anv: Allow compressed memtypes with default buffer types

Source 2 games segfault if certain buffers are not able to use the same
memory types as images. CS2 specifically expects this to be the case for
vertex and index buffers (VK_BUFFER_USAGE_2_INDEX_BUFFER_BIT,
VK_BUFFER_USAGE_2_VERTEX_BUFFER_BIT). I have not tested other Source 2
games to see how much the requirement differs for the usage (if at all).

Up until now, we've disabled CCS for the Source 2 engine with the
anv_disable_xe2_ccs driconf option. However, this option is not great
for performance. So, replace this with a new option to allow the same
memory types we use for images on buffers - anv_enable_buffer_comp.

Compression of buffers is generally not good for performance. I
collected the result of unconditionally enabling the feature in the
performance CI on BMG. I used the default configuration to average the
result of two runs of each trace.

The CI reports that 4 game traces would regress between 0.44-1.01% FPS
with buffer compression. However, the CI actually shows it to be
beneficial in three of our game traces:

* Cyberpunk-trace-dx12-1080p-high 106.51%
* Hitman3-trace-dx12-1080p-med    101.59%
* Blackops3-trace-dx11-1080p-high 100.44%

So, enable the option for the two games we already have driconf entries
for, Cyberpunk and Hitman3.

Of course, also enable the option for Source 2 games. Casey Bowman
reports that on BMG, some frame times drop from ~15ms to ~7ms in CS2.
This is in large part due to the removal of HiZ resolves, which is a
consequence of the game now using of HIZ_CCS_WT instead of plain HIZ.

Ref: https://gitlab.freedesktop.org/mesa/mesa/-/issues/11520
Acked-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32519>
This commit is contained in:
Nanley Chery
2024-12-05 13:03:33 -05:00
committed by Marge Bot
parent fec6ef6d7f
commit 483c40a21d
6 changed files with 26 additions and 14 deletions

View File

@@ -72,6 +72,9 @@ anv_get_buffer_memory_requirements(struct anv_device *device,
else if (usage & (VK_BUFFER_USAGE_2_RESOURCE_DESCRIPTOR_BUFFER_BIT_EXT |
VK_BUFFER_USAGE_2_SAMPLER_DESCRIPTOR_BUFFER_BIT_EXT))
memory_types = device->physical->memory.dynamic_visible_mem_types;
else if (device->physical->instance->enable_buffer_comp)
memory_types = device->physical->memory.default_buffer_mem_types |
device->physical->memory.compressed_mem_types;
else
memory_types = device->physical->memory.default_buffer_mem_types;

View File

@@ -17,7 +17,7 @@ static const driOptionDescription anv_dri_options[] = {
DRI_CONF_ANV_ASSUME_FULL_SUBGROUPS(0)
DRI_CONF_ANV_ASSUME_FULL_SUBGROUPS_WITH_BARRIER(false)
DRI_CONF_ANV_DISABLE_FCV(false)
DRI_CONF_ANV_DISABLE_XE2_CCS(false)
DRI_CONF_ANV_ENABLE_BUFFER_COMP(false)
DRI_CONF_ANV_EXTERNAL_MEMORY_IMPLICIT_SYNC(true)
DRI_CONF_ANV_FORCE_GUC_LOW_LATENCY(false)
DRI_CONF_ANV_SAMPLE_MASK_OUT_OPENGL_BEHAVIOUR(false)
@@ -171,8 +171,8 @@ anv_init_dri_options(struct anv_instance *instance)
instance->enable_tbimr = driQueryOptionb(&instance->dri_options, "intel_tbimr");
instance->disable_fcv =
driQueryOptionb(&instance->dri_options, "anv_disable_fcv");
instance->disable_xe2_ccs =
driQueryOptionb(&instance->dri_options, "anv_disable_xe2_ccs");
instance->enable_buffer_comp =
driQueryOptionb(&instance->dri_options, "anv_enable_buffer_comp");
instance->external_memory_implicit_sync =
driQueryOptionb(&instance->dri_options, "anv_external_memory_implicit_sync");
instance->compression_control_enabled =

