From: Matthew Auld <matthew.auld@intel.com>
To: intel-gfx@lists.freedesktop.org
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>,
Chris P Wilson <chris.p.wilson@intel.com>,
Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>,
dri-devel@lists.freedesktop.org,
John Harrison <John.C.Harrison@Intel.com>
Subject: [RFC PATCH 007/162] drm/i915: split wa_bb code to its own file
Date: Fri, 27 Nov 2020 12:04:43 +0000 [thread overview]
Message-ID: <20201127120718.454037-8-matthew.auld@intel.com> (raw)
In-Reply-To: <20201127120718.454037-1-matthew.auld@intel.com>
From: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Continuing the split of back-end independent code from the execlist
submission specific file.
Based on a patch by Chris Wilson.
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Chris P Wilson <chris.p.wilson@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
.../drm/i915/gt/intel_engine_workaround_bb.c | 335 ++++++++++++++++++
.../drm/i915/gt/intel_engine_workaround_bb.h | 14 +
.../drm/i915/gt/intel_execlists_submission.c | 327 +----------------
4 files changed, 352 insertions(+), 325 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.c
create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index f9ef5199b124..2445cc990e15 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -92,6 +92,7 @@ gt-y += \
gt/intel_engine_heartbeat.o \
gt/intel_engine_pm.o \
gt/intel_engine_user.o \
+ gt/intel_engine_workaround_bb.o \
gt/intel_execlists_submission.o \
gt/intel_ggtt.o \
gt/intel_ggtt_fencing.o \
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.c b/drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.c
new file mode 100644
index 000000000000..b03bdfc92bb2
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.c
@@ -0,0 +1,335 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2014 Intel Corporation
+ */
+
+#include "i915_drv.h"
+#include "intel_engine_types.h"
+#include "intel_engine_workaround_bb.h"
+#include "intel_execlists_submission.h" /* XXX */
+#include "intel_gpu_commands.h"
+#include "intel_gt.h"
+
+/*
+ * In this WA we need to set GEN8_L3SQCREG4[21:21] and reset it after
+ * PIPE_CONTROL instruction. This is required for the flush to happen correctly
+ * but there is a slight complication as this is applied in WA batch where the
+ * values are only initialized once so we cannot take register value at the
+ * beginning and reuse it further; hence we save its value to memory, upload a
+ * constant value with bit21 set and then we restore it back with the saved value.
+ * To simplify the WA, a constant value is formed by using the default value
+ * of this register. This shouldn't be a problem because we are only modifying
+ * it for a short period and this batch in non-premptible. We can ofcourse
+ * use additional instructions that read the actual value of the register
+ * at that time and set our bit of interest but it makes the WA complicated.
+ *
+ * This WA is also required for Gen9 so extracting as a function avoids
+ * code duplication.
+ */
+static u32 *
+gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
+{
+ /* NB no one else is allowed to scribble over scratch + 256! */
+ *batch++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = intel_gt_scratch_offset(engine->gt,
+ INTEL_GT_SCRATCH_FIELD_COHERENTL3_WA);
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
+
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_DC_FLUSH_ENABLE,
+ 0);
+
+ *batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = intel_gt_scratch_offset(engine->gt,
+ INTEL_GT_SCRATCH_FIELD_COHERENTL3_WA);
+ *batch++ = 0;
+
+ return batch;
+}
+
+/*
+ * Typically we only have one indirect_ctx and per_ctx batch buffer which are
+ * initialized at the beginning and shared across all contexts but this field
+ * helps us to have multiple batches at different offsets and select them based
+ * on a criteria. At the moment this batch always start at the beginning of the page
+ * and at this point we don't have multiple wa_ctx batch buffers.
+ *
+ * The number of WA applied are not known at the beginning; we use this field
+ * to return the no of DWORDS written.
+ *
+ * It is to be noted that this batch does not contain MI_BATCH_BUFFER_END
+ * so it adds NOOPs as padding to make it cacheline aligned.
+ * MI_BATCH_BUFFER_END will be added to perctx batch and both of them together
+ * makes a complete batch buffer.
