Merge branch 'main' into texture1d

Author: luciechoi

cmake/config-ix.cmake

@@ -573,7 +573,7 @@ string(REPLACE " " ";" LLVM_BINDINGS_LIST "${LLVM_BINDINGS}")
 
 # HLSL Change Begin - Metal IR Converter
 find_package(MetalIRConverter)
-if (METAL_IRCONVERTER_FOUND)
+if (MetalIRConverter_FOUND)
   set(ENABLE_METAL_CODEGEN On)
   message(STATUS "Enabling Metal Support")
   add_definitions(-DENABLE_METAL_CODEGEN)

cmake/modules/FindMetalIRConverter.cmake

@@ -1,16 +1,16 @@
-find_path(METAL_IRCONVERTER_INCLUDE_DIR metal_irconverter.h
+find_path(MetalIRConverter_INCLUDE_DIR metal_irconverter.h
           HINTS /usr/local/include/metal_irconverter
           DOC "Path to metal IR converter headers"
           )
 
-find_library(METAL_IRCONVERTER_LIB NAMES metalirconverter
+find_library(MetalIRConverter_LIB NAMES metalirconverter
   PATH_SUFFIXES lib
   )
 
 include(FindPackageHandleStandardArgs)
-FIND_PACKAGE_HANDLE_STANDARD_ARGS(METAL_IRCONVERTER
-                                    REQUIRED_VARS METAL_IRCONVERTER_LIB METAL_IRCONVERTER_INCLUDE_DIR)
+FIND_PACKAGE_HANDLE_STANDARD_ARGS(MetalIRConverter
+                                  REQUIRED_VARS MetalIRConverter_LIB MetalIRConverter_INCLUDE_DIR)
 
-message(STATUS "Metal IR Converter Include Dir: ${METAL_IRCONVERTER_INCLUDE_DIR}")
-message(STATUS "Metal IR Converter Library: ${METAL_IRCONVERTER_LIB}")
-mark_as_advanced(METAL_IRCONVERTER_LIB METAL_IRCONVERTER_INCLUDE_DIR)
+message(STATUS "Metal IR Converter Include Dir: ${MetalIRConverter_INCLUDE_DIR}")
+message(STATUS "Metal IR Converter Library: ${MetalIRConverter_LIB}")
+mark_as_advanced(MetalIRConverter_LIB MetalIRConverter_INCLUDE_DIR)

cmake/modules/HCT.cmake

@@ -67,6 +67,8 @@ function(add_hlsl_hctgen mode)
                        ${opcodes_json})
 
   get_filename_component(output_extension ${full_output} LAST_EXT)
+  string(SUBSTRING ${output_extension} 1 -1 output_extension_no_dot)
+  set(target_name "${mode}_${output_extension_no_dot}")
 
   if (CLANG_FORMAT_EXE AND output_extension MATCHES "\.h|\.cpp|\.inl")
     set(format_cmd COMMAND ${CLANG_FORMAT_EXE} -i ${temp_output})
@@ -132,8 +134,8 @@ function(add_hlsl_hctgen mode)
                      DEPENDS ${output}
                      COMMENT "Verifying clang-format results...")
 
-  add_custom_target(${mode}
+  add_custom_target(${target_name}
                     DEPENDS ${temp_output}.stamp)
 
-  add_dependencies(HCTGen ${mode})
+  add_dependencies(HCTGen ${target_name})
 endfunction()

docs/DXIL.rst

@@ -24,8 +24,7 @@ line upon naming the release. Refer to previous for appropriate section names.
 
