-
Notifications
You must be signed in to change notification settings - Fork 137
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
Fix ICE in ArithmeticIF for INT*8 type #845
Fix ICE in ArithmeticIF for INT*8 type #845
Conversation
When half precision reals were introduced the table (aif) used in the ILI codegen for ArithmeticIF was increased in size and the offset for INT*8 increased by one. But the entry for half precision was not inserted, this causes an ICE for INT*8 type. The fix is to introduce the entry for the half precision type before the INT*8 type.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We've had the folks here look at it, and they are OK with this change. Thanks, @kiranchandramohan !
Thanks @gklimowicz for having a look. Also, appreciate the quick response. |
@shivaramaarao FYI, I believe 859 (is it from AMD?) also addresses the same issue. But I feel this patch's approach of adding the missing entry is probably better than deleting the entry. Let me know in case you think differently. |
We checked this patch and it works fine with our testing. LGTM |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
@gklimowicz Since you have approved it previously is it safe to assume that you have run the tests? |
Yes, I probably ran them in December, but let me make sure and run them again specifically on Power. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This still works on Power and LLVM 9. Approved.
When half precision reals were introduced the table (aif) used in the
ILI codegen for ArithmeticIF was increased in size and the offset for
INT8 increased by one. But the entry for half precision was not
inserted, this causes an ICE for INT8 type. The fix is to introduce
the entry for the half precision type before the INT*8 type.
Fix for #844