commit 0f03cc378e6ce48f17a20e409f93bfc11345a6ed
parent d3cd95e055f50df97979bcca7bb26a3edd4b82ee
Author: Mattias Andrée <maandree@kth.se>
Date:   Tue, 16 May 2017 20:14:22 +0200
Fix errors and warnings and make the code cleaner
Signed-off-by: Mattias Andrée <maandree@kth.se>
Diffstat:
51 files changed, 469 insertions(+), 589 deletions(-)
diff --git a/Makefile b/Makefile
@@ -41,7 +41,9 @@ BIN =\
 	blind-to-video\
 	blind-translate\
 	blind-transpose\
-	blind-write-head
+	blind-write-head\
+	blind-kernel\
+	blind-temporal-mean
 
 # TODO Not tested yet (and doesn't have any manpages):
 #    blind-kernel
@@ -59,6 +61,7 @@ SRC = $(BIN:=.c) util.c stream.c
 
 HDR =\
 	arg.h\
+	common.h\
 	stream.h\
 	util.h\
 	util/to.h\
@@ -71,7 +74,8 @@ HDR =\
 	util/efflush.h\
 	util/efunc.h\
 	util/eprintf.h\
-	util/fshut.h
+	util/fshut.h\
+	video-math.h
 
 MISCFILES = Makefile config.mk LICENSE README TODO
 
diff --git a/config.mk b/config.mk
@@ -6,6 +6,6 @@ PREFIX = /usr/local
 MANPREFIX = $(PREFIX)/share/man
 
 # You may want to remove -DHAVE_PRCTL and -DHAVE_EPOLL from CPPFLAGS if you are not using Linux.
-CFLAGS   = -std=c99 -Wall -pedantic -O2
+CFLAGS   = -std=c11 -Wall -pedantic -O2
 CPPFLAGS = -D_DEFAULT_SOURCE -D_BSD_SOURCE -D_XOPEN_SOURCE=700 -D_FILE_OFFSET_BITS=64 -DHAVE_PRCTL -DHAVE_EPOLL
 LDFLAGS  = -lm -s
diff --git a/src/blind-arithm.c b/src/blind-arithm.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <math.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-axyz] operation right-hand-stream")
 
@@ -15,16 +11,16 @@ static int skip_z = 0;
 /* Because the syntax for a function returning a function pointer is disgusting. */
 typedef void (*process_func)(struct stream *left, struct stream *right, size_t n);
 
-#define LIST_OPERATORS(PIXFMT, TYPE, SUFFIX)\
-	X(add, *lh += rh,                                 PIXFMT, TYPE)\
-	X(sub, *lh -= rh,                                 PIXFMT, TYPE)\
-	X(mul, *lh *= rh,                                 PIXFMT, TYPE)\
-	X(div, *lh /= rh,                                 PIXFMT, TYPE)\
-	X(exp, *lh = pow##SUFFIX(*lh, rh),                PIXFMT, TYPE)\
-	X(log, *lh = log##SUFFIX(*lh) / log##SUFFIX(rh),  PIXFMT, TYPE)\
-	X(min, *lh = MIN(*lh, rh),                        PIXFMT, TYPE)\
-	X(max, *lh = MAX(*lh, rh),                        PIXFMT, TYPE)\
-	X(abs, *lh = fabs##SUFFIX(*lh - rh) + rh,         PIXFMT, TYPE)
+#define LIST_OPERATORS(PIXFMT, TYPE)\
+	X(add, *lh += rh,                  PIXFMT, TYPE)\
+	X(sub, *lh -= rh,                  PIXFMT, TYPE)\
+	X(mul, *lh *= rh,                  PIXFMT, TYPE)\
+	X(div, *lh /= rh,                  PIXFMT, TYPE)\
+	X(exp, *lh = pow(*lh, rh),         PIXFMT, TYPE)\
+	X(log, *lh = log2(*lh) / log2(rh), PIXFMT, TYPE)\
+	X(min, *lh = MIN(*lh, rh),         PIXFMT, TYPE)\
+	X(max, *lh = MAX(*lh, rh),         PIXFMT, TYPE)\
+	X(abs, *lh = abs(*lh - rh) + rh,   PIXFMT, TYPE)
 
