[Mesa-dev] [PATCH v2] ehl: disable repacking for compression for compatibilty
Anuj Phogat
anuj.phogat at gmail.com
Thu Jun 20 19:23:38 UTC 2019
+mesa-dev
On Thu, Jun 20, 2019 at 12:20 PM Anuj Phogat <anuj.phogat at gmail.com> wrote:
>
> I sent out comments on your older patch. They applies to this patch too.
> Split i965 and anv changes in separate patches.
>
> On Thu, Jun 20, 2019 at 11:25 AM Dongwon Kim <dongwon.kim at intel.com> wrote:
> >
> > Repacking components in certain pixel formats before compression
> > shouldn't be done for EHL to keep the compatibility with decompression
> > capability in its display controller.
> >
> > v2: applied to iris, i965 and anvil.
> >
> > Signed-off-by: Dongwon Kim <dongwon.kim at intel.com>
> > ---
> > src/gallium/drivers/iris/iris_state.c | 10 +++++++
> > src/intel/dev/gen_device_info.c | 4 +++
> > src/intel/dev/gen_device_info.h | 1 +
> > src/intel/genxml/gen11.xml | 30 +++++++++-----------
> > src/intel/vulkan/genX_state.c | 18 ++++++++++++
> > src/mesa/drivers/dri/i965/brw_defines.h | 1 +
> > src/mesa/drivers/dri/i965/brw_state_upload.c | 9 ++++++
> > 7 files changed, 57 insertions(+), 16 deletions(-)
> >
> > diff --git a/src/gallium/drivers/iris/iris_state.c b/src/gallium/drivers/iris/iris_state.c
> > index 23f8d290368..8744178e94d 100644
> > --- a/src/gallium/drivers/iris/iris_state.c
> > +++ b/src/gallium/drivers/iris/iris_state.c
> > @@ -749,6 +749,16 @@ iris_init_render_context(struct iris_screen *screen,
> > }
> > iris_emit_lri(batch, SLICE_COMMON_ECO_CHICKEN1, reg_val);
> >
> > + /* needed to disable replacking to keep the compatibility with
> > + * decompressing mechanism in display controller.
> > + */
> > + if (devinfo->need_disable_ccs_repack) {
> > + iris_pack_state(GENX(CACHE_MODE_0), ®_val, reg) {
> > + reg.DisableRepackingforCompression = true;
> > + reg.DisableRepackingforCompressionMask = true;
> > + }
> > + iris_emit_lri(batch, CACHE_MODE_0, reg_val);
> > + }
> >
> > // XXX: 3D_MODE?
> > #endif
> > diff --git a/src/intel/dev/gen_device_info.c b/src/intel/dev/gen_device_info.c
> > index fec6159fd37..ee3ad7ba085 100644
> > --- a/src/intel/dev/gen_device_info.c
> > +++ b/src/intel/dev/gen_device_info.c
> > @@ -64,6 +64,7 @@ gen_device_name_to_pci_device_id(const char *name)
> > { "cml", 0x9b41 },
> > { "cnl", 0x5a52 },
> > { "icl", 0x8a52 },
> > + { "ehl", 0x4500 },
> > };
> >
> > for (unsigned i = 0; i < ARRAY_SIZE(name_map); i++) {
> > @@ -958,6 +959,7 @@ static const struct gen_device_info gen_device_info_ehl_4x8 = {
> > [MESA_SHADER_GEOMETRY] = 1032,
> > },
> > },
> > + .need_disable_ccs_repack = 1,
> > .simulator_id = 28,
> > };
> >
> > @@ -978,6 +980,7 @@ static const struct gen_device_info gen_device_info_ehl_4x4 = {
> > [MESA_SHADER_GEOMETRY] = 1032,
> > },
> > },
> > + .need_disable_ccs_repack = 1,
> > .num_eu_per_subslice = 4,
> > .simulator_id = 28,
> > };
> > @@ -999,6 +1002,7 @@ static const struct gen_device_info gen_device_info_ehl_2x4 = {
> > [MESA_SHADER_GEOMETRY] = 1032,
> > },
> > },
> > + .need_disable_ccs_repack = 1,
> > .num_eu_per_subslice =4,
> > .simulator_id = 28,
> > };
> > diff --git a/src/intel/dev/gen_device_info.h b/src/intel/dev/gen_device_info.h
> > index af13615be2b..e90d31935f1 100644
> > --- a/src/intel/dev/gen_device_info.h
