Squashed 'data_compression' changes from 662a84b..013f3d7 (#494)

013f3d7 Deleting old decompression kernels
f98128b adding gzip multicore decompress design
94e8c13 latestUpdate description.json for vivado impl
c656ec7 Adding pass-phrase
bb65325 Adding pass-phrase
b51f610 create master branch from next branch
9d1709a Failed GUI cases fix
8d4b7ab Jenkins fix
4b0591a return value in host code
921bdb1 failing GUI cases fix
7fd33e4 missed case
a4756e4 missed case

Co-authored-by: sdausr <sdausr@xilinx.com>

Author: 2 people

data_compression/Jenkinsfile

@@ -1,5 +1,5 @@
 @Library('pipeline-library')_
 
-VitisLibPipeline (branch: 'next', libname: 'xf_compression', run_sw_in_pr: 'true',
+VitisLibPipeline (branch: 'master', libname: 'xf_compression', run_sw_in_pr: 'true',
 TARGETS: 'hls_csim:hls_csynth:hls_cosim:vitis_sw_emu:vitis_hw_emu',
-upstream_dependencies: 'xf_security,next,../security', email: 'heeran@xilinx.com', devtest: 'RunDeploy.sh', TOOLVERSION: '2021.2_stable_latest')
+upstream_dependencies: 'xf_security,master,../security', email: 'heeran@xilinx.com', devtest: 'RunDeploy.sh', TOOLVERSION: '2021.2_released')

data_compression/L1/include/hw/inflate.hpp

@@ -423,6 +423,76 @@ void kStreamReadZlibDecomp(hls::stream<ap_axiu<STREAM_WIDTH, 0, 0, 0> >& in,
     outEos << 1; // Terminate condition
 }
 
+template <int NUM_CORE>
+void axi2HlsDistributor(hls::stream<ap_axiu<64, 0, 0, 0> >& in, hls::stream<ap_uint<72> > out[NUM_CORE]) {
+    for (auto i = 0; i < NUM_CORE; i++) {
+        bool last = false;
+        ap_axiu<64, 0, 0, 0> tmp;
+        ap_uint<8> kp = 0;
+        ap_uint<8> cntr = 0;
+        ap_uint<72> tmpVal = 0;
+        while (last == false) {
+#pragma HLS PIPELINE II = 1
+            tmp = in.read();
+            tmpVal.range(71, 8) = tmp.data;
+            kp = tmp.keep;
+            tmpVal.range(7, 0) = tmp.keep;
+            if (kp == 0xFF)
+                tmpVal.range(7, 0) = 8;
+            else
+                break;
+            out[i] << tmpVal;
+            last = tmp.last;
+        }
+        if (kp != 0xFF) {
+            while (kp) {
+                cntr += (kp & 0x1);
+                kp >>= 1;
+            }
+            tmpVal.range(7, 0) = cntr;
+            out[i] << tmpVal;
+        }
+        out[i] << 0; // Terminate condition
+    }
+}
+
+template <int STREAM_WIDTH, int NUM_CORE>
+void hls2AxiMerger(hls::stream<ap_axiu<64, 0, 0, 0> >& outKStream, hls::stream<ap_uint<72> > outDataStream[NUM_CORE]) {
+    for (auto i = 0; i < NUM_CORE; i++) {
+        ap_uint<8> strb = 0;
+        ap_uint<64> data;
+        ap_uint<72> tmp;
+        ap_axiu<64, 0, 0, 0> t1;
+
+        tmp = outDataStream[i].read();
+
+        strb = tmp.range(7, 0);
+        t1.data = tmp.range(71, 8);
+        t1.strb = strb;
+        t1.keep = strb;
+        t1.last = 0;
+        if (strb == 0) {
+            t1.last = 1;
+            outKStream << t1;
+        }
+        while (strb != 0) {
+#pragma HLS PIPELINE II = 1
+            tmp = outDataStream[i].read();
+
+            strb = tmp.range(7, 0);
+            if (strb == 0) {
+                t1.last = 1;
+            }
+            outKStream << t1;
+
+            t1.data = tmp.range(71, 8);
+            t1.strb = strb;
+            t1.keep = strb;
+            t1.last = 0;
+        }
+    }
+}
+
 template <int STREAM_WIDTH>
 void kStreamWriteZlibDecomp(hls::stream<ap_axiu<STREAM_WIDTH, 0, 0, 0> >& outKStream,
                             hls::stream<ap_uint<STREAM_WIDTH + (STREAM_WIDTH / 8)> >& outDataStream) {
@@ -701,6 +771,42 @@ void inflateWithChkSum(hls::stream<ap_uint<16> >& inStream,
 
