[clang] [llvm] [DirectX] Start documenting DXIL Resource handling (PR #90553)
via cfe-commits
cfe-commits at lists.llvm.org
Mon Jul 15 10:03:36 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-directx
Author: Justin Bogner (bogner)
<details>
<summary>Changes</summary>
This adds a new document about DXIL Resource Handling. I've attempted to describe here how we intend to use TargetExtTypes to represent resources in LLVM IR and the various intrinsics we'll need to lower these through LLVM to DXIL.
For now this document is limited to the high level concepts and a few details on buffer types, and there are a number of TODOs in the document that we'll iterate on as we progress in the implementation.
---
Full diff: https://github.com/llvm/llvm-project/pull/90553.diff
4 Files Affected:
- (modified) clang/docs/HLSL/HLSLIRReference.rst (-9)
- (modified) clang/docs/HLSL/ResourceTypes.rst (+4-3)
- (added) llvm/docs/DirectX/DXILResources.rst (+468)
- (modified) llvm/docs/DirectXUsage.rst (+2-1)
``````````diff
diff --git a/clang/docs/HLSL/HLSLIRReference.rst b/clang/docs/HLSL/HLSLIRReference.rst
index c0d8d33f33bff..c0033946a6ec8 100644
--- a/clang/docs/HLSL/HLSLIRReference.rst
+++ b/clang/docs/HLSL/HLSLIRReference.rst
@@ -11,15 +11,6 @@ Introduction
The goal of this document is to provide a reference for all the special purpose
IR metadata and attributes used by the HLSL code generation path.
-IR Metadata
-===========
-
-``hlsl.uavs``
--------------
-
-The ``hlsl.uavs`` metadata is a list of all the external global variables that
-represent UAV resources.
-
Function Attributes
===================
diff --git a/clang/docs/HLSL/ResourceTypes.rst b/clang/docs/HLSL/ResourceTypes.rst
index aad7b3314f084..0d461ba3ee707 100644
--- a/clang/docs/HLSL/ResourceTypes.rst
+++ b/clang/docs/HLSL/ResourceTypes.rst
@@ -29,6 +29,7 @@ pointer of the template parameter type. The pointer is populated from a call to
data through until lowering in the backend.
Resource types are annotated with the ``HLSLResource`` attribute, which drives
-code generation for resource binding metadata. The ``hlsl`` metadata nodes are
-transformed in the backend to the binding information expected by the target
-runtime.
+code generation into target extension types in IR. These types are target
+specific and differ between DXIL and SPIR-V generation, providing the necessary
+information for the targets to generate binding metadata for their respective
+target runtimes.
diff --git a/llvm/docs/DirectX/DXILResources.rst b/llvm/docs/DirectX/DXILResources.rst
new file mode 100644
index 0000000000000..b7ea478373f50
--- /dev/null
+++ b/llvm/docs/DirectX/DXILResources.rst
@@ -0,0 +1,468 @@
+======================
+DXIL Resource Handling
+======================
+
+.. contents::
+ :local:
+
+.. toctree::
+ :hidden:
+
+Introduction
+============
+
+Resources in DXIL are represented via ``TargetExtType`` in LLVM IR and
+eventually lowered by the DirectX backend into metadata in DXIL.
+
+In DXC and DXIL, static resources are represented as lists of SRVs (Shader
+Resource Views), UAVs (Uniform Access Views), CBVs (Constant Bffer Views), and
+Samplers. This metadata consists of a "resource record ID" which uniquely
+identifies a resource and type information. As of shader model 6.6, there are
+also dynamic resources, which forgo the metadata and are described via
+``annotateHandle`` operations in the instruction stream instead.
+
+In LLVM we attempt to unify some of the alternative representations that are
+present in DXC, with the aim of making handling of resources in the middle end
+of the compiler simpler and more consistent.
+
+Resource Type Information and Properties
+========================================
+
+There are a number of properties associated with a resource in DXIL.
+
+`Resource ID`
+ An arbitrary ID that must be unique per resource type (SRV, UAV, etc).
+
+ In LLVM we don't bother representing this, instead opting to generate it at
+ DXIL lowering time.
+
+`Binding information`
+ Information about where the resource comes from. This is either (a) a
+ register space, lower bound in that space, and size of the binding, or (b)
+ an index into a dynamic resource heap.
+
+ In LLVM we represent binding information in the arguments of the
+ :ref:`handle creation intrinsics <dxil-resources-handles>`. When generating
+ DXIL we transform these calls to metadata, ``dx.op.createHandle``,
+ ``dx.op.createHandleFromBinding``, ``dx.op.createHandleFromHeap``, and
+ ``dx.op.createHandleForLib`` as needed.
