AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

tu: Re-emit visibility stream before each render pass

Browse Source 
When we set the visibility stream with CP_SET_PSEUDO_REG, it does two
things (or only one of the two, with concurrent binning):

- Set the "pseudo register" used by CP_SET_BIN_DATA5_OFFSET, which in
  turn is used when decoding the vis. streams.
- Set the VSC register used by the binning pass.

Preemption with skipsaverestore obliterates the second, but not the
first. This means that before running the binning pass, we have to
re-emit these registers. I *think* this is what the blob does on a7xx.
On a6xx, where the pseudo register doesn't exist, the blob seems to
re-emit the preamble every time we re-allocate the visibility streams,
but we don't support a6xx yet so we can defer making that decision.

Fixes supertuxkart under zink with preemption enabled in the kernel.

Fixes: 1d2b479a3b ("tu: Allow being preempted on a7xx")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31937>
This commit is contained in:
Connor Abbott
2024-11-01 14:44:40 +00:00
committed by Marge Bot
parent 2cadab5dcf
commit 423d472a4e
2 changed files with 32 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
src/freedreno/vulkan/tu_cmd_buffer.cc
View File

@@ -98,9 +98,8 @@ tu6_lazy_emit_tessfactor_addr(struct tu_cmd_buffer *cmd)
cmd->state.tessfactor_addr_set = true;
}
template <chip CHIP>
static void
tu6_lazy_emit_vsc(struct tu_cmd_buffer *cmd, struct tu_cs *cs)
tu6_lazy_init_vsc(struct tu_cmd_buffer *cmd)
{
struct tu_device *dev = cmd->device;
uint32_t num_vsc_pipes = dev->physical_device->info->num_vsc_pipes;
@@ -136,23 +135,30 @@ tu6_lazy_emit_vsc(struct tu_cmd_buffer *cmd, struct tu_cs *cs)
tu_get_scratch_bo(dev, size0 + num_vsc_pipes * 4, &vsc_bo);
cmd->vsc_draw_strm_va = vsc_bo->iova + cmd->vsc_prim_strm_pitch * num_vsc_pipes;
cmd->vsc_draw_strm_size_va = vsc_bo->iova + size0;
cmd->vsc_prim_strm_va = vsc_bo->iova;
}
template <chip CHIP>
static void
tu_emit_vsc(struct tu_cmd_buffer *cmd, struct tu_cs *cs)
{
if (CHIP == A6XX) {
tu_cs_emit_regs(cs,
A6XX_VSC_DRAW_STRM_SIZE_ADDRESS(.bo = vsc_bo, .bo_offset = size0));
A6XX_VSC_DRAW_STRM_SIZE_ADDRESS(.qword = cmd->vsc_draw_strm_size_va));
tu_cs_emit_regs(cs,
A6XX_VSC_PRIM_STRM_ADDRESS(.bo = vsc_bo));
A6XX_VSC_PRIM_STRM_ADDRESS(.qword = cmd->vsc_prim_strm_va));
tu_cs_emit_regs(
cs, A6XX_VSC_DRAW_STRM_ADDRESS(.bo = vsc_bo,
.bo_offset = cmd->vsc_prim_strm_pitch *
num_vsc_pipes));
cs, A6XX_VSC_DRAW_STRM_ADDRESS(.qword = cmd->vsc_draw_strm_va));
} else {
tu_cs_emit_pkt7(cs, CP_SET_PSEUDO_REG, 3 * 3);
tu_cs_emit(cs, A6XX_CP_SET_PSEUDO_REG__0_PSEUDO_REG(DRAW_STRM_ADDRESS));
tu_cs_emit_qw(cs, vsc_bo->iova + cmd->vsc_prim_strm_pitch * num_vsc_pipes);
tu_cs_emit_qw(cs, cmd->vsc_draw_strm_va);
tu_cs_emit(cs, A6XX_CP_SET_PSEUDO_REG__0_PSEUDO_REG(DRAW_STRM_SIZE_ADDRESS));
tu_cs_emit_qw(cs, vsc_bo->iova + size0);
tu_cs_emit_qw(cs, cmd->vsc_draw_strm_size_va);
tu_cs_emit(cs, A6XX_CP_SET_PSEUDO_REG__0_PSEUDO_REG(PRIM_STRM_ADDRESS));
tu_cs_emit_qw(cs, vsc_bo->iova);
tu_cs_emit_qw(cs, cmd->vsc_prim_strm_va);
}
cmd->vsc_initialized = true;
@@ -2121,9 +2127,23 @@ tu6_tile_render_begin(struct tu_cmd_buffer *cmd, struct tu_cs *cs,
if (use_hw_binning(cmd)) {
if (!cmd->vsc_initialized) {
tu6_lazy_emit_vsc<CHIP>(cmd, cs);
tu6_lazy_init_vsc(cmd);
}
/* We always emit VSC before each renderpass, because due to
* skipsaverestore the underlying VSC registers may have become
* invalid. Normally we'd need to WFI before setting these non-context
* registers, but we should be safe because we're only setting it to the
* same value it had before.
*
* TODO: On a6xx, we have to emit this 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.
*/
tu_emit_vsc<CHIP>(cmd, cs);
tu6_emit_bin_size<CHIP>(cs, tiling->tile0.width, tiling->tile0.height,
{
.render_mode = BINNING_PASS,

1
src/freedreno/vulkan/tu_cmd_buffer.h
View File

@@ -611,6 +611,7 @@ struct tu_cmd_buffer
uint32_t vsc_draw_strm_pitch;
uint32_t vsc_prim_strm_pitch;
uint64_t vsc_draw_strm_va, vsc_draw_strm_size_va, vsc_prim_strm_va;
bool vsc_initialized;
};
VK_DEFINE_HANDLE_CASTS(tu_cmd_buffer, vk.base, VkCommandBuffer,