Fix description of the half precision math functions (#860)

* Fix description of the half precision math functions

Likewise in already merged KhronosGroup/SYCL-Docs#306

* Fix of error value (by @bashbaug)

From discussion in #860

Author: Melirius

OpenCL_C.txt

@@ -4926,7 +4926,7 @@ The following table describes the following functions:
   * A subset of functions from <<table-builtin-math>> that are defined with
     the half_ prefix .
     These functions are implemented with a minimum of 10-bits of accuracy,
-    i.e. an ULP value \<= 8192 ulp.
+    i.e. the maximum error value \<= 8192 ulp.
   * A subset of functions from <<table-builtin-math>> that are defined with
     the native_ prefix.
     These functions may map to one or more native device instructions and

cxx/stdlib/math.txt

@@ -986,7 +986,7 @@ The maximum error is implementation-defined.
 This section describes the following functions:
 
   * A subset of functions from previous sections that are defined in the `cl::half_math` namespace.
-    These functions are implemented with a minimum of 10-bits of accuracy i.e. an ULP value \<= 8192 ulp.
+    These functions are implemented with a minimum of 10-bits of accuracy i.e. the maximum error \<= 8192 ulp.
   * half functions for following basic operations: divide and reciprocal.
   * Support for denormal values is optional for half_math:: functions. The `half_math::` functions may return any result allowed by the <<edge-case-behavior-in-flush-to-zero-mode, _Edge Case Behavior in Flush To Zero Mode_>> section, even when `-cl-denorms-are-zero` is not in force.