+
+`Type information`
+ The type of data that's accessible via the resource. For buffers and
+ textures this can be a simple type like ``float`` or ``float4``, a struct,
+ or raw bytes. For constant buffers this is just a size. For samplers this is
+ the kind of sampler.
+
+ In LLVM we embed this information as a parameter on the ``target()`` type of
+ the resource. See :ref:`dxil-resources-types-of-resource`.
+
+`Resource kind information`
+ The kind of resource. In HLSL we have things like ``ByteAddressBuffer``,
+ ``RWTexture2D``, and ``RasterizerOrderedStructuredBuffer``. These map to a
+ set of DXIL kinds like ``RawBuffer`` and ``Texture2D`` with fields for
+ certain properties such as ``IsUAV`` and ``IsROV``.
+
+ In LLVM we represent this in the ``target()`` type. We omit information
+ that's deriveable from the type information, but we do have fields to encode
+ ``IsWriteable``, ``IsROV``, and ``SampleCount`` when needed.
+
+.. note:: TODO: There are two fields in the DXIL metadata that are not
+ represented as part of the target type: ``IsGloballyCoherent`` and
+ ``HasCounter``.
+
+ Since these are derived from analysis, storing them on the type would mean
+ we need to change the type during the compiler pipeline. That just isn't
+ practical. It isn't entirely clear to me that we need to serialize this info
+ into the IR during the compiler pipeline anyway - we can probably get away
+ with an analysis pass that can calculate the information when we need it.
+
+ If analysis is insufficient we'll need something akin to ``annotateHandle``
+ (but limited to these two properties) or to encode these in the handle
+ creation.
+
+.. _dxil-resources-types-of-resource:
+
+Types of Resource
+=================
+
+We define a set of ``TargetExtTypes`` that is similar to the HLSL
+representations for the various resources, albeit with a few things
+parameterized. This is different than DXIL, as simplifying the types to
+something like "dx.srv" and "dx.uav" types would mean the operations on these
+types would have to be overly generic.
+
+Buffers
+-------
+
+.. code-block:: llvm
+
+ target("dx.TypedBuffer", ElementType, IsWriteable, IsROV)
+ target("dx.RawBuffer", ElementType, IsWriteable, IsROV)
+
+We need two separate buffer types to account for the differences between the
+16-byte `bufferLoad`_ / `bufferStore`_ operations that work on DXIL's
+TypedBuffers and the `rawBufferLoad`_ / `rawBufferStore`_ operations that are
+used for DXIL's RawBuffers and StructuredBuffers. We call the latter
+"RawBuffer" to match the naming of the operations, but it can represent both
+the Raw and Structured variants.
+
+For TypedBuffer, the element type must be an integer or floating point type.
+For RawBuffer the type can be an integer, floating point, or struct type.
+HLSL's ByteAddressBuffer is represented by an `i8` element type.
+
+These types are generally used by BufferLoad and BufferStore operations, as
+well as atomics.
+
+There are a few fields to describe variants of all of these types:
+
+.. list-table:: Buffer Fields
+ :header-rows: 1
+
+ * - Field
+ - Description
+ * - ElementType
+ - Type for a single element, such as ``i8``, ``v4f32``, or a structure
+ type.
+ * - IsWriteable
+ - Whether or not the field is writeable. This distinguishes SRVs (not
+ writeable) and UAVs (writeable).
+ * - IsROV
+ - Whether the UAV is a rasterizer ordered view. Always ``0`` for SRVs.
+
+.. _bufferLoad: https://github.com/microsoft/DirectXShaderCompiler/blob/main/docs/DXIL.rst#bufferload
+.. _bufferStore: https://github.com/microsoft/DirectXShaderCompiler/blob/main/docs/DXIL.rst#bufferstore
+.. _rawBufferLoad: https://github.com/microsoft/DirectXShaderCompiler/blob/main/docs/DXIL.rst#rawbufferload
+.. _rawBufferStore: https://github.com/microsoft/DirectXShaderCompiler/blob/main/docs/DXIL.rst#rawbufferstore
+
+Resource Operations
+===================
+
+.. _dxil-resources-handles:
+
+Resource Handles
+----------------
+
+We provide a few different ways to instantiate resources in the IR via the
+``llvm.dx.handle.*`` intrinsics. These intrinsics are overloaded on return
+type, returning an appropriate handle for the resource, and represent binding
+information in the arguments to the intrinsic.
