[mlir][gpu] Introduce gpu.global_id op

Introduce OpenCL-style global_id op and corresponding spirv lowering.

Differential Revision: https://reviews.llvm.org/D121548

mlir/include/mlir/Dialect/GPU/GPUOps.td

@@ -112,6 +112,21 @@ def GPU_SubgroupIdOp : GPU_Op<"subgroup_id", [NoSideEffect]>,
   let assemblyFormat = "attr-dict `:` type($result)";
 }
 
+def GPU_GlobalIdOp : GPU_IndexOp<"global_id"> {
+  let description = [{
+    Returns the unique global workitem/thread id, i.e., the unique index of the
+    current workitem/thread within all workgroups / grid along the x, y, or z
+    `dimension`.
+
+    Example:
+
+    ```mlir
+    %gidX = gpu.global_id x
+    ```
+  }];
+}
+
+
 def GPU_NumSubgroupsOp : GPU_Op<"num_subgroups", [NoSideEffect]>,
     Arguments<(ins)>, Results<(outs Index:$result)> {
   let description = [{

mlir/lib/Conversion/GPUToSPIRV/GPUToSPIRV.cpp

@@ -373,6 +373,8 @@ void mlir::populateGPUToSPIRVPatterns(SPIRVTypeConverter &typeConverter,
       LaunchConfigConversion<gpu::BlockDimOp, spirv::BuiltIn::WorkgroupSize>,
       LaunchConfigConversion<gpu::ThreadIdOp,
                              spirv::BuiltIn::LocalInvocationId>,
+      LaunchConfigConversion<gpu::GlobalIdOp,
+                             spirv::BuiltIn::GlobalInvocationId>,
       SingleDimLaunchConfigConversion<gpu::SubgroupIdOp,
                                       spirv::BuiltIn::SubgroupId>,
       SingleDimLaunchConfigConversion<gpu::NumSubgroupsOp,

mlir/test/Conversion/GPUToSPIRV/builtins.mlir

@@ -293,6 +293,79 @@ module attributes {gpu.container_module} {
   }
 }
 
+// -----
+
+module attributes {gpu.container_module} {
+  func @builtin() {
+    %c0 = arith.constant 1 : index
+    gpu.launch_func @kernels::@builtin_global_id_x
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
+    return
+  }
+
+  // CHECK-LABEL:  spv.module @{{.*}} Logical GLSL450
+  // CHECK: spv.GlobalVariable [[GLOBALINVOCATIONID:@.*]] built_in("GlobalInvocationId")
+  gpu.module @kernels {
+    gpu.func @builtin_global_id_x() kernel
+      attributes {spv.entry_point_abi = {local_size = dense<[16, 1, 1]>: vector<3xi32>}} {
+      // CHECK: [[ADDRESS:%.*]] = spv.mlir.addressof [[GLOBALINVOCATIONID]]
+      // CHECK-NEXT: [[VEC:%.*]] = spv.Load "Input" [[ADDRESS]]
+      // CHECK-NEXT: {{%.*}} = spv.CompositeExtract [[VEC]]{{\[}}0 : i32{{\]}}
+      %0 = gpu.global_id x
+      gpu.return
+    }
+  }
+}
+
+// -----
+
+module attributes {gpu.container_module} {
+  func @builtin() {
+    %c0 = arith.constant 1 : index
+    gpu.launch_func @kernels::@builtin_global_id_y
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
+    return
+  }
+
+  // CHECK-LABEL:  spv.module @{{.*}} Logical GLSL450
+  // CHECK: spv.GlobalVariable [[GLOBALINVOCATIONID:@.*]] built_in("GlobalInvocationId")
+  gpu.module @kernels {
+    gpu.func @builtin_global_id_y() kernel
+      attributes {spv.entry_point_abi = {local_size = dense<[16, 1, 1]>: vector<3xi32>}} {
+      // CHECK: [[ADDRESS:%.*]] = spv.mlir.addressof [[GLOBALINVOCATIONID]]
+      // CHECK-NEXT: [[VEC:%.*]] = spv.Load "Input" [[ADDRESS]]
+      // CHECK-NEXT: {{%.*}} = spv.CompositeExtract [[VEC]]{{\[}}1 : i32{{\]}}
+      %0 = gpu.global_id y
+      gpu.return
+    }
+  }
+}
+
+// -----
+
+module attributes {gpu.container_module} {
+  func @builtin() {
+    %c0 = arith.constant 1 : index
+    gpu.launch_func @kernels::@builtin_global_id_z
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
+    return
+  }
+
+  // CHECK-LABEL:  spv.module @{{.*}} Logical GLSL450
+  // CHECK: spv.GlobalVariable [[GLOBALINVOCATIONID:@.*]] built_in("GlobalInvocationId")
+  gpu.module @kernels {
+    gpu.func @builtin_global_id_z() kernel
+      attributes {spv.entry_point_abi = {local_size = dense<[16, 1, 1]>: vector<3xi32>}} {
+      // CHECK: [[ADDRESS:%.*]] = spv.mlir.addressof [[GLOBALINVOCATIONID]]
+      // CHECK-NEXT: [[VEC:%.*]] = spv.Load "Input" [[ADDRESS]]
+      // CHECK-NEXT: {{%.*}} = spv.CompositeExtract [[VEC]]{{\[}}2 : i32{{\]}}
+      %0 = gpu.global_id z
+      gpu.return
+    }
+  }
+}
+
+
 // -----
 
 module attributes {gpu.container_module} {

mlir/test/Dialect/GPU/ops.mlir

@@ -44,6 +44,10 @@ module attributes {gpu.container_module} {
       %gDimY = gpu.grid_dim y
       %gDimZ = gpu.grid_dim z
 
+      %gIdX = gpu.global_id x
+      %gIdY = gpu.global_id y
+      %gIdZ = gpu.global_id z
+
       %sgId = gpu.subgroup_id : index
       %numSg = gpu.num_subgroups : index
       %SgSi = gpu.subgroup_size : index