diff --git a/.gitignore b/.gitignore
index 131f9a9..2457d65 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,5 @@
 fio
 *.o
+*.d
 .depend
 cscope.out
diff --git a/HOWTO b/HOWTO
index f7948c3..1441734 100644
--- a/HOWTO
+++ b/HOWTO
@@ -1401,6 +1401,9 @@
 [netsplice] port=int
 [net] port=int	The TCP or UDP port to bind to or connect to.
 
+[netsplice] nodelay=bool
+[net] nodelay=bool	Set TCP_NODELAY on TCP connections.
+
 [netsplice] protocol=str
 [netsplice] proto=str
 [net] protocol=str
diff --git a/Makefile b/Makefile
index 0e79720..a14d1f3 100644
--- a/Makefile
+++ b/Makefile
@@ -128,10 +128,13 @@
 endif
 
 OBJS = $(SOURCE:.c=.o)
+
 FIO_OBJS = $(OBJS) fio.o
 GFIO_OBJS = $(OBJS) gfio.o graph.o tickmarks.o ghelpers.o goptions.o gerror.o \
 			gclient.o gcompat.o cairo_text_helpers.o printing.o
 
+-include $(OBJS:.o=.d)
+
 T_SMALLOC_OBJS = t/stest.o
 T_SMALLOC_OBJS += gettime.o mutex.o smalloc.o t/log.o
 T_SMALLOC_PROGS = t/stest
@@ -175,7 +178,7 @@
 mandir = $(prefix)/man
 endif
 
-all: .depend $(PROGS) $(SCRIPTS) FORCE
+all: $(PROGS) $(SCRIPTS) FORCE
 
 .PHONY: all install clean
 .PHONY: FORCE cscope
@@ -184,10 +187,16 @@
 	@$(SHELL) ./FIO-VERSION-GEN
 -include FIO-VERSION-FILE
 
-CFLAGS += -DFIO_VERSION='"$(FIO_VERSION)"'
+override CFLAGS += -DFIO_VERSION='"$(FIO_VERSION)"'
 
-.c.o: .depend FORCE
+.c.o: FORCE
 	$(QUIET_CC)$(CC) -o $@ $(CFLAGS) $(CPPFLAGS) -c $<
+	@$(CC) -MM $(CFLAGS) $(CPPFLAGS) $*.c > $*.d
+	@mv -f $*.d $*.d.tmp
+	@sed -e 's|.*:|$*.o:|' < $*.d.tmp > $*.d
+	@sed -e 's/.*://' -e 's/\\$$//' < $*.d.tmp | fmt -1 | \
+		sed -e 's/^ *//' -e 's/$$/:/' >> $*.d
+	@rm -f $*.d.tmp
 
 init.o: FIO-VERSION-FILE
 	$(QUIET_CC)$(CC) -o init.o $(CFLAGS) $(CPPFLAGS) -c init.c
@@ -237,13 +246,8 @@
 t/axmap: $(T_AXMAP_OBJS)
 	$(QUIET_CC)$(CC) $(LDFLAGS) $(CFLAGS) -o $@ $(T_AXMAP_OBJS) $(LIBS) $(LDFLAGS)
 
-.depend: $(SOURCE)
-	$(QUIET_DEP)$(CC) -MM $(CFLAGS) $(CPPFLAGS) $(SOURCE) 1> .depend
-
-$(PROGS): .depend
-
 clean: FORCE
-	-rm -f .depend $(GFIO_OBJS) $(OBJS) $(T_OBJS) $(PROGS) $(T_PROGS) core.* core gfio FIO-VERSION-FILE config-host.mak cscope.out
+	-rm -f .depend $(GFIO_OBJS) $(OBJS) $(T_OBJS) $(PROGS) $(T_PROGS) core.* core gfio FIO-VERSION-FILE config-host.mak cscope.out *.d
 
 cscope:
 	@cscope -b -R
@@ -254,9 +258,3 @@
 	$(INSTALL) -m 755 -d $(DESTDIR)$(mandir)/man1
 	$(INSTALL) -m 644 fio.1 $(DESTDIR)$(mandir)/man1
 	$(INSTALL) -m 644 fio_generate_plots.1 $(DESTDIR)$(mandir)/man1
-
-ifneq ($(wildcard .depend),)
-include .depend
-endif
-
-
diff --git a/README b/README
index c43b795..4c7b542 100644
--- a/README
+++ b/README
@@ -109,11 +109,22 @@
 Windows
 -------
 
