[Mesa-dev] [PATCH 4/5] anv: Port L3 cache programming from i965
Jordan Justen
jordan.l.justen at intel.com
Wed May 11 23:22:54 UTC 2016
Signed-off-by: Jordan Justen <jordan.l.justen at intel.com>
---
src/intel/vulkan/Makefile.sources | 4 +
src/intel/vulkan/anv_genX.h | 4 +-
src/intel/vulkan/anv_pipeline.c | 33 ++-
src/intel/vulkan/anv_private.h | 12 +-
src/intel/vulkan/gen7_cmd_buffer.c | 94 +------
src/intel/vulkan/gen8_cmd_buffer.c | 74 +----
src/intel/vulkan/genX_cmd_buffer.c | 2 +-
src/intel/vulkan/genX_l3.c | 541 +++++++++++++++++++++++++++++++++++++
src/intel/vulkan/genX_pipeline.c | 2 +
9 files changed, 593 insertions(+), 173 deletions(-)
create mode 100644 src/intel/vulkan/genX_l3.c
diff --git a/src/intel/vulkan/Makefile.sources b/src/intel/vulkan/Makefile.sources
index 182c1e1..6c6b29d 100644
--- a/src/intel/vulkan/Makefile.sources
+++ b/src/intel/vulkan/Makefile.sources
@@ -70,6 +70,7 @@ VULKAN_GENERATED_FILES := \
GEN7_FILES := \
genX_cmd_buffer.c \
+ genX_l3.c \
genX_pipeline.c \
gen7_cmd_buffer.c \
gen7_pipeline.c \
@@ -77,6 +78,7 @@ GEN7_FILES := \
GEN75_FILES := \
genX_cmd_buffer.c \
+ genX_l3.c \
genX_pipeline.c \
gen7_cmd_buffer.c \
gen7_pipeline.c \
@@ -84,6 +86,7 @@ GEN75_FILES := \
GEN8_FILES := \
genX_cmd_buffer.c \
+ genX_l3.c \
genX_pipeline.c \
gen8_cmd_buffer.c \
gen8_pipeline.c \
@@ -91,6 +94,7 @@ GEN8_FILES := \
GEN9_FILES := \
genX_cmd_buffer.c \
+ genX_l3.c \
genX_pipeline.c \
gen8_cmd_buffer.c \
gen8_pipeline.c \
diff --git a/src/intel/vulkan/anv_genX.h b/src/intel/vulkan/anv_genX.h
index 908a9e0..a5ec27d 100644
--- a/src/intel/vulkan/anv_genX.h
+++ b/src/intel/vulkan/anv_genX.h
@@ -42,8 +42,10 @@ void genX(cmd_buffer_set_subpass)(struct anv_cmd_buffer *cmd_buffer,
void genX(flush_pipeline_select_3d)(struct anv_cmd_buffer *cmd_buffer);
void genX(flush_pipeline_select_gpgpu)(struct anv_cmd_buffer *cmd_buffer);
+void genX(setup_pipeline_l3_config)(struct anv_pipeline *pipeline);
+
void genX(cmd_buffer_config_l3)(struct anv_cmd_buffer *cmd_buffer,
- bool enable_slm);
+ const struct anv_pipeline *pipeline);
void genX(cmd_buffer_flush_state)(struct anv_cmd_buffer *cmd_buffer);
void genX(cmd_buffer_flush_dynamic_state)(struct anv_cmd_buffer *cmd_buffer);
diff --git a/src/intel/vulkan/anv_pipeline.c b/src/intel/vulkan/anv_pipeline.c
index fcaa450..b774e0c 100644
--- a/src/intel/vulkan/anv_pipeline.c
+++ b/src/intel/vulkan/anv_pipeline.c
@@ -801,10 +801,34 @@ anv_pipeline_compile_cs(struct anv_pipeline *pipeline,
return VK_SUCCESS;
}
-static void
-gen7_compute_urb_partition(struct anv_pipeline *pipeline)
+
+void
+anv_setup_pipeline_l3_config(struct anv_pipeline *pipeline)
+{
+ const struct brw_device_info *devinfo = &pipeline->device->info;
+ switch (devinfo->gen) {
+ case 7:
+ if (devinfo->is_haswell)
+ gen75_setup_pipeline_l3_config(pipeline);
+ else
+ gen7_setup_pipeline_l3_config(pipeline);
+ break;
+ case 8:
+ gen8_setup_pipeline_l3_config(pipeline);
+ break;
+ case 9:
+ gen9_setup_pipeline_l3_config(pipeline);
+ break;
+ default:
+ unreachable("unsupported gen\n");
+ }
+}
+
+void
+anv_compute_urb_partition(struct anv_pipeline *pipeline)
{
const struct brw_device_info *devinfo = &pipeline->device->info;
+
bool vs_present = pipeline->active_stages & VK_SHADER_STAGE_VERTEX_BIT;
unsigned vs_size = vs_present ?