View File

@@ -2440,9 +2440,6 @@ anv_physical_device_try_create(struct vk_instance *vk_instance,
goto fail_fd;
}
if (devinfo.ver == 20 && instance->disable_xe2_ccs)
intel_debug |= DEBUG_NO_CCS;
/* Disable Wa_16013994831 on Gfx12.0 because we found other cases where we
* need to always disable preemption :
* - https://gitlab.freedesktop.org/mesa/mesa/-/issues/5963

View File

@@ -1137,7 +1137,10 @@ struct anv_physical_device {
uint32_t dynamic_visible_mem_types;
/** Mask of memory types of protected buffers/images */
uint32_t protected_mem_types;
/** Mask of memory types of compressed buffers/images */
/**
* Mask of memory types of compressed buffers/images. This is generally
* a win for images, but a loss for buffers.
*/
uint32_t compressed_mem_types;
} memory;
@@ -1304,7 +1307,7 @@ struct anv_instance {
unsigned force_vk_vendor;
bool has_fake_sparse;
bool disable_fcv;
bool disable_xe2_ccs;
bool enable_buffer_comp;
bool compression_control_enabled;
bool anv_fake_nonlocal_memory;
bool anv_upper_bound_descriptor_pool_sampler;

View File

@@ -1201,6 +1201,10 @@ TODO: document the other workarounds.
</application>
<application name="Cyberpunk 2077" executable="Cyberpunk2077.exe">
<option name="force_vk_vendor" value="-1" />
<!--
Cyberpunk 2077 gets a big speedup by enabling buffer compression.
-->
<option name="anv_enable_buffer_comp" value="true" />
</application>
<application name="Cyberpunk 2077" executable="Cyberpunk2077.exe">
<!--
@@ -1226,6 +1230,10 @@ TODO: document the other workarounds.
</application>
<application name="Hitman 3" executable="hitman3.exe">
<option name="force_vk_vendor" value="-1"/>
<!--
Hitman 3 gets a speedup by enabling buffer compression.
-->
<option name="anv_enable_buffer_comp" value="true" />
</application>
<application name="Hogwarts Legacy" executable="HogwartsLegacy.exe">
<option name="force_vk_vendor" value="-1" />
@@ -1278,11 +1286,12 @@ TODO: document the other workarounds.
<application name="Total War: WARHAMMER III" executable="TotalWarhammer3">
<option name="anv_fake_nonlocal_memory" value="true" />
</application>
<!-- Source2 games seem confused by the CCS-only memory type, so
temporarily disable CCS. More details on the issue:
<!-- Source2 games seem confused by the CCS-only memory type being
restricted to images, so allow it for buffers. More details on
the issue:
https://gitlab.freedesktop.org/mesa/mesa/-/issues/11520 -->
<engine engine_name_match="Source2">
<option name="anv_disable_xe2_ccs" value="true" />
<option name="anv_enable_buffer_comp" value="true" />
</engine>
<!--
Disable 16-bit feature on zink and angle so that GLES mediump doesn't

View File

@@ -827,9 +827,9 @@
DRI_CONF_OPT_B(anv_disable_fcv, def, \
"Disable FCV optimization")
#define DRI_CONF_ANV_DISABLE_XE2_CCS(def) \
DRI_CONF_OPT_B(anv_disable_xe2_ccs, def, \
"Disable CCS optimization on Xe2")
#define DRI_CONF_ANV_ENABLE_BUFFER_COMP(def) \
DRI_CONF_OPT_B(anv_enable_buffer_comp, def, \
"Enable CCS on buffers where possible")
#define DRI_CONF_ANV_EXTERNAL_MEMORY_IMPLICIT_SYNC(def) \
DRI_CONF_OPT_B(anv_external_memory_implicit_sync, def, "Implicit sync on external BOs")