mirror of
https://gitlab.freedesktop.org/mesa/mesa.git
synced 2025-12-31 22:50:08 +01:00
intel/bvh: Compile and adapt bvh shaders separately into Xe1/2 and Xe3+
This change separate the encode, header, and copy shader into versions for Xe1/2 and Xe3+, including adding compile options and handling 64bit version of instance leaf for Xe3+. Signed-off-by: Kevin Chuang <kaiwenjon23@gmail.com> Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33047>
This commit is contained in:
Kevin Chuang
Marge Bot
parent
36433e932b
commit
cbc8af4555
5 changed files with 110 additions and 59 deletions
|
|
@ -211,24 +211,46 @@ struct anv_internal_node {
|
|||
#define ANV_INSTANCE_ALL_AABB 0x40
|
||||
|
||||
struct instance_leaf_part0 {
|
||||
/* shader index (24-bits) for software instancing
|
||||
/* Xe1/2:
|
||||
* shader index (24-bits) for software instancing
|
||||
* geometry mask (8-bits) used for ray masking
|
||||
*
|
||||
* Xe3+:
|
||||
* instanceContribution: 24
|
||||
* geometry mask (8-bits): 8
|
||||
*/
|
||||
uint32_t shader_index_and_geom_mask;
|
||||
uint32_t DW0;
|
||||
|
||||
/* instance contribution to hit group index (24-bits)
|
||||
|
||||
/* Xe1/2:
|
||||
* instance contribution to hit group index (24-bits)
|
||||
* Padding (5-bits)
|
||||
* DisableOpacityCull (1-bit)
|
||||
* OpaqueGeometry (1-bit)
|
||||
* Padding (1-bit)
|
||||
*
|
||||
* Xe3+:
|
||||
* insFlags: 8
|
||||
* ComparisonMode: 1
|
||||
* ComparisonValue: 7
|
||||
* pad0: 8
|
||||
* subType: 4
|
||||
* pad1: 1
|
||||
* DisableOpacityCull: 1
|
||||
* OpaqueGeometry: 1
|
||||
* IgnoreRayMultiplier: 1
|
||||
*/
|
||||
uint32_t instance_contribution_and_geom_flags;
|
||||
uint32_t DW1;
|
||||
|
||||
/* 48 bit start node of the instanced object
|
||||
/* Xe1/2:
|
||||
* 48 bit start node of the instanced object
|
||||
* instFlags (8-bits)
|
||||
* Padding (16-bits)
|
||||
*
|
||||
* Xe3+:
|
||||
* 64 bit start node pointer
|
||||
*/
|
||||
uint64_t start_node_ptr_and_inst_flags;
|
||||
uint64_t QW_startNodePtr;
|
||||
|
||||
/* 1st row of Worl2Obj transform */
|
||||
float world2obj_vx_x;
|
||||
|
|
|
|||
|
|
@ -136,11 +136,15 @@ main(void)
|
|||
DEREF(instance_leaf).part1.bvh_ptr = blas_ptr;
|
||||
|
||||
/* set the startNodePtr to blas_ptr + ANV_HEADER_SIZE */
|
||||
uint64_t mask = 0x0000fffffffffffful;
|
||||
uint64_t new_startNodePtr = blas_ptr + ANV_RT_BVH_HEADER_SIZE;
|
||||
#if GFX_VERx10 >= 300
|
||||
DEREF(instance_leaf).part0.QW_startNodePtr = new_startNodePtr;
|
||||
#else
|
||||
uint64_t mask = 0x0000fffffffffffful;
|
||||
/* clear bits and set */
|
||||
DEREF(instance_leaf).part0.start_node_ptr_and_inst_flags =
|
||||
(DEREF(instance_leaf).part0.start_node_ptr_and_inst_flags & ~mask) | (new_startNodePtr & mask);
|
||||
DEREF(instance_leaf).part0.QW_startNodePtr =
|
||||
(DEREF(instance_leaf).part0.QW_startNodePtr & ~mask) | (new_startNodePtr & mask);
|
||||
#endif
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -29,7 +29,7 @@ layout(push_constant) uniform CONSTS {
|
|||
encode_args args;
|
||||
};
|
||||
|
||||
uint64_t
|
||||
uint32_t
|
||||
get_instance_flag(uint32_t src)
|
||||
{
|
||||
uint32_t flags = src & 0xff;
|
||||
|
|
@ -129,20 +129,6 @@ encode_leaf_node(uint32_t type, uint64_t src_node, uint64_t dst_node, REF(anv_ac
|
|||
uint64_t start_node_ptr = uint64_t(src.base_ptr) + DEREF(blas_header).rootNodeOffset;
|
||||
|
||||
uint32_t sbt_offset_and_flags = src.sbt_offset_and_flags;
|
||||
|
||||
uint32_t shader_index_and_geom_mask = 0;
|
||||
shader_index_and_geom_mask |= (src.custom_instance_and_mask & 0xff000000);
|
||||
DEREF(dst_instance).part0.shader_index_and_geom_mask = shader_index_and_geom_mask;
|
||||
|
||||
uint32_t instance_contribution_and_geom_flags = 0;
|
||||
instance_contribution_and_geom_flags |= src.sbt_offset_and_flags & 0xffffff;
|
||||
instance_contribution_and_geom_flags |= (1 << 29);
|
||||
instance_contribution_and_geom_flags |=
|
||||
(get_instance_flag(src.sbt_offset_and_flags >> 24) == ANV_INSTANCE_FLAG_FORCE_OPAQUE ?
