Add a set of self-microbenchmarks (#51)

Co-authored-by: Lauri Kasanen <cand@gmx.com>

common/network/GetAPIMessager.cxx

@@ -21,6 +21,7 @@
 #include <inttypes.h>
 #include <network/GetAPI.h>
 #include <rfb/ConnParams.h>
+#include <rfb/EncodeManager.h>
 #include <rfb/LogWriter.h>
 #include <rfb/JpegCompressor.h>
 #include <rfb/xxhash.h>
@@ -32,10 +33,6 @@ using namespace rfb;
 
 static LogWriter vlog("GetAPIMessager");
 
-PixelBuffer *progressiveBilinearScale(const PixelBuffer *pb,
-					const uint16_t tgtw, const uint16_t tgth,
-					const float tgtdiff);
-
 struct TightJPEGConfiguration {
     int quality;
     int subsampling;

common/rfb/CMakeLists.txt

@@ -48,6 +48,7 @@ set(RFB_SOURCES
   Security.cxx
   SecurityServer.cxx
   SecurityClient.cxx
+  SelfBench.cxx
   SSecurityPlain.cxx
   SSecurityStack.cxx
   SSecurityVncAuth.cxx

common/rfb/EncodeManager.cxx

@@ -895,7 +895,7 @@ void EncodeManager::updateVideoStats(const std::vector<Rect> &rects, const Pixel
   }
 }
 
-static PixelBuffer *nearestScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
+PixelBuffer *rfb::nearestScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
                                  const float diff)
 {
   ManagedPixelBuffer *newpb = new ManagedPixelBuffer(pb->getPF(), w, h);
@@ -920,7 +920,7 @@ static PixelBuffer *nearestScale(const PixelBuffer *pb, const uint16_t w, const
   return newpb;
 }
 
-static PixelBuffer *bilinearScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
+PixelBuffer *rfb::bilinearScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
                                  const float diff)
 {
   ManagedPixelBuffer *newpb = new ManagedPixelBuffer(pb->getPF(), w, h);
@@ -968,7 +968,7 @@ static PixelBuffer *bilinearScale(const PixelBuffer *pb, const uint16_t w, const
   return newpb;
 }
 
-PixelBuffer *progressiveBilinearScale(const PixelBuffer *pb,
+PixelBuffer *rfb::progressiveBilinearScale(const PixelBuffer *pb,
                                  const uint16_t tgtw, const uint16_t tgth,
                                  const float tgtdiff)
 {

common/rfb/EncodeManager.h

@@ -215,6 +215,13 @@ namespace rfb {
       virtual rdr::U8* getBufferRW(const Rect& r, int* stride);
     };
   };
+
+  PixelBuffer *nearestScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
+                            const float diff);
+  PixelBuffer *bilinearScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
+                            const float diff);
+  PixelBuffer *progressiveBilinearScale(const PixelBuffer *pb, const uint16_t w, const uint16_t h,
+                            const float diff);
 }
 