+ */
+static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
+{
+ /* WaDisableCtxRestoreArbitration:bdw,chv */
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
+
+ /* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw */
+ if (IS_BROADWELL(engine->i915))
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
+
+ /* WaClearSlmSpaceAtContextSwitch:bdw,chv */
+ /* Actual scratch location is at 128 bytes offset */
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_FLUSH_L3 |
+ PIPE_CONTROL_STORE_DATA_INDEX |
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_QW_WRITE,
+ LRC_PPHWSP_SCRATCH_ADDR);
+
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+
+ /* Pad to end of cacheline */
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
+
+ /*
+ * MI_BATCH_BUFFER_END is not required in Indirect ctx BB because
+ * execution depends on the length specified in terms of cache lines
+ * in the register CTX_RCS_INDIRECT_CTX
+ */
+
+ return batch;
+}
+
+struct lri {
+ i915_reg_t reg;
+ u32 value;
+};
+
+static u32 *emit_lri(u32 *batch, const struct lri *lri, unsigned int count)
+{
+ GEM_BUG_ON(!count || count > 63);
+
+ *batch++ = MI_LOAD_REGISTER_IMM(count);
+ do {
+ *batch++ = i915_mmio_reg_offset(lri->reg);
+ *batch++ = lri->value;
+ } while (lri++, --count);
+ *batch++ = MI_NOOP;
+
+ return batch;
+}
+
+static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
+{
+ static const struct lri lri[] = {
+ /* WaDisableGatherAtSetShaderCommonSlice:skl,bxt,kbl,glk */
+ {
+ COMMON_SLICE_CHICKEN2,
+ __MASKED_FIELD(GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE,
+ 0),
+ },
+
+ /* BSpec: 11391 */
+ {
+ FF_SLICE_CHICKEN,
+ __MASKED_FIELD(FF_SLICE_CHICKEN_CL_PROVOKING_VERTEX_FIX,
+ FF_SLICE_CHICKEN_CL_PROVOKING_VERTEX_FIX),
+ },
+
+ /* BSpec: 11299 */
+ {
+ _3D_CHICKEN3,
+ __MASKED_FIELD(_3D_CHICKEN_SF_PROVOKING_VERTEX_FIX,
+ _3D_CHICKEN_SF_PROVOKING_VERTEX_FIX),
+ }
+ };
+
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
+
+ /* WaFlushCoherentL3CacheLinesAtContextSwitch:skl,bxt,glk */
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
+
+ /* WaClearSlmSpaceAtContextSwitch:skl,bxt,kbl,glk,cfl */
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_FLUSH_L3 |
+ PIPE_CONTROL_STORE_DATA_INDEX |
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_QW_WRITE,
+ LRC_PPHWSP_SCRATCH_ADDR);
+
+ batch = emit_lri(batch, lri, ARRAY_SIZE(lri));
+
+ /* WaMediaPoolStateCmdInWABB:bxt,glk */
+ if (HAS_POOLED_EU(engine->i915)) {
+ /*
+ * EU pool configuration is setup along with golden context
+ * during context initialization. This value depends on
+ * device type (2x6 or 3x6) and needs to be updated based
+ * on which subslice is disabled especially for 2x6
+ * devices, however it is safe to load default
+ * configuration of 3x6 device instead of masking off
+ * corresponding bits because HW ignores bits of a disabled
+ * subslice and drops down to appropriate config. Please
+ * see render_state_setup() in i915_gem_render_state.c for
+ * possible configurations, to avoid duplication they are
+ * not shown here again.
+ */
+ *batch++ = GEN9_MEDIA_POOL_STATE;
+ *batch++ = GEN9_MEDIA_POOL_ENABLE;
+ *batch++ = 0x00777000;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+ }
+
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+
+ /* Pad to end of cacheline */
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
+
+ return batch;
+}
+
+static u32 *
+gen10_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
+{
+ int i;
+
+ /*
+ * WaPipeControlBefore3DStateSamplePattern: cnl
+ *
+ * Ensure the engine is idle prior to programming a
+ * 3DSTATE_SAMPLE_PATTERN during a context restore.
+ */
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_CS_STALL,
+ 0);
+ /*
+ * WaPipeControlBefore3DStateSamplePattern says we need 4 dwords for
+ * the PIPE_CONTROL followed by 12 dwords of 0x0, so 16 dwords in
+ * total. However, a PIPE_CONTROL is 6 dwords long, not 4, which is
+ * confusing. Since gen8_emit_pipe_control() already advances the
+ * batch by 6 dwords, we advance the other 10 here, completing a
+ * cacheline. It's not clear if the workaround requires this padding
+ * before other commands, or if it's just the regular padding we would
+ * already have for the workaround bb, so leave it here for now.
