1 # What's new in vkd3d 1.11 (5 Mar 2024)
5 - Descriptor updates happen asynchronously on an internal worker thread, for
6 a minor performance improvement in applications that update many
9 - When the VK_EXT_mutable_descriptor_type extension is available, libvkd3d
10 will make more efficient use of descriptor pools and sets.
12 - When the VK_EXT_shader_viewport_index_layer extension is available,
13 libvkd3d supports indexing viewport and render target arrays from vertex
14 and tessellation evaluation shaders.
16 - Support for standard (i.e., black and white) border colours is
19 - The GetResourceAllocationInfo1() method of the ID3D12Device4 interface is
22 - The ID3D12Device7 interface is supported.
24 - The ID3D12Resource2 interface is supported.
26 - Several new feature queries are supported:
27 - D3D12_FEATURE_D3D12_OPTIONS6
28 - D3D12_FEATURE_D3D12_OPTIONS7
29 - D3D12_FEATURE_D3D12_OPTIONS8
30 - D3D12_FEATURE_D3D12_OPTIONS9
31 - D3D12_FEATURE_D3D12_OPTIONS10
32 - D3D12_FEATURE_D3D12_OPTIONS11
33 - D3D12_FEATURE_D3D12_OPTIONS12
34 - D3D12_FEATURE_D3D12_OPTIONS13
38 - Initial support for compiling legacy Direct3D bytecode to SPIR-V.
40 - Experimental support for compiling DirectX Intermediate Language (DXIL) to
41 SPIR-V and Direct3D shader assembly. Being an experimental feature, this
42 requires building vkd3d with the ‘-DVKD3D_SHADER_UNSUPPORTED_DXIL’
43 preprocessor option. Note that enabling this feature will affect the
44 capabilities reported by libvkd3d as well, and may cause previously
45 working applications to break due to attempting to use incomplete DXIL
46 support. No API or ABI stability guarantees are provided for experimental
49 - New features for the HLSL source type:
50 - Initial support for the ‘fx_2_0’, ‘fx_4_0’, ‘fx_4_1’, and ‘fx_5_0’
51 profiles, using the new ‘VKD3D_SHADER_TARGET_FX’ target type.
52 - Support for ‘Buffer’ resources.
53 - The acos(), asin(), atan(), and atan2() intrinsic functions are
55 - Explicit register assignment using the ‘register()’ keyword in shader
56 model 1-3 profiles. This was previously only supported in shader model
58 - Casts from integer to floating-point types in shader model 1-3 profiles.
59 - Support for various input/output semantics:
60 - SV_InstanceID in shader model 4+ fragment shaders.
61 - SV_PrimitiveID in shader model 4+ fragment shaders. In previous
62 versions this was only supported in shader model 4+ geometry shaders.
63 - SV_RenderTargetArrayIndex in shader model 4+ vertex and fragment shaders.
64 - SV_ViewportArrayIndex in shader model 4+ vertex and fragment shaders.
65 - Support for various rasteriser-ordered view types. Specifically:
66 - RasterizerOrderedBuffer
67 - RasterizerOrderedStructuredBuffer
68 - RasterizerOrderedTexture1D
69 - RasterizerOrderedTexture1DArray
70 - RasterizerOrderedTexture2D
71 - RasterizerOrderedTexture2DArray
72 - RasterizerOrderedTexture3D
74 - New features for the SPIR-V target type:
75 - Support for globally coherent unordered access views. These have the
76 ‘globallycoherent’ storage class in HLSL, and the ‘_glc’ suffix in
78 - Support for thread group unordered access view barriers. This
79 corresponds to ‘sync_ugroup’ instructions in Direct3D assembly.
80 - When the SPV_EXT_viewport_index_layer extension is supported, vertex and
81 tessellation evaluation shaders can write render target and viewport
82 array indices. This corresponds to the ‘SV_RenderTargetArrayIndex’ and
83 ‘SV_ViewportArrayIndex’ HLSL output semantics.
86 - The VKD3D_SHADER_COMPILE_OPTION_FEATURE compile option can be used to
87 specify features available in the target environment. The
88 VKD3D_SHADER_COMPILE_OPTION_FEATURE_INT64 flag indicates support for
89 64-bit integer types in the SPIR-V target environment. The
90 VKD3D_SHADER_COMPILE_OPTION_FEATURE_FLOAT64 flag indicates support for
91 64-bit floating-point types in the SPIR-V target environment. For
92 backward compatibility, VKD3D_SHADER_API_VERSION_1_10 and earlier also
93 imply support for 64-bit floating-point types.
94 - The VKD3D_SHADER_SPIRV_EXTENSION_EXT_VIEWPORT_INDEX_LAYER enumeration
95 value indicates support for the SPV_EXT_viewport_index_layer extension
96 in the SPIR-V target environment.
100 - When available, the following Vulkan extensions are enabled by
101 D3D12CreateDeviceVKD3D() and D3D12CreateDevice():
102 - VK_KHR_android_surface
103 - VK_KHR_wayland_surface
104 - VK_KHR_win32_surface
105 - VK_KHR_xlib_surface
106 - VK_EXT_metal_surface
109 Previous versions of vkd3d-utils enabled VK_KHR_xcb_surface and
110 VK_MVK_macos_surface. In practice this means that
111 D3D12CreateDevice()/D3D12CreateDeviceVKD3D() can be used on the
112 corresponding additional window systems.
115 - D3DReflect() is used to retrieve information about shaders. It currently
116 supports retrieving information about input, output, and patch constant
117 parameters using the ID3D12ShaderReflection interface.
118 - D3DDisassemble() is used to disassemble legacy Direct3D bytecode (shader
119 model 1-3) and ‘Tokenized Program Format’ (shader model 4 and 5)
124 - The new ‘fx’ target is used for outputting Direct3D effects when compiling
125 HLSL ‘fx_2_0’, ‘fx_4_0’, ‘fx_4_1’, and ‘fx_5_0’ profiles.
129 - The minimum required version of Vulkan-Headers for this release is version
132 # What's new in vkd3d 1.10 (6 Dec 2023)
136 - Creating pipeline state objects from pipeline state stream descriptions is
138 - Depth-bounds testing is implemented.
139 - When the VK_KHR_maintenance2 extension is available, libvkd3d will
140 explicitly specify the usage flags of Vulkan image views. This is
141 particularly useful on MoltenVK, where 2D-array views of 3D textures are
142 subject to usage restrictions.
143 - The D3D12_FORMAT_SUPPORT2_UAV_TYPED_LOAD and/or
144 D3D12_FORMAT_SUPPORT2_UAV_TYPED_STORE feature flags are reported for
145 UAV formats when the ‘shaderStorageImageReadWithoutFormat’ and/or
146 ‘shaderStorageImageWriteWithoutFormat’ Vulkan device features are
148 - The ID3D12Device5 interface is supported.
149 - The ID3D12GraphicsCommandList5 interface is supported.
150 - The ID3D12Resource1 interface is supported.
154 - New features for the HLSL source type:
155 - Support for the following intrinsic functions:
157 - degrees() and radians()
160 - tex2Dlod(), tex2Dproj(), texCUBEproj(), and tex3Dproj()
161 - Constant folding support for more expression types. In particular:
162 - ternary operators and branches
163 - reciprocal square roots
165 - logical ‘not’ on booleans
166 - bitwise complements
168 - ceil(), floor(), frac(), and saturate()
169 - Support for dynamic indexing of arrays.
170 - Support for ‘break’ and ‘continue’ statements.
171 - Support for ‘switch’ statements.
172 - The ‘linear’, ‘centroid’, and ‘noperspective’ interpolation modifiers
174 - The ‘RWTexture1DArray’ and ‘RWTexture2DArray’ unordered access view
176 - ‘\[loop\]’ attributes are accepted on loops.
177 - u/U and l/L suffixes on integer constants.
179 - Floating-point values are explicitly clamped to the upper and lower bounds
180 of the target type by ‘ftoi’ and ‘ftou’ instructions when targeting
181 SPIR-V. Similarly, NaNs are flushed to zero. Some hardware/drivers would
182 already do this implicitly, but behaviour for such inputs is undefined as
183 far as SPIR-V is concerned.
185 - The VKD3D_SHADER_CONFIG environment variable can be used to modify the
186 behaviour of libvkd3d-shader at run-time, analogous to the existing
187 VKD3D_CONFIG environment variable for libvkd3d. See the README for a list
188 of supported options.
190 - When scanning legacy Direct3D bytecode using vkd3d_shader_scan(),
191 descriptor information for shader model 2 and 3 combined resource-sampler
192 pairs is returned in the vkd3d_shader_scan_descriptor_info structure.
193 Note that this information is not yet available for shader model 1
194 sources, although this will likely be added in a future release.
196 - The Direct3D shader assembly target supports the ‘rasteriser ordered view’
197 flag (‘_rov’) on unordered access view declarations.
200 - The VKD3D_SHADER_COMPILE_OPTION_BACKWARD_COMPATIBILITY compile
201 option can be used to specify backward compatibility options. The
202 VKD3D_SHADER_COMPILE_OPTION_BACKCOMPAT_MAP_SEMANTIC_NAMES flag is
203 the only currently supported flag, and can be used to specify that
204 shader model 1-3 semantic names should be mapped to their shader model
205 4+ system value equivalents when compiling HLSL sources.