get_vs_prog_data(pipeline)->base.urb_entry_size : 1;
@@ -828,7 +852,7 @@ gen7_compute_urb_partition(struct anv_pipeline *pipeline)
unsigned chunk_size_bytes = 8192;
/* Determine the size of the URB in chunks. */
- unsigned urb_chunks = devinfo->urb.size * 1024 / chunk_size_bytes;
+ unsigned urb_chunks = pipeline->urb.total_size * 1024 / chunk_size_bytes;
/* Reserve space for push constants */
unsigned push_constant_kb;
@@ -1196,7 +1220,8 @@ anv_pipeline_init(struct anv_pipeline *pipeline,
assert(extra->disable_vs);
}
- gen7_compute_urb_partition(pipeline);
+ anv_setup_pipeline_l3_config(pipeline);
+ anv_compute_urb_partition(pipeline);
const VkPipelineVertexInputStateCreateInfo *vi_info =
pCreateInfo->pVertexInputState;
diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_private.h
index fb308eb..8d7a5ae 100644
--- a/src/intel/vulkan/anv_private.h
+++ b/src/intel/vulkan/anv_private.h
@@ -52,6 +52,8 @@ typedef struct xcb_connection_t xcb_connection_t;
typedef uint32_t xcb_visualid_t;
typedef uint32_t xcb_window_t;
+struct anv_l3_config;
+
#define VK_PROTOTYPES
#include <vulkan/vulkan.h>
#include <vulkan/vulkan_intel.h>
@@ -1158,7 +1160,7 @@ struct anv_attachment_state {
struct anv_cmd_state {
/* PIPELINE_SELECT.PipelineSelection */
uint32_t current_pipeline;
- uint32_t current_l3_config;
+ const struct anv_l3_config * current_l3_config;
uint32_t vb_dirty;
anv_cmd_dirty_mask_t dirty;
anv_cmd_dirty_mask_t compute_dirty;
@@ -1413,6 +1415,8 @@ struct anv_pipeline {
uint32_t start[MESA_SHADER_GEOMETRY + 1];
uint32_t size[MESA_SHADER_GEOMETRY + 1];
uint32_t entries[MESA_SHADER_GEOMETRY + 1];
+ const struct anv_l3_config * l3_config;
+ uint32_t total_size;
} urb;
VkShaderStageFlags active_stages;
@@ -1548,6 +1552,12 @@ anv_format_is_depth_or_stencil(const struct anv_format *format)
return format->has_depth || format->has_stencil;
}
+void
+anv_compute_urb_partition(struct anv_pipeline *pipeline);
+
+void
+anv_setup_pipeline_l3_config(struct anv_pipeline *pipeline);
+
/**
* Subsurface of an anv_image.
*/
diff --git a/src/intel/vulkan/gen7_cmd_buffer.c b/src/intel/vulkan/gen7_cmd_buffer.c
index 479790e..437779a 100644
--- a/src/intel/vulkan/gen7_cmd_buffer.c
+++ b/src/intel/vulkan/gen7_cmd_buffer.c
@@ -284,97 +284,6 @@ flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer)
}
void
-genX(cmd_buffer_config_l3)(struct anv_cmd_buffer *cmd_buffer, bool enable_slm)
-{
- /* References for GL state:
- *
- * - commits e307cfa..228d5a3
- * - src/mesa/drivers/dri/i965/gen7_l3_state.c
- */
-
- uint32_t l3cr2_slm, l3cr2_noslm;
- anv_pack_struct(&l3cr2_noslm, GENX(L3CNTLREG2),
- .URBAllocation = 24,
- .ROAllocation = 0,
- .DCAllocation = 16);
- anv_pack_struct(&l3cr2_slm, GENX(L3CNTLREG2),
- .SLMEnable = 1,
- .URBAllocation = 16,
- .URBLowBandwidth = 1,
- .ROAllocation = 0,
- .DCAllocation = 8);
- const uint32_t l3cr2_val = enable_slm ? l3cr2_slm : l3cr2_noslm;
- bool changed = cmd_buffer->state.current_l3_config != l3cr2_val;
-
- if (changed) {
- /* According to the hardware docs, the L3 partitioning can only be
- * changed while the pipeline is completely drained and the caches are
- * flushed, which involves a first PIPE_CONTROL flush which stalls the
- * pipeline...
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
- pc.DCFlushEnable = true;
- pc.CommandStreamerStallEnable = true;
- pc.PostSyncOperation = NoWrite;
- }
-
- /* ...followed by a second pipelined PIPE_CONTROL that initiates
- * invalidation of the relevant caches. Note that because RO
- * invalidation happens at the top of the pipeline (i.e. right away as
- * the PIPE_CONTROL command is processed by the CS) we cannot combine it
- * with the previous stalling flush as the hardware documentation
- * suggests, because that would cause the CS to stall on previous
- * rendering *after* RO invalidation and wouldn't prevent the RO caches
- * from being polluted by concurrent rendering before the stall
- * completes. This intentionally doesn't implement the SKL+ hardware
- * workaround suggesting to enable CS stall on PIPE_CONTROLs with the
- * texture cache invalidation bit set for GPGPU workloads because the
- * previous and subsequent PIPE_CONTROLs already guarantee that there is
- * no concurrent GPGPU kernel execution (see SKL HSD 2132585).