|
||||
ANV_GEOMETRY_FLAG_OPAQUE : 0) << 30;
|
||||
DEREF(dst_instance).part0.instance_contribution_and_geom_flags =
|
||||
instance_contribution_and_geom_flags;
|
||||
|
||||
uint32_t instance_flags = DEREF(blas_header).instance_flags;
|
||||
if (((sbt_offset_and_flags >> 24) & (VK_GEOMETRY_INSTANCE_FORCE_OPAQUE_BIT_KHR |
|
||||
VK_GEOMETRY_INSTANCE_FORCE_NO_OPAQUE_BIT_KHR)) != 0) {
|
||||
|
|
@ -152,9 +138,39 @@ encode_leaf_node(uint32_t type, uint64_t src_node, uint64_t dst_node, REF(anv_ac
|
|||
VK_GEOMETRY_INSTANCE_FORCE_NO_OPAQUE_BIT_KHR);
|
||||
}
|
||||
|
||||
DEREF(dst_instance).part0.start_node_ptr_and_inst_flags =
|
||||
#if GFX_VERx10 >= 300
|
||||
DEREF(dst_instance).part0.QW_startNodePtr = start_node_ptr;
|
||||
uint32_t instance_contribution_and_geom_mask = 0;
|
||||
instance_contribution_and_geom_mask |= src.sbt_offset_and_flags & 0xffffff;
|
||||
instance_contribution_and_geom_mask |= (src.custom_instance_and_mask & 0xff000000);
|
||||
DEREF(dst_instance).part0.DW0 = instance_contribution_and_geom_mask;
|
||||
|
||||
uint32_t inst_flags_and_the_rest = 0;
|
||||
inst_flags_and_the_rest |= get_instance_flag(instance_flags | (src.sbt_offset_and_flags >> 24));
|
||||
inst_flags_and_the_rest |= (1 << 29);
|
||||
inst_flags_and_the_rest |=
|
||||
((get_instance_flag(src.sbt_offset_and_flags >> 24) & ANV_INSTANCE_FLAG_FORCE_OPAQUE) != 0 ?
|
||||
ANV_GEOMETRY_FLAG_OPAQUE : 0) << 30;
|
||||
|
||||
DEREF(dst_instance).part0.DW1 = inst_flags_and_the_rest;
|
||||
|
||||
#else
|
||||
uint32_t shader_index_and_geom_mask = 0;
|
||||
shader_index_and_geom_mask |= (src.custom_instance_and_mask & 0xff000000);
|
||||
DEREF(dst_instance).part0.DW0 = shader_index_and_geom_mask;
|
||||
|
||||
uint32_t instance_contribution_and_geom_flags = 0;
|
||||
instance_contribution_and_geom_flags |= src.sbt_offset_and_flags & 0xffffff;
|
||||
instance_contribution_and_geom_flags |= (1 << 29);
|
||||
instance_contribution_and_geom_flags |=
|
||||
((get_instance_flag(src.sbt_offset_and_flags >> 24) & ANV_INSTANCE_FLAG_FORCE_OPAQUE) != 0 ?
