diff options
| -rw-r--r-- | test/com/jogamp/opencl/CLCommandQueueTest.java | 295 |
1 files changed, 152 insertions, 143 deletions
diff --git a/test/com/jogamp/opencl/CLCommandQueueTest.java b/test/com/jogamp/opencl/CLCommandQueueTest.java index cafdbc1..2147a6b 100644 --- a/test/com/jogamp/opencl/CLCommandQueueTest.java +++ b/test/com/jogamp/opencl/CLCommandQueueTest.java @@ -63,61 +63,61 @@ public class CLCommandQueueTest { CLContext context = CLContext.create(); - CLBuffer<ByteBuffer> clBufferA = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLBuffer<ByteBuffer> clBufferB = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLBuffer<ByteBuffer> clBufferC = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLBuffer<ByteBuffer> clBufferD = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - - fillBuffer(clBufferA.buffer, 12345); - fillBuffer(clBufferB.buffer, 67890); - - CLProgram program = context.createProgram(getClass().getResourceAsStream("testkernels.cl")).build(); - CLKernel vectorAddKernel = program.createCLKernel("VectorAddGM").setArg(3, elements); - CLCommandQueue queue = context.getDevices()[0].createCommandQueue(); - - out.println(queue); - - final CLEventList events = new CLEventList(2); + try{ - out.println(events); + CLBuffer<ByteBuffer> clBufferA = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + CLBuffer<ByteBuffer> clBufferB = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + CLBuffer<ByteBuffer> clBufferC = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + CLBuffer<ByteBuffer> clBufferD = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - assertEquals(0, events.size()); + fillBuffer(clBufferA.buffer, 12345); + fillBuffer(clBufferB.buffer, 67890); - queue.putWriteBuffer(clBufferA, false, events) // write A - .putWriteBuffer(clBufferB, false, events);// write B + CLProgram program = context.createProgram(getClass().getResourceAsStream("testkernels.cl")).build(); + CLKernel vectorAddKernel = program.createCLKernel("VectorAddGM").setArg(3, elements); + CLCommandQueue queue = context.getDevices()[0].createCommandQueue(); - out.println(events); + out.println(queue); - assertEquals(2, events.size()); - queue.putWaitForEvents(events, true); + final CLEventList events = new CLEventList(2); - events.release(); - assertEquals(0, events.size()); + out.println(events); - vectorAddKernel.setArgs(clBufferA, clBufferB, clBufferC); // C = A+B - queue.put1DRangeKernel(vectorAddKernel, 0, elements, groupSize, events); + assertEquals(0, events.size()); - vectorAddKernel.setArgs(clBufferA, clBufferB, clBufferD); // D = A+B - queue.put1DRangeKernel(vectorAddKernel, 0, elements, groupSize, events); + queue.putWriteBuffer(clBufferA, false, events) // write A + .putWriteBuffer(clBufferB, false, events);// write B - assertEquals(2, events.size()); - queue.putWaitForEvent(events, 0, false) - .putWaitForEvent(events, 1, true); + out.println(events); - queue.putReadBuffer(clBufferC, false) - .putReadBuffer(clBufferD, true); + assertEquals(2, events.size()); + queue.putWaitForEvents(events, true); - events.release(); + events.release(); + assertEquals(0, events.size()); - checkIfEqual(clBufferC.buffer, clBufferD.buffer, elements); + vectorAddKernel.setArgs(clBufferA, clBufferB, clBufferC); // C = A+B + queue.put1DRangeKernel(vectorAddKernel, 0, elements, groupSize, events); + vectorAddKernel.setArgs(clBufferA, clBufferB, clBufferD); // D = A+B + queue.put1DRangeKernel(vectorAddKernel, 0, elements, groupSize, events); - context.release(); + assertEquals(2, events.size()); + queue.putWaitForEvent(events, 0, false) + .putWaitForEvent(events, 1, true); + queue.putReadBuffer(clBufferC, false) + .putReadBuffer(clBufferD, true); - out.println("results are valid"); + events.release(); + checkIfEqual(clBufferC.buffer, clBufferD.buffer, elements); + out.println("results are valid"); + }finally{ + context.release(); + } } + @Test public void profilingEventsTest() throws IOException { @@ -127,44 +127,48 @@ public class CLCommandQueueTest { CLContext context = CLContext.create(); - CLBuffer<ByteBuffer> clBufferA = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLBuffer<ByteBuffer> clBufferB = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLBuffer<ByteBuffer> clBufferC = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + try { - fillBuffer(clBufferA.buffer, 12345); - fillBuffer(clBufferB.buffer, 67890); + CLBuffer<ByteBuffer> clBufferA = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + CLBuffer<ByteBuffer> clBufferB = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + CLBuffer<ByteBuffer> clBufferC = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLProgram program = context.createProgram(getClass().getResourceAsStream("testkernels.cl")).build(); - CLKernel vectorAddKernel = program.createCLKernel("VectorAddGM").setArg(3, elements); - CLCommandQueue queue = context.getDevices()[0].createCommandQueue(Mode.PROFILING_MODE); + fillBuffer(clBufferA.buffer, 12345); + fillBuffer(clBufferB.buffer, 67890); - out.println(queue); + CLProgram program = context.createProgram(getClass().getResourceAsStream("testkernels.cl")).build(); + CLKernel vectorAddKernel = program.createCLKernel("VectorAddGM").setArg(3, elements); + CLCommandQueue queue = context.getDevices()[0].createCommandQueue(Mode.PROFILING_MODE); - queue.putWriteBuffer(clBufferA, true) // write A - .putWriteBuffer(clBufferB, true);// write B + out.println(queue); - final CLEventList events = new CLEventList(1); + queue.putWriteBuffer(clBufferA, true) // write A + .putWriteBuffer(clBufferB, true);// write B - assertEquals(0, events.size()); + final CLEventList events = new CLEventList(1); - vectorAddKernel.setArgs(clBufferA, clBufferB, clBufferC); // C = A+B - queue.put1DRangeKernel(vectorAddKernel, 0, elements, groupSize, events); + assertEquals(0, events.size()); - assertEquals(1, events.size()); - CLEvent probe = events.getEvent(0); - out.println(probe); + vectorAddKernel.setArgs(clBufferA, clBufferB, clBufferC); // C = A+B + queue.put1DRangeKernel(vectorAddKernel, 0, elements, groupSize, events); - queue.putWaitForEvents(events, true); - assertEquals(CLEvent.ExecutionStatus.COMPLETE, probe.getStatus()); + assertEquals(1, events.size()); + CLEvent probe = events.getEvent(0); + out.println(probe); - out.println(probe); - long time = probe.getProfilingInfo(CLEvent.ProfilingCommand.END) - - probe.getProfilingInfo(CLEvent.ProfilingCommand.START); - out.println("time: "+time); - assertTrue(time > 0); + queue.putWaitForEvents(events, true); + assertEquals(CLEvent.ExecutionStatus.COMPLETE, probe.getStatus()); - events.release(); - context.release(); + out.println(probe); + long time = probe.getProfilingInfo(CLEvent.ProfilingCommand.END) + - probe.getProfilingInfo(CLEvent.ProfilingCommand.START); + out.println("time: "+time); + assertTrue(time > 0); + + events.release(); + }finally{ + context.release(); + } } @@ -261,26 +265,29 @@ public class CLCommandQueueTest { CLContext context = CLContext.create(); - final CLUserEvent customEvent = CLUserEvent.create(context); + try{ - final CountDownLatch countdown = new CountDownLatch(1); - customEvent.registerCallback(new CLEventListener() { + final CLUserEvent customEvent = CLUserEvent.create(context); - public void eventStateChanged(CLEvent event, int status) { - out.println("event received: "+event); - assertEquals(event, customEvent); - countdown.countDown(); - } + final CountDownLatch countdown = new CountDownLatch(1); + customEvent.registerCallback(new CLEventListener() { - }); + public void eventStateChanged(CLEvent event, int status) { + out.println("event received: "+event); + assertEquals(event, customEvent); + countdown.countDown(); + } - customEvent.setStatus(ExecutionStatus.COMPLETE); - countdown.await(2, TimeUnit.SECONDS); - assertEquals(countdown.getCount(), 0); + }); - customEvent.release(); + customEvent.setStatus(ExecutionStatus.COMPLETE); + countdown.await(2, TimeUnit.SECONDS); + assertEquals(countdown.getCount(), 0); - context.release(); + customEvent.release(); + }finally{ + context.release(); + } } @@ -293,101 +300,103 @@ public class CLCommandQueueTest { CLContext context = CLContext.create(); - CLDevice[] devices = context.getDevices(); + try{ - // ignore this test if we can't test in parallel - if (devices.length < 2) { - out.println("aborting test... need at least 2 devices"); - context.release(); - return; - } + CLDevice[] devices = context.getDevices(); - final CLBuffer<ByteBuffer> clBufferC = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - final CLBuffer<ByteBuffer> clBufferD = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + // ignore this test if we can't test in parallel + if (devices.length < 2) { + out.println("aborting test... need at least 2 devices"); + return; + } - final CLBuffer<ByteBuffer> clBufferA1 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - final CLBuffer<ByteBuffer> clBufferB1 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - final CLBuffer<ByteBuffer> clBufferA2 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - final CLBuffer<ByteBuffer> clBufferB2 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + final CLBuffer<ByteBuffer> clBufferC = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + final CLBuffer<ByteBuffer> clBufferD = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - CLProgram program = context.createProgram(getClass().getResourceAsStream("testkernels.cl")).build(); + final CLBuffer<ByteBuffer> clBufferA1 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + final CLBuffer<ByteBuffer> clBufferB1 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + final CLBuffer<ByteBuffer> clBufferA2 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); + final CLBuffer<ByteBuffer> clBufferB2 = context.createByteBuffer(elements * SIZEOF_INT, Mem.READ_ONLY); - //two independent kernel instances - final CLKernel vectorAddKernel1 = program.createCLKernel("VectorAddGM").setArg(3, elements); - final CLKernel vectorAddKernel2 = program.createCLKernel("VectorAddGM").setArg(3, elements); + CLProgram program = context.createProgram(getClass().getResourceAsStream("testkernels.cl")).build(); - final CLCommandQueue queue1 = devices[0].createCommandQueue(); - final CLCommandQueue queue2 = devices[1].createCommandQueue(); + //two independent kernel instances + final CLKernel vectorAddKernel1 = program.createCLKernel("VectorAddGM").setArg(3, elements); + final CLKernel vectorAddKernel2 = program.createCLKernel("VectorAddGM").setArg(3, elements); - out.println(queue1); - out.println(queue2); + final CLCommandQueue queue1 = devices[0].createCommandQueue(); + final CLCommandQueue queue2 = devices[1].createCommandQueue(); - fillBuffer(clBufferC.buffer, 12345); + out.println(queue1); + out.println(queue2); + fillBuffer(clBufferC.buffer, 12345); - final MultiQueueBarrier barrier = new MultiQueueBarrier(2); - Thread thread1 = new Thread("C") { + final MultiQueueBarrier barrier = new MultiQueueBarrier(2); - @Override - public void run() { + Thread thread1 = new Thread("C") { - fillBuffer(clBufferA1.buffer, 12345); - fillBuffer(clBufferB1.buffer, 67890); + @Override + public void run() { -// System.out.println("C buffer"); - queue1.putWriteBuffer(clBufferA1, false) // write A - .putWriteBuffer(clBufferB1, false); // write B + fillBuffer(clBufferA1.buffer, 12345); + fillBuffer(clBufferB1.buffer, 67890); -// System.out.println("C args"); - vectorAddKernel1.setArgs(clBufferA1, clBufferB1, clBufferC); // C = A+B + // System.out.println("C buffer"); + queue1.putWriteBuffer(clBufferA1, false) // write A + .putWriteBuffer(clBufferB1, false); // write B -// System.out.println("C kernels"); - CLEventList events1 = new CLEventList(2); - queue1.put1DRangeKernel(vectorAddKernel1, 0, elements, groupSize, events1) - .putReadBuffer(clBufferC, false, events1); + // System.out.println("C args"); + vectorAddKernel1.setArgs(clBufferA1, clBufferB1, clBufferC); // C = A+B - barrier.waitFor(queue1, events1); + // System.out.println("C kernels"); + CLEventList events1 = new CLEventList(2); + queue1.put1DRangeKernel(vectorAddKernel1, 0, elements, groupSize, events1) + .putReadBuffer(clBufferC, false, events1); - } - }; + barrier.waitFor(queue1, events1); - Thread thread2 = new Thread("D") { + } + }; - @Override - public void run() { + Thread thread2 = new Thread("D") { - fillBuffer(clBufferA2.buffer, 12345); - fillBuffer(clBufferB2.buffer, 67890); + @Override + public void run() { -// System.out.println("D buffer"); - queue2.putWriteBuffer(clBufferA2, false) // write A - .putWriteBuffer(clBufferB2, false); // write B + fillBuffer(clBufferA2.buffer, 12345); + fillBuffer(clBufferB2.buffer, 67890); -// System.out.println("D args"); - vectorAddKernel2.setArgs(clBufferA2, clBufferB2, clBufferD); // D = A+B + // System.out.println("D buffer"); + queue2.putWriteBuffer(clBufferA2, false) // write A + .putWriteBuffer(clBufferB2, false); // write B -// System.out.println("D kernels"); - CLEventList events2 = new CLEventList(2); - queue2.put1DRangeKernel(vectorAddKernel2, 0, elements, groupSize, events2) - .putReadBuffer(clBufferD, false, events2); + // System.out.println("D args"); + vectorAddKernel2.setArgs(clBufferA2, clBufferB2, clBufferD); // D = A+B - barrier.waitFor(queue2, events2); + // System.out.println("D kernels"); + CLEventList events2 = new CLEventList(2); + queue2.put1DRangeKernel(vectorAddKernel2, 0, elements, groupSize, events2); + queue2.putReadBuffer(clBufferD, false, events2); - } - }; + barrier.waitFor(queue2, events2); - out.println("starting threads"); - thread1.start(); - thread2.start(); - barrier.await(); - out.println("done"); + } + }; - checkIfEqual(clBufferC.buffer, clBufferD.buffer, elements); + out.println("starting threads"); + thread1.start(); + thread2.start(); + assertTrue(barrier.await(5, TimeUnit.SECONDS)); + out.println("done"); - context.release(); + checkIfEqual(clBufferC.buffer, clBufferD.buffer, elements); + out.println("results are valid"); - out.println("results are valid"); + }finally{ + context.release(); + } } } |