radeonsi: don't use the SPI barrier management bug workaround
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
This commit is contained in:
parent
3045c5f274
commit
3fea237c85
|
@ -7971,6 +7971,11 @@ static bool si_shader_select_ps_parts(struct si_screen *sscreen,
|
|||
void si_multiwave_lds_size_workaround(struct si_screen *sscreen,
|
||||
unsigned *lds_size)
|
||||
{
|
||||
/* If tessellation is all offchip and on-chip GS isn't used, this
|
||||
* workaround is not needed.
|
||||
*/
|
||||
return;
|
||||
|
||||
/* SPI barrier management bug:
|
||||
* Make sure we have at least 4k of LDS in use to avoid the bug.
|
||||
* It applies to workgroup sizes of more than one wavefront.
|
||||
|
|
Loading…
Reference in New Issue