运行 Java 杂志文章中的 Concurnas GPU 内核

Question

我正在尝试运行 introductory article 中给出的示例 GPU 内核。我将“图形处理单元”部分的代码片段合并到以下程序中：

gpukernel 1 twoArrayOp(global in A float[], global in B float[], global out result float[]) {
    idx = get_global_id(0)
    result[idx] = A[idx]**2 + B[idx] + 10
}

def runKernel() {
    //select a GPU device...
    device = gpus.GPU().getGPUDevices()[0].devices[0]
    
    //create three arrays of size 10 on this GPU, two as input
    inGPU1 = device.makeOffHeapArrayIn(float[].class, 10)
    inGPU2 = device.makeOffHeapArrayIn(float[].class, 10)
    //and one as output
    result = device.makeOffHeapArrayOut(float[].class, 10)
    
    //now write to the arrays on the GPU
    c1 := inGPU1.writeToBuffer([ 1.f 2 3 4 5 6 7 8 9 10])
    c2 := inGPU2.writeToBuffer([ 1.f 2 1 2 3 1 2 1 2 1])
    
    inst = twoArrayOp(inGPU1, inGPU2, result)
    compute := device.exe(inst, [10], c1, c2)//run 10 cores to process twoArrayOp
    ret = result.readFromBuffer(compute)
    return ret
}

h := runKernel()
System.out.println(h)

运行带有 Java 14 和 Intel UHD Graphics 620 GPU 的 Windows 系统上的代码我收到以下错误：

PS C:\Users\johannes\Documents> concc gpu.conc; conc gpu
C:\Users\johannes\Documents\gpu.conc line 0:0

Bug Report: Internal compiler error on compilation of: C:\Users\johannes\Documents\gpu.conc due to java.lang.IndexOutOfBoundsException: Index 0 out of bounds for length 0: Index 0 out of bounds for length 0
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
File name: C:\Users\johannes\Documents\gpu.conc

Compiler build number: 1.14.020
Exception in phase: VectorizedRedirector
Exception at line: 27
Exception: java.lang.IndexOutOfBoundsException: Index 0 out of bounds for length 0
Exception message: Index 0 out of bounds for length 0
Exception stacktrace:
java.lang.IndexOutOfBoundsException: Index 0 out of bounds for length 0
        at java.base/java.lang.Throwable.init(Throwable.java:272)
        at java.base/java.lang.Exception.init(Exception.java:67)
        at java.base/java.lang.RuntimeException.init(RuntimeException.java:63)
        at java.base/java.lang.IndexOutOfBoundsException.init(IndexOutOfBoundsException.java:55)
        at java.base/jdk.internal.util.Preconditions.outOfBounds(Preconditions.java:64)
        at java.base/jdk.internal.util.Preconditions.outOfBoundsCheckIndex(Preconditions.java:70)
        at java.base/jdk.internal.util.Preconditions.checkIndex(Preconditions.java:248)
        at java.base/java.util.Objects.checkIndex(Objects.java:373)
        at java.base/java.util.ArrayList.get(ArrayList.java:427)
        at com.concurnas.compiler.visitors.VectorizedRedirector.remapVectorization(VectorizedRedirector.java:1139)
        at com.concurnas.compiler.visitors.VectorizedRedirector.visit(VectorizedRedirector.java:1327)
        at com.concurnas.compiler.ast.FuncInvoke.accept(FuncInvoke.java:125)
        at com.concurnas.compiler.visitors.VectorizedRedirector.visit(VectorizedRedirector.java:1948)
        at com.concurnas.compiler.ast.DotOperator.accept(DotOperator.java:312)
        at com.concurnas.compiler.visitors.AbstractVisitor.visit(AbstractVisitor.java:556)
        at com.concurnas.compiler.ast.DuffAssign.accept(DuffAssign.java:27)
        at com.concurnas.compiler.visitors.AbstractErrorRaiseVisitor.visit(AbstractErrorRaiseVisitor.java:32)
        at com.concurnas.compiler.ast.LineHolder.accept(LineHolder.java:26)
        at com.concurnas.compiler.visitors.AbstractVisitor.visit(AbstractVisitor.java:740)
        at com.concurnas.compiler.visitors.VectorizedRedirector.visit(VectorizedRedirector.java:2042)
        at com.concurnas.compiler.ModuleCompiler.tryToProgressCompilation(ModuleCompiler.java:520)
        at com.concurnas.compiler.MainLoop.compileFile(MainLoop.java:384)
        at com.concurnas.concc.Concc.doMain(Concc.java:228)
        at com.concurnas.concc.Concc.doit(Concc.java:145)
        at com.concurnas.concc.Concc.main(Concc.java:89)

Source:
    1: gpukernel 1 twoArrayOp(global in A float[], global in B float[], global out result float[]) {
    2:     idx = get_global_id(0)
    3:     result[idx] = A[idx]**2 + B[idx] + 10
    4: }
    5:
    6: def runKernel() {
    7:     //select a GPU device...
    8:     device = gpus.GPU().getGPUDevices()[0].devices[0]
    9:
   10:     //create three arrays of size 10 on this GPU, two as input
   11:     inGPU1 = device.makeOffHeapArrayIn(float[].class, 10)
   12:     inGPU2 = device.makeOffHeapArrayIn(float[].class, 10)
   13:     //and one as output
   14:     result = device.makeOffHeapArrayOut(float[].class, 10)
   15:
   16:     //now write to the arrays on the GPU
   17:     c1 := inGPU1.writeToBuffer([ 1.f 2 3 4 5 6 7 8 9 10])
   18:     c2 := inGPU2.writeToBuffer([ 1.f 2 1 2 3 1 2 1 2 1])
   19:
   20:     inst = twoArrayOp(inGPU1, inGPU2, result)
   21:     compute := device.exe(inst, [10], c1, c2)//run 10 cores to process twoArrayOp
   22:     ret = result.readFromBuffer(compute)
   23:     return ret
   24: }
   25:
   26: h := runKernel()
-->27: System.out.println(h)

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Cannot find entry-point class to load: gpu

如何显示运行 GPU 内核的结果？

Answer 1

我在 Discord 上联系了该语言的设计者，他说这是一个错误，他将在下一个版本中修复。

运行 Java 杂志文章中的 Concurnas GPU 内核

Run the Concurnas GPU Kernel in the Java Magazine article

jvm

gpu