206 - The VKD3D_SHADER_COMPILE_OPTION_FRAGMENT_COORDINATE_ORIGIN compile
207 option can be used to specify the origin of fragment coordinates for
208 SPIR-V targets. This is especially useful in OpenGL environments, where
209 the origin may be different than in Direct3D or Vulkan environments.
210 - The vkd3d_shader_scan_combined_resource_sampler_info structure
211 extends the vkd3d_shader_compile_info structure, and can be used to
212 retrieve information about the combined resource-sampler pairs used by a
213 shader. This is especially useful when compiling shaders for usage in
214 environments without separate binding points for samplers and resources,
216 - vkd3d_shader_free_scan_combined_resource_sampler_info() is used
217 to free vkd3d_shader_scan_combined_resource_sampler_info
222 - Passing the D3DCOMPILE_ENABLE_BACKWARDS_COMPATIBILITY flag to
223 D3DCompile() and D3DCompile2() will enable mapping shader model 1-3
224 semantic names to their shader model 4+ system value equivalents.
227 - D3DGetBlobPart() is used to retrieve specific parts of DXBC blobs.
228 - D3DGetDebugInfo() is used to retrieve debug information from DXBC blobs.
229 - D3DGetInputAndOutputSignatureBlob() is used to retrieve input and output
230 signatures from DXBC blobs.
231 - D3DGetInputSignatureBlob() is used to retrieve input signatures from
233 - D3DGetOutputSignatureBlob() is used to retrieve output signatures from
235 - D3DStripShader() is used to remove specific parts from DXBC blobs.
239 - The ‘--fragment-coordinate-origin’ option can be used to specify the
240 origin of fragment coordinates for SPIR-V targets.
242 - The ‘--semantic-compat-map’ option can be used to specify that shader
243 model 1-3 semantic names should be mapped to their shader model 4+ system
244 value equivalents when compiling HLSL sources.
248 - The ‘--list’ and ‘--list-data’ options now also output the offsets of
249 sections inside the input data.
253 - The minimum required version of Vulkan-Headers for this release is version
256 - When available, the libEGL and libOpenGL libraries are used to run the
257 vkd3d tests in additional configurations. These libraries are not used by
260 - The SONAME_LIBDXCOMPILER configure variable can be used specify the
261 shared object name of the dxcompiler library. When available, it's used to
262 run the vkd3d tests in additional configurations. The dxcompiler library
263 is not used by vkd3d itself.
266 # What's new in vkd3d 1.9 (21 Sep 2023)
270 - Copying between depth/stencil and colour formats in
271 ID3D12GraphicsCommandList::CopyResource() is supported.
272 - The ID3D12Fence1 interface is supported.
277 - vkd3d_shader_scan() supports retrieving descriptor information for ‘d3dbc’
278 shaders. This is one of the requirements for eventual SPIR-V generation from
281 - New features for the HLSL source type:
282 - Support for the following intrinsic functions:
284 - ddx_coarse() and ddy_coarse()
285 - ddx_fine() and ddy_fine()
286 - tex1D(), tex2D(), texCUBE(), and tex3D()
287 - Constant folding support for more expression types. In particular:
288 - comparison operators
289 - floating-point min() and max()
290 - logical ‘and’ and ‘or’
294 - Support for multi-sample texture object declarations without explicit
295 sample counts in shader model 4.1 and later shaders.
296 - Support for using constant expressions as sample counts in multi-sample
297 texture object declarations.
298 - Support for variable initialisers using variables declared earlier in the
299 same declaration list. E.g., ‘float a = 1, b = a, c = b + 1;’.
300 - The GetDimensions() texture object method is implemented.
301 - Matrix swizzles are implemented.
302 - Parser support for if-statement attributes like ‘[branch]’ and
304 - Support for the ‘inline’ function modifier.
306 - Previously, vkd3d_shader_compile() would in some cases return VKD3D_OK
307 despite compilation failing when targeting legacy Direct3D bytecode. These
308 cases have been fixed.
310 - Various HLSL preprocessor fixes for edge cases related to stringification.
312 - SPIR-V target support for the ‘linear noperspective centroid’ input
316 - The vkd3d_shader_scan_signature_info structure extends the
317 vkd3d_shader_compile_info structure, and can be used to retrieve
318 descriptions of ‘dxbc-tpf’ and ‘d3dbc’ shader inputs and outputs.
319 - vkd3d_shader_free_scan_signature_info() is used to free
320 vkd3d_shader_scan_signature_info structures.
321 - The VKD3D_SHADER_COMPILE_OPTION_PACK_MATRIX_ORDER compile option can be
322 used to specify the default matrix packing order for HLSL sources.
323 - The vkd3d_shader_varying_map_info structure extends the
324 vkd3d_shader_compile_info structure, and can be used to specify a mapping
325 between the outputs of a shader stage and the inputs of the next shader
327 - vkd3d_shader_build_varying_map() is used to build a mapping between the
328 outputs of a shader stage and the inputs of the next shader stage.
