Re: aapoint issue

[Ian Romanick]

Can you test 
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32313 ?

On 11/14/24 12:57 AM, Erik Faye-Lund wrote:
Yeah, that looks like a mistake, indeed. Do you mind submitting a patch
on GitLab for it?
The problem seems to originate from this commit (CCed ian):
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20869/diffs?commit_id=d3a95f0f713ff3a0216f0dfa634798a1db55ef27

On Thu, 2024-11-14 at 03:31 +0000, Anddy Ren(WH-RD) wrote:



Hi,
  
I'm experiencing a crash while verifying aapoint functionality, and
after debugging, I found that there may be a problem with the
following statement.
  
In nir_draw_helpers.c,
  
static void
nir_lower_aapoint_impl(nir_function_impl *impl, lower_aapoint *state,
                        nir_alu_type bool_type)
{
…………
    switch (bool_type) {
    case nir_type_bool1:
       sel =nir_b32csel(b, nir_fge(b, k, dist), coverage,
chan_val_one);
       break;
    case nir_type_bool32:
       sel = nir_b32csel(b, nir_fge32(b, k, dist), coverage,
chan_val_one);
       break;
    …………
  
The bool_type passed to this function is nir_type_bool1, when I
change nir_b32csel to nir_bcsel, the test passes. Is it a bug? Can
you help confirm if it's okay to change it this way?
  
Best Regards,
Anddy Ren
  


保密声明：
本邮件含有保密或专有信息，仅供指定收件人使用。严禁对本邮件或其内容做任何未经授权的查阅、使用、复制或转发。

CONFIDENTIAL NOTE:

This email contains confidential or legally privileged information
and is for the sole use of its intended recipient. Any unauthorized
review, use, copying or forwarding of this email or the content of
this email is strictly prohibited.