改critical零拷贝，性能+5%

.gitignore

@@ -3,3 +3,21 @@
 /input3/
 /PngSuite-2017jul19.zip
 /input/
+/hs_err_pid22004.log
+/hs_err_pid22004.mdmp
+/hs_err_pid24748.log
+/hs_err_pid24748.mdmp
+/hs_err_pid31056.log
+/hs_err_pid31056.mdmp
+/hs_err_pid31060.log
+/hs_err_pid31060.mdmp
+/hs_err_pid33172.log
+/hs_err_pid33172.mdmp
+/hs_err_pid33180.log
+/hs_err_pid33180.mdmp
+/hs_err_pid33300.log
+/hs_err_pid33300.mdmp
+/hs_err_pid33776.log
+/hs_err_pid33776.mdmp
+/hs_err_pid5920.log
+/hs_err_pid5920.mdmp

src/org/kne/codec/kif/KIFCodec.java

@@ -66,7 +66,7 @@ public class KIFCodec {
 		CompletableFuture<KIFCompressedImageBlock>f2=f1.thenApplyAsync((pipeline)->{
 			return pipeline.encodeStage2();
 		},cpupool);*/
-		return getBestPool().submit(()->{
+		return gpupool.submit(()->{
 			KIFEncodePipeline kip=new KIFEncodePipeline(input.get(), mode);
 			kip.encodeStage1();
 			return kip.encodeStage2();
@@ -122,7 +122,7 @@ public class KIFCodec {
 				throw new RuntimeException(e);
 			}
 		},gpupool);*/
-		return getBestPool().submit(()->{
+		return gpupool.submit(()->{
 			KIFDecodePipeline kip=new KIFDecodePipeline(input.get());
 			try {
 				kip.decodeStage1();

src/org/kne/codec/kif/KIFDecodePipeline.java

@@ -53,9 +53,6 @@ public class KIFDecodePipeline {
 		    return decodeStage2CPU();
 		}
 	}
-	private static ThreadLocal<MemorySegment>m1=ThreadLocal.withInitial(()->{
-		return ((ArenaImpl)Arena.ofAuto()).allocateNoInit(1024*1024*4,4);
-	});
 
 	private static ThreadLocalOpenCLMemory tlm0=new ThreadLocalOpenCLMemory(1024*1024*4);
 	private static ThreadLocalOpenCLMemory tlm1=new ThreadLocalOpenCLMemory(1024*1024*4);
@@ -69,13 +66,10 @@ public class KIFDecodePipeline {
 			OpenCLMemory ocm2 = null; 
 			OpenCLMemory ocm3 = null; 
 		       int[]des;
-			MemorySegment dataserpseg= m1.get();
+		     
-		      tdb.mark("堆外内存分配");
+				ocm1=tlm0.get(oct.getOpenCLContext());
-			dataserpseg.asByteBuffer().put(dataserp).flip();
+			      tdb.mark("分配显存");
-		      tdb.mark("堆内存->堆外内存");
+				oct.getOpenCLCommandQueue().executeWriteBuffer(ocm1, MemorySegment.ofArray(dataserp), true,0,0,dataserp.length);
-			ocm1=tlm0.get(oct.getOpenCLContext());
-		      tdb.mark("分配显存");
-			oct.getOpenCLCommandQueue().executeWriteBuffer(ocm1, dataserpseg, false,0,0,dataserp.length);
 		      tdb.mark("内存->显存");
 			
 			OpenCLMemory dataser;

src/org/kne/codec/kif/KIFEncodePipeline.java

@@ -80,10 +80,6 @@ public class KIFEncodePipeline {
 	}
 
 
-	private static ThreadLocal<MemorySegment>m1=ThreadLocal.withInitial(()->{
-		return ((ArenaImpl)Arena.ofAuto()).allocateNoInit(1024*1024*4,4);
-	});
-
 	private static ThreadLocalOpenCLMemory tlm0=new ThreadLocalOpenCLMemory(1024*1024*4);
 	private static ThreadLocalOpenCLMemory tlmpred1=new ThreadLocalOpenCLMemory(1024*1024*4);
 	private static ThreadLocalOpenCLMemory tlmpred2=new ThreadLocalOpenCLMemory(1024*1024*4);
@@ -96,7 +92,6 @@ public class KIFEncodePipeline {
 			OpenCLMemory ocm1 = null; 
 			OpenCLMemory ocm2 = null; 
 			OpenCLMemory ocm3 = null; 
-			MemorySegment dataserpseg= m1.get();
 		      tdb.mark("堆外内存分配");
 	        //MemorySegment.copy(pixels,0,dataserpseg,0, pixels.byteSize());
 		      tdb.mark("堆内存->堆外内存");
@@ -170,10 +165,8 @@ public class KIFEncodePipeline {
 	        serialData2=MembandBoost.allocateUninitializedByteArray(width*height*4);
 		      tdb.mark("分配堆内存");
 
-	        oct.getOpenCLCommandQueue().executeReadBuffer(dataser, dataserpseg, true,0,0,width*height*4);
+	        oct.getOpenCLCommandQueue().executeReadBuffer(dataser, MemorySegment.ofArray(serialData2), true,0,0,width*height*4);
 		      tdb.mark("显存->内存");
-	        MemorySegment.copy(dataserpseg, ValueLayout.JAVA_BYTE, 0, serialData2, 0,width*height*4);
-		      tdb.mark("堆外内存->堆内存");
 			
 	}
 	public KIFCompressedImageBlock encodeStage2() {