commit dcc8ff5c9d69365c49d231013fd0024c0679cd9d
parent 7733817ef7d7744ed46cd0d784d0b4983aff41aa
Author: Shimmy Xu <shimmy.xu@shimmy1996.com>
Date: Sun, 7 Nov 2021 16:46:13 -0600
Add 1.3.1
Diffstat:
1 file changed, 28 insertions(+), 0 deletions(-)
diff --git a/README.org b/README.org
@@ -542,6 +542,34 @@ Try Ramanujan's treatment for \(Q(n)\)!
***** 19
Watson's lemma! o7
***** 20
+** INACTIVE 1.3'
+Reading the Fascicle 1 instead for 1.3.1 - 1.3.2. For 1.3.3 reading the original + program listing in the MMIX Supplement.
+*** 1.3.1'
+Note how in (6) the load behavior is defined: when the load width increases, we can include bytes before or after A. Also note the sign behavior.
+
+Why does it overflow in =STB/STW=? Because neither a byte (8 bits) or a wide (16 bits) can represent the value \(-65536 = -2^{16}\), while a tetra (32 bits) or a octa (64 bits) can in signed mode.
+
+Note the use of rH and rD in unsigned arithmetic.
+
+The bitwise operations are indeed bitwise! Bytewise operations feels like AVX.
+
+For floating point numbers, notice that the normal case includes an implicit \(1\cdot 2^{E-1023}\) (probably because 0 has special representation). Also note that there are actually \(2^{52} - 1\) possible representations for NaN (which seems to be used as a way to sneak in extra information).
+
+The \(\pi\), \(\mu\), and \(v\) denotes the time taken for each operation.
+
+**** 32
+**** 33
+**** 34
+**** 35
+**** 36
+**** 37
+**** 57
+
+** INACTIVE 1.4'
+Reading Fascicle 1 for 1.4.1 - 1.4.3.
+*** 1.4.2
+**** 21
+1.3.1
** TODO Chapter 2: Information structures
**** 2.3.4.4
***** INACTIVE 29