 #endif

common/rfb/SelfBench.cxx

@@ -0,0 +1,197 @@
+/* Copyright (C) 2021 Kasm Web
+ *
+ * This is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This software is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this software; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307,
+ * USA.
+ */
+
+#include <rfb/ComparingUpdateTracker.h>
+#include <rfb/EncodeManager.h>
+#include <rfb/LogWriter.h>
+#include <rfb/SConnection.h>
+#include <rfb/ServerCore.h>
+#include <rfb/PixelBuffer.h>
+#include <rfb/TightJPEGEncoder.h>
+#include <rfb/TightWEBPEncoder.h>
+#include <rfb/util.h>
+#include <sys/time.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+using namespace rfb;
+static LogWriter vlog("SelfBench");
+
+static const PixelFormat pfRGBX(32, 24, false, true, 255, 255, 255, 0, 8, 16);
+
+#define RUNS 64
+
+#define W 1600
+#define H 1200
+
+void SelfBench() {
+
+	unsigned i, runs;
+	struct timeval start;
+
+	ManagedPixelBuffer f1(pfRGBX, W, H);
+	ManagedPixelBuffer f2(pfRGBX, W, H);
+	ManagedPixelBuffer screen(pfRGBX, W, H);
+
+	int stride;
+	rdr::U8 *f1ptr = f1.getBufferRW(f1.getRect(), &stride);
+	rdr::U8 *f2ptr = f2.getBufferRW(f2.getRect(), &stride);
+	rdr::U8 * const screenptr = screen.getBufferRW(screen.getRect(), &stride);
+
+	rdr::U8 * const f1orig = f1ptr;
+	rdr::U8 * const f2orig = f2ptr;
+
+	for (i = 0; i < W * H * 4; i += 4) {
+		f1ptr[0] = rand();
+		f1ptr[1] = rand();
+		f1ptr[2] = rand();
+
+		f2ptr[0] = rand();
+		f2ptr[1] = rand();
+		f2ptr[2] = rand();
+
+		f1ptr += 4;
+		f2ptr += 4;
+	}
+
+	vlog.info("Running micro-benchmarks (single-threaded, runs depending on task)");
+
+	// Encoding
+	std::vector<uint8_t> vec;
+
+	TightJPEGEncoder jpeg(NULL);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		jpeg.compressOnly(&f1, 8, vec, false);
+	}
+	vlog.info("Jpeg compression at quality 8 took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		jpeg.compressOnly(&f1, 4, vec, false);
+	}
+	vlog.info("Jpeg compression at quality 4 took %u ms (%u runs)", msSince(&start), runs);
+
+
+	TightWEBPEncoder webp(NULL);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS / 8;
+	for (i = 0; i < runs; i++) {
+		webp.compressOnly(&f1, 8, vec, false);
+	}
+	vlog.info("Webp compression at quality 8 took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS / 4;
+	for (i = 0; i < runs; i++) {
+		webp.compressOnly(&f1, 4, vec, false);
+	}
+	vlog.info("Webp compression at quality 4 took %u ms (%u runs)", msSince(&start), runs);
+
+	// Scaling
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		PixelBuffer *pb = nearestScale(&f1, W * 0.8, H * 0.8, 0.8);
+		delete pb;
+	}
+	vlog.info("Nearest scaling to 80%% took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		PixelBuffer *pb = nearestScale(&f1, W * 0.4, H * 0.4, 0.4);
+		delete pb;
+	}
+	vlog.info("Nearest scaling to 40%% took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		PixelBuffer *pb = bilinearScale(&f1, W * 0.8, H * 0.8, 0.8);
+		delete pb;
+	}
+	vlog.info("Bilinear scaling to 80%% took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		PixelBuffer *pb = bilinearScale(&f1, W * 0.4, H * 0.4, 0.4);
+		delete pb;
+	}
+	vlog.info("Bilinear scaling to 40%% took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		PixelBuffer *pb = progressiveBilinearScale(&f1, W * 0.8, H * 0.8, 0.8);
+		delete pb;
+	}
+	vlog.info("Progressive bilinear scaling to 80%% took %u ms (%u runs)", msSince(&start), runs);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		PixelBuffer *pb = progressiveBilinearScale(&f1, W * 0.4, H * 0.4, 0.4);
+		delete pb;
+	}
+	vlog.info("Progressive bilinear scaling to 40%% took %u ms (%u runs)", msSince(&start), runs);
+
+	// Analysis
+	ComparingUpdateTracker *comparer = new ComparingUpdateTracker(&screen);
+	Region cursorReg;
+
+	Server::detectScrolling.setParam(false);
+	Server::detectHorizontal.setParam(false);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		memcpy(screenptr, i % 2 ? f1orig : f2orig, W * H * 4);
+		comparer->compare(true, cursorReg);
+	}
+	vlog.info("Analysis took %u ms (%u runs) (incl. memcpy overhead)", msSince(&start), runs);
+
+	Server::detectScrolling.setParam(true);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS;
+	for (i = 0; i < runs; i++) {
+		memcpy(screenptr, i % 2 ? f1orig : f2orig, W * H * 4);
+		comparer->compare(false, cursorReg);
+	}
+	vlog.info("Analysis w/ scroll detection took %u ms (%u runs) (incl. memcpy overhead)", msSince(&start), runs);
+
+	Server::detectHorizontal.setParam(true);
+	delete comparer;
+	comparer = new ComparingUpdateTracker(&screen);
+
+	gettimeofday(&start, NULL);
+	runs = RUNS / 2;
+	for (i = 0; i < runs; i++) {
+		memcpy(screenptr, i % 2 ? f1orig : f2orig, W * H * 4);
+		comparer->compare(false, cursorReg);
+	}
+	vlog.info("Analysis w/ horizontal scroll detection took %u ms (%u runs) (incl. memcpy overhead)", msSince(&start), runs);
+
+	exit(0);
+}

common/rfb/ServerCore.cxx

@@ -113,6 +113,10 @@ rfb::BoolParameter rfb::Server::ignoreClientSettingsKasm
 ("IgnoreClientSettingsKasm",
  "Ignore the additional client settings exposed in Kasm.",
  false);
+rfb::BoolParameter rfb::Server::selfBench
+("SelfBench",
+ "Run self-benchmarks and exit.",
+ false);
 rfb::IntParameter rfb::Server::dynamicQualityMin
 ("DynamicQualityMin",
  "The minimum dynamic JPEG quality, 0 = low, 9 = high",

common/rfb/ServerCore.h

@@ -74,6 +74,7 @@ namespace rfb {
     static BoolParameter detectScrolling;
     static BoolParameter detectHorizontal;
     static BoolParameter ignoreClientSettingsKasm;
+    static BoolParameter selfBench;
     static PresetParameter preferBandwidth;
 
   };

common/rfb/VNCServerST.cxx

@@ -76,6 +76,8 @@ static LogWriter slog("VNCServerST");
 LogWriter VNCServerST::connectionsLog("Connections");
 EncCache VNCServerST::encCache;
 
+void SelfBench();
+
 //
 // -=- VNCServerST Implementation
 //
@@ -212,6 +214,9 @@ VNCServerST::VNCServerST(const char* name_, SDesktop* desktop_)
   }
 
   trackingClient[0] = 0;
+
+  if (Server::selfBench)
+    SelfBench();
 }
 
 VNCServerST::~VNCServerST()

unix/xserver/hw/vnc/Xvnc.man

@@ -317,6 +317,10 @@ Log clipboard and keyboard actions. Info logs just clipboard direction and size,
 verbose adds the contents for both.
 .
 .TP
+.B \-selfBench
+Run a set of self-benchmarks and exit.
+.
+.TP
 .B \-noWebsocket
 Disable websockets and expose a traditional VNC port (5901, etc.).
 .