> > +++ b/src/intel/dev/gen_device_info.h
> > @@ -74,6 +74,7 @@ struct gen_device_info
> > bool has_surface_tile_offset;
> > bool supports_simd16_3src;
> > bool has_resource_streamer;
> > + bool need_disable_ccs_repack;
> >
> > /**
> > * \name Intel hardware quirks
> > diff --git a/src/intel/genxml/gen11.xml b/src/intel/genxml/gen11.xml
> > index 1579345f69f..c1774501f4c 100644
> > --- a/src/intel/genxml/gen11.xml
> > +++ b/src/intel/genxml/gen11.xml
> > @@ -6818,30 +6818,28 @@
> > </register>
> >
> > <register name="CACHE_MODE_0" length="1" num="0x7000">
> > - <field name="Null tile fix disable" start="0" end="0" type="bool"/>
> > + <field name="Disable Byte sharing for 3D TYF LOD1 surfaces for 32/64/128 bpp" start="0" end="0" type="bool"/>
> > <field name="Disable clock gating in the pixel backend" start="1" end="1" type="bool"/>
> > - <field name="Hierarchical Z RAW Stall Optimization Disable" start="2" end="2" type="bool"/>
> > + <field name="Hierarchical Z Disable" start="3" end="3" type="bool"/>
> > <field name="RCC Eviction Policy" start="4" end="4" type="bool"/>
> > - <field name="STC PMA Optimization Enable" start="5" end="5" type="bool"/>
> > - <field name="Sampler L2 Request Arbitration" start="6" end="7" type="uint">
> > - <value name="Round Robin" value="0"/>
> > - <value name="Fetch are Highest Priority" value="1"/>
> > - <value name="Constants are Highest Priority" value="2"/>
> > - </field>
> > + <field name="STC PMA Optimization Disable" start="5" end="5" type="bool"/>
> > + <field name="STC Read-Hit Wonly Optimization Disable" start="6" end="6" type="bool"/>
> > + <field name="Depth Related Cache Pipelined Flush Disable" start="8" end="8" type="bool"/>
> > <field name="Sampler L2 TLB Prefetch Enable" start="9" end="9" type="bool"/>
> > - <field name="Sampler Set Remapping for 3D Disable" start="11" end="11" type="bool"/>
> > + <field name="RCZ PMA Not-Promoted Allocation stall optimization Disable due to change in depth parameters" start="10" end="10" type="bool"/>
> > <field name="MSAA Compression Plane Number Threshold for eLLC" start="12" end="14" type="uint"/>
> > - <field name="Sampler L2 Disable" start="15" end="15" type="bool"/>
> > - <field name="Null tile fix disable Mask" start="16" end="16" type="bool"/>
> > + <field name="Disable Repacking for Compression" start="15" end="15" type="bool"/>
> > + <field name="Disable Byte sharing for 3D TYF LOD1 surfaces for 32/64/128 bpp Mask" start="16" end="16" type="bool"/>
> > <field name="Disable clock gating in the pixel backend Mask" start="17" end="17" type="bool"/>
> > - <field name="Hierarchical Z RAW Stall Optimization Disable Mask" start="18" end="18" type="bool"/>
> > + <field name="Hierarchical Z Disable Mask" start="19" end="19" type="bool"/>
> > <field name="RCC Eviction Policy Mask" start="20" end="20" type="bool"/>
> > - <field name="STC PMA Optimization Enable Mask" start="21" end="21" type="bool"/>
> > - <field name="Sampler L2 Request Arbitration Mask" start="22" end="23" type="uint"/>
> > + <field name="STC PMA Optimization Disable Mask" start="21" end="21" type="bool"/>
> > + <field name="STC Read-Hit Wonly Optimization Disable Mask" start="22" end="22" type="bool"/>