 #define C(CH, CHI, ALGO, TYPE)\
 	(!skip_##CH ? ((lh = ((TYPE *)(left->buf + i)) + (CHI),\
@@ -49,8 +45,8 @@ typedef void (*process_func)(struct stream *left, struct stream *right, size_t n
 			C(a, 3, ALGO, TYPE);\
 		}\
 	}
-LIST_OPERATORS(xyza, double,)
-LIST_OPERATORS(xyzaf, float, f)
+LIST_OPERATORS(xyza, double)
+LIST_OPERATORS(xyzaf, float)
 #undef X
 
 #if defined(__GNUC__) && !defined(__clang__)
@@ -62,7 +58,7 @@ get_process_xyza(const char *operation)
 {
 #define X(NAME, ALGO, PIXFMT, TYPE)\
 	if (!strcmp(operation, #NAME)) return process_##PIXFMT##_##NAME;
-	LIST_OPERATORS(xyza, double,)
+	LIST_OPERATORS(xyza, double)
 #undef X
 	eprintf("algorithm not recognised: %s\n", operation);
 	return NULL;
@@ -73,7 +69,7 @@ get_process_xyzaf(const char *operation)
 {
 #define X(NAME, ALGO, PIXFMT, TYPE)\
 	if (!strcmp(operation, #NAME)) return process_##PIXFMT##_##NAME;
-	LIST_OPERATORS(xyzaf, float, f)
+	LIST_OPERATORS(xyzaf, float)
 #undef X
 	eprintf("algorithm not recognised: %s\n", operation);
 	return NULL;
diff --git a/src/blind-colour-ciexyz.c b/src/blind-colour-ciexyz.c
@@ -1,5 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "util.h"
+#include "common.h"
 
 USAGE("(X Y Z | Y)")
 
diff --git a/src/blind-colour-srgb.c b/src/blind-colour-srgb.c
@@ -1,5 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "util.h"
+#include "common.h"
 
 USAGE("[-d depth] [-l] red green blue")
 
@@ -26,9 +26,9 @@ main(int argc, char *argv[])
 
 	max   = 1ULL << (depth - 1);
 	max  |= max - 1;
-	red   = etolf_arg("the red value",   argv[0]) / max;
-	green = etolf_arg("the green value", argv[1]) / max;
-	blue  = etolf_arg("the blue value",  argv[2]) / max;
+	red   = etolf_arg("the red value",   argv[0]) / (double)max;
+	green = etolf_arg("the green value", argv[1]) / (double)max;
+	blue  = etolf_arg("the blue value",  argv[2]) / (double)max;
 	if (!linear) {
 		red   = srgb_decode(red);
 		green = srgb_decode(green);
diff --git a/src/blind-compress.c b/src/blind-compress.c
@@ -1,6 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
 USAGE("")
 
diff --git a/src/blind-concat.c b/src/blind-concat.c
@@ -1,12 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#if defined(HAVE_EPOLL)
-# include <sys/epoll.h>
-#endif
-#include <sys/mman.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-o output-file [-j jobs]] first-stream ... last-stream")
 
@@ -97,9 +90,9 @@ concat_to_file_parallel(int argc, char *argv[], char *output_file, size_t jobs)
 #else
 	struct epoll_event *events;
 	struct stream *streams;
-	size_t *ptrs;
+	off_t *ptrs, ptr;
 	char head[STREAM_HEAD_MAX];
-	size_t ptr, frames = 0, next = 0, j;
+	size_t frames = 0, next = 0, j;
 	ssize_t headlen;
 	int fd, i, n, pollfd;
 
@@ -123,14 +116,14 @@ concat_to_file_parallel(int argc, char *argv[], char *output_file, size_t jobs)
 	SPRINTF_HEAD_ZN(head, frames, streams->width, streams->height, streams->pixfmt, &headlen);
 
 	echeck_dimensions(streams, WIDTH | HEIGHT, NULL);
-	ptr = (size_t)headlen;
+	ptr = (off_t)headlen;
 	for (i = 0; i < argc; i++) {
 		ptrs[i] = ptr;
-		ptr += streams->frames * streams->frame_size;
+		ptr += (off_t)streams->frames * (off_t)streams->frame_size;
 	}
 	if (ftruncate(fd, (off_t)ptr))
 		eprintf("ftruncate %s:", output_file);
-        fadvise_random(fd, (size_t)headlen, 0);
+        fadvise_random(fd, (off_t)headlen, 0);
 
 	pollfd = epoll_create1(0);
 	if (pollfd == -1)
@@ -139,7 +132,7 @@ concat_to_file_parallel(int argc, char *argv[], char *output_file, size_t jobs)
 	epwriteall(fd, head, (size_t)headlen, 0, output_file);
 	for (i = 0; i < argc; i++) {
 		epwriteall(fd, streams[i].buf, streams[i].ptr, ptrs[i], output_file);
-		ptrs[i] += streams[i].ptr;
+		ptrs[i] += (off_t)(streams[i].ptr);
 		streams[i].ptr = 0;
 	}
 