+ */
+ for (i = 0; i < 10; i++)
+ *batch++ = MI_NOOP;
+
+ /* Pad to end of cacheline */
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
+
+ return batch;
+}
+
+#define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
+
+static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
+{
+ struct drm_i915_gem_object *obj;
+ struct i915_vma *vma;
+ int err;
+
+ obj = i915_gem_object_create_shmem(engine->i915, CTX_WA_BB_OBJ_SIZE);
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
+
+ vma = i915_vma_instance(obj, &engine->gt->ggtt->vm, NULL);
+ if (IS_ERR(vma)) {
+ err = PTR_ERR(vma);
+ goto err;
+ }
+
+ err = i915_ggtt_pin(vma, NULL, 0, PIN_HIGH);
+ if (err)
+ goto err;
+
+ engine->wa_ctx.vma = vma;
+ return 0;
+
+err:
+ i915_gem_object_put(obj);
+ return err;
+}
+
+typedef u32 *(*wa_bb_func_t)(struct intel_engine_cs *engine, u32 *batch);
+
+int intel_init_workaround_bb(struct intel_engine_cs *engine)
+{
+ struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx;
+ struct i915_wa_ctx_bb *wa_bb[2] = { &wa_ctx->indirect_ctx,
+ &wa_ctx->per_ctx };
+ wa_bb_func_t wa_bb_fn[2];
+ void *batch, *batch_ptr;
+ unsigned int i;
+ int ret;
+
+ if (engine->class != RENDER_CLASS)
+ return 0;
+
+ switch (INTEL_GEN(engine->i915)) {
+ case 12:
+ case 11:
+ return 0;
+ case 10:
+ wa_bb_fn[0] = gen10_init_indirectctx_bb;
+ wa_bb_fn[1] = NULL;
+ break;
+ case 9:
+ wa_bb_fn[0] = gen9_init_indirectctx_bb;
+ wa_bb_fn[1] = NULL;
+ break;
+ case 8:
+ wa_bb_fn[0] = gen8_init_indirectctx_bb;
+ wa_bb_fn[1] = NULL;
+ break;
+ default:
+ MISSING_CASE(INTEL_GEN(engine->i915));
+ return 0;
+ }
+
+ ret = lrc_setup_wa_ctx(engine);
+ if (ret) {
+ drm_dbg(&engine->i915->drm,
+ "Failed to setup context WA page: %d\n", ret);
+ return ret;
+ }
+
+ batch = i915_gem_object_pin_map(wa_ctx->vma->obj, I915_MAP_WB);
+
+ /*
+ * Emit the two workaround batch buffers, recording the offset from the
+ * start of the workaround batch buffer object for each and their
+ * respective sizes.
+ */
+ batch_ptr = batch;
+ for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
+ wa_bb[i]->offset = batch_ptr - batch;
+ if (GEM_DEBUG_WARN_ON(!IS_ALIGNED(wa_bb[i]->offset,
+ CACHELINE_BYTES))) {
+ ret = -EINVAL;
+ break;
+ }
+ if (wa_bb_fn[i])
+ batch_ptr = wa_bb_fn[i](engine, batch_ptr);
+ wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
+ }
+ GEM_BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
+
+ __i915_gem_object_flush_map(wa_ctx->vma->obj, 0, batch_ptr - batch);
+ __i915_gem_object_release_map(wa_ctx->vma->obj);
+ if (ret)
+ intel_fini_workaround_bb(engine);
+
+ return ret;
+}
+
+void intel_fini_workaround_bb(struct intel_engine_cs *engine)
+{
+ i915_vma_unpin_and_release(&engine->wa_ctx.vma, 0);
+}
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.h b/drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.h
new file mode 100644
index 000000000000..88771d77fd42
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_engine_workaround_bb.h
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2014 Intel Corporation
+ */
+
+#ifndef __INTEL_ENGINE_WORKAROUND_BB_H__
+#define __INTEL_ENGINE_WORKAROUND_BB_H__
+
+struct intel_engine_cs;
+
+int intel_init_workaround_bb(struct intel_engine_cs *engine);
+void intel_fini_workaround_bb(struct intel_engine_cs *engine);
+
+#endif /* __INTEL_ENGINE_WORKAROUND_BB_H__ */
diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
index 9069a456d2f7..1cc93ea6b7f0 100644
--- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
+++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
@@ -116,6 +116,7 @@
#include "intel_breadcrumbs.h"
#include "intel_context.h"
#include "intel_engine_pm.h"
+#include "intel_engine_workaround_bb.h"
#include "intel_execlists_submission.h"
#include "intel_gt.h"
#include "intel_gt_pm.h"
@@ -3695,330 +3696,6 @@ static int execlists_request_alloc(struct i915_request *request)
return 0;
}
-/*
- * In this WA we need to set GEN8_L3SQCREG4[21:21] and reset it after
- * PIPE_CONTROL instruction. This is required for the flush to happen correctly
- * but there is a slight complication as this is applied in WA batch where the
- * values are only initialized once so we cannot take register value at the
- * beginning and reuse it further; hence we save its value to memory, upload a
- * constant value with bit21 set and then we restore it back with the saved value.
- * To simplify the WA, a constant value is formed by using the default value
- * of this register. This shouldn't be a problem because we are only modifying
- * it for a short period and this batch in non-premptible. We can ofcourse
- * use additional instructions that read the actual value of the register
- * at that time and set our bit of interest but it makes the WA complicated.
- *
- * This WA is also required for Gen9 so extracting as a function avoids
- * code duplication.
- */
-static u32 *
-gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
-{
- /* NB no one else is allowed to scribble over scratch + 256! */
- *batch++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
- *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
- *batch++ = intel_gt_scratch_offset(engine->gt,
- INTEL_GT_SCRATCH_FIELD_COHERENTL3_WA);
- *batch++ = 0;
-
- *batch++ = MI_LOAD_REGISTER_IMM(1);
- *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
- *batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
-
- batch = gen8_emit_pipe_control(batch,
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_DC_FLUSH_ENABLE,
- 0);
-
- *batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
- *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
- *batch++ = intel_gt_scratch_offset(engine->gt,
- INTEL_GT_SCRATCH_FIELD_COHERENTL3_WA);
- *batch++ = 0;
-
- return batch;
-}
-
-/*
- * Typically we only have one indirect_ctx and per_ctx batch buffer which are
- * initialized at the beginning and shared across all contexts but this field
- * helps us to have multiple batches at different offsets and select them based
- * on a criteria. At the moment this batch always start at the beginning of the page
- * and at this point we don't have multiple wa_ctx batch buffers.
