Sh

Message893

Your Queries (edit)

Issues
Show Unassigned
Show Bite-sized
Show All
Search

Login



Register
Lost your login?

Help
Roundup docs

Author francois
Recipients
Date 2006-02-15.22:17:14
Content
It works on ATI and it seems to work on NVIDIA 7174 drivers (using either the
ARB or GLSL backends).
It also works on the latest NVIDIA drivers using GLSL.
If we generate the same ARB code than on ATI (using CMP instead of COND, 2D
textures instead of RECT), it still doesn't work on NVIDIA with the given drivers.
It does work on ARB when we remove the amount of dead code that gets eliminated
by commenting out the hasDecl() line.  The extra code that stays there doesn't
affect anything except that it may change the register allocation.
This is starting to smell like a driver bug...
History
Date User Action Args
2006-02-15 22:17:15francoislinkissue320 messages
2006-02-15 22:17:15francoiscreate

Related topics

Frequently Asked Questions

What is this page?

This page is message 893 from the Sh issue tracker at libsh.org. It shows the message author, date, content, and the history of actions recorded for the message.

Which issue does this message belong to?

According to the history table, message 893 is linked to issue 320. You can follow the “issue320 messages” link to see the full thread.

Who wrote this message and when?

The message was authored by francois and dated 2006-02-15 22:17:14, as shown in the message details table.

What is the message about?

The message discusses testing Sh output on ATI and NVIDIA drivers, differences between the ARB and GLSL backends, and the effect of dead-code elimination and register allocation — the author suggests it may be a driver bug.

How can I find other messages or issues?

Use the sidebar links to show unassigned, bite-sized, or all issues, search the tracker, or jump to a specific issue by number.