-On Windows MinGW (http://www.mingw.org/) is required in order to
-build fio. To create an MSI installer package install WiX 3.6 from
-http://wix.sourceforge.net/releases/ and run dobuild.cmd from the
+On Windows Cygwin (http://www.cygwin.com/) is required in order to
+build fio. To create an MSI installer package install WiX 3.7 from
+http://wixtoolset.org and run dobuild.cmd from the
 os/windows directory.
 
+How to compile FIO on 64-bit Windows:
+
+ 1. Install Cygwin (http://www.cygwin.com/setup.exe). Install 'make' and all
+    packages starting with 'mingw64-i686' and 'mingw64-x86_64'.
+ 2. Download ftp://sourceware.org/pub/pthreads-win32/prebuilt-dll-2-9-1-release/dll/x64/pthreadGC2.dll
+    and copy to the fio source directory.
+ 3. Open the Cygwin Terminal.
+ 4. Go to the fio directory (source files).
+ 5. Run 'make clean'.
+ 6. Run 'make'.
+
 
 Command line
 ------------
diff --git a/arch/arch-x86-common.h b/arch/arch-x86-common.h
index d533d22..78fd40c 100644
--- a/arch/arch-x86-common.h
+++ b/arch/arch-x86-common.h
@@ -1,6 +1,8 @@
 #ifndef FIO_ARCH_X86_COMMON
 #define FIO_ARCH_X86_COMMON
 
+#include <string.h>
+
 static inline void do_cpuid(unsigned int *eax, unsigned int *ebx,
 			    unsigned int *ecx, unsigned int *edx)
 {
@@ -10,9 +12,20 @@
 		: "memory");
 }
 
+static inline void cpuid(unsigned int op,
+			 unsigned int *eax, unsigned int *ebx,
+			 unsigned int *ecx, unsigned int *edx)
+{
+	*eax = op;
+	*ecx = 0;
+	do_cpuid(eax, ebx, ecx, edx);
+}
+
 #define ARCH_HAVE_INIT
+
 extern int tsc_reliable;
-static inline int arch_init(char *envp[])
+
+static inline int arch_init_intel(unsigned int level)
 {
 	unsigned int eax, ebx, ecx = 0, edx;
 
@@ -29,7 +42,38 @@
 	 */
 	eax = 0x80000007;
 	do_cpuid(&eax, &ebx, &ecx, &edx);
-	tsc_reliable = edx & (1U << 8);
+	return edx & (1U << 8);
+}
+
+static inline int arch_init_amd(unsigned int level)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(0x80000000, &eax, &ebx, &ecx, &edx);
+	if (eax < 0x80000007)
+		return 0;
+
+	cpuid(0x80000007, &eax, &ebx, &ecx, &edx);
+	if (edx & (1 << 8))
+		return 1;
+
+	return 0;
+}
+
+static inline int arch_init(char *envp[])
+{
+	unsigned int level;
+	char str[12];
+
+	cpuid(0, &level, (unsigned int *) &str[0],
+			 (unsigned int *) &str[8],
+			 (unsigned int *) &str[4]);
+
+	if (!strcmp(str, "GenuineIntel"))
+		tsc_reliable = arch_init_intel(level);
+	else if (!strcmp(str, "AuthenticAMD"))
+		tsc_reliable = arch_init_amd(level);
+
 	return 0;
 }
 
diff --git a/backend.c b/backend.c
index 49d6bc7..592ea8b 100644
--- a/backend.c
+++ b/backend.c
@@ -813,7 +813,7 @@
 
 		i = td->cur_depth;
 		if (i) {
-			ret = io_u_queued_complete(td, i, NULL);
+			ret = io_u_queued_complete(td, i, bytes_done);
 			if (td->o.fill_device && td->error == ENOSPC)
 				td->error = 0;
 		}
@@ -1017,8 +1017,19 @@
 		return 1;
 	}
 
