According to RadeonSI, this seems to be required by the hardware to avoid GPU hangs. I think I just forgot to set that bit when I implemented VK_EXT_transform_feedback.
This fixes a GPU hang with Space Engineers and DXVK. Cc: "19.0" "19.1" <[email protected]> Signed-off-by: Samuel Pitoiset <[email protected]> --- src/amd/vulkan/radv_cmd_buffer.c | 1 + src/amd/vulkan/radv_private.h | 1 + src/amd/vulkan/si_cmd_buffer.c | 7 +++++++ 3 files changed, 9 insertions(+) diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c index 963118215fa..b04c998fac2 100644 --- a/src/amd/vulkan/radv_cmd_buffer.c +++ b/src/amd/vulkan/radv_cmd_buffer.c @@ -2202,6 +2202,7 @@ radv_emit_draw_registers(struct radv_cmd_buffer *cmd_buffer, ia_multi_vgt_param = si_get_ia_multi_vgt_param(cmd_buffer, draw_info->instance_count > 1, draw_info->indirect, + !!draw_info->strmout_buffer, draw_info->indirect ? 0 : draw_info->count); if (state->last_ia_multi_vgt_param != ia_multi_vgt_param) { diff --git a/src/amd/vulkan/radv_private.h b/src/amd/vulkan/radv_private.h index 352072f8ff3..df85d0cf889 100644 --- a/src/amd/vulkan/radv_private.h +++ b/src/amd/vulkan/radv_private.h @@ -1160,6 +1160,7 @@ void si_write_scissors(struct radeon_cmdbuf *cs, int first, const VkViewport *viewports, bool can_use_guardband); uint32_t si_get_ia_multi_vgt_param(struct radv_cmd_buffer *cmd_buffer, bool instanced_draw, bool indirect_draw, + bool count_from_stream_output, uint32_t draw_vertex_count); void si_cs_emit_write_event_eop(struct radeon_cmdbuf *cs, enum chip_class chip_class, diff --git a/src/amd/vulkan/si_cmd_buffer.c b/src/amd/vulkan/si_cmd_buffer.c index e75c6d127d6..e73c13762e5 100644 --- a/src/amd/vulkan/si_cmd_buffer.c +++ b/src/amd/vulkan/si_cmd_buffer.c @@ -561,6 +561,7 @@ radv_prims_for_vertices(struct radv_prim_vertex_count *info, unsigned num) uint32_t si_get_ia_multi_vgt_param(struct radv_cmd_buffer *cmd_buffer, bool instanced_draw, bool indirect_draw, + bool count_from_stream_output, uint32_t draw_vertex_count) { enum chip_class chip_class = cmd_buffer->device->physical_device->rad_info.chip_class; @@ -622,6 +623,12 @@ si_get_ia_multi_vgt_param(struct radv_cmd_buffer *cmd_buffer, (instanced_draw || indirect_draw)) partial_vs_wave = true; + /* Hardware requirement when drawing primitives from a stream + * output buffer. + */ + if (count_from_stream_output) + wd_switch_on_eop = true; + /* If the WD switch is false, the IA switch must be false too. */ assert(wd_switch_on_eop || !ia_switch_on_eop); } -- 2.21.0 _______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