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
- pc.TextureCacheInvalidationEnable = true;
- pc.ConstantCacheInvalidationEnable = true;
- pc.InstructionCacheInvalidateEnable = true;
- pc.StateCacheInvalidationEnable = true;
- pc.PostSyncOperation = NoWrite;
- }
-
- /* Now send a third stalling flush to make sure that invalidation is
- * complete when the L3 configuration registers are modified.
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
- pc.DCFlushEnable = true;
- pc.CommandStreamerStallEnable = true;
- pc.PostSyncOperation = NoWrite;
- }
-
- anv_finishme("write GEN7_L3SQCREG1");
- anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
- lri.RegisterOffset = GENX(L3CNTLREG2_num);
- lri.DataDWord = l3cr2_val;
- }
-
- uint32_t l3cr3_slm, l3cr3_noslm;
- anv_pack_struct(&l3cr3_noslm, GENX(L3CNTLREG3),
- .ISAllocation = 8,
- .CAllocation = 4,
- .TAllocation = 8);
- anv_pack_struct(&l3cr3_slm, GENX(L3CNTLREG3),
- .ISAllocation = 8,
- .CAllocation = 8,
- .TAllocation = 8);
- const uint32_t l3cr3_val = enable_slm ? l3cr3_slm : l3cr3_noslm;
- anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
- lri.RegisterOffset = GENX(L3CNTLREG3_num);
- lri.DataDWord = l3cr3_val;
- }
-
- cmd_buffer->state.current_l3_config = l3cr2_val;
- }
-}
-
-void
genX(cmd_buffer_flush_compute_state)(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline;
@@ -383,8 +292,7 @@ genX(cmd_buffer_flush_compute_state)(struct anv_cmd_buffer *cmd_buffer)
assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT);
- bool needs_slm = cs_prog_data->base.total_shared > 0;
- genX(cmd_buffer_config_l3)(cmd_buffer, needs_slm);
+ genX(cmd_buffer_config_l3)(cmd_buffer, pipeline);
genX(flush_pipeline_select_gpgpu)(cmd_buffer);
diff --git a/src/intel/vulkan/gen8_cmd_buffer.c b/src/intel/vulkan/gen8_cmd_buffer.c
index d078dac..693d2a6 100644
--- a/src/intel/vulkan/gen8_cmd_buffer.c
+++ b/src/intel/vulkan/gen8_cmd_buffer.c
@@ -91,77 +91,6 @@ gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer)
}
#endif
-void
-genX(cmd_buffer_config_l3)(struct anv_cmd_buffer *cmd_buffer, bool enable_slm)
-{
- /* References for GL state:
- *
- * - commits e307cfa..228d5a3
- * - src/mesa/drivers/dri/i965/gen7_l3_state.c
- */
-
- uint32_t l3cr_slm, l3cr_noslm;
- anv_pack_struct(&l3cr_noslm, GENX(L3CNTLREG),
- .URBAllocation = 48,
- .AllAllocation = 48);
- anv_pack_struct(&l3cr_slm, GENX(L3CNTLREG),
- .SLMEnable = 1,
- .URBAllocation = 16,
- .AllAllocation = 48);
- const uint32_t l3cr_val = enable_slm ? l3cr_slm : l3cr_noslm;
- bool changed = cmd_buffer->state.current_l3_config != l3cr_val;
-
- if (changed) {
- /* According to the hardware docs, the L3 partitioning can only be
- * changed while the pipeline is completely drained and the caches are
- * flushed, which involves a first PIPE_CONTROL flush which stalls the
- * pipeline...
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
- pc.DCFlushEnable = true;
- pc.PostSyncOperation = NoWrite;
- pc.CommandStreamerStallEnable = true;
- }
-
- /* ...followed by a second pipelined PIPE_CONTROL that initiates
- * invalidation of the relevant caches. Note that because RO
- * invalidation happens at the top of the pipeline (i.e. right away as
- * the PIPE_CONTROL command is processed by the CS) we cannot combine it
- * with the previous stalling flush as the hardware documentation
- * suggests, because that would cause the CS to stall on previous
- * rendering *after* RO invalidation and wouldn't prevent the RO caches
- * from being polluted by concurrent rendering before the stall
- * completes. This intentionally doesn't implement the SKL+ hardware
- * workaround suggesting to enable CS stall on PIPE_CONTROLs with the
- * texture cache invalidation bit set for GPGPU workloads because the
- * previous and subsequent PIPE_CONTROLs already guarantee that there is
- * no concurrent GPGPU kernel execution (see SKL HSD 2132585).
