mirror of
https://github.com/capstone-engine/llvm-capstone.git
synced 2024-12-02 10:49:22 +00:00
9048ea28da
This reverts commit 5561e17411
The logic was moved from cmake into lit fixing the issue that lead to the revert and potentially others with multi-config cmake generators
Differential Revision: https://reviews.llvm.org/D143925
57 lines
2.3 KiB
MLIR
57 lines
2.3 KiB
MLIR
// RUN: mlir-vulkan-runner %s --shared-libs=%vulkan-runtime-wrappers,%mlir_runner_utils --entry-point-result=void | FileCheck %s
|
|
|
|
// CHECK: Compute shader execution time
|
|
// CHECK: Command buffer submit time
|
|
// CHECK: Wait idle time
|
|
|
|
module attributes {
|
|
gpu.container_module,
|
|
spirv.target_env = #spirv.target_env<
|
|
#spirv.vce<v1.0, [Shader], [SPV_KHR_storage_buffer_storage_class]>, #spirv.resource_limits<>>
|
|
} {
|
|
gpu.module @kernels {
|
|
gpu.func @kernel_add(%arg0 : memref<16384xf32>, %arg1 : memref<16384xf32>, %arg2 : memref<16384xf32>)
|
|
kernel attributes { spirv.entry_point_abi = #spirv.entry_point_abi<workgroup_size = [128, 1, 1]>} {
|
|
%bid = gpu.block_id x
|
|
%tid = gpu.thread_id x
|
|
%cst = arith.constant 128 : index
|
|
%b = arith.muli %bid, %cst : index
|
|
%0 = arith.addi %b, %tid : index
|
|
%1 = memref.load %arg0[%0] : memref<16384xf32>
|
|
%2 = memref.load %arg1[%0] : memref<16384xf32>
|
|
%3 = arith.addf %1, %2 : f32
|
|
memref.store %3, %arg2[%0] : memref<16384xf32>
|
|
gpu.return
|
|
}
|
|
}
|
|
|
|
func.func @main() {
|
|
%arg0 = memref.alloc() : memref<16384xf32>
|
|
%arg1 = memref.alloc() : memref<16384xf32>
|
|
%arg2 = memref.alloc() : memref<16384xf32>
|
|
%0 = arith.constant 0 : i32
|
|
%1 = arith.constant 1 : i32
|
|
%2 = arith.constant 2 : i32
|
|
%value0 = arith.constant 0.0 : f32
|
|
%value1 = arith.constant 1.1 : f32
|
|
%value2 = arith.constant 2.2 : f32
|
|
%arg3 = memref.cast %arg0 : memref<16384xf32> to memref<?xf32>
|
|
%arg4 = memref.cast %arg1 : memref<16384xf32> to memref<?xf32>
|
|
%arg5 = memref.cast %arg2 : memref<16384xf32> to memref<?xf32>
|
|
call @fillResource1DFloat(%arg3, %value1) : (memref<?xf32>, f32) -> ()
|
|
call @fillResource1DFloat(%arg4, %value2) : (memref<?xf32>, f32) -> ()
|
|
call @fillResource1DFloat(%arg5, %value0) : (memref<?xf32>, f32) -> ()
|
|
|
|
%cst1 = arith.constant 1 : index
|
|
%cst128 = arith.constant 128 : index
|
|
gpu.launch_func @kernels::@kernel_add
|
|
blocks in (%cst128, %cst1, %cst1) threads in (%cst128, %cst1, %cst1)
|
|
args(%arg0 : memref<16384xf32>, %arg1 : memref<16384xf32>, %arg2 : memref<16384xf32>)
|
|
%arg6 = memref.cast %arg5 : memref<?xf32> to memref<*xf32>
|
|
return
|
|
}
|
|
func.func private @fillResource1DFloat(%0 : memref<?xf32>, %1 : f32)
|
|
func.func private @printMemrefF32(%ptr : memref<*xf32>)
|
|
}
|
|
|