 #### Experimental Shader Model 6.10
 
-- Moved Linear Algebra (Cooperative Vector) DXIL Opcodes to experimental Shader Model 6.10
-- The Cooperative Vectors API was moved to `coopvec.h` header and under the `dx::coopvec` namespace.
+- Removed experimental Cooperative Vector, this has been replaced by LinAlg matrix.
 - Implement GetGroupWaveIndex and GetGroupWaveCount in experimental Shader Model 6.10.
   - [proposal](https://github.com/microsoft/hlsl-specs/blob/main/proposals/0048-group-wave-index.md)
   - GetGroupWaveIndex: New intrinsic for Compute, Mesh, Amplification and Node shaders which returns the index of the wave within the thread group that the the thread is executing.

docs/ReleaseNotes.md

@@ -204,6 +204,13 @@ enum class MatrixScope : uint32_t {
   ThreadGroup = 2,
 };
 
+enum class LinalgMatrixLayout : uint32_t {
+  RowMajor = 0,
+  ColumnMajor = 1,
+  MulOptimal = 2,
+  OuterProductOptimal = 3,
+};
+
 // Must match D3D_INTERPOLATION_MODE
 enum class InterpolationMode : uint8_t {
   Undefined = 0,
@@ -524,9 +531,8 @@ static const OpCodeTableID TableID = OpCodeTableID::ExperimentalOps;
 // Enumeration for ExperimentalOps DXIL operations
 enum class OpCode : unsigned {
   //
-  ReservedD1 = 30, // reserved
-  ReservedD2 = 31, // reserved
-  ReservedD3 = 32, // reserved
+  ReservedE0 = 31, // reserved
+  ReservedE1 = 32, // reserved
 
   // Debugging
   DebugBreak = 33,        // triggers a breakpoint if a debugger is attached
@@ -545,6 +551,8 @@ enum class OpCode : unsigned {
       9, // returns committed triangle vertices in object space as <9 x float>
 
   // Linear Algebra Operations
+  LinAlgConvert =
+      30, // Convert vector components from one interpretation to another
   LinAlgCopyConvertMatrix =
       13, // Converts and copies the element and use type of the source matrix
           // to the destination matrix with optional transpose
@@ -647,6 +655,10 @@ enum class OpCode : unsigned {
   ReservedC7 = 300,  // reserved
   ReservedC8 = 301,  // reserved
   ReservedC9 = 302,  // reserved
+  ReservedD0 = 305,  // reserved
+  ReservedD1 = 306,  // reserved
+  ReservedD2 = 307,  // reserved
+  ReservedD3 = 308,  // reserved
 
   // Amplification shader instructions
   DispatchMesh = 173, // Amplification shader intrinsic DispatchMesh
@@ -876,19 +888,6 @@ enum class OpCode : unsigned {
   CreateHandleForLib =
       160, // create resource handle from resource struct for library
 
-  // Linear Algebra Operations
-  MatVecMul =
-      305, // Multiplies a MxK dimension matrix and a K sized input vector
-  MatVecMulAdd = 306, // multiplies a MxK dimension matrix and a K sized input
-                      // vector and adds an M-sized bias vector
-  OuterProductAccumulate =
-      307, // Computes the outer product between column vectors and an MxN
-           // matrix is accumulated component-wise atomically (with device
-           // scope) in memory
-  VectorAccumulate = 308, // Accumulates the components of a vector
-                          // component-wise atomically (with device scope) to
-                          // the corresponding elements of an array in memory
-
   // Mesh shader instructions
   EmitIndices = 169, // emit a primitive's vertex indices in a mesh shader
   GetMeshPayload =
@@ -1351,12 +1350,13 @@ enum class OpCode : unsigned {
       ExperimentalOps,
       LinAlgMatrixOuterProduct), // Outer products an M sized vector and a N
                                  // sized vector producing an MxN matrix
-  // ReservedD1 = 0x8000001E, 2147483678U, -2147483618
-  EXP_OPCODE(ExperimentalOps, ReservedD1), // reserved
-  // ReservedD2 = 0x8000001F, 2147483679U, -2147483617
-  EXP_OPCODE(ExperimentalOps, ReservedD2), // reserved
-  // ReservedD3 = 0x80000020, 2147483680U, -2147483616
-  EXP_OPCODE(ExperimentalOps, ReservedD3), // reserved
+  // LinAlgConvert = 0x8000001E, 2147483678U, -2147483618
+  EXP_OPCODE(ExperimentalOps, LinAlgConvert), // Convert vector components from
+                                              // one interpretation to another
+  // ReservedE0 = 0x8000001F, 2147483679U, -2147483617
+  EXP_OPCODE(ExperimentalOps, ReservedE0), // reserved
+  // ReservedE1 = 0x80000020, 2147483680U, -2147483616
+  EXP_OPCODE(ExperimentalOps, ReservedE1), // reserved
   // DebugBreak = 0x80000021, 2147483681U, -2147483615
   EXP_OPCODE(ExperimentalOps,
              DebugBreak), // triggers a breakpoint if a debugger is attached
@@ -1522,6 +1522,7 @@ enum class OpCodeClass : unsigned {
   CreateHandleForLib,
 