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
- pc.TextureCacheInvalidationEnable = true,
- pc.ConstantCacheInvalidationEnable = true,
- pc.InstructionCacheInvalidateEnable = true,
- pc.StateCacheInvalidationEnable = true,
- pc.PostSyncOperation = NoWrite;
- }
-
- /* Now send a third stalling flush to make sure that invalidation is
- * complete when the L3 configuration registers are modified.
- */
- anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
- pc.DCFlushEnable = true;
- pc.PostSyncOperation = NoWrite;
- pc.CommandStreamerStallEnable = true;
- }
-
- anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
- lri.RegisterOffset = GENX(L3CNTLREG_num);
- lri.DataDWord = l3cr_val;
- }
- cmd_buffer->state.current_l3_config = l3cr_val;
- }
-}
-
static void
__emit_genx_sf_state(struct anv_cmd_buffer *cmd_buffer)
{
@@ -448,8 +377,7 @@ genX(cmd_buffer_flush_compute_state)(struct anv_cmd_buffer *cmd_buffer)
assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT);
- bool needs_slm = cs_prog_data->base.total_shared > 0;
- genX(cmd_buffer_config_l3)(cmd_buffer, needs_slm);
+ genX(cmd_buffer_config_l3)(cmd_buffer, pipeline);
genX(flush_pipeline_select_gpgpu)(cmd_buffer);
diff --git a/src/intel/vulkan/genX_cmd_buffer.c b/src/intel/vulkan/genX_cmd_buffer.c
index 0a5c404..daa1884 100644
--- a/src/intel/vulkan/genX_cmd_buffer.c
+++ b/src/intel/vulkan/genX_cmd_buffer.c
@@ -340,7 +340,7 @@ genX(cmd_buffer_flush_state)(struct anv_cmd_buffer *cmd_buffer)
* but at least 3-D will. In the mean time, I'm going to make this
* gen8+ only so that we can get Haswell working again.
*/
- genX(cmd_buffer_config_l3)(cmd_buffer, false);
+ genX(cmd_buffer_config_l3)(cmd_buffer, pipeline);
#endif
genX(flush_pipeline_select_3d)(cmd_buffer);
diff --git a/src/intel/vulkan/genX_l3.c b/src/intel/vulkan/genX_l3.c
new file mode 100644
index 0000000..0d36e3c
--- /dev/null
+++ b/src/intel/vulkan/genX_l3.c
@@ -0,0 +1,541 @@
+/*
+ * Copyright (c) 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include "anv_private.h"
+
+#include "genxml/gen_macros.h"
+#include "genxml/genX_pack.h"
+
+/**
+ * Chunk of L3 cache reserved for some specific purpose.
+ */
+enum anv_l3_partition {
+ /** Shared local memory. */
+ L3P_SLM = 0,
+ /** Unified return buffer. */
+ L3P_URB,
+ /** Union of DC and RO. */
+ L3P_ALL,
+ /** Data cluster RW partition. */
+ L3P_DC,
+ /** Union of IS, C and T. */
+ L3P_RO,
+ /** Instruction and state cache. */
+ L3P_IS,
+ /** Constant cache. */
+ L3P_C,
+ /** Texture cache. */
+ L3P_T,
+ /** Number of supported L3 partitions. */
+ NUM_L3P
+};
+
+/**
+ * L3 configuration represented as the number of ways allocated for each
+ * partition. \sa get_l3_way_size().
+ */
+struct anv_l3_config {
+ unsigned n[NUM_L3P];
+};
+
+#if GEN_GEN == 7
+
+/**
+ * IVB/HSW validated L3 configurations. The first entry will be used as
+ * default by gen7_restore_default_l3_config(), otherwise the ordering is
+ * unimportant.
+ */
+static const struct anv_l3_config ivb_l3_configs[] = {
+ /* SLM URB ALL DC RO IS C T */
+ {{ 0, 32, 0, 0, 32, 0, 0, 0 }},
+ {{ 0, 32, 0, 16, 16, 0, 0, 0 }},
+ {{ 0, 32, 0, 4, 0, 8, 4, 16 }},
+ {{ 0, 28, 0, 8, 0, 8, 4, 16 }},
+ {{ 0, 28, 0, 16, 0, 8, 4, 8 }},
+ {{ 0, 28, 0, 8, 0, 16, 4, 8 }},
+ {{ 0, 28, 0, 0, 0, 16, 4, 16 }},
+ {{ 0, 32, 0, 0, 0, 16, 0, 16 }},
+ {{ 0, 28, 0, 4, 32, 0, 0, 0 }},
+ {{ 16, 16, 0, 16, 16, 0, 0, 0 }},
+ {{ 16, 16, 0, 8, 0, 8, 8, 8 }},
+ {{ 16, 16, 0, 4, 0, 8, 4, 16 }},
+ {{ 16, 16, 0, 4, 0, 16, 4, 8 }},
+ {{ 16, 16, 0, 0, 32, 0, 0, 0 }},
+ {{ 0 }}
+};
+
+#endif
+
+#if GEN_GEN == 7 && !GEN_IS_HASWELL
+
+/**
+ * VLV validated L3 configurations. \sa ivb_l3_configs.