+
+The three operations we need are ``llvm.dx.handle.fromBinding``,
+``llvm.dx.handle.fromHeap``, and ``llvm.dx.handle.fromPointer``. These are
+rougly equivalent to the DXIL operations ``dx.op.createHandleFromBinding``,
+``dx.op.createHandleFromHeap``, and ``dx.op.createHandleForLib``, but they fold
+the subsequent ``dx.op.annotateHandle`` operation in. Note that we don't have
+an analogue for `dx.op.createHandle`_, since ``dx.op.createHandleFromBinding``
+subsumes it.
+
+.. _dx.op.createHandle: https://github.com/microsoft/DirectXShaderCompiler/blob/main/docs/DXIL.rst#resource-handles
+
+.. list-table:: ``@llvm.dx.handle.fromBinding``
+ :header-rows: 1
+
+ * - Argument
+ -
+ - Type
+ - Description
+ * - Return value
+ -
+ - A ``target()`` type
+ - A handle which can be operated on
+ * - ``%reg_space``
+ - 1
+ - ``i32``
+ - Register space ID in the root signature for this resource.
+ * - ``%lower_bound``
+ - 2
+ - ``i32``
+ - Lower bound of the binding in its register space.
+ * - ``%range_size``
+ - 3
+ - ``i32``
+ - Range size of the binding.
+ * - ``%index``
+ - 4
+ - ``i32``
+ - Index of the resource to access.
+ * - ``%non-uniform``
+ - 5
+ - i1
+ - Must be ``true`` if the resource index may be non-uniform.
+
+.. note:: TODO: Can we drop the uniformity bit? I suspect we can derive it from
+ uniformity analysis...
+
+Examples:
+
+.. code-block:: llvm
+
+ ; RWBuffer<float4> Buf : register(u5, space3)
+ %buf = call target("dx.Buffer", <4 x float>, 1, 0)
+ @llvm.dx.handle.fromBinding.tdx.Buffer_v4f32_1_0(
+ i32 3, i32 5, i32 1, i32 0, i1 false)
+
+ ; RWBuffer<uint> Buf : register(u7, space2)
+ %buf = call target("dx.Buffer", i32, 1, 0)
+ @llvm.dx.handle.fromBinding.tdx.Buffer_i32_1_0t(
+ i32 2, i32 7, i32 1, i32 0, i1 false)
+
+ ; Buffer<uint4> Buf[24] : register(t3, space5)
+ %buf = call target("dx.Buffer", <4 x i32>, 0, 0)
+ @llvm.dx.handle.fromBinding.tdx.Buffer_v4i32_0_0t(
+ i32 2, i32 7, i32 24, i32 0, i1 false)
+
+ ; struct S { float4 a; uint4 b; };
+ ; StructuredBuffer<S> Buf : register(t2, space4)
+ %buf = call target("dx.Buffer", {<4 x f32>, <4 x i32>}, 0, 0)
+ @llvm.dx.handle.fromBinding.tdx.Buffer_sl_v4f32v4i32s_0_0t(
+ i32 4, i32 2, i32 1, i32 0, i1 false)
+
+ ; ByteAddressBuffer Buf : register(t8, space1)
+ %buf = call target("dx.Buffer", i8, 0, 0)
+ @llvm.dx.handle.fromBinding.tdx.Buffer_i8_0_0t(
+ i32 1, i32 8, i32 1, i32 0, i1 false)
+
+.. list-table:: ``@llvm.dx.handle.fromHeap``
+ :header-rows: 1
+
+ * - Argument
+ -
+ - Type
+ - Description
+ * - Return value
+ -
+ - A ``target()`` type
+ - A handle which can be operated on
+ * - ``%index``
+ - 0
+ - ``i32``
+ - Index of the resource to access.
+ * - ``%non-uniform``
+ - 1
+ - i1
+ - Must be ``true`` if the resource index may be non-uniform.
+
+Examples:
+
+.. code-block:: llvm
+
+ ; RWStructuredBuffer<float4> Buf = ResourceDescriptorHeap[2];
+ declare
+ target("dx.Buffer", <4 x float>, 1, 0)
+ @llvm.dx.handle.fromHeap.tdx.Buffer_v4f32_1_0(
+ i32 %index, i1 %non_uniform)
+ ; ...
+ %buf = call target("dx.Buffer", <4 x f32>, 1, 0)
+ @llvm.dx.handle.fromHeap.tdx.Buffer_v4f32_1_0(
+ i32 2, i1 false)
+
+Buffer Loads and Stores
+-----------------------
+
+*relevant types: Buffers*
+
+We separate loading from buffers into two operations, ``llvm.dx.bufferLoad``
+and ``llvm.dx.bufferLoadComponent``. Store operations consist of their inverse,
+``llvm.dx.bufferStore`` and ``llvm.dx.bufferStoreComponent``. These map to the
+DXIL `rawBufferLoad`_ and `rawBufferStore`_ operations (and their older non-raw
+counterparts).