- *
- * The number of WA applied are not known at the beginning; we use this field
- * to return the no of DWORDS written.
- *
- * It is to be noted that this batch does not contain MI_BATCH_BUFFER_END
- * so it adds NOOPs as padding to make it cacheline aligned.
- * MI_BATCH_BUFFER_END will be added to perctx batch and both of them together
- * makes a complete batch buffer.
- */
-static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
-{
- /* WaDisableCtxRestoreArbitration:bdw,chv */
- *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
-
- /* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw */
- if (IS_BROADWELL(engine->i915))
- batch = gen8_emit_flush_coherentl3_wa(engine, batch);
-
- /* WaClearSlmSpaceAtContextSwitch:bdw,chv */
- /* Actual scratch location is at 128 bytes offset */
- batch = gen8_emit_pipe_control(batch,
- PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_STORE_DATA_INDEX |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE,
- LRC_PPHWSP_SCRATCH_ADDR);
-
- *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
-
- /* Pad to end of cacheline */
- while ((unsigned long)batch % CACHELINE_BYTES)
- *batch++ = MI_NOOP;
-
- /*
- * MI_BATCH_BUFFER_END is not required in Indirect ctx BB because
- * execution depends on the length specified in terms of cache lines
- * in the register CTX_RCS_INDIRECT_CTX
- */
-
- return batch;
-}
-
-struct lri {
- i915_reg_t reg;
- u32 value;
-};
-
-static u32 *emit_lri(u32 *batch, const struct lri *lri, unsigned int count)
-{
- GEM_BUG_ON(!count || count > 63);
-
- *batch++ = MI_LOAD_REGISTER_IMM(count);
- do {
- *batch++ = i915_mmio_reg_offset(lri->reg);
- *batch++ = lri->value;
- } while (lri++, --count);
- *batch++ = MI_NOOP;
-
- return batch;
-}
-
-static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
-{
- static const struct lri lri[] = {
- /* WaDisableGatherAtSetShaderCommonSlice:skl,bxt,kbl,glk */
- {
- COMMON_SLICE_CHICKEN2,
- __MASKED_FIELD(GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE,
- 0),
- },
-
- /* BSpec: 11391 */
- {
- FF_SLICE_CHICKEN,
- __MASKED_FIELD(FF_SLICE_CHICKEN_CL_PROVOKING_VERTEX_FIX,
- FF_SLICE_CHICKEN_CL_PROVOKING_VERTEX_FIX),
- },
-
- /* BSpec: 11299 */
- {
- _3D_CHICKEN3,
- __MASKED_FIELD(_3D_CHICKEN_SF_PROVOKING_VERTEX_FIX,
- _3D_CHICKEN_SF_PROVOKING_VERTEX_FIX),
- }
- };
-
- *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
-
- /* WaFlushCoherentL3CacheLinesAtContextSwitch:skl,bxt,glk */
- batch = gen8_emit_flush_coherentl3_wa(engine, batch);
-
- /* WaClearSlmSpaceAtContextSwitch:skl,bxt,kbl,glk,cfl */
- batch = gen8_emit_pipe_control(batch,
- PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_STORE_DATA_INDEX |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE,
- LRC_PPHWSP_SCRATCH_ADDR);
-
- batch = emit_lri(batch, lri, ARRAY_SIZE(lri));
-
- /* WaMediaPoolStateCmdInWABB:bxt,glk */
- if (HAS_POOLED_EU(engine->i915)) {
- /*
- * EU pool configuration is setup along with golden context
- * during context initialization. This value depends on
- * device type (2x6 or 3x6) and needs to be updated based
- * on which subslice is disabled especially for 2x6
- * devices, however it is safe to load default
- * configuration of 3x6 device instead of masking off
- * corresponding bits because HW ignores bits of a disabled
- * subslice and drops down to appropriate config. Please
- * see render_state_setup() in i915_gem_render_state.c for
- * possible configurations, to avoid duplication they are
- * not shown here again.
- */
- *batch++ = GEN9_MEDIA_POOL_STATE;
- *batch++ = GEN9_MEDIA_POOL_ENABLE;
- *batch++ = 0x00777000;
- *batch++ = 0;
- *batch++ = 0;
- *batch++ = 0;
- }
-
- *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
-
- /* Pad to end of cacheline */
- while ((unsigned long)batch % CACHELINE_BYTES)
- *batch++ = MI_NOOP;
-
- return batch;
-}
-
-static u32 *
-gen10_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
-{
- int i;
-
- /*
- * WaPipeControlBefore3DStateSamplePattern: cnl
- *
- * Ensure the engine is idle prior to programming a
- * 3DSTATE_SAMPLE_PATTERN during a context restore.