329 - The VKD3D_SHADER_DESCRIPTOR_INFO_FLAG_RAW_BUFFER flag returned as part of
330 the vkd3d_shader_descriptor_info structure indicates the descriptor refers
331 to a byte-addressed (‘raw’) buffer resource.
336 - The ‘--matrix-storage-order’ option can used to specify the default matrix
337 storage order for HLSL sources.
342 - vkd3d-dxbc is a new utility that can be used to inspect the contents of DXBC
346 # What's new in vkd3d 1.8 (22 Jun 2023)
350 - Performance improvements have been made to the code that handles descriptor
351 updates. In some applications the improvement can be quite significant.
353 - Host-visible descriptor heaps are persistently mapped on creation. Some
354 applications access resource data from the CPU after calling Unmap(), and
355 that's supposed to work in practice.
357 - 1-dimensional texture unordered-access views and shader resource views are
360 - Shader resource view, unordered access view, and constant buffer view root
361 descriptors with NULL GPU addresses are supported.
363 - Direct3D 12 descriptor heap destruction is delayed until all contained
364 resources are destroyed.
369 - New features for the HLSL source type:
370 - Support for the ternary conditional operator "?:".
371 - Support for "discard" statements.
372 - Support for the "packoffset" keyword.
373 - Support for semantics on array types.
374 - Support for RWBuffer loads and stores.
375 - Register allocation for arrays and structures of resources and samplers
377 - Support for the SV_IsFrontFace pixel shader system-value semantics.
378 - Support for using constant expressions as array sizes and indices.
379 - Support for dynamic selection of vector components.
380 - Support for the following intrinsic functions:
386 - log(), log2(), and log10()
389 - The SampleBias(), SampleCmp(), SampleCmpLevelZero(), and SampleGrad()
390 texture object methods are implemented.
391 - Support for the case-insensitive variants of the "vector" and "matrix"
393 - Parser support for the "unroll" loop attribute. A warning is output for
394 "unroll" without iteration count, and an error is output when an iteration
395 count is specified. Actual unrolling is not implemented yet.
396 - Parser support for RWStructuredBuffer resources.
397 - Parser support for SamplerComparisonState objects. Note that outputting
398 compiled effects is not supported yet, but parsing these allows shaders
399 containing SamplerComparisonState state objects to be compiled.
401 - More improvements to HLSL support for the Direct3D shader model 1/2/3
404 - The section alignment of DXBC blobs produced by
405 vkd3d_shader_serialize_dxbc() matches those produced by d3dcompiler more
408 - The "main" function for shaders produced by the SPIR-V target is always
409 terminated, even when the source was a TPF shader without explicit "ret"
412 - Relative addressing of shader input registers is supported by SPIR-V
416 # What's new in vkd3d 1.7 (23 Mar 2023)
420 - New features for the HLSL source type:
421 - Support for calling user-defined functions.
422 - Support for array parameters to user-defined functions.
423 - Much improved support for the Direct3D shader model 1/2/3 profiles.
424 - Support for the SV_DispatchThreadID, SV_GroupID, and SV_GroupThreadID
425 compute shader system-value semantics.
426 - Support for the optional "offset" parameter of the texture object Load()
428 - Support for the all() intrinsic function.
429 - Support for the distance() intrinsic function.
430 - Support for the exp() and exp2() intrinsic functions.
431 - Support for the frac() intrinsic function.
432 - Support for the lit() intrinsic function.
433 - Support for the reflect() intrinsic function.
434 - Support for the sin() and cos() intrinsic functions.
435 - Support for the smoothstep() intrinsic function.
436 - Support for the sqrt() and rsqrt() intrinsic functions.
437 - Support for the step() intrinsic function.
438 - Support for the transpose() intrinsic function.
439 - Support for the case-insensitive variants of the "float" and "dword" data
441 - Partial support for minimum precision data types like "min16float". These
442 are currently interpreted as their regular counterparts.
443 - Improved constant propagation support, in particular to constant
444 propagation through swizzles.
446 - HLSL static variables are now properly zero-initialised.
448 - The Direct3D shader model 4 and 5 disassembler outputs sample counts for
449 multi-sampled resource declarations.
452 - vkd3d_shader_parse_dxbc() provides support for parsing DXBC blobs.
453 - vkd3d_shader_serialize_dxbc() provides support for serialising DXBC blobs.
454 - vkd3d_shader_free_dxbc() is used to free vkd3d_shader_dxbc_desc
455 structures, as returned by vkd3d_shader_parse_dxbc().
