[doc] Update OTBN style guide.
Add some advice on inline data, and update the variable-time example to
show good practice with regard to the OTBN loop stack.
Signed-off-by: Jade Philipoom <jadep@google.com>
diff --git a/doc/rm/otbn_style_guide.md b/doc/rm/otbn_style_guide.md
index a137b7c..bcdb807 100644
--- a/doc/rm/otbn_style_guide.md
+++ b/doc/rm/otbn_style_guide.md
@@ -133,6 +133,17 @@
ret
```
+### Alignment Directives
+
+Big-number load/store instructions (`bn.lid`, `bn.sid`) require DMEM addresses
+to be 256-bit aligned; use `.balign 32` for any big-number inline binary data
+to ensure this. For non-big-number data, `.balign 4` (32-bit alignment) is
+recommended.
+
+### Inline Binary Directives
+
+Prefer `.word` over alternatives such as `.quad`.
+
## Secure Coding for Cryptography
The following guidelines address cryptography-specific concerns for OTBN assembly.
@@ -216,18 +227,20 @@
li x12, 28
/* Check if all limbs are equal. */
+ li x2, 1
loopi 8, 5
+ /* If x2 is 0, skip to the end of the loop. (It's still necessary to
+ complete the same number of loop iterations to avoid polluting the loop
+ stack, but we can skip all instructions except the last.) */
+ beq x2, x0, loop_end
/* w27 <= A[i] */
bn.movr x11, x9++
/* w28 <= B[i] */
bn.movr x12, x10++
/* x2 <= (w27 == w28) */
jal x1, eq_256
- /* If x2 is 0, exit loop early. */
- beq x2, x0, eq_2048_var_done
+loop_end:
nop
-eq_2048_var_done:
-
ret
```
