From f13576b5d273ef9175e938b15f55bb1ead22fb1d Mon Sep 17 00:00:00 2001
From: Kartik Agaram <vc@akkartik.com>
Date: Mon, 5 Oct 2020 10:16:53 -0700
Subject: 6957

The final fix to the raytracing program involves rounding modes. It turns
out x86 processors round floats by default, unlike C which has trained
me to expect truncation. Rather than mess with the MXCSR register, I added
another instruction for truncation. Now milestone 3 emits perfectly correct
results.
---
 html/mu_instructions.html | 9 +++++++++
 1 file changed, 9 insertions(+)

(limited to 'html')

diff --git a/html/mu_instructions.html b/html/mu_instructions.html
index 8f21aa08..3f25242f 100644
--- a/html/mu_instructions.html
+++ b/html/mu_instructions.html
@@ -351,10 +351,19 @@ var/xreg <span class="SpecialChar">&lt;-</span> convert var2/reg2     =&gt; <spa
 var/xreg <span class="SpecialChar">&lt;-</span> convert var2          =&gt; <span class="Constant">&quot;f3 0f 2a/convert-to-float *(ebp+&quot;</span> var2.stack-offset <span class="Constant">&quot;) &quot;</span> xreg <span class="Constant">&quot;/x32&quot;</span>
 var/xreg <span class="SpecialChar">&lt;-</span> convert *var2/reg2    =&gt; <span class="Constant">&quot;f3 0f 2a/convert-to-float *&quot;</span> reg2 <span class="Constant">&quot; &quot;</span> xreg <span class="Constant">&quot;/x32&quot;</span>
 
+Converting floats to ints performs rounding by default. (We don't mess with the
+MXCSR control register.)
+
 var/reg <span class="SpecialChar">&lt;-</span> convert var2/xreg2     =&gt; <span class="Constant">&quot;f3 0f 2d/convert-to-int %&quot;</span> xreg2 <span class="Constant">&quot; &quot;</span> reg <span class="Constant">&quot;/r32&quot;</span>
 var/reg <span class="SpecialChar">&lt;-</span> convert var2           =&gt; <span class="Constant">&quot;f3 0f 2d/convert-to-int *(ebp+&quot;</span> var2.stack-offset <span class="Constant">&quot;) &quot;</span> reg <span class="Constant">&quot;/r32&quot;</span>
 var/reg <span class="SpecialChar">&lt;-</span> convert *var2/reg2     =&gt; <span class="Constant">&quot;f3 0f 2d/convert-to-int *&quot;</span> reg2 <span class="Constant">&quot; &quot;</span> reg <span class="Constant">&quot;/r32&quot;</span>
 
+There's a separate instruction for truncating the fractional part.
+
+var/reg <span class="SpecialChar">&lt;-</span> truncate var2/xreg2     =&gt; <span class="Constant">&quot;f3 0f 2c/truncate-to-int %&quot;</span> xreg2 <span class="Constant">&quot; &quot;</span> reg <span class="Constant">&quot;/r32&quot;</span>
+var/reg <span class="SpecialChar">&lt;-</span> truncate var2           =&gt; <span class="Constant">&quot;f3 0f 2c/truncate-to-int *(ebp+&quot;</span> var2.stack-offset <span class="Constant">&quot;) &quot;</span> reg <span class="Constant">&quot;/r32&quot;</span>
+var/reg <span class="SpecialChar">&lt;-</span> truncate *var2/reg2     =&gt; <span class="Constant">&quot;f3 0f 2c/truncate-to-int *&quot;</span> reg2 <span class="Constant">&quot; &quot;</span> reg <span class="Constant">&quot;/r32&quot;</span>
+
 There are no instructions accepting floating-point literals. To obtain integer
 literals in floating-point registers, copy them to general-purpose registers
 and then convert them to floating-point.
-- 
cgit 1.4.1-2-gfad0