456 - The VKD3D_SHADER_COMPILE_OPTION_WRITE_TESS_GEOM_POINT_SIZE compile option
457 can be used to specify whether SPIR-V shaders targeting Vulkan
458 environments should write point sizes for geometry and tessellation
459 shaders. If left unspecified, point sizes will be written.
462 # What's new in vkd3d 1.6 (7 Dec 2022)
466 - New features for the HLSL source type:
467 - Initial support for compute shaders.
468 - Improved support for initialisation and assignment of compound objects
469 like structures and arrays, including casts and implicit conversions.
470 - Support for loads and stores of texture resource unordered-access views.
471 - Support for function attributes. In particular, the required "numthreads"
472 attribute for compute shader entry points is now supported.
473 - Support for the asuint() intrinsic function.
474 - Support for the length() intrinsic function.
475 - Support for the normalize() intrinsic function.
476 - Support for integer division and modulus.
477 - Support for taking the absolute value of integers.
478 - Support for floating-point modulus.
482 - The VKD3D_SHADER_DESCRIPTOR_INFO_FLAG_UAV_ATOMICS descriptor info flag is
483 used to indicate that atomic operations are used on unordered-access view
489 - vkd3d debug output is prefixed with "vkd3d:" in order to make it easier to
490 distinguish from output produced by applications or other libraries.
495 - The demos now use libvkd3d-shader to compile HLSL shaders at run-time.
498 # What's new in vkd3d 1.5 (22 Sep 2022)
502 - New features for the HLSL source type:
503 - Improved support for HLSL object types (like e.g. ‘Texture2D’) inside
504 structures and arrays.
505 - Implicitly sized array initialisers.
506 - Support for the dot() intrinsic function.
507 - Support for the ldexp() intrinsic function.
508 - Support for the lerp() intrinsic function.
509 - Support for the logical ‘and’, ‘or’, and ‘not’ operators in shader model 4
511 - Support for casts from ‘bool’ types in shader model 4 and 5 targets.
512 - Constant folding for integer bitwise operations.
513 - Constant folding for integer min() and max().
516 - The VKD3D_SHADER_COMPILE_OPTION_TYPED_UAV compile option can be used to
517 specify the SPIR-V format to use for typed unordered access view loads.
518 When set to ‘Unknown’, and the ‘shaderStorageImageReadWithoutFormat’
519 feature is enabled in the target environment, this allows typed loads from
520 multicomponent format unordered access views. If left unspecified, a R32
521 format will be used, like in previous versions of libvkd3d-shader.
524 # What's new in vkd3d 1.4 (22 Jun 2022)
528 - A new descriptor heap implementation using the VK_EXT_descriptor_indexing
529 extension. In particular, the new implementation is more efficient when
530 large descriptor heaps are used by multiple command lists. The new
531 ‘virtual_heaps’ configuration option can be used to select the original
532 implementation even when the VK_EXT_descriptor_indexing extension is
535 - A new fence implementation using the VK_KHR_timeline_semaphore extension.
536 The new implementation addresses a number of edge cases the original
537 implementation was unable to, as well as being somewhat more efficient.
539 - When the VK_EXT_robustness2 extension is available, it is used to implement
540 null views. This more accurately matches Direct3D 12 behaviour. For example,
541 all reads from such a null view return zeroes, while that isn't necessarily
542 the case for out-of-bounds reads with the original implementation.
545 - vkd3d_set_log_callback() allows writing log output via a custom callback.
546 This can be used to integrate vkd3d's log output with other logging
552 - New features for the HLSL source type:
553 - Support for integer arithmetic, bitwise and shift operations.
554 - Support for matrix and vector subscripting.
555 - Support for the mul() intrinsic function.
556 - Support for matrix copying, casting, and entry-wise operations.
557 - Support for complex initialisers.
558 - Support for the ‘nointerpolation’ modifier. This modifier is applied by
559 default to integer variables.
560 - Support for the SV_VertexID semantic.
561 - Support for matrix-typed varyings.
562 - Constant folding for a number of operators.
563 - Copy propagation across branches and loops. This allows use of non-numeric
564 variables anywhere in a program, as well as more optimised code for
565 accessing numeric variables within branches and loops.
567 - The disassembler supports the shader model 5 ‘msad’ instruction.
570 - vkd3d_shader_set_log_callback() allows writing log output via a custom
577 - vkd3d_utils_set_log_callback() allows writing log output via a custom
583 - The minimum required version of Vulkan-Headers and SPIRV-Headers for this
584 release is version 1.2.139.
586 - The SONAME_LIBVULKAN configure variable can be used to specify the shared
587 object name of the Vulkan library. Because vkd3d loads the Vulkan library
588 dynamically, specifying this removes the need for a Vulkan import library at
591 - The ‘crosstests’ target no longer builds Win32/PE demos or tests when these
592 were not enabled at configure time.
