src_elwidth)
 
+        # note that here we would now do saturation if it was enabled.
+        ... saturation adjustment...
+
         # takes care of inserting memory-read (now correctly byteswapped)
         # into regfile underlying LE-defined order, into the right place
         # within the NEON-like register, respecting destination element