-
Notifications
You must be signed in to change notification settings - Fork 194
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.
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
test_printf producing false expected results #2009
Comments
@coldav could you tell which subtest this is? This should be printed before the line starting with |
float... |
With a little bit more digging if we disable double support we do get the 128 answer. We think it's acceptable to do the conversion to double though so both the 123 and 128 are correct. |
Thanks, I figured this had to be related to I'm not sure if this is a recent regression though; the
That seems like a recent regression indeed. I can put up a fix unless you already have a patch for it. |
Sorry, I meant to reply. Thanks for the patch. This should stay open for the original problem though. |
Can we make this change, for now at least? We're encountering the same problem on our devices that support fp64. |
I get the same failure on Linux x86-64 Gcc 13.2.0 with Clvk (top of tree), and NVIDIA Quadro P1000 Driver 525.147.5.0 (2204418368) for completeness, subtests 14 and 15 have the same failure for the same value.
|
I posted #2015 which does that. |
Thanks, I approved #2015. We're not meeting for the next couple of weeks, and this should be non-contentious, so as soon as somebody else signs off on it we should merge it. |
@bashbaug I approved as well |
Thanks! @coldav let us know if this doesn't fix your problem for some reason. |
in the oneAPI Construction Kit, our overnight testing on 10th shows repeated fails on the test_printf test:
15)testing printf("%+#21.15E",789456123.0)
verifyOutputBuffer failed with kernel:
__kernel void test3(void)
{
printf("%+#21.15E\n",789456123.0);
}
expected: +7.894561280000000E+08
got: +7.894561230000000E+08
This was run on x86 in our "host" target.
I believe the 123 part is actually correct, so the test seems wrong.
#1988 maybe related as it was committed around the same time.
To show with the oneAPI Construction Kit.
The text was updated successfully, but these errors were encountered: