Expand intro and add request for PR

Author: danielwe

docs/src/index.md

@@ -20,7 +20,7 @@ Pages = [
     "examples/special_functions.md",
     "examples/sum_of_squared_error.md",
     "vectorized_convenience_functions.md",
-    "performance_accuracy_fastmath.md",
+    "performance_accuracy.md",
     "future_work.md",
     "devdocs/overview.md",
     "devdocs/loopset_structure.md",

docs/src/performance_accuracy_fastmath.md renamed to docs/src/performance_accuracy.md

@@ -1,6 +1,6 @@
-## Performance vs. accuracy and comparison with `@fastmath`
+## Performance-accuracy tradeoff
 
-_From the Julia slack #performance-helpdesk 2022-03-30 03:06 UTC_
+_`@turbo` performs arithmetic transformations and substitutes standard special functions with implementations that lend themselves better to vectorization but sometimes achieve slightly lower accuracy. The following is a discussion from the Julia slack #performance-helpdesk on 2022-03-30, regarding this tradeoff, how it compares to similar transformations by `@fastmath`, and how to preserve full accuracy while vectorizing. It is included here for the benefit of anyone looking for information about this. PRs welcome to turn this page into a self-contained discussion or FAQ section._
 
 **Daniel Wennberg**