- */
- batch = gen8_emit_pipe_control(batch,
- PIPE_CONTROL_CS_STALL,
- 0);
- /*
- * WaPipeControlBefore3DStateSamplePattern says we need 4 dwords for
- * the PIPE_CONTROL followed by 12 dwords of 0x0, so 16 dwords in
- * total. However, a PIPE_CONTROL is 6 dwords long, not 4, which is
- * confusing. Since gen8_emit_pipe_control() already advances the
- * batch by 6 dwords, we advance the other 10 here, completing a
- * cacheline. It's not clear if the workaround requires this padding
- * before other commands, or if it's just the regular padding we would
- * already have for the workaround bb, so leave it here for now.
- */
- for (i = 0; i < 10; i++)
- *batch++ = MI_NOOP;
-
- /* Pad to end of cacheline */
- while ((unsigned long)batch % CACHELINE_BYTES)
- *batch++ = MI_NOOP;
-
- return batch;
-}
-
-#define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
-
-static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
-{
- struct drm_i915_gem_object *obj;
- struct i915_vma *vma;
- int err;
-
- obj = i915_gem_object_create_shmem(engine->i915, CTX_WA_BB_OBJ_SIZE);
- if (IS_ERR(obj))
- return PTR_ERR(obj);
-
- vma = i915_vma_instance(obj, &engine->gt->ggtt->vm, NULL);
- if (IS_ERR(vma)) {
- err = PTR_ERR(vma);
- goto err;
- }
-
- err = i915_ggtt_pin(vma, NULL, 0, PIN_HIGH);
- if (err)
- goto err;
-
- engine->wa_ctx.vma = vma;
- return 0;
-
-err:
- i915_gem_object_put(obj);
- return err;
-}
-
-static void lrc_destroy_wa_ctx(struct intel_engine_cs *engine)
-{
- i915_vma_unpin_and_release(&engine->wa_ctx.vma, 0);
-}
-
-typedef u32 *(*wa_bb_func_t)(struct intel_engine_cs *engine, u32 *batch);
-
-static int intel_init_workaround_bb(struct intel_engine_cs *engine)
-{
- struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx;
- struct i915_wa_ctx_bb *wa_bb[2] = { &wa_ctx->indirect_ctx,
- &wa_ctx->per_ctx };
- wa_bb_func_t wa_bb_fn[2];
- void *batch, *batch_ptr;
- unsigned int i;
- int ret;
-
- if (engine->class != RENDER_CLASS)
- return 0;
-
- switch (INTEL_GEN(engine->i915)) {
- case 12:
- case 11:
- return 0;
- case 10:
- wa_bb_fn[0] = gen10_init_indirectctx_bb;
- wa_bb_fn[1] = NULL;
- break;
- case 9:
- wa_bb_fn[0] = gen9_init_indirectctx_bb;
- wa_bb_fn[1] = NULL;
- break;
- case 8:
- wa_bb_fn[0] = gen8_init_indirectctx_bb;
- wa_bb_fn[1] = NULL;
- break;
- default:
- MISSING_CASE(INTEL_GEN(engine->i915));
- return 0;
- }
-
- ret = lrc_setup_wa_ctx(engine);
- if (ret) {
- drm_dbg(&engine->i915->drm,
- "Failed to setup context WA page: %d\n", ret);
- return ret;
- }
-
- batch = i915_gem_object_pin_map(wa_ctx->vma->obj, I915_MAP_WB);
-
- /*
- * Emit the two workaround batch buffers, recording the offset from the
- * start of the workaround batch buffer object for each and their
- * respective sizes.
