Make takfp benchmark work with OpenDylan

README.rst

@@ -15,18 +15,18 @@ In general, we want each example to be one file (if possible).
 TODO
 ----
 
-* ``fasta``, ``pidigits``, ``spellcheck``, ``takfp``, ``wc`` and ``wordfreq``
+* ``fasta``, ``pidigits``, ``spellcheck``, ``wc`` and ``wordfreq``
   have to be converted from GD supported single-file format to a dual-file
   format compilable under OpenDylan.
 * ``binary-trees``, ``fannkuch``, ``fasta``, ``fibonacci``, ``harmonic``,
-  ``mandelbrot``, ``nestedloop`` and ``strcat`` should provide a default
-  command line argument value.
+  ``mandelbrot``, ``nestedloop``, ``strcat`` and ``takfp`` should provide a
+  default command line argument value.
 * ``nsieve`` provides an invalid default and should check for valid argument.
 * ``nsieve`` and ``sieve`` benchmarks don't seem to work properly.
 * ``fasta`` benchmark generates invalid output.
-* ``k-nucleotide``, ``n-body`` and ``spectralnorm`` benchmarks don't provide exactly
-  expected output, due to lack of precision configuration support in the
-  control-string of ``format-out``.
+* ``k-nucleotide``, ``n-body``, ``spectralnorm`` and ``takfp`` benchmarks
+  don't provide exactly expected output, due to lack of precision
+  configuration support in the control-string of ``format-out``.
 * ``moments`` benchmark doesn't work due to lack of support for ``string-to-float``.
 * ``pidigits`` benchmark doesn't work due to lack of extended integer support.
 * Many benchmarks can still be optimized.

takfp.dylan

@@ -1,5 +1,15 @@
 module: takfp
 
+define library takfp
+  use common-dylan;
+  use io;
+end library;
+
+define module takfp
+  use common-dylan, exclude: { format-to-string };
+  use format-out;
+end module;
+
 define constant <fp> = <single-float>;
 
 begin
@@ -13,6 +23,9 @@ begin
         end method;
 
   let n = application-arguments()[0].string-to-integer;
-  format-out("%.1f\n", tak(3.0s0 * n, 2.0s0 * n, 1.0s0 * n));
+  // FIXME: "%.1f" is not supported as control-string, as a result 7 decimal
+  // digits are printed instead of 1.
+  //format-out("%.1f\n", tak(3.0s0 * n, 2.0s0 * n, 1.0s0 * n));
+  format-out("%=\n", tak(3.0s0 * n, 2.0s0 * n, 1.0s0 * n));
 end;
 

takfp.lid

@@ -0,0 +1,3 @@
+library: takfp
+executable: takfp
+files: takfp