+
+We opt for two different intrinsics to best support the two main ways of
+accessing buffer data.
+
+The ``llvm.dx.bufferLoad`` intrinsic can return either a single element of a
+buffer or a vector of consecutive elements. This makes accessing buffers of
+scalars and simple vectors like `float4` simple, and is also convenient when
+loading an entire `struct`. The variant with a vector of elements returned is
+most useful for raw buffers, where we can load a number of bytes and `bitcast`
+to the appropriate type, but can also be used to preserve the information that
+we're loading data in bulk if needed.
+
+The ``llvm.dx.bufferLoadComponent`` intrinsic has an extra index so that it can
+be used to access specific struct elements or a particular component of a
+simple vector type, like ``x`` of a ``float4``. This API gives the same
+flexibility as the DXIL ``rawBufferLoad`` operation, but in a slightly more
+readable way since it avoids ``undef`` values and bit masks, as well as using
+an index instead of a byte offset.
+
+The types involved in the store intrinsics match the load intrinsics.
+
+When lowering these to the DXIL operations we need to pay attention to the DXIL
+version and the type of data in the buffer. Post DXIL 1.2 structured and byte
+address (or raw) buffers prefer RawBufferLoad and Store, whereas TypedBuffer
+always uses BufferLoad and Store. For the raw operations, we need to provide an
+alignment, but this can be derived from the buffer types in the LLVM
+intrinsics.
+
+.. note:: TODO: Can we always derive the alignment late, or do we need to
+ parametrize these ops?
+
+.. note:: TODO: We need to account for `CheckAccessFullyMapped`_ here.
+
+ In DXIL the load operations always return an ``i32`` status value, but this
+ isn't very ergonomic when it isn't used. We can (1) bite the bullet and have
+ the loads return `{%ret_type, %i32}` all the time, (2) create a variant or
+ update the signature iff the status is used, or (3) hide this in a sideband
+ channel somewhere. I'm leaning towards (2), but could probably be convinced
+ that the ugliness of (1) is worth the simplicity.
+
+.. _CheckAccessFullyMapped: https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/checkaccessfullymapped
+
+.. list-table:: ``@llvm.dx.bufferLoad``
+ :header-rows: 1
+
+ * - Argument
+ -
+ - Type
+ - Description
+ * - Return value
+ -
+ - The element type of the buffer, or a vector of the same.
+ - The data loaded from the buffer
+ * - ``%buffer``
+ - 0
+ - ``target(dx.Buffer, ...)``
+ - The buffer to load from
+ * - ``%index``
+ - 1
+ - ``i32``
+ - Index into the buffer
+
+Examples:
+
+.. code-block:: llvm
+
+ ; Load from a buffer containing float4
+ %ret = call <4 x float> @llvm.dx.bufferLoad.v4f32.tdx.Buffer_v4f32_0_0t(
+ target("dx.Buffer", <4 x f32>, 0, 0) %buffer, i32 %index)
+ ; Load a single element from a buffer containing float
+ %ret = call float @llvm.dx.bufferLoad.f32.tdx.Buffer_f32_0_0t(
+ target("dx.Buffer", f32, 0, 0) %buffer, i32 %index)
+ ; Load 4 elements from a buffer containing float
+ %ret = call <4 x float> @llvm.dx.bufferLoad.v4f32.tdx.Buffer_f32_0_0t(
+ target("dx.Buffer", f32, 0, 0) %buffer, i32 %index)
+ ; Load a struct
+ %ret = call {f32, i32}
+ @llvm.dx.bufferLoad.sl_f32i32s.tdx.Buffer_sl_f32i32s_0_0t(
+ target("dx.Buffer", {f32, i32}, 0, 0) %buffer, i32 %index)
+
+ ; Load an i32 from a byte address buffer
+ %ret = call <4 x i8> @llvm.dx.bufferLoad.v4i8.tdx.Buffer_i8_0_0t(
+ target("dx.Buffer", i8, 0, 0) %buffer, i32 %index)
+ %cast = bitcast <4 x i8> %ret to i32
+
+.. list-table:: ``@llvm.dx.bufferLoadComponent``
+ :header-rows: 1
+
+ * - Argument
+ -
+ - Type
+ - Description
+ * - Return value
+ -
+ - The type of the component
+ - The data loaded from the buffer
+ * - ``%buffer``
+ - 0
+ - ``target(dx.Buffer, ...)``
+ - The buffer to load from
+ * - ``%index``
+ - 1
+ - ``i32``
+ - Index into the buffer
+ * - ``%component``
+ - 2
+ - ``i32``
+ - Index of the component to access. Must be constant for buffers
+ containing structs.