-	if (ddir_rw_sum(td->io_bytes) < td->o.size)
+	if (ddir_rw_sum(td->io_bytes) < td->o.size) {
+		uint64_t diff;
+
+		/*
+		 * If the difference is less than the minimum IO size, we
+		 * are done.
+		 */
+		diff = td->o.size - ddir_rw_sum(td->io_bytes);
+		if (diff < td->o.rw_min_bs)
+			return 0;
+
 		return 1;
+	}
 
 	return 0;
 }
diff --git a/cconv.c b/cconv.c
index 3a8572e..e2548a8 100644
--- a/cconv.c
+++ b/cconv.c
@@ -182,7 +182,6 @@
 	o->trim_batch = le32_to_cpu(top->trim_batch);
 	o->trim_zero = le32_to_cpu(top->trim_zero);
 	o->clat_percentiles = le32_to_cpu(top->clat_percentiles);
-	o->overwrite_plist = le32_to_cpu(top->overwrite_plist);
 	o->continue_on_error = le32_to_cpu(top->continue_on_error);
 	o->cgroup_weight = le32_to_cpu(top->cgroup_weight);
 	o->cgroup_nodelete = le32_to_cpu(top->cgroup_nodelete);
@@ -328,7 +327,6 @@
 	top->trim_batch = cpu_to_le32(o->trim_batch);
 	top->trim_zero = cpu_to_le32(o->trim_zero);
 	top->clat_percentiles = cpu_to_le32(o->clat_percentiles);
-	top->overwrite_plist = cpu_to_le32(o->overwrite_plist);
 	top->continue_on_error = cpu_to_le32(o->continue_on_error);
 	top->cgroup_weight = cpu_to_le32(o->cgroup_weight);
 	top->cgroup_nodelete = cpu_to_le32(o->cgroup_nodelete);
diff --git a/configure b/configure
index d226960..cab3da8 100755
--- a/configure
+++ b/configure
@@ -195,9 +195,6 @@
   fi
   output_sym "CONFIG_LITTLE_ENDIAN"
   output_sym "CONFIG_64BIT_LLP64"
-  output_sym "CONFIG_CLOCK_GETTIME"
-  output_sym "CONFIG_CLOCK_MONOTONIC"
-  output_sym "CONFIG_GETTIMEOFDAY"
   output_sym "CONFIG_FADVISE"
   output_sym "CONFIG_SOCKLEN_T"
   output_sym "CONFIG_POSIX_FALLOCATE"
@@ -206,6 +203,7 @@
   output_sym "CONFIG_RUSAGE_THREAD"
   output_sym "CONFIG_WINDOWSAIO"
   output_sym "CONFIG_FDATASYNC"
+  output_sym "CONFIG_CLOCK_MONOTONIC"
   output_sym "CONFIG_GETTIMEOFDAY"
   output_sym "CONFIG_CLOCK_GETTIME"
   output_sym "CONFIG_SCHED_IDLE"
diff --git a/eta.c b/eta.c
index 238a0af..6f897a4 100644
--- a/eta.c
+++ b/eta.c
@@ -285,7 +285,8 @@
 	static struct timeval rate_prev_time, disp_prev_time;
 
 	if (!force) {
-		if (output_format != FIO_OUTPUT_NORMAL)
+		if (output_format != FIO_OUTPUT_NORMAL &&
+		    f_out == stdout)
 			return 0;
 		if (temp_stall_ts || eta_print == FIO_ETA_NEVER)
 			return 0;
diff --git a/fio.1 b/fio.1
index c665591..586edcc 100644
--- a/fio.1
+++ b/fio.1
@@ -1168,6 +1168,9 @@
 .BI (net,netsplice)port \fR=\fPint
 The TCP or UDP port to bind to or connect to.
 .TP
+.BI (net,netsplice)nodelay \fR=\fPbool
+Set TCP_NODELAY on TCP connections.
+.TP
 .BI (net,netsplice)protocol \fR=\fPstr "\fR,\fP proto" \fR=\fPstr
 The network protocol to use. Accepted values are:
 .RS
diff --git a/fio.h b/fio.h
index 5bfa438..2d75213 100644
--- a/fio.h
+++ b/fio.h
@@ -361,7 +361,6 @@
 extern int nr_clients;
 extern int log_syslog;
 extern const char fio_version_string[];
-extern const fio_fp64_t def_percentile_list[FIO_IO_U_LIST_MAX_LEN];
 
 extern struct thread_data *threads;
 
diff --git a/init.c b/init.c
index f5a1693..25c8d78 100644
--- a/init.c
+++ b/init.c
@@ -66,26 +66,6 @@
 static char cmd_optstr[256];
 static int did_arg;
 
