Differences

This shows you the differences between two versions of the page.

--- tangle [2026/05/03 16:45] – created - external edit 127.0.0.1
+++ tangle [2026/05/05 22:26] (current) – admin
@@ Line 26: / Line 26: @@
 <code>
-g<del> -O3 -std=c</del>17 -static -o tangle tangle.cpp float256.cpp -lm
+g++ -O3 -std=c++17 -static -o tangle tangle.cpp float256.cpp -lm
 </code>
@@ Line 116: / Line 116: @@
 Two custom fixed-point types provide the extended precision, both implemented as one IEEE double (carrying the sign, exponent, and top 52 mantissa bits) plus additional unsigned 64-bit integers ("uint64_t") words that extend the mantissa:
-- **float128** — double + 1×uint64_t, giving ~117 bits of mantissa   precision. Used for orient2d, where the 2×2 determinant structure   requires at most ~106 bits for an exact sign determination.
+  * **float128** — double + 1×uint64_t, giving ~117 bits of mantissa   precision. Used for orient2d, where the 2×2 determinant structure   requires at most ~106 bits for an exact sign determination.
-- **float256** — double + 3×uint64_t, giving ~245 bits of mantissa   precision. Used for inCircle, where the 3×3 determinant with squared   terms requires up to ~212 bits.
+  * **float256** — double + 3×uint64_t, giving ~245 bits of mantissa   precision. Used for inCircle, where the 3×3 determinant with squared   terms requires up to ~212 bits.
 This two-tier approach (double → float128 or float256) avoids the robustness failures that plague naive floating-point implementations while keeping the common case fast. Over 99% of predicate calls are resolved at double precision, and the few that fall through use the narrowest exact type sufficient for the predicate.