+ */
+static const struct anv_l3_config vlv_l3_configs[] = {
+ /* SLM URB ALL DC RO IS C T */
+ {{ 0, 64, 0, 0, 32, 0, 0, 0 }},
+ {{ 0, 80, 0, 0, 16, 0, 0, 0 }},
+ {{ 0, 80, 0, 8, 8, 0, 0, 0 }},
+ {{ 0, 64, 0, 16, 16, 0, 0, 0 }},
+ {{ 0, 60, 0, 4, 32, 0, 0, 0 }},
+ {{ 32, 32, 0, 16, 16, 0, 0, 0 }},
+ {{ 32, 40, 0, 8, 16, 0, 0, 0 }},
+ {{ 32, 40, 0, 16, 8, 0, 0, 0 }},
+ {{ 0 }}
+};
+
+#endif
+
+#if GEN_GEN == 8
+
+/**
+ * BDW validated L3 configurations. \sa ivb_l3_configs.
+ */
+static const struct anv_l3_config bdw_l3_configs[] = {
+ /* SLM URB ALL DC RO IS C T */
+ {{ 0, 48, 48, 0, 0, 0, 0, 0 }},
+ {{ 0, 48, 0, 16, 32, 0, 0, 0 }},
+ {{ 0, 32, 0, 16, 48, 0, 0, 0 }},
+ {{ 0, 32, 0, 0, 64, 0, 0, 0 }},
+ {{ 0, 32, 64, 0, 0, 0, 0, 0 }},
+ {{ 24, 16, 48, 0, 0, 0, 0, 0 }},
+ {{ 24, 16, 0, 16, 32, 0, 0, 0 }},
+ {{ 24, 16, 0, 32, 16, 0, 0, 0 }},
+ {{ 0 }}
+};
+
+#endif
+
+#if GEN_GEN == 8 || GEN_GEN == 9
+
+/**
+ * CHV/SKL validated L3 configurations. \sa ivb_l3_configs.
+ */
+static const struct anv_l3_config chv_l3_configs[] = {
+ /* SLM URB ALL DC RO IS C T */
+ {{ 0, 48, 48, 0, 0, 0, 0, 0 }},
+ {{ 0, 48, 0, 16, 32, 0, 0, 0 }},
+ {{ 0, 32, 0, 16, 48, 0, 0, 0 }},
+ {{ 0, 32, 0, 0, 64, 0, 0, 0 }},
+ {{ 0, 32, 64, 0, 0, 0, 0, 0 }},
+ {{ 32, 16, 48, 0, 0, 0, 0, 0 }},
+ {{ 32, 16, 0, 16, 32, 0, 0, 0 }},
+ {{ 32, 16, 0, 32, 16, 0, 0, 0 }},
+ {{ 0 }}
+};
+
+#endif
+
+/**
+ * Return a zero-terminated array of validated L3 configurations for the
+ * specified device.
+ */
+static inline const struct anv_l3_config *
+get_l3_configs(const struct brw_device_info *devinfo)
+{
+ assert(devinfo->gen == GEN_GEN);
+#if GEN_IS_HASWELL
+ return ivb_l3_configs;
+#elif GEN_GEN == 7
+ return (devinfo->is_baytrail ? vlv_l3_configs : ivb_l3_configs);
+#elif GEN_GEN == 8
+ return (devinfo->is_cherryview ? chv_l3_configs : bdw_l3_configs);
+#elif GEN_GEN == 9
+ return chv_l3_configs;
+#else
+#error GEN not supported
+#endif
+}
+
+/**
+ * Return the size of an L3 way in KB.
+ */
+static unsigned
+get_l3_way_size(const struct brw_device_info *devinfo)
+{
+ if (devinfo->is_baytrail)
+ return 2;
+
+ else if (devinfo->is_cherryview || devinfo->gt == 1)
+ return 4;
+
+ else
+ return 8 * devinfo->num_slices;
+}
+
+/**
+ * L3 configuration represented as a vector of weights giving the desired
+ * relative size of each partition. The scale is arbitrary, only the ratios
+ * between weights will have an influence on the selection of the closest L3
+ * configuration.
+ */
+struct anv_l3_weights {
+ float w[NUM_L3P];
+};
+
+/**
+ * L1-normalize a vector of L3 partition weights.
+ */
+static struct anv_l3_weights
+norm_l3_weights(struct anv_l3_weights w)
+{
+ float sz = 0;
+
+ for (unsigned i = 0; i < NUM_L3P; i++)
+ sz += w.w[i];
+
+ for (unsigned i = 0; i < NUM_L3P; i++)
+ w.w[i] /= sz;
+
+ return w;
+}
+
+/**
+ * Get the relative partition weights of the specified L3 configuration.