> > + <field name="Depth Related Cache Pipelined Flush Disable Mask" start="24" end="24" type="bool"/>
> > <field name="Sampler L2 TLB Prefetch Enable Mask" start="25" end="25" type="bool"/>
> > - <field name="Sampler Set Remapping for 3D Disable Mask" start="27" end="27" type="bool"/>
> > + <field name="RCZ PMA Not-Promoted Allocation stall optimization Disable due to change in depth parameters Mask" start="26" end="26" type="bool"/>
> > <field name="MSAA Compression Plane Number Threshold for eLLC Mask" start="28" end="30" type="uint"/>
> > - <field name="Sampler L2 Disable Mask" start="31" end="31" type="bool"/>
> > + <field name="Disable Repacking for Compression Mask" start="31" end="31" type="bool"/>
> > </register>
> >
> > <register name="CACHE_MODE_1" length="1" num="0x7004">
> > diff --git a/src/intel/vulkan/genX_state.c b/src/intel/vulkan/genX_state.c
> > index 21b8cd648d4..412faac9e52 100644
> > --- a/src/intel/vulkan/genX_state.c
> > +++ b/src/intel/vulkan/genX_state.c
> > @@ -225,6 +225,24 @@ genX(init_device_state)(struct anv_device *device)
> > }
> > #endif
> >
> > +#if GEN_GEN >= 11
> > + /* needed to disable replacking to keep the compatibility with
> > + * decompressing mechanism in display controller.
> > + */
> > + if (device->info.need_disable_ccs_repack) {
> > + uint32_t cache_mode_0;
> > + anv_pack_struct(&cache_mode_0,
> > + GENX(CACHE_MODE_0),
> > + .DisableRepackingforCompression = true,
> > + .DisableRepackingforCompressionMask = true);
> > +
> > + anv_batch_emit(&batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
> > + lri.RegisterOffset = GENX(CACHE_MODE_0_num);
> > + lri.DataDWord = cache_mode_0;
> > + }
> > + }
> > +#endif
> > +
> > /* Set the "CONSTANT_BUFFER Address Offset Disable" bit, so
> > * 3DSTATE_CONSTANT_XS buffer 0 is an absolute address.
> > *
> > diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
> > index 17bca1991f1..e8507b7e5ff 100644
> > --- a/src/mesa/drivers/dri/i965/brw_defines.h
> > +++ b/src/mesa/drivers/dri/i965/brw_defines.h
> > @@ -1576,6 +1576,7 @@ enum brw_pixel_shader_coverage_mask_mode {
> > # define GEN9_PARTIAL_RESOLVE_DISABLE_IN_VC (1 << 1)
> > # define GEN8_HIZ_PMA_MASK_BITS \
> > REG_MASK(GEN8_HIZ_NP_PMA_FIX_ENABLE | GEN8_HIZ_NP_EARLY_Z_FAILS_DISABLE)
> > +# define GEN11_DISABLE_REPACKING_FOR_COMPRESSION (1 << 15)
> >
> > #define GEN7_GT_MODE 0x7008
> > # define GEN9_SUBSLICE_HASHING_8x8 (0 << 8)
> > diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
> > index 938b9defeda..8395688c793 100644
> > --- a/src/mesa/drivers/dri/i965/brw_state_upload.c
> > +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
> > @@ -121,6 +121,15 @@ brw_upload_initial_gpu_state(struct brw_context *brw)
> > REG_MASK(GEN11_STATE_CACHE_REDIRECT_TO_CS_SECTION_ENABLE));
> > }
> >
> > + /* needed to disable replacking to keep the compatibility with
> > + * decompressing mechanism in display controller.
> > + */
> > + if (devinfo->need_disable_ccs_repack) {
> > + brw_load_register_imm32(brw, GEN7_CACHE_MODE_0,
> > + GEN11_DISABLE_REPACKING_FOR_COMPRESSION |
> > + REG_MASK(GEN11_DISABLE_REPACKING_FOR_COMPRESSION));
> > + }
> > +
> > if (devinfo->gen == 10 || devinfo->gen == 11) {
> > /* From gen10 workaround table in h/w specs:
> > *
> > --
> > 2.17.1
> >
More information about the mesa-dev
mailing list