- */
- batch_ptr = batch;
- for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
- wa_bb[i]->offset = batch_ptr - batch;
- if (GEM_DEBUG_WARN_ON(!IS_ALIGNED(wa_bb[i]->offset,
- CACHELINE_BYTES))) {
- ret = -EINVAL;
- break;
- }
- if (wa_bb_fn[i])
- batch_ptr = wa_bb_fn[i](engine, batch_ptr);
- wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
- }
- GEM_BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
-
- __i915_gem_object_flush_map(wa_ctx->vma->obj, 0, batch_ptr - batch);
- __i915_gem_object_release_map(wa_ctx->vma->obj);
- if (ret)
- lrc_destroy_wa_ctx(engine);
-
- return ret;
-}
-
static void reset_csb_pointers(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -4707,7 +4384,7 @@ static void execlists_release(struct intel_engine_cs *engine)
execlists_shutdown(engine);
intel_engine_cleanup_common(engine);
- lrc_destroy_wa_ctx(engine);
+ intel_fini_workaround_bb(engine);
}
static void
--
2.26.2
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
next prev parent reply other threads:[~2020-11-27 12:08 UTC|newest]
Thread overview: 208+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-11-27 12:04 [RFC PATCH 000/162] DG1 + LMEM enabling Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 001/162] drm/i915/selftest: also consider non-contiguous objects Matthew Auld
2020-11-27 19:44 ` Chris Wilson
2020-11-27 12:04 ` [RFC PATCH 002/162] drm/i915/selftest: assert we get 2M GTT pages Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 003/162] drm/i915/selftest: handle local-memory in perf_memcpy Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 004/162] drm/i915/gt: Move move context layout registers and offsets to lrc_reg.h Matthew Auld
2020-11-27 19:55 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:04 ` [RFC PATCH 005/162] drm/i915/gt: Rename lrc.c to execlists_submission.c Matthew Auld
2020-11-27 19:56 ` Chris Wilson
2020-11-27 12:04 ` [RFC PATCH 006/162] drm/i915: split gen8+ flush and bb_start emission functions to their own file Matthew Auld
2020-11-27 19:58 ` Chris Wilson
2020-11-27 12:04 ` Matthew Auld [this message]
2020-11-27 12:04 ` [RFC PATCH 008/162] HAX drm/i915: Work around the selftest timeline lock splat workaround Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 009/162] drm/i915: Introduce drm_i915_lock_isolated Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 010/162] drm/i915: Lock hwsp objects isolated for pinning at create time Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 011/162] drm/i915: Pin timeline map after first timeline pin, v5 Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 012/162] drm/i915: Move cmd parser pinning to execbuffer Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 013/162] drm/i915: Add missing -EDEADLK handling to execbuf pinning, v2 Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 014/162] drm/i915: Ensure we hold the object mutex in pin correctly v2 Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 015/162] drm/i915: Add gem object locking to madvise Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 016/162] drm/i915: Move HAS_STRUCT_PAGE to obj->flags Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 017/162] drm/i915: Rework struct phys attachment handling Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 018/162] drm/i915: Convert i915_gem_object_attach_phys() to ww locking, v2 Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 019/162] drm/i915: make lockdep slightly happier about execbuf Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 020/162] drm/i915: Disable userptr pread/pwrite support Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 021/162] drm/i915: No longer allow exporting userptr through dma-buf Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 022/162] drm/i915: Reject more ioctls for userptr Matthew Auld
2020-11-27 12:04 ` [RFC PATCH 023/162] drm/i915: Reject UNSYNCHRONIZED for userptr, v2 Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 024/162] drm/i915: Make compilation of userptr code depend on MMU_NOTIFIER Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 025/162] drm/i915: Fix userptr so we do not have to worry about obj->mm.lock, v5 Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 026/162] drm/i915: Flatten obj->mm.lock Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 027/162] drm/i915: Populate logical context during first pin Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 028/162] drm/i915: Make ring submission compatible with obj->mm.lock removal, v2 Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 029/162] drm/i915: Handle ww locking in init_status_page Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 030/162] drm/i915: Rework clflush to work correctly without obj->mm.lock Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 031/162] drm/i915: Pass ww ctx to intel_pin_to_display_plane Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 032/162] drm/i915: Add object locking to vm_fault_cpu Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 033/162] drm/i915: Move pinning to inside engine_wa_list_verify() Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 034/162] drm/i915: Take reservation lock around i915_vma_pin Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 035/162] drm/i915: Make intel_init_workaround_bb more compatible with ww locking Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 036/162] drm/i915: Make __engine_unpark() compatible with ww locking v2 Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 037/162] drm/i915: Take obj lock around set_domain ioctl Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 038/162] drm/i915: Defer pin calls in buffer pool until first use by caller Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 039/162] drm/i915: Fix pread/pwrite to work with new locking rules Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 040/162] drm/i915: Fix workarounds selftest, part 1 Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 041/162] drm/i915: Prepare for obj->mm.lock removal Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 042/162] drm/i915: Add igt_spinner_pin() to allow for ww locking around spinner Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 043/162] drm/i915: Add ww locking around vm_access() Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 044/162] drm/i915: Increase ww locking for perf Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 045/162] drm/i915: Lock ww in ucode objects correctly Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 046/162] drm/i915: Add ww locking to dma-buf ops Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 047/162] drm/i915: Add missing ww lock in intel_dsb_prepare Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 048/162] drm/i915: Fix ww locking in shmem_create_from_object Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 049/162] drm/i915: Use a single page table lock for each gtt Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 