On Thu, 20 Mar 2025 13:30:27 +0100 (CET)
Richard Biener <[email protected]> wrote:
> @@ -4126,7 +4137,11 @@ count: %empty { $$ = 0; }
> if( e ) { // verify not floating point with nonzero fraction
> auto field = cbl_field_of(e);
> assert(is_literal(field));
> - if( field->data.value_of() !=
> size_t(field->data.value_of()) ) {
> + REAL_VALUE_TYPE vi;
> + HOST_WIDE_INT vii = real_to_integer (TREE_REAL_CST_PTR
> (field->data.value_of()));
> + real_from_integer (&vi, VOIDmode, vii, SIGNED);
> + if( !real_identical (TREE_REAL_CST_PTR
> (field->data.value_of()),
> + &vi) ) {
> nmsg++;
> error_msg(@NAME, "invalid PICTURE count '(%s)'",
> field->data.initial );
I just want to verify this still does what we want. We had
> - if( field->data.value_of() != size_t(field->data.value_of()) ) {
When cbl_field_t::data::value_of returned _Float128, that line compared a
_Float128 to its value as a size_t. If the number was negative, had a
fractional component, or was too large, the test failed.
Now cbl_field_t::data::value_of returns a tree. Steps:
1. produce HOST_WIDE_INT vii from the tree
2. set REAL_VALUE_TYPE vi from vii
3. use real_identical to compare vi to the original
The comment for real_identical says it's a bitwise comparison. That's not the
same as saying the floating point value can be represented exactly as a size_t.
If that's guaranteed to work, great, thanks. If not, how do I test the sign,
magnitude and fraction of the value?
<future work>
Until now, numeric literals in the parser were represented as _Float128 and
in the CDF (cdf.y) as int64_t. Neither is strictly correct. ISO says numeric
literals are fixed-point, not floating:
"The value of a fixed-point numeric literal is the algebraic quantity
represented by the characters in the fixed-point numeric literal. The size of a
fixed-point numeric literal is equal to the number of digits in the string of
characters in the literal."
That means
01 ONE-GRAND CONSTANT 10 * 10 * 10.
should be a fixed-point value of 3 bytes with scaling factor of 0. As of now,
I believe we're using a tree of REAL_CST, but in the future I guess we should
switch to FIXED_CST.
</future work>
--jkl