 } // namespace details
 
+template <int NUM_CORE,
+          int DECODER,
+          int PARALLEL_BYTES,
+          int FILE_FORMAT,
+          bool LOW_LATENCY = false,
+          int HISTORY_SIZE = (32 * 1024)>
+void inflateMultiCores(hls::stream<ap_axiu<64, 0, 0, 0> >& inaxistream,
+                       hls::stream<ap_axiu<64, 0, 0, 0> >& outaxistream) {
+    constexpr int c_parallelBit = PARALLEL_BYTES * 8;
+
+    hls::stream<ap_uint<72> > axi2HlsStrm[NUM_CORE];
+    hls::stream<ap_uint<72> > inflateOut[NUM_CORE];
+    hls::stream<ap_uint<16> > hlsDownStrm[NUM_CORE];
+    hls::stream<bool> hlsEos[NUM_CORE];
+
+#pragma HLS STREAM variable = axi2HlsStrm depth = 4096
+#pragma HLS STREAM variable = inflateOut depth = 4096
+#pragma HLS STREAM variable = hlsDownStrm depth = 32
+#pragma HLS STREAM variable = hlsEos depth = 32
+
+#pragma HLS BIND_STORAGE variable = axi2HlsStrm type = FIFO impl = URAM
+#pragma HLS BIND_STORAGE variable = inflateOut type = fifo impl = URAM
+
+#pragma HLS dataflow disable_start_propagation
+    details::axi2HlsDistributor<NUM_CORE>(inaxistream, axi2HlsStrm);
+
+    for (auto i = 0; i < NUM_CORE; i++) {
+#pragma HLS UNROLL
+        details::bufferDownsizer<NUM_CORE, PARALLEL_BYTES>(axi2HlsStrm[i], hlsDownStrm[i], hlsEos[i]);
+        details::inflateMultiByteCore<DECODER, PARALLEL_BYTES, FILE_FORMAT, LOW_LATENCY, HISTORY_SIZE>(
+            hlsDownStrm[i], hlsEos[i], inflateOut[i]);
+    }
+
+    details::hls2AxiMerger<c_parallelBit, NUM_CORE>(outaxistream, inflateOut);
+}
+
 template <int DECODER, int PARALLEL_BYTES, int FILE_FORMAT, bool LOW_LATENCY = false, int HISTORY_SIZE = (32 * 1024)>
 void inflateMultiByte(hls::stream<ap_axiu<16, 0, 0, 0> >& inaxistream,
                       hls::stream<ap_axiu<PARALLEL_BYTES * 8, 0, 0, 0> >& outaxistream) {
@@ -714,7 +820,6 @@ void inflateMultiByte(hls::stream<ap_axiu<16, 0, 0, 0> >& inaxistream,
 #pragma HLS STREAM variable = axi2HlsStrm depth = 32
 #pragma HLS STREAM variable = axi2HlsEos depth = 32
 #pragma HLS STREAM variable = inflateOut depth = 32
-//#pragma HLS STREAM variable = inflateOut depth = 4096
 
 #pragma HLS BIND_STORAGE variable = axi2HlsStrm type = FIFO impl = SRL
 #pragma HLS BIND_STORAGE variable = axi2HlsEos type = FIFO impl = SRL

data_compression/L1/include/hw/stream_downsizer.hpp

@@ -99,6 +99,43 @@ void bufferDownsizer(hls::stream<ap_uint<IN_DATAWIDTH> >& inStream,
     outStream << outVal;
 }
 