050/162] drm/i915/selftests: Prepare huge_pages testcases for obj->mm.lock removal Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 051/162] drm/i915/selftests: Prepare client blit " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 052/162] drm/i915/selftests: Prepare coherency tests " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 053/162] drm/i915/selftests: Prepare context " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 054/162] drm/i915/selftests: Prepare dma-buf " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 055/162] drm/i915/selftests: Prepare execbuf " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 056/162] drm/i915/selftests: Prepare mman testcases " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 057/162] drm/i915/selftests: Prepare object tests " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 058/162] drm/i915/selftests: Prepare object blit " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 059/162] drm/i915/selftests: Prepare igt_gem_utils " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 060/162] drm/i915/selftests: Prepare context selftest " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 061/162] drm/i915/selftests: Prepare hangcheck " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 062/162] drm/i915/selftests: Prepare execlists " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 063/162] drm/i915/selftests: Prepare mocs tests " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 064/162] drm/i915/selftests: Prepare ring submission " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 065/162] drm/i915/selftests: Prepare timeline tests " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 066/162] drm/i915/selftests: Prepare i915_request " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 067/162] drm/i915/selftests: Prepare memory region " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 068/162] drm/i915/selftests: Prepare cs engine " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 069/162] drm/i915/selftests: Prepare gtt " Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 070/162] drm/i915: Finally remove obj->mm.lock Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 071/162] drm/i915: Keep userpointer bindings if seqcount is unchanged, v2 Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 072/162] drm/i915: Avoid some false positives in assert_object_held() Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 073/162] drm/i915: Reference contending lock objects Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 074/162] drm/i915: Break out dma_resv ww locking utilities to separate files Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 075/162] drm/i915: Introduce a for_i915_gem_ww(){} Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 076/162] drm/i915: Untangle the vma pages_mutex Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 077/162] drm/i915/fbdev: Use lmem physical addresses for fb_mmap() on discrete Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 078/162] drm/i915: Return error value when bo not in LMEM for discrete Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 079/162] drm/i915/dmabuf: Disallow LMEM objects from dma-buf Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 080/162] drm/i915/lmem: Fail driver init if LMEM training failed Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 081/162] HAX drm/i915/lmem: support CPU relocations Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 082/162] HAX drm/i915/lmem: support pread and pwrite Matthew Auld
2020-11-27 12:05 ` [RFC PATCH 083/162] drm/i915: Update the helper to set correct mapping Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 084/162] drm/i915: introduce kernel blitter_context Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 085/162] drm/i915/region: support basic eviction Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 086/162] drm/i915: Add blit functions that can be called from within a WW transaction Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 087/162] drm/i915: Delay publishing objects on the eviction lists Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 088/162] drm/i915: support basic object migration Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 089/162] drm/i915/dg1: Fix occasional migration error Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 090/162] drm/i915/query: Expose memory regions through the query uAPI Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 091/162] drm/i915: Store gt in memory region Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 092/162] drm/i915/uapi: introduce drm_i915_gem_create_ext Matthew Auld
2020-11-27 13:25 ` [Intel-gfx] " Chris Wilson
2020-12-01 15:06 ` Thomas Hellström (Intel)
2020-11-27 19:21 ` Chris Wilson
2020-12-01 12:55 ` Chris Wilson
2020-12-01 13:43 ` Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 093/162] drm/i915/lmem: allocate cmd ring in lmem Matthew Auld
2020-11-27 13:27 ` Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 094/162] drm/i915/dg1: Do not check r->sgt.pfn for NULL Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 095/162] drm/i915/dg1: Introduce dmabuf mmap to LMEM Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 096/162] drm/i915: setup the LMEM region Matthew Auld
2020-11-30 10:14 ` Jani Nikula
2020-11-27 12:06 ` [RFC PATCH 097/162] drm/i915: Distinction of memory regions Matthew Auld
2020-11-27 13:30 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 098/162] drm/i915/gtt: map the PD up front Matthew Auld
2020-11-27 13:31 ` Chris Wilson
2021-01-12 10:47 ` [Intel-gfx] " Matthew Auld
2021-01-12 14:33 ` Daniel Vetter
2020-11-27 12:06 ` [RFC PATCH 099/162] drm/i915/gtt/dgfx: place the PD in LMEM Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 100/162] drm/i915/gtt: make flushing conditional Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 101/162] drm/i915/gtt/dg1: add PTE_LM plumbing for PPGTT Matthew Auld
2020-11-27 13:35 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 102/162] drm/i915/gtt/dg1: add PTE_LM plumbing for GGTT Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 103/162] drm/i915: allocate context from LMEM Matthew Auld
2020-11-27 13:37 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 104/162] drm/i915: move engine scratch to LMEM Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 105/162] drm/i915: Provide a way to disable PCIe relaxed write ordering Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 106/162] drm/i915: i915 returns -EBUSY on thread contention Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 107/162] drm/i915: setup GPU device lmem region Matthew Auld