   // Linear Algebra Operations
+  LinAlgConvert,
   LinAlgCopyConvertMatrix,
   LinAlgFillMatrix,
   LinAlgMatVecMul,
@@ -1541,10 +1542,6 @@ enum class OpCodeClass : unsigned {
   LinAlgMatrixSetElement,
   LinAlgMatrixStoreToDescriptor,
   LinAlgMatrixStoreToMemory,
-  MatVecMul,
-  MatVecMulAdd,
-  OuterProductAccumulate,
-  VectorAccumulate,
 
   // Mesh shader instructions
   EmitIndices,
@@ -1731,7 +1728,7 @@ enum class OpCodeClass : unsigned {
   NodeOutputIsValid,
   OutputComplete,
 
-  NumOpClasses = 225, // exclusive last value of enumeration
+  NumOpClasses = 222, // exclusive last value of enumeration
 };
 // OPCODECLASS-ENUM:END
 
@@ -1911,29 +1908,6 @@ const unsigned kHitObjectTraceRay_RayDescOpIdx = 7;
 const unsigned kHitObjectTraceRay_PayloadOpIdx = 15;
 const unsigned kHitObjectTraceRay_NumOp = 16;
 
-// MatVec Ops
-const unsigned kMatVecMulInputVectorIdx = 1;
-const unsigned kMatVecMulIsInputUnsignedIdx = 2;
-const unsigned kMatVecMulInputInterpretationIdx = 3;
-const unsigned kMatVecMulMatrixBufferIdx = 4;
-const unsigned kMatVecMulMatrixOffsetIdx = 5;
-const unsigned kMatVecMulMatrixInterpretationIdx = 6;
-const unsigned kMatVecMulMatrixMIdx = 7;
-const unsigned kMatVecMulMatrixKIdx = 8;
-const unsigned kMatVecMulMatrixLayoutIdx = 9;
-const unsigned kMatVecMulMatrixTransposeIdx = 10;
-const unsigned kMatVecMulMatrixStrideIdx = 11;
-const unsigned kMatVecMulIsOutputUnsignedIdx = 12;
-
-// MatVecAdd
-const unsigned kMatVecMulAddBiasInterpretation = 14;
-const unsigned kMatVecMulAddIsOutputUnsignedIdx = 15;
-
-// Outer Product Accumulate
-const unsigned kOuterProdAccMatrixInterpretation = 5;
-const unsigned kOuterProdAccMatrixLayout = 6;
-const unsigned kOuterProdAccMatrixStride = 7;
-
 // TODO: add operand index for all the OpCodeClass.
 } // namespace OperandIndex
 
@@ -2506,13 +2480,6 @@ extern const char *kHostLayoutTypePrefix;
 
 extern const char *kWaveOpsIncludeHelperLanesString;
 
-enum class LinalgMatrixLayout : uint32_t {
-  RowMajor = 0,
-  ColumnMajor = 1,
-  MulOptimal = 2,
-  OuterProductOptimal = 3,
-};
-
 } // namespace DXIL
 
 } // namespace hlsl