[Intel-xe] [PATCH 2/2] drm/xe: Emit a render cache flush after each rcs/ccs batch
Thomas Hellström
thomas.hellstrom at linux.intel.com
Wed Jun 7 17:47:29 UTC 2023
We need to flush render caches before fence signalling, where we might
release the memory for reuse. We can't rely on userspace doing this,
so flush render caches after the batch, but before user fence- and
dma_fence signalling.
Copy the cache flush from i915, but omit PIPE_CONTROL_FLUSH_L3, since it
should be implied by the other flushes. Also omit
PIPE_CONTROL_TLB_INVALIDATE since there should be no apparent need to
invalidate TLB after batch completion.
Signed-off-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>
---
drivers/gpu/drm/xe/regs/xe_gpu_commands.h | 3 ++
drivers/gpu/drm/xe/xe_ring_ops.c | 35 +++++++++++++++++++++++
drivers/gpu/drm/xe/xe_wa_oob.rules | 1 +
3 files changed, 39 insertions(+)
diff --git a/drivers/gpu/drm/xe/regs/xe_gpu_commands.h b/drivers/gpu/drm/xe/regs/xe_gpu_commands.h
index d2d41f717525..dd3408fd3d33 100644
--- a/drivers/gpu/drm/xe/regs/xe_gpu_commands.h
+++ b/drivers/gpu/drm/xe/regs/xe_gpu_commands.h
@@ -66,6 +66,9 @@
#define PVC_MS_MOCS_INDEX_MASK GENMASK(6, 1)
#define GFX_OP_PIPE_CONTROL(len) ((0x3<<29)|(0x3<<27)|(0x2<<24)|((len)-2))
+
+#define PIPE_CONTROL0_HDC_PIPELINE_FLUSH REG_BIT(9) /* gen12 */
+
#define PIPE_CONTROL_COMMAND_CACHE_INVALIDATE (1<<29)
#define PIPE_CONTROL_TILE_CACHE_FLUSH (1<<28)
#define PIPE_CONTROL_AMFS_FLUSH (1<<25)
diff --git a/drivers/gpu/drm/xe/xe_ring_ops.c b/drivers/gpu/drm/xe/xe_ring_ops.c
index c20fe41c0729..91511f72d971 100644
--- a/drivers/gpu/drm/xe/xe_ring_ops.c
+++ b/drivers/gpu/drm/xe/xe_ring_ops.c
@@ -5,6 +5,7 @@
#include "xe_ring_ops.h"
+#include "generated/xe_wa_oob.h"
#include "regs/xe_gpu_commands.h"
#include "regs/xe_gt_regs.h"
#include "regs/xe_lrc_layout.h"
@@ -16,6 +17,7 @@
#include "xe_sched_job.h"
#include "xe_vm_types.h"
#include "xe_vm.h"
+#include "xe_wa.h"
/*
* 3D-related flags that can't be set on _engines_ that lack access to the 3D
@@ -147,6 +149,37 @@ static int emit_store_imm_ppgtt_posted(u64 addr, u64 value,
return i;
}
+static int emit_render_cache_flush(struct xe_sched_job *job, u32 *dw, int i)
+{
+ struct xe_gt *gt = job->engine->gt;
+ bool lacks_render = !(gt->info.engine_mask & XE_HW_ENGINE_RCS_MASK);
+ u32 flags;
+
+ flags = (PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_TILE_CACHE_FLUSH |
+ PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
+ PIPE_CONTROL_DEPTH_CACHE_FLUSH |
+ PIPE_CONTROL_DC_FLUSH_ENABLE |
+ PIPE_CONTROL_FLUSH_ENABLE);
+
+ if (XE_WA(gt, 1409600907))
+ flags |= PIPE_CONTROL_DEPTH_STALL;
+
+ if (lacks_render)
+ flags &= ~PIPE_CONTROL_3D_ARCH_FLAGS;
+ else if (job->engine->class == XE_ENGINE_CLASS_COMPUTE)
+ flags &= ~PIPE_CONTROL_3D_ENGINE_FLAGS;
+
+ dw[i++] = GFX_OP_PIPE_CONTROL(6) | PIPE_CONTROL0_HDC_PIPELINE_FLUSH;
+ dw[i++] = flags;
+ dw[i++] = 0;
+ dw[i++] = 0;
+ dw[i++] = 0;
+ dw[i++] = 0;
+
+ return i;
+}
+
static int emit_pipe_imm_ggtt(u32 addr, u32 value, bool stall_only, u32 *dw,
int i)
{
@@ -279,6 +312,8 @@ static void __emit_job_gen12_render_compute(struct xe_sched_job *job,
i = emit_bb_start(batch_addr, ppgtt_flag, dw, i);
+ i = emit_render_cache_flush(job, dw, i);
+
if (job->user_fence.used)
i = emit_store_imm_ppgtt_posted(job->user_fence.addr,
job->user_fence.value,
diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules b/drivers/gpu/drm/xe/xe_wa_oob.rules
index 1ecb10390b28..15c23813398a 100644
--- a/drivers/gpu/drm/xe/xe_wa_oob.rules
+++ b/drivers/gpu/drm/xe/xe_wa_oob.rules
@@ -14,3 +14,4 @@
SUBPLATFORM(DG2, G12)
18020744125 PLATFORM(PVC)
1509372804 PLATFORM(PVC), GRAPHICS_STEP(A0, C0)
+1409600907 GRAPHICS_VERSION_RANGE(1200, 1250)
--
2.39.2
More information about the Intel-xe
mailing list