595 # What's new in vkd3d 1.3 (2 Mar 2022)
599 - Newly implemented Direct3D 12 features:
600 - Root signature support for unbounded descriptor tables.
601 - Unordered-access view counters in pixel shaders. These were previously
602 only supported in compute shaders.
603 - Output merger logical operations.
604 - Retrieving CPU/GPU timestamp calibration values. This requires support for
605 the VK_EXT_calibrated_timestamps extension.
606 - The ‘mirror_once’ texture addressing mode. This requires support for the
607 VK_KHR_sampler_mirror_clamp_to_edge extension.
610 - The vkd3d_host_time_domain_info structure extends the
611 vkd3d_instance_create_info structure, and can be used to specify how to
612 convert between timestamps and tick counts. If left unspecified, a tick
613 is assumed to take 100 nanoseconds.
621 - Initial support for HLSL compilation and preprocessing. This is an ongoing
622 effort; although support for many features is already implemented, support
623 for many more isn't yet.
624 - Support for disassembling Direct3D byte-code shaders to Direct3D assembly.
625 - Support for parsing the legacy Direct3D byte-code format used by Direct3D
626 shader model 1, 2, and 3 shaders. In the current vkd3d-shader release,
627 only Direct3D assembly is supported as a target for these; we intend to
628 support SPIR-V as a target in a future release.
630 - New features for the SPIR-V target:
631 - Support for various aspects of Direct3D shader model 5.1 descriptor
632 arrays, including unbounded descriptor arrays, UAV counter arrays, dynamic
633 indexing of descriptor arrays, and non-uniform indexing of descriptor
634 arrays. With the exception of some special cases, this requires support
635 for the SPV_EXT_descriptor_indexing extension in the target environment.
636 - Support for double precision floating-point operations.
637 - Support for indirect addressing of tessellation control shader inputs.
638 - Stencil export. I.e., writing stencil values from shaders. This requires
639 support for the SPV_EXT_shader_stencil_export extension in the target
641 - Support for the Direct3D shader model 4+ ‘precise’ modifier.
642 - Support for Direct3D shader model 4+ global resource memory barriers.
645 - vkd3d_shader_preprocess() provides support for preprocessing shaders.
646 - The vkd3d_shader_preprocess_info structure extends the
647 vkd3d_shader_compile_info structure, and can be used to specify
648 preprocessing parameters like preprocessor macro definitions.
649 - The vkd3d_shader_hlsl_source_info structure extends the
650 vkd3d_shader_compile_info structure, and can be used to specify HLSL
651 compilation parameters like the target profile and entry point.
652 - The vkd3d_shader_descriptor_offset_info structure extends the
653 vkd3d_shader_interface_info structure, and can be used to specify offsets
654 into descriptor arrays referenced by shader interface bindings. This
655 allows mapping multiple descriptor arrays in a shader to a single binding
656 point in the target environment, and helps with mapping between the
657 Direct3D 12 and Vulkan binding models.
658 - The VKD3D_SHADER_COMPILE_OPTION_API_VERSION compile option can
659 be used to specify the version of the libvkd3d-shader API the
660 application is targetting. If left unspecified,
661 VKD3D_SHADER_API_VERSION_1_2 will be used.
663 - Various shader translation fixes, for tessellation shaders in particular.
668 - New source and target types:
669 - The ‘hlsl’ source type specifies High Level Shader Language source code.
670 - The ‘d3d-asm’ target type specifies Direct3D assembly shaders.
671 - The ‘d3dbc’ format specifies legacy Direct3D byte-code, which is used for
672 Direct3D shader model 1, 2, and 3 shaders.
673 - The existing ‘dxbc-tpf’ format can now also be used as a target format.
675 - New command line options:
676 - ‘-E’ can be used to specify the input should only be preprocessed.
677 - ‘-e’/‘--entry’ can be used to specify the entry point for HLSL and/or
679 - ‘-p’/‘--profile’ can be used to specify the target profile for HLSL
682 - When no source type is explicitly specified, vkd3d-compiler will attempt to
683 determine the source type from the provided input. Note that this is
684 intended as a convenience for interactive usage only, and the heuristics
685 used are subject to future change. Non-interactive usage of vkd3d-compiler,
686 for example in build scripts, should always explicitly specify source and
689 - When no target type is explicitly specified, a default will be chosen based
690 on the source type. Like the earlier mentioned source type detection, this
691 is intended for interactive usage only.
693 - vkd3d-compiler will default to colour output if it can determine that the
694 output is a colour-capable teleprinter.
696 - New environment variables:
697 - NO_COLOUR/NO_COLOR can be used to disable default colour output.
698 See the README for more detailed descriptions and how to use these.
704 - D3DCreateBlob() provides support for creating ID3DBlob objects.
705 - D3DPreprocess() provides support for preprocessing HLSL source code.