2020-11-30 11:18 ` Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 108/162] drm/i915: Fix object page offset within a region Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 109/162] drm/i915: add i915_gem_object_is_devmem() function Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 110/162] drm/i915: finish memory region support for stolen objects Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 111/162] drm/i915/lmem: support optional CPU clearing for special internal use Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 112/162] drm/i915/guc: put all guc objects in lmem when available Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 113/162] drm/i915: Create stolen memory region from local memory Matthew Auld
2020-12-07 13:39 ` [Intel-gfx] " Jani Nikula
2020-11-27 12:06 ` [RFC PATCH 114/162] drm/i915/lmem: Bypass aperture when lmem is available Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 115/162] drm/i915/lmem: reset the lmem buffer created by fbdev Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 116/162] drm/i915/dsb: Enable lmem for dsb Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 117/162] drm/i915: Reintroduce mem->reserved Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 118/162] drm/i915/dg1: Reserve first 1MB of local memory Matthew Auld
2020-11-27 13:52 ` [Intel-gfx] " Chris Wilson
2020-11-30 11:09 ` Matthew Auld
2020-11-30 11:22 ` Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 119/162] drm/i915/dg1: Read OPROM via SPI controller Matthew Auld
2020-11-30 10:16 ` [Intel-gfx] " Jani Nikula
2020-11-27 12:06 ` [RFC PATCH 120/162] drm/i915/oprom: Basic sanitization Matthew Auld
2020-11-30 10:24 ` [Intel-gfx] " Jani Nikula
2020-11-27 12:06 ` [RFC PATCH 121/162] drm/i915: WA for zero memory channel Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 122/162] drm/i915/dg1: Compute MEM Bandwidth using MCHBAR Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 123/162] drm/i915/dg1: Double memory bandwidth available Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 124/162] drm/i915/lmem: allocate HWSP in lmem Matthew Auld
2020-11-27 13:55 ` [Intel-gfx] " Chris Wilson
2020-11-30 17:17 ` Matthew Auld
2020-11-30 17:35 ` Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 125/162] drm/i915/lmem: Limit block size to 4G Matthew Auld
2020-11-27 14:02 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 126/162] drm/i915/gem: Update shmem available memory Matthew Auld
2020-11-27 14:04 ` Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 127/162] drm/i915: Allow non-uniform subslices in gen12+ Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 128/162] drm/i915/dg1: intel_memory_region_evict() changes for eviction Matthew Auld
2020-11-27 14:07 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 129/162] drm/i915/dg1: i915_gem_object_memcpy(..) infrastructure Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 130/162] drm/i915/dg1: Eviction logic Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 131/162] drm/i915/dg1: Add enable_eviction modparam Matthew Auld
2020-11-30 12:20 ` Jani Nikula
2020-11-27 12:06 ` [RFC PATCH 132/162] drm/i915/dg1: Add lmem_size modparam Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 133/162] drm/i915/dg1: Track swap in/out stats via debugfs Matthew Auld
2020-11-27 14:09 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 134/162] drm/i915/dg1: Measure swap in/out timing stats Matthew Auld
2020-11-27 14:11 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 135/162] drm/i915: define intel_partial_pages_for_sg_table Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 136/162] drm/i915: create and destroy dummy vma Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 137/162] drm/i915: blt copy between objs using pre-created vma windows Matthew Auld
2020-11-27 14:19 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 138/162] drm/i915/dg1: Eliminate eviction mutex Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 139/162] drm/i915/dg1: Keep engine awake across whole blit Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 140/162] drm/i915: window_blt_copy is used for swapin and swapout Matthew Auld
2020-11-27 14:20 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 141/162] drm/i915: Lmem eviction statistics by category Matthew Auld
2020-11-27 14:21 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:06 ` [RFC PATCH 142/162] drm/i915/gem/selftest: test and measure window based blt cpy Matthew Auld
2020-11-27 12:06 ` [RFC PATCH 143/162] drm/i915: suspend/resume eviction Matthew Auld
2020-11-27 14:22 ` Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 144/162] drm/i915: Reset blitter context when unpark engine Matthew Auld
2020-11-27 14:26 ` Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 145/162] drm/i915/dg1: Add dedicated context for blitter eviction Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 146/162] drm/i915/pm: suspend and restore ppgtt mapping Matthew Auld
2020-11-27 14:29 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 147/162] drm/i915/gt: Allocate default ctx objects in SMEM Matthew Auld
2020-11-27 14:30 ` Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 148/162] drm/i915: suspend/resume enable blitter eviction Matthew Auld
2020-11-27 14:32 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 149/162] drm/i915: suspend/resume handling of perma-pinned objects Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 150/162] drm/i915: need consider system BO snoop for dgfx Matthew Auld
2020-11-27 14:36 ` Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 151/162] drm/i915: move eviction to prepare hook Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 152/162] drm/i915: Perform execbuffer object locking as a separate step Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 153/162] drm/i915: Implement eviction locking v2 Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 154/162] drm/i915: Support ww eviction Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 155/162] drm/i915: Use a ww transaction in the fault handler Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 156/162] drm/i915: Use a ww transaction in i915_gem_object_pin_map_unlocked() Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 157/162] drm/i915: Improve accuracy of eviction stats Matthew Auld
2020-11-27 14:40 ` [Intel-gfx] " Chris Wilson
2020-11-30 10:36 ` Tvrtko Ursulin
2020-11-27 12:07 ` [RFC PATCH 158/162] drm/i915: Support ww locks in suspend/resume Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 159/162] drm/i915/dg1: Fix mapping type for default state object Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 160/162] drm/i915/dg1: Fix GPU hang due to shmemfs page drop Matthew Auld
2020-11-27 14:44 ` [Intel-gfx] " Chris Wilson
2020-11-27 12:07 ` [RFC PATCH 161/162] drm/i915/dg1: allow pci to auto probe Matthew Auld
2020-11-27 12:07 ` [RFC PATCH 162/162] drm/i915: drop fake lmem Matthew Auld
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20201127120718.454037-8-matthew.auld@intel.com \
--to=matthew.auld@intel.com \
--cc=John.C.Harrison@Intel.com \
--cc=chris.p.wilson@intel.com \
--cc=daniele.ceraolospurio@intel.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.org \
--cc=tvrtko.ursulin@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).