+template <int NUM_CORE, int SWIDTH>
+void bufferDownsizer(hls::stream<ap_uint<72> >& inStream,
+                     hls::stream<ap_uint<16> >& outStream,
+                     hls::stream<bool>& outEos) {
+    constexpr int16_t c_factor = 4;  // = 64 / 16
+    constexpr int16_t c_outWord = 2; // = 16 / 2
+    ap_uint<16> outVal;
+
+    ap_uint<SWIDTH> dsize = 0;
+    ap_uint<SWIDTH> cntr = 0;
+    auto inVal = inStream.read();
+    // proceed further if valid size
+    ap_uint<SWIDTH> inSize = inVal.range(7, 0);
+    // if (inSize == 0) break;
+    auto outSizeV = ((inSize - 1) / c_outWord) + 1;
+downsizer_assign:
+    while (inSize > 0) {
+#pragma HLS PIPELINE II = 1
+        outVal = inVal.range(23, 8);
+        inVal >>= 16;
+        outStream << outVal;
+
+        outEos << 0;
+        dsize++;
+        if (dsize == outSizeV) {
+            inVal = inStream.read();
+            inSize = inVal.range(7, 0);
+
+            dsize = 0;
+            outSizeV = ((inSize - 1) / c_outWord) + 1;
+        }
+    }
+    // Block end Condition
+    outEos << 1;
+    outStream << 0;
+}
+
 template <int IN_DATAWIDTH, int OUT_DATAWIDTH, int SIZE_DWIDTH = 4>
 void bufferDownsizer(hls::stream<ap_uint<IN_DATAWIDTH + SIZE_DWIDTH> >& inStream,
                      hls::stream<IntVectorStream_dt<OUT_DATAWIDTH, 1> >& outStream) {
@@ -380,8 +417,6 @@ void streamDownSizerP2PComp(hls::stream<ap_uint<IN_WIDTH> >& inStream,
         // Send ouputSize of the module
         outStreamSize << size;
 
-    // printf("[ %s ] sizeOutputV %d input_size %d size_4m_mm2s %d \n", __FUNCTION__, sizeOutputV, input_size, size);
-
     conv512toV:
         for (int i = 0; i < sizeOutputV; i++) {
 #pragma HLS PIPELINE II = 1

data_compression/L1/tests/gzipc_16KB/description.json

@@ -52,7 +52,7 @@
                     "hls_vivado_syn": 16384, 
                     "hls_csim": 10240, 
                     "hls_cosim": 16384, 
-                    "hls_vivado_impl": 16384, 
+                    "hls_vivado_impl": 32768, 
                     "hls_csynth": 10240
                 }, 
                 "max_time_min": 300

data_compression/L2/demos/gzip/Makefile

@@ -109,14 +109,14 @@ endif
 ifneq (,$(shell echo $(XPLATFORM) | awk '/vck190/'))
 HOST_SRCS += $(CUR_DIR)/src/host.cpp $(XFLIB_DIR)/common/libs/compress/gzipOCLHost.cpp $(XFLIB_DIR)/common/libs/compress/gzipBase.cpp $(XFLIB_DIR)/common/libs/compress/gzipApp.cpp $(XFLIB_DIR)/common/libs/compress/compressApp.cpp $(XFLIB_DIR)/common/libs/xcl2/xcl2.cpp $(XFLIB_DIR)/common/libs/cmdparser/cmdlineparser.cpp $(XFLIB_DIR)/common/libs/logger/logger.cpp $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7/crc32.c $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7/adler32.c 
 CXXFLAGS +=  -D GZIP_MULTICORE_COMPRESS -D PARALLEL_BLOCK=1 -D GZIP_MODE=1
-CXXFLAGS +=  -I $(SYSROOT)/usr/include -I $(XFLIB_DIR)/L1/include/hw -I $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7 -I $(XFLIB_DIR)/common/libs/compress/ -I $(XFLIB_DIR)/common/libs/xcl2 -I $(XFLIB_DIR)/common/libs/cmdparser -I $(XFLIB_DIR)/common/libs/logger
+CXXFLAGS +=  -I $(SYSROOT)/usr/include -I $(XFLIB_DIR)/L1/include/hw -I $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7 -I $(XFLIB_DIR)/common/libs/compress/ -I $(XFLIB_DIR)/L2/tests/src -I $(XFLIB_DIR)/common/libs/xcl2 -I $(XFLIB_DIR)/common/libs/cmdparser -I $(XFLIB_DIR)/common/libs/logger
 CXXFLAGS += --sysroot=$(SYSROOT)
 LDFLAGS +=  -L $(SYSROOT)/usr/lib -L ${SYSROOT}/opt/xilinx/xrt/lib
 