706 - D3DCompile() and D3DCompile2() provide support for compiling HLSL source
712 - The ‘gears’ and ‘triangle’ demo applications are installed as ‘vkd3d-gears’
713 and ‘vkd3d-triangle’. These were originally intended more as documentation
714 than as end-user executables, but there's some value in using them for
715 diagnostic purposes, much like e.g. ‘glxgears’.
717 - The VULKAN_LIBS configure variable is used when detecting the Vulkan
720 - Builds for the Microsoft Windows target platform no longer require support
721 for POSIX threads. Windows synchronisation primitives are used instead.
723 - If ncurses is available, it will be use by vkd3d-compiler to determine the
724 capabilities of the connected teleprinter, if any.
727 # What's new in vkd3d 1.2 (22 Sep 2020)
731 - Newly implemented Direct3D 12 features:
733 - Reserved resources.
734 - Instance data step rates. This requires the
735 VK_EXT_vertex_attribute_divisor extension.
736 - ‘Private data’ APIs for all interfaces.
737 - Shader-resource view component mappings.
738 - Indirect indexed draws.
739 - Indirect draws with a count buffer. This requires the
740 VK_KHR_draw_indirect_count extension.
741 - Stream output and stream output queries. This requires the
742 VK_EXT_transform_feedback extension.
743 - Predicated/conditional rendering.
745 - Depth rendering without a pixel shader.
746 - Depth clipping. This requires the VK_EXT_depth_clip_enable extension.
747 - Rasteriser discard.
748 - Dual-source blending.
749 - Mapping placed resources.
750 - The ReadFromSubresource() and WriteToSubresource() ID3D12Resource methods.
751 - Simultaneous access to resources from multiple queues.
752 - Null-views. I.e., views without an underlying resource.
753 - Several more feature support queries.
756 - vkd3d_serialize_versioned_root_signature() and
757 vkd3d_create_versioned_root_signature_deserializer() provide support for
758 versioned root signatures.
759 - The vkd3d_application_info structure extends the
760 vkd3d_instance_create_info structure, and can be used to pass information
761 about the application to libvkd3d. It is analogous to the
762 VkApplicationInfo structure in Vulkan. Its ‘api_version’ field should be
763 set to the version of the libvkd3d API that the application targets.
764 - The vkd3d_optional_device_extensions_info structure extends the
765 vkd3d_device_create_info structure, and can be used to pass a list of
766 device extensions to enable only when available to libvkd3d. It is
767 analogous to the vkd3d_optional_instance_extensions_info structure for
770 - New environment variables:
771 - VKD3D_CONFIG can be used to set options that change the behaviour of
773 - VKD3D_TEST_BUG can be used to disable bug_if() conditions in the test
775 - VKD3D_TEST_FILTER can be used to control which tests are run.
776 - VKD3D_VULKAN_DEVICE can be used to override the Vulkan physical device
778 See the README for more detailed descriptions and how to use these.
780 - When the VK_KHR_dedicated_allocation extension is available, dedicated
781 allocations may be used for committed resources.
783 - When the VK_KHR_image_format_list extension is available, it will be used to
784 inform the driver about the view formats that a particular mutable Vulkan
785 image can be used with. This improves performance on some Vulkan
788 - When the VK_EXT_debug_marker extension is available, object names set with
789 the ID3D12Object SetName() method will be propagated to the underlying
792 - Unordered-access view clears are supported on more formats. Previously these
793 were limited to integer formats for texture resources, and single component
794 integer formats for buffer resources.
796 - When the D24_UNORM_S8_UINT format is not supported by the Vulkan
797 implementation, the D32_SFLOAT_S8_UINT format will be used instead to
798 implement the D24_UNORM_S8_UINT and related DXGI formats. This is required
799 because the DXGI D24_UNORM_S8_UINT format is mandatory, while the Vulkan
800 D24_UNORM_S8_UINT format is optional.
807 - libvkd3d-shader is now available as a public instead of an internal library.
810 - Tessellation shaders.
811 - Root signature version 1.1 serialisation, deserialisation, and conversion.
812 - Multi-sample masks.
813 - Per-sample shading.
814 - Early depth/stencil test.
815 - Conservative depth output.
816 - Dual-source blending.
820 - Support for OpenGL SPIR-V target environments. This allows SPIR-V produced
821 by libvkd3d-shader to be used with GL_ARB_gl_spirv. This includes support
822 for OpenGL atomic counters and combined samplers.
824 - Preliminary support for shader model 5.1 shaders. This is still a work in
825 progress. Notably, support for resource arrays is not yet implemented.
827 - When the SPV_EXT_demote_to_helper_invocation is available, it will be used
828 to implement the ‘discard’ shader instruction instead of using SpvOpKill. In
829 particular, this ensures the ‘deriv_rtx’ and ‘deriv_rty’ instruction return
830 accurate results after a (conditional) ‘discard’ instruction.
