Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update long double test with correct values #4060

Merged
merged 4 commits into from
Mar 8, 2024
Merged

Update long double test with correct values #4060

Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
bde3a01
Update long double test with correct values
byrnHDF Mar 4, 2024
21236ab
Committing clang-format changes
github-actions[bot] Mar 4, 2024
6a4c71f
Correct name of macro
byrnHDF Mar 4, 2024
92a4762
remove incorrect comment
byrnHDF Mar 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
303 changes: 303 additions & 0 deletions test/dsets.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5681,6 +5681,308 @@ test_multiopen(hid_t file)
return FAIL;
} /* end test_multiopen() */

/*-------------------------------------------------------------------------
* Function: test_floattypes
*
* Purpose: Make some datasets with various float types.
*
* Return: Success: 0
*
* Failure: -1
*
*-------------------------------------------------------------------------
*/
static herr_t
test_floattypes(hid_t file)
{
hid_t dataset = H5I_INVALID_HID;
hid_t datatype = H5I_INVALID_HID;
hid_t space = H5I_INVALID_HID;
const hsize_t size[2] = {2, 5};
size_t i, j;
size_t precision, offset;

puts("Testing float datatypes");

/* float */
{
/* orig_data[] are initialized to be within the range that can be represented by
* dataset datatype (no precision loss during datatype conversion)
*/
float orig_data[2][5] = {{188384.0F, 19.103516F, -1.0831790e9F, -84.242188F, 5.2045898F},
{-49140.0F, 2350.25F, -3.2110596e-1F, 6.4998865e-5F, -0.0F}};
float new_data[2][5];

TESTING(" float (setup)");

/* Define user-defined single-precision floating-point type for dataset */
datatype = H5Tcopy(H5T_IEEE_F32BE);
if (H5Tset_fields(datatype, (size_t)26, (size_t)20, (size_t)6, (size_t)7, (size_t)13) < 0)
goto error;
offset = 7;
if (H5Tset_offset(datatype, offset) < 0)
goto error;
precision = 20;
if (H5Tset_precision(datatype, precision) < 0)
goto error;
if (H5Tset_size(datatype, (size_t)4) < 0)
goto error;
if (H5Tset_ebias(datatype, (size_t)31) < 0)
goto error;

/* Create the data space */
if ((space = H5Screate_simple(2, size, NULL)) < 0)
goto error;

/* Create the dataset */
if ((dataset =
H5Dcreate2(file, "float_type)", datatype, space, H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT)) < 0)
goto error;
PASSED();

/*----------------------------------------------------------------------
* STEP 1: Test writing to it.
*----------------------------------------------------------------------
*/
TESTING(" float (write)");

if (H5Dwrite(dataset, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL, H5P_DEFAULT, orig_data) < 0)
goto error;

PASSED();

/*----------------------------------------------------------------------
* STEP 2: Try to read the data we just wrote.
*----------------------------------------------------------------------
*/
TESTING(" float (read)");

/* Read the dataset back */
if (H5Dread(dataset, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL, H5P_DEFAULT, new_data) < 0)
goto error;

/* Check that the values read are the same as the values written
* Assume size of int = size of float
*/
for (i = 0; i < (size_t)size[0]; i++) {
for (j = 0; j < (size_t)size[1]; j++) {
if (isnan(orig_data[i][j]))
continue; /* skip if value is NaN */
if (!H5_FLT_ABS_EQUAL(new_data[i][j], orig_data[i][j])) {
H5_FAILED();
printf(" Read different values than written.\n");
printf(" At index %lu,%lu\n", (unsigned long)i, (unsigned long)j);
goto error;
}
}
}

PASSED();

/*----------------------------------------------------------------------
* Cleanup
*----------------------------------------------------------------------
*/
if (H5Tclose(datatype) < 0)
goto error;
if (H5Sclose(space) < 0)
goto error;
if (H5Dclose(dataset) < 0)
goto error;
}

/* double */
{
double orig_data[2][5] = {
{(double)1.6081706885101836e+60L, (double)-255.32099170994480, (double)1.2677579992621376e-61L,
(double)64568.289448797700, (double)-1.0619721778839084e-75L},
{(double)2.1499497833454840e+56L, (double)6.6562295504670740e-3, (double)-1.5747263393432150,
(double)1.0711093225222612, (double)-9.8971679387636870e-1}};
double new_data[2][5];

TESTING(" double (setup)");

/* Define user-defined double-precision floating-point type for dataset */
datatype = H5Tcopy(H5T_IEEE_F64BE);
if (H5Tset_fields(datatype, (size_t)55, (size_t)46, (size_t)9, (size_t)5, (size_t)41) < 0)
goto error;
offset = 5;
if (H5Tset_offset(datatype, offset) < 0)
goto error;
precision = 51;
if (H5Tset_precision(datatype, precision) < 0)
goto error;
if (H5Tset_size(datatype, (size_t)8) < 0)
goto error;
if (H5Tset_ebias(datatype, (size_t)255) < 0)
goto error;

/* Create the data space */
if ((space = H5Screate_simple(2, size, NULL)) < 0)
goto error;

/* Create the dataset */
if ((dataset =
H5Dcreate2(file, "double_type", datatype, space, H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT)) < 0)
goto error;

PASSED();

/*----------------------------------------------------------------------
* STEP 1: Test writing to it.
*----------------------------------------------------------------------
*/
TESTING(" double (write)");

if (H5Dwrite(dataset, H5T_NATIVE_DOUBLE, H5S_ALL, H5S_ALL, H5P_DEFAULT, orig_data) < 0)
goto error;
PASSED();

/*----------------------------------------------------------------------
* STEP 2: Try to read the data we just wrote.
*----------------------------------------------------------------------
*/
TESTING(" double (read)");

/* Read the dataset back */
if (H5Dread(dataset, H5T_NATIVE_DOUBLE, H5S_ALL, H5S_ALL, H5P_DEFAULT, new_data) < 0)
goto error;

/* Check that the values read are the same as the values written
* Assume size of long long = size of double
*/
for (i = 0; i < (size_t)size[0]; i++) {
for (j = 0; j < (size_t)size[1]; j++) {
if (isnan(orig_data[i][j]))
continue; /* skip if value is NaN */
if (!H5_DBL_ABS_EQUAL(new_data[i][j], orig_data[i][j])) {
H5_FAILED();
printf(" Read different values than written.\n");
printf(" At index %lu,%lu\n", (unsigned long)i, (unsigned long)j);
goto error;
}
}
}

PASSED();

/*----------------------------------------------------------------------
* Cleanup
*----------------------------------------------------------------------
*/
if (H5Tclose(datatype) < 0)
goto error;
if (H5Sclose(space) < 0)
goto error;
if (H5Dclose(dataset) < 0)
goto error;
}
#if H5_SIZEOF_LONG_DOUBLE != H5_SIZEOF_DOUBLE
/* long double */
{
long double orig_data[2][5] = {
{(long double)1.6081706885101836e+600L, (long double)-255.3209917099448032099170994480,
(long double)1.2677579992621376e-610L, (long double)64568.289448797700289448797700,
(long double)-1.0619721778839084e-750L},
{(long double)2.1499497833454840991499497833454840e+560L,
(long double)6.6562295504670740996562295504670740e-3,
(long double)-1.5747263393432150995747263393432150,
(long double)1.0711093225222612990711093225222612,
(long double)-9.8971679387636870998971679387636870e-1}};
Copy link
Contributor

@brtnfld brtnfld Mar 4, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why hard code the values as opposed to using a math function? If not, how were these values chosen? They can go much bigger/smaller than this.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It was just to test values that would be past the double range of values.
A programmatic way would be useful in the future.

long double new_data[2][5];

TESTING(" long double (setup)");