 else 
 HOST_SRCS += $(CUR_DIR)/src/host.cpp $(XFLIB_DIR)/common/libs/compress/gzipOCLHost.cpp $(XFLIB_DIR)/common/libs/compress/gzipBase.cpp $(XFLIB_DIR)/common/libs/compress/gzipApp.cpp $(XFLIB_DIR)/common/libs/compress/compressApp.cpp $(XFLIB_DIR)/common/libs/xcl2/xcl2.cpp $(XFLIB_DIR)/common/libs/cmdparser/cmdlineparser.cpp $(XFLIB_DIR)/common/libs/logger/logger.cpp $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7/crc32.c $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7/adler32.c 
 CXXFLAGS +=  -D GZIP_MULTICORE_COMPRESS -D PARALLEL_BLOCK=1 -D GZIP_MODE=1
-CXXFLAGS +=  -I $(XFLIB_DIR)/L1/include/hw -I $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7 -I $(XFLIB_DIR)/common/libs/compress/ -I $(XFLIB_DIR)/common/libs/xcl2 -I $(XFLIB_DIR)/common/libs/cmdparser -I $(XFLIB_DIR)/common/libs/logger
+CXXFLAGS +=  -I $(XFLIB_DIR)/L1/include/hw -I $(XFLIB_DIR)/common/thirdParty/zlib-1.2.7 -I $(XFLIB_DIR)/common/libs/compress/ -I $(XFLIB_DIR)/L2/tests/src -I $(XFLIB_DIR)/common/libs/xcl2 -I $(XFLIB_DIR)/common/libs/cmdparser -I $(XFLIB_DIR)/common/libs/logger
 
 endif
 EXE_NAME := xil_gzip

data_compression/L2/demos/gzip/description.json

@@ -50,7 +50,8 @@
             "includepaths": [
                 "LIB_DIR/L1/include/hw",
                 "LIB_DIR/common/thirdParty/zlib-1.2.7",
-                "LIB_DIR/common/libs/compress/"
+                "LIB_DIR/common/libs/compress/",
+                "LIB_DIR/L2/tests/src"
             ], 
             "symbols": [
                 "GZIP_MULTICORE_COMPRESS", 

data_compression/L2/include/gzip_decompress_multicore.hpp

@@ -0,0 +1,57 @@
+/*
+ * (c) Copyright 2019-2021 Xilinx, Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ */
+
+#ifndef _XFCOMPRESSION_ZLIB_DECOMPRESS_QUAD_HPP_
+#define _XFCOMPRESSION_ZLIB_DECOMPRESS_QUAD_HPP_
+
+#include <stdio.h>
+#include <stdint.h>
+#include <assert.h>
+#include <ap_int.h>
+#include "hls_stream.h"
+#include "huffman_decoder.hpp"
+#include "zlib_specs.hpp"
+#include "lz_decompress.hpp"
+#include "inflate.hpp"
+#include "stream_downsizer.hpp"
+#include "ap_axi_sdata.h"
+
+// use dynamic decoder by default
+#ifndef DECODER_TYPE
+#define DECODER_TYPE c_fullDecoder
+#endif
+
+// by default disable low latency model
+#ifndef LL_MODEL
+#define LL_MODEL true
+#endif
+
+#ifndef NUM_CORE
+#define NUM_CORE 4
+#endif
+
+#ifndef MULTIPLE_BYTES
+#define MULTIPLE_BYTES 8
+#endif
+
+#define LZ_MAX_OFFSET_LIMIT 32768
+#define LOW_OFFSET 10
+
+extern "C" {
+void xilDecompress(hls::stream<ap_axiu<64, 0, 0, 0> >& inaxistreamd, hls::stream<ap_axiu<64, 0, 0, 0> >& outaxistreamd);
+}
+#endif // _XFCOMPRESSION_ZLIB_DECOMPRESS_QUAD_HPP_