intel/fs/gen12: Fix hangs with per-sample SIMD32 fragment shader dispatch.

The Gen12 docs are rather contradictory regarding the dispatch
configurations supported by the fragment shader -- The same table
present in previous generations seems to imply that only one dispatch
mode can be enabled when doing per-sample shading, but a restriction
documented in the 3DSTATE_PS_BODY page implies the opposite: That
SIMD32 can only be used in combination with some other dispatch mode.

The latter seems to match the behavior of real hardware as I could
tell from my testing: A bunch of multisample test-cases that do
per-sample shading hang if we only provide a SIMD32 shader.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit is contained in:
Francisco Jerez
2020-04-07 17:22:10 -07:00
parent 35ee6b3d36
commit d6ae079771

View File

@@ -8793,13 +8793,20 @@ brw_compile_fs(const struct brw_compiler *compiler, void *log_data,
if (prog_data->persample_dispatch) {
/* Starting with SandyBridge (where we first get MSAA), the different
* pixel dispatch combinations are grouped into classifications A
* through F (SNB PRM Vol. 2 Part 1 Section 7.7.1). On all hardware
* through F (SNB PRM Vol. 2 Part 1 Section 7.7.1). On most hardware
* generations, the only configurations supporting persample dispatch
* are are this in which only one dispatch width is enabled.
* are those in which only one dispatch width is enabled.
*
* The Gen12 hardware spec has a similar dispatch grouping table, but
* the following conflicting restriction applies (from the page on
* "Structure_3DSTATE_PS_BODY"), so we need to keep the SIMD16 shader:
*
* "SIMD32 may only be enabled if SIMD16 or (dual)SIMD8 is also
* enabled."
*/
if (simd32_cfg || simd16_cfg)
simd8_cfg = NULL;
if (simd32_cfg)
if (simd32_cfg && devinfo->gen < 12)
simd16_cfg = NULL;
}