-const fio_fp64_t def_percentile_list[FIO_IO_U_LIST_MAX_LEN] = {
-	{ .u.f	=  1.00 },
-	{ .u.f	=  5.00 },
-	{ .u.f	= 10.00 },
-	{ .u.f	= 20.00 },
-	{ .u.f	= 30.00 },
-	{ .u.f	= 40.00 },
-	{ .u.f	= 50.00 },
-	{ .u.f	= 60.00 },
-	{ .u.f	= 70.00 },
-	{ .u.f	= 80.00 },
-	{ .u.f	= 90.00 },
-	{ .u.f	= 95.00 },
-	{ .u.f	= 99.00 },
-	{ .u.f	= 99.50 },
-	{ .u.f	= 99.90 },
-	{ .u.f	= 99.95 },
-	{ .u.f	= 99.99 },
-};
-
 #define FIO_CLIENT_FLAG		(1 << 16)
 
 /*
@@ -889,10 +869,7 @@
 	td->mutex = fio_mutex_init(FIO_MUTEX_LOCKED);
 
 	td->ts.clat_percentiles = td->o.clat_percentiles;
-	if (td->o.overwrite_plist)
-		memcpy(td->ts.percentile_list, td->o.percentile_list, sizeof(td->o.percentile_list));
-	else
-		memcpy(td->ts.percentile_list, def_percentile_list, sizeof(def_percentile_list));
+	memcpy(td->ts.percentile_list, td->o.percentile_list, sizeof(td->o.percentile_list));
 
 	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 		td->ts.clat_stat[i].min_val = ULONG_MAX;
diff --git a/options.c b/options.c
index c39a6b4..63293e0 100644
--- a/options.c
+++ b/options.c
@@ -2851,8 +2851,8 @@
 		.lname	= "Completion latency percentile list",
 		.type	= FIO_OPT_FLOAT_LIST,
 		.off1	= td_var_offset(percentile_list),
-		.off2   = td_var_offset(overwrite_plist),
 		.help	= "Specify a custom list of percentiles to report",
+		.def    = "1:5:10:20:30:40:50:60:70:80:90:95:99:99.5:99.9:99.95:99.99",
 		.maxlen	= FIO_IO_U_LIST_MAX_LEN,
 		.minfp	= 0.0,
 		.maxfp	= 100.0,
@@ -2947,6 +2947,8 @@
 		.off1	= td_var_offset(unified_rw_rep),
 		.help	= "Unify reporting across data direction",
 		.def	= "0",
+		.category = FIO_OPT_C_GENERAL,
+		.group	= FIO_OPT_G_INVALID,
 	},
 	{
 		.name	= "continue_on_error",
diff --git a/os/os.h b/os/os.h
index 1e924d3..4416ae4 100644
--- a/os/os.h
+++ b/os/os.h
@@ -127,7 +127,11 @@
 #endif
 
 #ifndef FIO_PREFERRED_CLOCK_SOURCE
+#ifdef CONFIG_CLOCK_GETTIME
 #define FIO_PREFERRED_CLOCK_SOURCE	CS_CGETTIME
+#else
+#define FIO_PREFERRED_CLOCK_SOURCE	CS_GTOD
+#endif
 #endif
 
 #ifndef FIO_MAX_JOBS
diff --git a/parse.c b/parse.c
index a0b18e2..7652a4d 100644
--- a/parse.c
+++ b/parse.c
@@ -10,11 +10,13 @@
 #include <limits.h>
 #include <stdlib.h>
 #include <math.h>
+#include <float.h>
 
 #include "parse.h"
 #include "debug.h"
 #include "options.h"
 #include "minmax.h"
+#include "lib/ieee754.h"
 
 static struct fio_option *__fio_options;
 
