fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-26 18:58:13 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

tu/a6xx: Emit VSC addresses for each bin to restore after preemption

Browse source 
KGSL unconditionally supports preemption so we cannot ignore it.

On a6xx, we have to emit VSC addresses per-bin or make the amble include
these registers, because CP_SET_BIN_DATA5_OFFSET will use the
register instead of the pseudo register and its value won't survive
across preemptions. The blob seems to take the second approach and
emits the preamble lazily. We chose the per-bin approach but blob's
should be a better one.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12627

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33580>
This commit is contained in:
Danylo Piliaiev 2025-02-17 14:52:34 +01:00 • committed by Marge Bot
parent 9427df23b4
commit 49191f46e6
1 changed files with 6 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
src/freedreno/vulkan/tu_cmd_buffer.cc
View file

@ -1093,6 +1093,10 @@ tu6_emit_tile_select(struct tu_cmd_buffer *cmd,
tu_cs_emit(cs, A6XX_CP_SET_MARKER_0_MODE(RM6_BIN_RENDER_START) | tu_cs_emit(cs, A6XX_CP_SET_MARKER_0_MODE(RM6_BIN_RENDER_START) |
A6XX_CP_SET_MARKER_0_USES_GMEM); A6XX_CP_SET_MARKER_0_USES_GMEM);
if (CHIP == A6XX && cmd->device->physical_device->has_preemption) {
tu_emit_vsc<CHIP>(cmd, &cmd->cs);
}
tu6_emit_bin_size<CHIP>( tu6_emit_bin_size<CHIP>(
cs, tiling->tile0.width, tiling->tile0.height, cs, tiling->tile0.width, tiling->tile0.height,
{ {
@ -2183,7 +2187,8 @@ tu6_tile_render_begin(struct tu_cmd_buffer *cmd, struct tu_cs *cs,
* these registers, because CP_SET_BIN_DATA5_OFFSET will use the * these registers, because CP_SET_BIN_DATA5_OFFSET will use the
* register instead of the pseudo register and its value won't survive * register instead of the pseudo register and its value won't survive
* across preemptions. The blob seems to take the second approach and * across preemptions. The blob seems to take the second approach and
* emits the preamble lazily. * emits the preamble lazily. We chose the per-bin approach but blob's
* should be a better one.
*/ */
tu_emit_vsc<CHIP>(cmd, cs); tu_emit_vsc<CHIP>(cmd, cs);