+
+Examples:
+
+.. code-block:: llvm
+
+ ; Load the `y` component from a buffer containing float4
+ %ret = call float @llvm.dx.bufferLoadComponent.f32.tdx.Buffer_v4f32_0_0t(
+ target("dx.Buffer", <4 x f32>, 0, 0) %buffer, i32 %index, i32 1)
+ ; Load the double from a struct containing an int, a float, and a double
+ %ret = call f64 @llvm.dx.bufferLoad.f64.tdx.Buffer_sl_i32f32f64s_0_0t(
+ target("dx.Buffer", {i32, f32, f64}, 0, 0) %buffer, i32 %index, i32 2)
+
+.. list-table:: ``@llvm.dx.bufferStore``
+ :header-rows: 1
+
+ * - Argument
+ -
+ - Type
+ - Description
+ * - Return value
+ -
+ - ``void``
+ -
+ * - ``%buffer``
+ - 0
+ - ``target(dx.Buffer, ...)``
+ - The buffer to store into
+ * - ``%index``
+ - 1
+ - ``i32``
+ - Index into the buffer
+ * - ``%data``
+ - 2
+ - The element type of the buffer, or a vector of the same.
+ - The data to store
+
+Examples:
+
+.. code-block:: llvm
+
+ call void @llvm.dx.bufferStore.tdx.Buffer_v4f32_1_0t.v4f32(
+ target("dx.Buffer", <4 x f32>, 1, 0) %buf, i32 %index, <4 x f32> %data)
+ call void @llvm.dx.bufferStore.tdx.Buffer_f32_1_0t.v4f32(
+ target("dx.Buffer", f32, 1, 0) %buf, i32 %index, <4 x f32> %data)
+ call void @llvm.dx.bufferStore.tdx.Buffer_f32_1_0t.f32(
+ target("dx.Buffer", f32, 1, 0) %buf, i32 %index, f32 %data)
+
+ %vec = bitcast f32 %data to <4 x i8>
+ call void @llvm.dx.bufferStore.tdx.Buffer_i8_1_0t.v4i8(
+ target("dx.Buffer", i8, 1, 0) %buf, i32 %index, <4 x i8> %vec)
+
+.. list-table:: ``@llvm.dx.bufferStoreComponent``
+ :header-rows: 1
+
+ * - Argument
+ -
+ - Type
+ - Description
+ * - Return value
+ -
+ - ``void``
+ -
+ * - ``%buffer``
+ - 0
+ - ``target(dx.Buffer, ...)``
+ - The buffer to store into
+ * - ``%index``
+ - 1
+ - ``i32``
+ - Index into the buffer
+ * - ``%component``
+ - 2
+ - ``i32``
+ - Index of the component to store. Must be constant for buffers containing
+ structs.
+ * - ``%data``
+ - 3
+ - The element type of the buffer, or a vector of the same.
+ - The data to store
+
+Examples:
+
+.. code-block:: llvm
+
+ ; Store the `y` component from a buffer containing float4
+ call void @llvm.dx.bufferStoreComponent.tdx.Buffer_v4f32_0_0t.f32(
+ target("dx.Buffer", <4 x f32>, 0, 0) %buf, i32 %index, i32 1, f32 %data)
+ ; Store the float from a struct containing an int and a float
+ call void @llvm.dx.bufferLoad.tdx.Buffer_sl_i32f32f64s_0_0t.f32(
+ target("dx.Buffer", {i32, f32}, 0, 0) %buf, i32 %index, i32 1, f32 %data)
+
diff --git a/llvm/docs/DirectXUsage.rst b/llvm/docs/DirectXUsage.rst
index 3e0ce40be569b..4d8f49bd224cb 100644
--- a/llvm/docs/DirectXUsage.rst
+++ b/llvm/docs/DirectXUsage.rst
@@ -13,9 +13,10 @@ User Guide for the DirectX Target
.. toctree::
:hidden:
- DirectX/DXILArchitecture
DirectX/DXContainer
+ DirectX/DXILArchitecture
DirectX/DXILOpTableGenDesign
+ DirectX/DXILResources
Introduction
============
``````````
</details>
https://github.com/llvm/llvm-project/pull/90553
More information about the cfe-commits
mailing list