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.

1 files changed, 9 insertions, 0 deletions

diff --git a/mu_instructions b/mu_instructions
index 9e7f26bc..f93bb685 100644
--- a/mu_instructions
+++ b/mu_instructions
@@ -327,10 +327,19 @@ var/xreg <- convert var2/reg2     => "f3 0f 2a/convert-to-float %" reg2 " " xreg
 var/xreg <- convert var2          => "f3 0f 2a/convert-to-float *(ebp+" var2.stack-offset ") " xreg "/x32"
 var/xreg <- convert *var2/reg2    => "f3 0f 2a/convert-to-float *" reg2 " " xreg "/x32"
 
+Converting floats to ints performs rounding by default. (We don't mess with the
+MXCSR control register.)
+
 var/reg <- convert var2/xreg2     => "f3 0f 2d/convert-to-int %" xreg2 " " reg "/r32"
 var/reg <- convert var2           => "f3 0f 2d/convert-to-int *(ebp+" var2.stack-offset ") " reg "/r32"
 var/reg <- convert *var2/reg2     => "f3 0f 2d/convert-to-int *" reg2 " " reg "/r32"
 
+There's a separate instruction for truncating the fractional part.
+
+var/reg <- truncate var2/xreg2     => "f3 0f 2c/truncate-to-int %" xreg2 " " reg "/r32"
+var/reg <- truncate var2           => "f3 0f 2c/truncate-to-int *(ebp+" var2.stack-offset ") " reg "/r32"
+var/reg <- truncate *var2/reg2     => "f3 0f 2c/truncate-to-int *" reg2 " " reg "/r32"
+
 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.