|
||||
ANV_GEOMETRY_FLAG_OPAQUE : 0) << 30;
|
||||
DEREF(dst_instance).part0.DW1 = instance_contribution_and_geom_flags;
|
||||
|
||||
DEREF(dst_instance).part0.QW_startNodePtr =
|
||||
(start_node_ptr & ((1ul << 48) - 1)) |
|
||||
(get_instance_flag(instance_flags | (src.sbt_offset_and_flags >> 24)) << 48);
|
||||
(uint64_t(get_instance_flag(instance_flags | (src.sbt_offset_and_flags >> 24))) << 48);
|
||||
#endif
|
||||
|
||||
mat4 transform = mat4(src.otw_matrix);
|
||||
|
||||
|
|
|
|||
|
|
@ -4,21 +4,16 @@
|
|||
|
||||
# source file, output name, defines
|
||||
bvh_shaders = [
|
||||
[
|
||||
'encode.comp',
|
||||
'encode',
|
||||
[],
|
||||
],
|
||||
[
|
||||
'header.comp',
|
||||
'header',
|
||||
[],
|
||||
],
|
||||
[
|
||||
'copy.comp',
|
||||
'copy',
|
||||
[]
|
||||
],
|
||||
'encode.comp',
|
||||
'header.comp',
|
||||
'copy.comp',
|
||||
]
|
||||
|
||||
# A mapping: [filename version, GFX_VERx10 define version]
|
||||
gfx_versions = [
|
||||
['125', '125'],
|
||||
['20', '200'],
|
||||
['30', '300']
|
||||
]
|
||||
|
||||
anv_bvh_include_dir = dir_source_root + '/src/intel/vulkan/bvh'
|
||||
|
|
@ -29,21 +24,28 @@ anv_bvh_includes = files(
|
|||
'anv_bvh.h',
|
||||
)
|
||||
|
||||
foreach s : bvh_shaders
|
||||
command = [
|
||||
prog_glslang, '-V', '-I' + vk_bvh_include_dir, '-I' + anv_bvh_include_dir, '--target-env', 'spirv1.5', '-x', '-o', '@OUTPUT@', '@INPUT@'
|
||||
]
|
||||
command += glslang_quiet
|
||||
foreach shader : bvh_shaders
|
||||
foreach gfx: gfx_versions
|
||||
file_name_ver = gfx[0]
|
||||
define_ver = gfx[1]
|
||||
command = [
|
||||
prog_glslang, '-V', '-I' + vk_bvh_include_dir, '-I' + anv_bvh_include_dir,
|
||||
'--target-env', 'spirv1.5', '-x', '-o', '@OUTPUT@', '@INPUT@',
|
||||
'-DGFX_VERx10=' + define_ver # so that we can use this macro inside shaders
|
||||
]
|
||||
command += glslang_quiet
|
||||
|
||||
foreach define : s[2]
|
||||
command += '-D' + define
|
||||
endforeach
|
||||
shader_name = shader.split('.')[0]
|
||||
output_name = 'gfx' + file_name_ver + '_' + shader_name + '.spv.h'
|
||||
# By doing this, encode.comp with DGFX_VERx10=300 will be compiled to gfx30_encode.spv.h
|
||||
# So the genX(encode).spv.h in genX_acceleration_structure.c can find the right file
|
||||
|
||||
bvh_spv += custom_target(
|
||||
s[1] + '.spv.h',
|
||||
input : s[0],
|
||||
output : s[1] + '.spv.h',
|
||||
command : command,
|
||||
depend_files: [vk_bvh_includes, anv_bvh_includes],
|
||||
)
|
||||
bvh_spv += custom_target(
|
||||
output_name,
|
||||
input : shader,
|
||||
output : output_name,
|
||||
command : command,
|
||||
depend_files: [vk_bvh_includes, anv_bvh_includes],
|
||||
)
|
||||
endforeach
|
||||
endforeach
|
||||
|
|
|
|||
|
|
@ -229,16 +229,23 @@ debug_record_as_to_bvh_dump(struct anv_cmd_buffer *cmd_buffer,
|
|||
}
|
||||
}
|
||||
|
||||
#define STRINGIFY_HELPER(x) #x
|
||||
#define STRINGIFY(x) STRINGIFY_HELPER(x)
|
||||
|
||||
#define ENCODE_SPV_PATH STRINGIFY(bvh/genX(encode).spv.h)
|
||||
#define HEADER_SPV_PATH STRINGIFY(bvh/genX(header).spv.h)
|
||||
#define COPY_SPV_PATH STRINGIFY(bvh/genX(copy).spv.h)
|
||||
|
||||
static const uint32_t encode_spv[] = {
|
||||
#include "bvh/encode.spv.h"
|
||||
#include ENCODE_SPV_PATH
|
||||
};
|
||||
|
||||
static const uint32_t header_spv[] = {
|
||||
#include "bvh/header.spv.h"
|
||||
#include HEADER_SPV_PATH
|
||||
};
|
||||
|
||||
static const uint32_t copy_spv[] = {
|
||||
#include "bvh/copy.spv.h"
|
||||
#include COPY_SPV_PATH
|
||||
};
|
||||
|
||||
static VkResult
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue