This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
[PATCH] Improve gen-libm-test.pl LIT() application
- From: "Paul E. Murphy" <murphyp at linux dot vnet dot ibm dot com>
- To: "libc-alpha at sourceware dot org" <libc-alpha at sourceware dot org>
- Cc: Joseph Myers <joseph at codesourcery dot com>
- Date: Thu, 4 Aug 2016 10:05:59 -0500
- Subject: [PATCH] Improve gen-libm-test.pl LIT() application
- Authentication-results: sourceware.org; auth=none
When bootstrapping float128, this exposed a number of areas where
the L suffix is incorrectly applied to simple expressions when it
should be applied to each constant in the expression.
In order to stave off more macros in libm-test.inc, apply_lit is
made slightly more intelligent. It will now split most basic
expressions and apply LIT() individually to each token. As noted,
it is not a perfect solution, but compromises correctness,
readability, and simplicity.
The above is problematic when the L real suffix is not the most
expressive modifier, and the compiler complains (i.e ppc64) or
silently truncates a value (i.e ppc64).
* math/gen-libm-test.pl (apply_lit): Rewrite to apply
LIT() to individual constants in simple expressions.
(_apply_lit): Rename replaced version, and use it to
apply to what appears to be a token.
---
math/gen-libm-test.pl | 40 +++++++++++++++++++++++++++++++++++++++-
1 file changed, 39 insertions(+), 1 deletion(-)
diff --git a/math/gen-libm-test.pl b/math/gen-libm-test.pl
index aa66e76..cb0a02b 100755
--- a/math/gen-libm-test.pl
+++ b/math/gen-libm-test.pl
@@ -163,7 +163,7 @@ sub show_exceptions {
# Apply the LIT(x) macro to a literal floating point constant
# and strip any existing suffix.
-sub apply_lit {
+sub _apply_lit {
my ($lit) = @_;
my $exp_re = "([+-])?[[:digit:]]+";
# Don't wrap something that does not look like a:
@@ -180,6 +180,44 @@ sub apply_lit {
return "LIT (${lit})";
}
+
+# Apply LIT macro to individual tokens within an
+# expression. This is only meant to work with
+# the very simple expressions encountered like
+# A (op B)*. It will not split all literals
+# correctly, but suffices for this usage without
+# a substantially more complex tokenizer.
+sub apply_lit {
+ my ($lit) = @_;
+ my @toks = ();
+
+ # There are some constant expressions embedded in the
+ # test cases. To avoid writing a more complex lexer,
+ # we apply some fixups, and split based on simple
+ # operators, unfixup, then apply LIT as needed.
+
+ # This behaves poorly on inputs like 0x0e+1.0f,
+ # or MAX_EXP+1, but ultimately doesn't break
+ # anything... for now.
+ $lit =~ s/([[:xdigit:]])[pP]\+/$1,/g;
+ $lit =~ s/([[:xdigit:]])[pP]\-/$1#/g;
+ $lit =~ s/([0-9])[eE]\+/$1'/g;
+ $lit =~ s/([0-9])[eE]\-/$1"/g;
+
+ # Split into tokenish looking things
+ @toks = split (/([\*\-\+\/])/, $lit);
+
+ # Remove fixups and apply LIT() if needed.
+ foreach (@toks) {
+ $_ =~ s/,/p+/g;
+ $_ =~ s/#/p-/g;
+ $_ =~ s/'/e+/g;
+ $_ =~ s/"/e-/g;
+ $_ = _apply_lit ($_);
+ }
+ return join ('', @toks);
+}
+
# Parse the arguments to TEST_x_y
sub parse_args {
my ($file, $descr, $args) = @_;
--
2.4.11