+ */
+static struct anv_l3_weights
+get_config_l3_weights(const struct anv_l3_config *cfg)
+{
+ if (cfg) {
+ struct anv_l3_weights w;
+
+ for (unsigned i = 0; i < NUM_L3P; i++)
+ w.w[i] = cfg->n[i];
+
+ return norm_l3_weights(w);
+ } else {
+ const struct anv_l3_weights w = { { 0 } };
+ return w;
+ }
+}
+
+/**
+ * Distance between two L3 configurations represented as vectors of weights.
+ * Usually just the L1 metric except when the two configurations are
+ * considered incompatible in which case the distance will be infinite. Note
+ * that the compatibility condition is asymmetric -- They will be considered
+ * incompatible whenever the reference configuration \p w0 requires SLM, DC,
+ * or URB but \p w1 doesn't provide it.
+ */
+static float
+diff_l3_weights(struct anv_l3_weights w0, struct anv_l3_weights w1)
+{
+ if ((w0.w[L3P_SLM] && !w1.w[L3P_SLM]) ||
+ (w0.w[L3P_DC] && !w1.w[L3P_DC] && !w1.w[L3P_ALL]) ||
+ (w0.w[L3P_URB] && !w1.w[L3P_URB])) {
+ return HUGE_VALF;
+
+ } else {
+ float dw = 0;
+
+ for (unsigned i = 0; i < NUM_L3P; i++)
+ dw += fabs(w0.w[i] - w1.w[i]);
+
+ return dw;
+ }
+}
+
+/**
+ * Return the closest validated L3 configuration for the specified device and
+ * weight vector.
+ */
+static const struct anv_l3_config *
+get_l3_config(const struct brw_device_info *devinfo, struct anv_l3_weights w0)
+{
+ const struct anv_l3_config *const cfgs = get_l3_configs(devinfo);
+ const struct anv_l3_config *cfg_best = NULL;
+ float dw_best = HUGE_VALF;
+
+ for (const struct anv_l3_config *cfg = cfgs; cfg->n[L3P_URB]; cfg++) {
+ const float dw = diff_l3_weights(w0, get_config_l3_weights(cfg));
+
+ if (dw < dw_best) {
+ cfg_best = cfg;
+ dw_best = dw;
+ }
+ }
+
+ return cfg_best;
+}
+
+/**
+ * Return a reasonable default L3 configuration for the specified device based
+ * on whether SLM and DC are required. In the non-SLM non-DC case the result
+ * is intended to approximately resemble the hardware defaults.
+ */
+static struct anv_l3_weights
+get_default_l3_weights(const struct brw_device_info *devinfo,
+ bool needs_dc, bool needs_slm)
+{
+ struct anv_l3_weights w = {{ 0 }};
+
+ w.w[L3P_SLM] = needs_slm;
+ w.w[L3P_URB] = 1.0;
+
+ if (devinfo->gen >= 8) {
+ w.w[L3P_ALL] = 1.0;
+ } else {
+ w.w[L3P_DC] = needs_dc ? 0.1 : 0;
+ w.w[L3P_RO] = devinfo->is_baytrail ? 0.5 : 1.0;
+ }
+
+ return norm_l3_weights(w);
+}
+
+/**
+ * Calculate the desired L3 partitioning based on the current state of the
+ * pipeline. For now this simply returns the conservative defaults calculated
+ * by get_default_l3_weights(), but we could probably do better by gathering
+ * more statistics from the pipeline state (e.g. guess of expected URB usage
+ * and bound surfaces), or by using feed-back from performance counters.
+ */
+static struct anv_l3_weights
+get_pipeline_state_l3_weights(const struct anv_pipeline *pipeline)
+{
+ bool needs_dc = false, needs_slm = false;
+
+ for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
+ const struct brw_stage_prog_data *prog_data = pipeline->prog_data[i];
+
+ needs_dc |= pipeline->needs_data_cache;
+ needs_slm |= prog_data && prog_data->total_shared;
+ }
+
+ return get_default_l3_weights(&pipeline->device->info,
+ needs_dc, needs_slm);
+}
+
+#define emit_lri(batch, reg, imm) \
+ anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_IMM), lri) { \
+ lri.RegisterOffset = __anv_reg_num(reg); \
+ lri.DataDWord = imm; \
+ }
+
+#define IVB_L3SQCREG1_SQGHPCI_DEFAULT 0x00730000
+#define VLV_L3SQCREG1_SQGHPCI_DEFAULT 0x00d30000
+#define HSW_L3SQCREG1_SQGHPCI_DEFAULT 0x00610000
+
+/**
+ * Program the hardware to use the specified L3 configuration.
+ */
+static void
+setup_l3_config(struct anv_cmd_buffer *cmd_buffer/*, struct brw_context *brw*/,
+ const struct anv_l3_config *cfg)
+{
+ const bool has_slm = cfg->n[L3P_SLM];
+
+ /* According to the hardware docs, the L3 partitioning can only be changed
+ * while the pipeline is completely drained and the caches are flushed,
+ * which involves a first PIPE_CONTROL flush which stalls the pipeline...