@@ -155,14 +148,14 @@ concat_to_file_parallel(int argc, char *argv[], char *output_file, size_t jobs)
 	jobs = j;
 
 	while (jobs) {
-		n = epoll_wait(pollfd, events, jobs, -1);
+		n = epoll_wait(pollfd, events, (int)jobs, -1);
 		if (n < 0)
 			eprintf("epoll_wait:");
 		for (i = 0; i < n; i++) {
 			j = events[i].data.u64;
 			if (streams[j].ptr || eread_stream(streams + j, SIZE_MAX)) {
 				epwriteall(fd, streams[j].buf, streams[j].ptr, ptrs[j], output_file);
-				ptrs[j] += streams[j].ptr;
+				ptrs[j] += (off_t)(streams[j].ptr);
 				streams[j].ptr = 0;
 				continue;
 			}
diff --git a/src/blind-convert.c b/src/blind-convert.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <alloca.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("pixel-format ...")
 
@@ -20,10 +16,10 @@ static void (*outconv)(double *xyzas, size_t n);
 			interm = buf;\
 			n = stream->ptr / stream->pixel_size;\
 			for (m = n; m--; in += 4, interm += 4) { \
-				interm[0] = (TYPE)(in[0]);\
-				interm[1] = (TYPE)(in[1]);\
-				interm[2] = (TYPE)(in[2]);\
-				interm[3] = (TYPE)(in[3]);\
+				interm[0] = (double)(in[0]);\
+				interm[1] = (double)(in[1]);\
+				interm[2] = (double)(in[2]);\
+				interm[3] = (double)(in[3]);\
 			}\
 			outconv(buf, n);\
 			n *= stream->pixel_size;\
diff --git a/src/blind-crop.c b/src/blind-crop.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-s | -S | -t] width height left top")
 
diff --git a/src/blind-cut.c b/src/blind-cut.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("start-point (end-point | 'end') file")
 
diff --git a/src/blind-decompress.c b/src/blind-decompress.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("")
 
diff --git a/src/blind-dissolve.c b/src/blind-dissolve.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-r]")
 
@@ -55,7 +52,7 @@ main(int argc, char *argv[])
 
 	fprint_stream_head(stdout, &stream);
 	efflush(stdout, "<stdout>");
-	fm_double = fm = stream.frames - 1;
+	fm_double = (double)(fm = stream.frames - 1);
 	fm_float = (float)fm_double;
 	process_each_frame_segmented(&stream, STDOUT_FILENO, "<stdout>", process);
 	return 0;
diff --git a/src/blind-extend.c b/src/blind-extend.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-l left] [-r right] [-a above] [-b below] [-t]")
 
diff --git a/src/blind-flip.c b/src/blind-flip.c
@@ -1,6 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
 USAGE("")
 
diff --git a/src/blind-flop.c b/src/blind-flop.c
@@ -1,6 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
 USAGE("")
 
diff --git a/src/blind-from-image.c b/src/blind-from-image.c
@@ -1,10 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <arpa/inet.h>
-#include <inttypes.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-h] [-f | -p]")
 
@@ -24,7 +19,7 @@ get_value_u8(char** bufp)
 {
 	uint8_t value = *(uint8_t *)(*bufp);
 	*bufp += 1;
-	return value / value_max;
+	return (double)value / (double)value_max;
 }
 
 static double
@@ -32,7 +27,7 @@ get_value_u16(char** bufp)
 {
 	uint16_t value = ntohs(*(uint16_t *)(*bufp));
 	*bufp += 2;
-	return value / value_max;
+	return (double)value / (double)value_max;
 }
 
 static double
@@ -40,7 +35,7 @@ get_value_u32(char** bufp)
 {
 	uint32_t value = ntohl(*(uint32_t *)(*bufp));
 	*bufp += 4;
-	return value / value_max;
+	return (double)value / (double)value_max;
 }
 
