From: lkcl <lkcl@web>
Date: Thu, 28 Oct 2021 16:55:56 +0000 (+0100)
Subject: (no commit message)
X-Git-Tag: opf_rfc_ls005_v1~3499
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=7aa542b5bcebd34169ac5f53ac77b3240315efbf;p=libreriscv.git

---

diff --git a/3d_gpu/architecture/dynamic_simd/shape.mdwn b/3d_gpu/architecture/dynamic_simd/shape.mdwn
index 57d514c4e..6817c65ae 100644
--- a/3d_gpu/architecture/dynamic_simd/shape.mdwn
+++ b/3d_gpu/architecture/dynamic_simd/shape.mdwn
@@ -420,6 +420,20 @@ Divide on the other hand (and right-shift) will only work (when
 both elwidth and fixed-width are set) if there is
 no rounding (no bits lost due to the division / right-shift).
 
+Summary:
+
+* Fixed_width Priority
+  - all operations (add, mul, sub, shift, div)
+    expected to work (caveat: layout() should check rounding)
+* Element-width Priority
+  - all operations expected to work (no caveats)
+* Fixed-and-Elwidth (no priority)
+  - multiply and left-shift always expected to work
+  - divide and right-shift expected to work (caveat: no bits
+    lost due to rounding)
+  - add and subtract **not** expected to work (ambiguous):
+    exception must be raised
+
 ## Arithmetic of two SimdShapes
 
 With some thought it becomes clear that when performing operations