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.DCFlushEnable = true;
+ pc.PostSyncOperation = NoWrite;
+ pc.CommandStreamerStallEnable = true;
+ }
+
+ /* ...followed by a second pipelined PIPE_CONTROL that initiates
+ * invalidation of the relevant caches. Note that because RO invalidation
+ * happens at the top of the pipeline (i.e. right away as the PIPE_CONTROL
+ * command is processed by the CS) we cannot combine it with the previous
+ * stalling flush as the hardware documentation suggests, because that
+ * would cause the CS to stall on previous rendering *after* RO
+ * invalidation and wouldn't prevent the RO caches from being polluted by
+ * concurrent rendering before the stall completes. This intentionally
+ * doesn't implement the SKL+ hardware workaround suggesting to enable CS
+ * stall on PIPE_CONTROLs with the texture cache invalidation bit set for
+ * GPGPU workloads because the previous and subsequent PIPE_CONTROLs
+ * already guarantee that there is no concurrent GPGPU kernel execution
+ * (see SKL HSD 2132585).
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.TextureCacheInvalidationEnable = true;
+ pc.ConstantCacheInvalidationEnable = true;
+ pc.InstructionCacheInvalidateEnable = true;
+ pc.StateCacheInvalidationEnable = true;
+ pc.PostSyncOperation = NoWrite;
+ }
+
+ /* Now send a third stalling flush to make sure that invalidation is
+ * complete when the L3 configuration registers are modified.
+ */
+ anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) {
+ pc.DCFlushEnable = true;
+ pc.PostSyncOperation = NoWrite;
+ pc.CommandStreamerStallEnable = true;
+ }
+
+#if GEN_GEN >= 8
+
+ assert(!cfg->n[L3P_IS] && !cfg->n[L3P_C] && !cfg->n[L3P_T]);
+
+ uint32_t l3cr;
+ anv_pack_struct(&l3cr, GENX(L3CNTLREG),
+ .SLMEnable = has_slm,
+ .URBAllocation = cfg->n[L3P_URB],
+ .ROAllocation = cfg->n[L3P_RO],
+ .DCAllocation = cfg->n[L3P_DC],
+ .AllAllocation = cfg->n[L3P_ALL]);
+
+ /* Set up the L3 partitioning. */
+ emit_lri(&cmd_buffer->batch, GENX(L3CNTLREG), l3cr);
+
+#else
+
+ const bool has_dc = cfg->n[L3P_DC] || cfg->n[L3P_ALL];
+ const bool has_is = cfg->n[L3P_IS] || cfg->n[L3P_RO] || cfg->n[L3P_ALL];
+ const bool has_c = cfg->n[L3P_C] || cfg->n[L3P_RO] || cfg->n[L3P_ALL];
+ const bool has_t = cfg->n[L3P_T] || cfg->n[L3P_RO] || cfg->n[L3P_ALL];
+
+ assert(!cfg->n[L3P_ALL]);
+
+ /* When enabled SLM only uses a portion of the L3 on half of the banks,
+ * the matching space on the remaining banks has to be allocated to a
+ * client (URB for all validated configurations) set to the
+ * lower-bandwidth 2-bank address hashing mode.
+ */
+ const struct brw_device_info *devinfo = &cmd_buffer->device->info;
+ const bool urb_low_bw = has_slm && !devinfo->is_baytrail;
+ assert(!urb_low_bw || cfg->n[L3P_URB] == cfg->n[L3P_SLM]);
+
+ /* Minimum number of ways that can be allocated to the URB. */
+ const unsigned n0_urb = (devinfo->is_baytrail ? 32 : 0);
+ assert(cfg->n[L3P_URB] >= n0_urb);
+
+ uint32_t l3sqcr1, l3cr2, l3cr3;
+ anv_pack_struct(&l3sqcr1, GENX(L3SQCREG1),
+ .ConvertDC_UC = !has_dc,
+ .ConvertIS_UC = !has_is,
+ .ConvertC_UC = !has_c,
+ .ConvertT_UC = !has_t);
+ l3sqcr1 |=
+ GEN_IS_HASWELL ? HSW_L3SQCREG1_SQGHPCI_DEFAULT :
+ devinfo->is_baytrail ? VLV_L3SQCREG1_SQGHPCI_DEFAULT :
+ IVB_L3SQCREG1_SQGHPCI_DEFAULT;
+
+ anv_pack_struct(&l3cr2, GENX(L3CNTLREG2),
+ .SLMEnable = has_slm,
+ .URBLowBandwidth = urb_low_bw,
+ .URBAllocation = cfg->n[L3P_URB],
+#if !GEN_IS_HASWELL
+ .ALLAllocation = cfg->n[L3P_ALL],
+#endif
+ .ROAllocation = cfg->n[L3P_RO],
+ .DCAllocation = cfg->n[L3P_DC]);
+
+ anv_pack_struct(&l3cr3, GENX(L3CNTLREG3),
+ .ISAllocation = cfg->n[L3P_IS],
+ .ISLowBandwidth = 0,
+ .CAllocation = cfg->n[L3P_C],
+ .CLowBandwidth = 0,
+ .TAllocation = cfg->n[L3P_T],
+ .TLowBandwidth = 0);
+
+ /* Set up the L3 partitioning. */
+ emit_lri(&cmd_buffer->batch, GENX(L3SQCREG1), l3sqcr1);
+ emit_lri(&cmd_buffer->batch, GENX(L3CNTLREG2), l3cr2);
+ emit_lri(&cmd_buffer->batch, GENX(L3CNTLREG3), l3cr3);
+
+#if GEN_IS_HASWELL
+ if (cmd_buffer->device->instance->physicalDevice.cmd_parser_version >= 4) {
+ /* Enable L3 atomics on HSW if we have a DC partition, otherwise keep
+ * them disabled to avoid crashing the system hard.