 static double
@@ -56,7 +51,7 @@ get_value_u64(char** bufp)
 	value |= (uint64_t)(buf[6]) <<  8;
 	value |= (uint64_t)(buf[7]);
 	*bufp += 8;
-	return value / value_max;
+	return (double)value / (double)value_max;
 }
 
 static void
@@ -96,13 +91,13 @@ static size_t
 pam_head(int fd, const char *fname)
 {
 	size_t ptr;
-	ssize_t r;
+	size_t r;
 	char *p;
 	unsigned long long int maxval = UINT8_MAX;
 	for (ptr = 0;;) {
 		if (!(r = eread(fd, buf + ptr, sizeof(buf) - 1, fname)))
 			eprintf("%s\n", conv_fail_msg);
-		ptr += (size_t)r;
+		ptr += r;
 		for (;;) {
 			p = memchr(buf, '\n', ptr);
 			if (!p) {
@@ -163,8 +158,8 @@ header_done:
 		pixel_size = sizeof(uint64_t);
 		get_value = get_value_u64;
 	}
-	value_max = maxval;
-	pixel_size *= (with_colour ? 3 : 1) + with_alpha;
+	value_max = (double)maxval;
+	pixel_size *= (size_t)((with_colour ? 3 : 1) + with_alpha);
 	convert = from_srgb;
 	return ptr;
 }
diff --git a/src/blind-from-portable.c b/src/blind-from-portable.c
@@ -1,17 +1,12 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <alloca.h>
-#include <math.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-s]")
 
 #define USING_BINARY32 0
 #define USING_BINARY64 0
 
-#define CONV(ITYPE, SITYPE, OTYPE, EXPONENT, HA2EXPONENT, FRACTION, SUFFIX)\
+#define CONV(ITYPE, SITYPE, OTYPE, EXPONENT, HA2EXPONENT, FRACTION)\
 	do {\
 		static int cache_i = 0;\
 		static ITYPE cache_in[] = {0, 0, 0, 0};\
@@ -24,6 +19,7 @@ USAGE("[-s]")
 			cache_i &= 3;\
 			return ret;\
 		}\
+		cache_in[cache_i] = portable;\
 		signb = portable >> (EXPONENT + FRACTION);\
 		exponent = (portable >> FRACTION) ^ (signb << EXPONENT);\
 		fraction = portable & (((ITYPE)1 << FRACTION) - 1);\
@@ -33,8 +29,8 @@ USAGE("[-s]")
 			} else {\
 				sexponent = 1 - HA2EXPONENT - FRACTION;\
 				dexponent = (OTYPE)sexponent;\
-				ret = (ITYPE)fraction;\
-				ret *= pow##SUFFIX((OTYPE)2.0, dexponent);\
+				ret = (OTYPE)fraction;\
+				ret *= pow((OTYPE)2.0, dexponent);\
 			}\
 		} else if (exponent + 1 == (ITYPE)1 << EXPONENT) {\
 			ret = (OTYPE)(fraction ? NAN : INFINITY);\
@@ -43,8 +39,8 @@ USAGE("[-s]")
 			sexponent = (SITYPE)exponent;\
 			sexponent -= HA2EXPONENT + FRACTION;\
 			dexponent = (OTYPE)sexponent;\
-			ret = (ITYPE)fraction;\
-			ret *= pow##SUFFIX((OTYPE)2.0, dexponent);\
+			ret = (OTYPE)fraction;\
+			ret *= pow((OTYPE)2.0, dexponent);\
 		}\
 		ret = signb ? -ret : ret;\
 		cache_out[cache_i++] = ret;\
@@ -78,8 +74,8 @@ USAGE("[-s]")
 			eprintf("%s: incomplete frame\n", stream->file);\
 	} while (0)
 
-static double conv_double(uint64_t portable) {CONV(uint64_t, int64_t, double, 11, 1023, 52,);}
-static float  conv_float (uint32_t portable) {CONV(uint32_t, int32_t, float, 8, 127, 23, f);}
+static double conv_double(uint64_t portable) {CONV(uint64_t, int64_t, double, 11, 1023, 52);}
+static float  conv_float (uint32_t portable) {CONV(uint32_t, int32_t, float, 8, 127, 23);}
 
 static void process_xyza (struct stream *stream, int strict) {PROCESS(uint64_t, double, 64);}
 static void process_xyzaf(struct stream *stream, int strict) {PROCESS(uint32_t, float, 32);}
diff --git a/src/blind-from-text.c b/src/blind-from-text.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("")
 
diff --git a/src/blind-from-video.c b/src/blind-from-video.c
@@ -1,10 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <sys/mman.h>
-#include <sys/stat.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-F pixel-format] [-r frame-rate] [-w width -h height] [-dL] input-file output-file")
 