/* Define user-defined quad-precision floating-point type for dataset */
datatype = H5Tcopy(H5T_NATIVE_LDOUBLE);
precision = 128;
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The precision is actually 123 bits here. I'm surprised setting it to 128 doesn't cause conversion accuracy issues, but we should still set it to 123 bits here so we aren't lying.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I could try, but I don't remember if I had problems if it didn't match the sign pos. Maybe because of the offset?

if (H5Tset_precision(datatype, precision) < 0)
goto error;
if (H5Tset_fields(datatype, (size_t)127, (size_t)112, (size_t)15, (size_t)5, (size_t)107) < 0)
goto error;
offset = 5;
if (H5Tset_offset(datatype, offset) < 0)
goto error;
if (H5Tset_size(datatype, (size_t)16) < 0)
goto error;
if (H5Tset_ebias(datatype, (size_t)255) < 0)
goto error;

/* Create the data space */
if ((space = H5Screate_simple(2, size, NULL)) < 0)
goto error;

/* Create the dataset */
if ((dataset = H5Dcreate2(file, "long_double_type", datatype, space, H5P_DEFAULT, H5P_DEFAULT,
H5P_DEFAULT)) < 0)
goto error;

PASSED();

/*----------------------------------------------------------------------
* STEP 1: Test writing to it.
*----------------------------------------------------------------------
*/
TESTING(" long double (write)");

if (H5Dwrite(dataset, H5T_NATIVE_LDOUBLE, H5S_ALL, H5S_ALL, H5P_DEFAULT, orig_data) < 0)
goto error;
PASSED();

/*----------------------------------------------------------------------
* STEP 2: Try to read the data we just wrote.
*----------------------------------------------------------------------
*/
TESTING(" long double (read)");

/* Read the dataset back */
if (H5Dread(dataset, H5T_NATIVE_LDOUBLE, H5S_ALL, H5S_ALL, H5P_DEFAULT, new_data) < 0)
goto error;

/* Check that the values read are the same as the values written
* Assume size of long long = size of double
jhendersonHDF marked this conversation as resolved.
Show resolved Hide resolved
*/
for (i = 0; i < (size_t)size[0]; i++) {
for (j = 0; j < (size_t)size[1]; j++) {
if (isnan(orig_data[i][j]))
continue; /* skip if value is NaN */
if (!H5_LDBL_ABS_EQUAL(new_data[i][j], orig_data[i][j])) {
Copy link
Contributor

@brtnfld brtnfld Mar 4, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are we sure that is checking the complete precision? I would think it would fail for the cases where the complete precision was not initialized. Also, do we expect NaN values? I would think that would be an error.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This just follows the other float and double parts.
These would be issues that should be noted.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These macros just get the difference of the two float, double, or long double values and then take the absolute value and compare it to the relevant epsilon value. It should be fine for checking the data.

H5_FAILED();
printf(" Read different values than written.\n");
printf(" At index %lu,%lu\n", (unsigned long)i, (unsigned long)j);
goto error;
}
}
}

PASSED();

/*----------------------------------------------------------------------
* Cleanup
*----------------------------------------------------------------------
*/
if (H5Tclose(datatype) < 0)
goto error;
if (H5Sclose(space) < 0)
goto error;
if (H5Dclose(dataset) < 0)
goto error;
}
#endif

return SUCCEED;

error:
H5E_BEGIN_TRY
{
H5Tclose(datatype);
H5Sclose(space);
H5Dclose(dataset);
}
H5E_END_TRY
return FAIL;
} /* end test_floattypes() */

/*-------------------------------------------------------------------------
* Function: test_types
*
Expand Down Expand Up @@ -15583,6 +15885,7 @@ main(void)
nerrors += (test_scaleoffset_double_2(file) < 0 ? 1 : 0);
nerrors += (test_multiopen(file) < 0 ? 1 : 0);
nerrors += (test_types(file) < 0 ? 1 : 0);
nerrors += (test_floattypes(file) < 0 ? 1 : 0);
nerrors += (test_userblock_offset(envval, my_fapl, new_format) < 0 ? 1 : 0);

if (driver_is_default_compatible) {
Expand Down
Loading