[kd-11]

Closes #17091
Closes #17099

[digant73]

Not yet tested this PR but it should still fail. I fixed the issue passing the base pointer (_ptr is increased in the loop so it will point to the end of the buffer instead of base address) in:

_vkCmdDrawMultiIndexedEXT(*m_current_command_buffer, subranges_count, _ptr, 1, 0, sizeof(u32) * 3, nullptr);

so something like:

auto _ptr = m_multidraw_parameters_buffer.data(), _ptr_base = _ptr;
...
...

_vkCmdDrawMultiIndexedEXT(*m_current_command_buffer, subranges_count, _ptr_base, 1, 0, sizeof(u32) * 3, nullptr);

[digant73]

I confirm the issue even on this PR (wrong base address). No issue passing the base address

[digant73]

also closes #17099

[kd-11]

I confirm the issue even on this PR (wrong base address). No issue passing the base address

The struct was also constructed incorrectly, so even if you pass base pointer you would still have missing draws and flicker.
That whole changeset was an untested disaster.

[digant73]

I see. I limited my test on GT5 only. Regressions are very dangerous if not immediately fixed with so many PRs continuously merged. New bugs over regressed code could be extremely difficult to be properly discovered. For that reason I'm focused on that testing. Sorry for stressing you even on Sunday