@@ -47,12 +49,12 @@
 static void show_option_range(struct fio_option *o,
 				int (*logger)(const char *format, ...))
 {
-	if (o->type == FIO_OPT_FLOAT_LIST) {
-		if (isnan(o->minfp) && isnan(o->maxfp))
+	if (o->type == FIO_OPT_FLOAT_LIST){
+		if (o->minfp == DBL_MIN && o->maxfp == DBL_MAX)
 			return;
 
 		logger("%20s: min=%f", "range", o->minfp);
-		if (!isnan(o->maxfp))
+		if (o->maxfp != DBL_MAX)
 			logger(", max=%f", o->maxfp);
 		logger("\n");
 	} else {
@@ -361,7 +363,7 @@
 			   int first, int more, int curr)
 {
 	int il, *ilp;
-	double *flp;
+	fio_fp64_t *flp;
 	long long ull, *ullp;
 	long ul1, ul2;
 	double uf;
@@ -499,12 +501,6 @@
 		break;
 	}
 	case FIO_OPT_FLOAT_LIST: {
-
-		if (first) {
-			ul2 = 1;
-			ilp = td_var(data, o->off2);
-			*ilp = ul2;
-		}
 		if (curr >= o->maxlen) {
 			log_err("the list exceeding max length %d\n",
 					o->maxlen);
@@ -514,19 +510,19 @@
 			log_err("not a floating point value: %s\n", ptr);
 			return 1;
 		}
-		if (!isnan(o->maxfp) && uf > o->maxfp) {
+		if (uf > o->maxfp) {
 			log_err("value out of range: %f"
 				" (range max: %f)\n", uf, o->maxfp);
 			return 1;
 		}
-		if (!isnan(o->minfp) && uf < o->minfp) {
+		if (uf < o->minfp) {
 			log_err("value out of range: %f"
 				" (range min: %f)\n", uf, o->minfp);
 			return 1;
 		}
 
 		flp = td_var(data, o->off1);
-		flp[curr] = uf;
+		flp[curr].u.f = uf;
 
 		break;
 	}
@@ -1094,11 +1090,8 @@
 			o->maxval = UINT_MAX;
 	}
 	if (o->type == FIO_OPT_FLOAT_LIST) {
-#ifndef NAN
-#define NAN __builtin_nanf("")
-#endif
-		o->minfp = NAN;
-		o->maxfp = NAN;
+		o->minfp = DBL_MIN;
+		o->maxfp = DBL_MAX;
 	}
 	if (o->type == FIO_OPT_STR_SET && o->def) {
 		log_err("Option %s: string set option with"
diff --git a/stat.c b/stat.c
index 26b3133..b16c55c 100644
--- a/stat.c
+++ b/stat.c
@@ -1169,10 +1169,7 @@
 		ts = &threadstats[j];
 
 		ts->clat_percentiles = td->o.clat_percentiles;
-		if (td->o.overwrite_plist)
-			memcpy(ts->percentile_list, td->o.percentile_list, sizeof(td->o.percentile_list));
-		else
-			memcpy(ts->percentile_list, def_percentile_list, sizeof(def_percentile_list));
+		memcpy(ts->percentile_list, td->o.percentile_list, sizeof(td->o.percentile_list));
 
 		idx++;
 		ts->members++;
diff --git a/thread_options.h b/thread_options.h
index ae83f08..11e7af7 100644
--- a/thread_options.h
+++ b/thread_options.h
@@ -197,7 +197,6 @@
 	unsigned int trim_zero;
 	unsigned long long trim_backlog;
 	unsigned int clat_percentiles;
-	unsigned int overwrite_plist;
 	fio_fp64_t percentile_list[FIO_IO_U_LIST_MAX_LEN];
 
 	char *read_iolog_file;
@@ -397,7 +396,6 @@
 	uint32_t trim_zero;
 	uint64_t trim_backlog;
 	uint32_t clat_percentiles;
-	uint32_t overwrite_plist;
 	fio_fp64_t percentile_list[FIO_IO_U_LIST_MAX_LEN];
 
 	uint8_t read_iolog_file[FIO_TOP_STR_MAX];