@@ -76,21 +71,23 @@ get_metadata(char *file, size_t *width, size_t *height)
 		exit(1);
 }
 
-#define CONVERT_SEGMENT(TYPE, SUFFIX)\
+#define CONVERT_SEGMENT(TYPE)\
 	do {\
 		typedef TYPE pixel_t[4];\
 		size_t i, ptr;\
 		TYPE y, u, v, max = (TYPE)UINT16_MAX;\
 		TYPE r, g, b;\
 		pixel_t pixels[1024];\
+		uint16_t *pix;\
 		if (draft) {\
 			for (ptr = i = 0; ptr < n; ptr += 8) {\
+				pix = (uint16_t *)(buf + ptr);\
 				pixels[i][3] = 1;\
-				y = (long int)(le16toh(((uint16_t *)(buf + ptr))[1])) -  16L * 256L;\
-				u = (long int)(le16toh(((uint16_t *)(buf + ptr))[2])) - 128L * 256L;\
-				v = (long int)(le16toh(((uint16_t *)(buf + ptr))[3])) - 128L * 256L;\
-				scaled_yuv_to_ciexyz##SUFFIX(y, u, v, pixels[i] + 0,\
-							     pixels[i] + 1, pixels[i] + 2);\
+				y = (TYPE)((long int)(le16toh(pix[1])) -  16L * 256L);\
+				u = (TYPE)((long int)(le16toh(pix[2])) - 128L * 256L);\
+				v = (TYPE)((long int)(le16toh(pix[3])) - 128L * 256L);\
+				scaled_yuv_to_ciexyz(y, u, v, pixels[i] + 0,\
+						     pixels[i] + 1, pixels[i] + 2);\
 				if (++i == 1024) {\
 					i = 0;\
 					ewriteall(fd, pixels, sizeof(pixels), file);\
@@ -98,15 +95,16 @@ get_metadata(char *file, size_t *width, size_t *height)
 			}\
 		} else {\
 			for (ptr = i = 0; ptr < n; ptr += 8) {\
-				pixels[i][3] = le16toh(((uint16_t *)(buf + ptr))[0]) / max;\
-				y = ((long int)le16toh(((uint16_t *)(buf + ptr))[1]) -  16L * 256L) / max;\
-				u = ((long int)le16toh(((uint16_t *)(buf + ptr))[2]) - 128L * 256L) / max;\
-				v = ((long int)le16toh(((uint16_t *)(buf + ptr))[3]) - 128L * 256L) / max;\
-				yuv_to_srgb##SUFFIX(y, u, v, &r, &g, &b);\
-				r = srgb_decode##SUFFIX(r);\
-				g = srgb_decode##SUFFIX(g);\
-				b = srgb_decode##SUFFIX(b);\
-				srgb_to_ciexyz##SUFFIX(r, g, b, pixels[i] + 0, pixels[i] + 1, pixels[i] + 2);\
+				pix = (uint16_t *)(buf + ptr);\
+				pixels[i][3] = le16toh(pix[0]) / max;\
+				y = (TYPE)((long int)le16toh(pix[1]) -  16L * 256L) / max;\
+				u = (TYPE)((long int)le16toh(pix[2]) - 128L * 256L) / max;\
+				v = (TYPE)((long int)le16toh(pix[3]) - 128L * 256L) / max;\
+				yuv_to_srgb(y, u, v, &r, &g, &b);\
+				r = srgb_decode(r);\
+				g = srgb_decode(g);\
+				b = srgb_decode(b);\
+				srgb_to_ciexyz(r, g, b, pixels[i] + 0, pixels[i] + 1, pixels[i] + 2);\
 				if (++i == 1024) {\
 					i = 0;\
 					ewriteall(fd, pixels, sizeof(pixels), file);\
@@ -117,17 +115,8 @@ get_metadata(char *file, size_t *width, size_t *height)
 			ewriteall(fd, pixels, i * sizeof(*pixels), file);\
 	} while (0)
 