832 - Support for using SPIR-V specialisation constants for shader parameters.
834 - Support for more shader instructions:
844 - When built against SPIRV-Tools, libvkd3d-shader can produce SPIR-V shaders
847 - libvkd3d-shader now has its own environment variable (VKD3D_SHADER_DEBUG) to
848 control debug output.
850 - Various shader translation fixes.
855 - When supported by libvkd3d-shader, text form SPIR-V is available as a target
856 format, in addition to the existing binary form SPIR-V target format.
858 - Input from standard input, and output to standard output is supported.
863 - To specify the libvkd3d API version to use when creating vkd3d instances,
864 define VKD3D_UTILS_API_VERSION to the desired version before including
865 vkd3d_utils.h. If VKD3D_UTILS_API_VERSION is not explicitly defined,
866 VKD3D_API_VERSION_1_0 will be used.
868 - Support for versioned root signatures is provided by the
869 D3D12SerializeVersionedRootSignature() and
870 D3D12CreateVersionedRootSignatureDeserializer() entry points.
875 - The minimum required version of Vulkan-Headers and SPIRV-Headers for this
876 release is version 1.1.113.
878 - The minimum required version of widl for this release is version 3.20.
880 - If doxygen is available, it will be used to build API documentation. By
881 default, documentation will be generated in HTML and PDF formats.
883 - If debug logs are not required or desired, defining VKD3D_NO_TRACE_MESSAGES
884 and VKD3D_NO_DEBUG_MESSAGES will prevent them from being included in the
885 build. For example, a release build may want to configure with
886 ‘CPPFLAGS="-DNDEBUG -DVKD3D_NO_TRACE_MESSAGES -DVKD3D_NO_DEBUG_MESSAGES"’.
888 - Microsoft Windows is now a supported target platform. To create a build for
889 Windows, either cross-compile by configuring with an appropriate --host
890 option like for example ‘--host=x86_64-w64-mingw32’, or build on Windows
891 itself using an environment like MSYS2 or Cygwin.
894 # What's new in vkd3d 1.1 (5 Oct 2018)
898 - Initial support for memory heaps and placed resources.
900 - Improved support for resource views.
902 - ClearUnorderedAccessViewUint() is implemented for textures.
904 - Blend factor is implemented.
906 - Performance improvements.
908 - A new interface is available for enabling additional Vulkan instance
911 - A new public function is available for mapping VkFormats to DXGI_FORMATs.
913 - Support for more DXGI formats.
920 - Support for geometry shaders.
922 - Pretty printing is implemented for shader code extracted from DXBC.
924 - Clip and cull distances are supported.
926 - Support for more shader instructions:
936 - Texel offsets are supported.
938 - Various shader translation fixes.
943 - Vulkan WSI extensions are detected at runtime.
948 - Demos are not built by default.
950 - libxcb is now an optional dependency required only for demos.
952 - MoltenVK is supported.
955 # What's included in vkd3d 1.0 (23 May 2018)
959 - libvkd3d is the main component of the vkd3d project. It's a 3D graphics
960 library built on top of Vulkan with an API very similar to Direct3D 12.
962 - A significant number of Direct3D 12 features are implemented, including:
963 - Graphics and compute pipelines.
964 - Command lists, command allocators and command queues.
965 - Descriptors and descriptor heaps.
967 - Constant buffer, shader resource, unordered access, render target and depth
971 - Descriptors copying.
972 - Committed resources.
974 - Queries and query heaps.
977 - Basic support for indirect draws and dispatches.
978 - Basic support for command signatures.
979 - Various Clear*() methods.
980 - Various Copy*() methods.
985 - libvkd3d-shader is a library which translates shader model 4 and 5 bytecode
986 to SPIR-V. In this release, libvkd3d-shader is an internal library. Its API
987 isn't set in stone yet.
989 - Vertex, pixel and compute shaders are supported. Also, very simple geometry
990 shaders should be translated correctly. Issues are expected when trying to
991 translate tessellation shaders or more complex geometry shaders.
993 - A significant number of shader instructions are supported in this release,
995 - Arithmetic instructions.
997 - Comparison instructions.
998 - Control flow instructions.
999 - Sample, gather and load instructions.
1000 - Atomic instructions.
1003 - Root signature serialization and deserialization is implemented.
1005 - Shader model 4 and 5 bytecode parser is imported from wined3d.
1007 - Shader model 5.1 is not supported yet.
1012 - libvkd3d-utils contains simple implementations of various functions which
1013 might be useful for source ports of Direct3D 12 applications.
1018 - A simple hello triangle Direct3D 12 demo.
1019 - A Direct3D 12 port of glxgears.