+ */
+ uint32_t scratch1, chicken3;
+ anv_pack_struct(&scratch1, GENX(SCRATCH1),
+ .L3AtomicDisable = !has_dc);
+ anv_pack_struct(&chicken3, GENX(CHICKEN3),
+ .L3AtomicDisable = !has_dc);
+ emit_lri(&cmd_buffer->batch, GENX(SCRATCH1), scratch1);
+ emit_lri(&cmd_buffer->batch, GENX(CHICKEN3), chicken3);
+ }
+#endif
+
+#endif
+
+}
+
+/**
+ * Return the unit brw_context::urb::size is expressed in, in KB. \sa
+ * brw_device_info::urb::size.
+ */
+static unsigned
+get_urb_size_scale(const struct brw_device_info *devinfo)
+{
+ return (devinfo->gen >= 8 ? devinfo->num_slices : 1);
+}
+
+void
+genX(setup_pipeline_l3_config)(struct anv_pipeline *pipeline)
+{
+ const struct anv_l3_weights w = get_pipeline_state_l3_weights(pipeline);
+ const struct brw_device_info *devinfo = &pipeline->device->info;
+ const struct anv_l3_config *const cfg = get_l3_config(devinfo, w);
+ pipeline->urb.l3_config = cfg;
+
+ unsigned sz = cfg->n[L3P_URB] * get_l3_way_size(devinfo);
+
+#if GEN_GEN == 9
+ /* From the SKL "L3 Allocation and Programming" documentation:
+ *
+ * "URB is limited to 1008KB due to programming restrictions. This is not
+ * a restriction of the L3 implementation, but of the FF and other clients.
+ * Therefore, in a GT4 implementation it is possible for the programmed
+ * allocation of the L3 data array to provide 3*384KB=1152KB for URB, but
+ * only 1008KB of this will be used."
+ */
+ sz = MIN2(1008, sz);
+#endif
+
+ pipeline->urb.total_size = sz / get_urb_size_scale(devinfo);
+}
+
+/**
+ * Print out the specified L3 configuration.
+ */
+static void
+dump_l3_config(const struct anv_l3_config *cfg)
+{
+ fprintf(stderr, "SLM=%d URB=%d ALL=%d DC=%d RO=%d IS=%d C=%d T=%d\n",
+ cfg->n[L3P_SLM], cfg->n[L3P_URB], cfg->n[L3P_ALL],
+ cfg->n[L3P_DC], cfg->n[L3P_RO],
+ cfg->n[L3P_IS], cfg->n[L3P_C], cfg->n[L3P_T]);
+}
+
+void
+genX(cmd_buffer_config_l3)(struct anv_cmd_buffer *cmd_buffer,
+ const struct anv_pipeline *pipeline)
+{
+ struct anv_cmd_state *state = &cmd_buffer->state;
+ const struct anv_l3_config *const cfg = pipeline->urb.l3_config;
+ assert(cfg);
+ if (cfg != state->current_l3_config) {
+ setup_l3_config(cmd_buffer, cfg);
+ state->current_l3_config = cfg;
+
+ if (unlikely(INTEL_DEBUG & DEBUG_L3)) {
+ fprintf(stderr, "L3 config transition: ");
+ dump_l3_config(cfg);
+ }
+ }
+}
diff --git a/src/intel/vulkan/genX_pipeline.c b/src/intel/vulkan/genX_pipeline.c
index 2a41b2d..918a9a4 100644
--- a/src/intel/vulkan/genX_pipeline.c
+++ b/src/intel/vulkan/genX_pipeline.c
@@ -84,6 +84,8 @@ genX(compute_pipeline_create)(
pipeline->use_repclear = false;
+ anv_setup_pipeline_l3_config(pipeline);
+
const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline);
const struct brw_stage_prog_data *prog_data = &cs_prog_data->base;
--
2.8.1
More information about the mesa-dev
mailing list