-static void
-convert_segment_xyza(char *buf, size_t n, int fd, const char *file)
-{
-	CONVERT_SEGMENT(double,);
-}
-
-static void
-convert_segment_xyzaf(char *buf, size_t n, int fd, const char *file)
-{
-	CONVERT_SEGMENT(float, _f);
-}
+static void convert_segment_xyza (char *buf, size_t n, int fd, const char *file) {CONVERT_SEGMENT(double);}
+static void convert_segment_xyzaf(char *buf, size_t n, int fd, const char *file) {CONVERT_SEGMENT(float);}
 
 static void
 convert(const char *infile, int outfd, const char *outfile, size_t width, size_t height, const char *frame_rate)
@@ -136,7 +125,6 @@ convert(const char *infile, int outfd, const char *outfile, size_t width, size_t
 	const char *cmd[13];
 	int status, pipe_rw[2];
 	size_t i = 0, n, ptr;
-	ssize_t r;
 	pid_t pid;
 
 	cmd[i++] = "ffmpeg";
@@ -166,9 +154,9 @@ convert(const char *infile, int outfd, const char *outfile, size_t width, size_t
 	close(pipe_rw[1]);
 
 	for (ptr = 0;;) {
-		if (!(r = eread(pipe_rw[0], buf + ptr, sizeof(buf) - ptr, "<subprocess>")))
+		if (!(n = eread(pipe_rw[0], buf + ptr, sizeof(buf) - ptr, "<subprocess>")))
 			break;
-		ptr += (size_t)r;
+		ptr += n;
 		n = ptr - (ptr % 8);
 		convert_segment(buf, n, outfd, outfile);
 		memmove(buf, buf + n, ptr -= n);
@@ -253,7 +241,7 @@ main(int argc, char *argv[])
 	}
 
 	if (skip_length) {
-		SPRINTF_HEAD_ZN(head, frames, width, height, pixfmt, &headlen);
+		SPRINTF_HEAD_ZN(head, 0, width, height, pixfmt, &headlen);
 		ewriteall(outfd, head, (size_t)headlen, outfile);
 	}
 
diff --git a/src/blind-gauss-blur.c b/src/blind-gauss-blur.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <math.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-j jobs] [-s spread | -s 'auto'] [-achvy] sd-stream")
 
@@ -62,13 +58,13 @@ static size_t spread = 0;
 	if (auto_spread && spread < 1)\
 		spread = 1;
 
-#define BLUR_PIXEL(START, LOOP, DISTANCE, SUFFIX)\
+#define BLUR_PIXEL(TYPE, START, LOOP, DISTANCE)\
 	if (k[0] == k[1] && k[1] == k[2]) {\
 		START;\
 		for (LOOP) {\
-			d = (DISTANCE);\
+			d = (TYPE)(DISTANCE);\
 			d *= d;\
-			m = c[0] * exp##SUFFIX(d * k[0]);\
+			m = c[0] * exp(d * k[0]);\
 			img[i2][0] += clr[i1][0] * m;\
 			img[i2][1] += clr[i1][1] * m;\
 			img[i2][2] += clr[i1][2] * m;\
@@ -87,11 +83,11 @@ static size_t spread = 0;
 				continue;\
 			START;\
 			for (LOOP) {\
-				d = (DISTANCE);\
+				d = (TYPE)(DISTANCE);\
 				d *= d;\
-				m = c[i] * exp##SUFFIX(d * k[i]);\
+				m = c[i] * exp(d * k[i]);\
 				img[i2][i] += clr[i1][i] * m;\
-				img[i2][3] += clr[i1][3] * m / blurred;\
+				img[i2][3] += clr[i1][3] * m / (TYPE)blurred;\
 			}\
 		}\
 	}
@@ -104,7 +100,7 @@ static size_t spread = 0;
 	img[i1][2] = clr[i1][2];\
 	img[i1][3] = clr[i1][3];
 
-#define BLUR(TYPE, DIR, SETSTART, SETEND, START, LOOP, DISTANCE, SUFFIX)\
+#define BLUR(TYPE, DIR, SETSTART, SETEND, START, LOOP, DISTANCE)\
 	do {\
 		memset(img, 0, colour->frame_size);\
 		start = 0, end = colour->height;\
@@ -117,14 +113,14 @@ static size_t spread = 0;
 					SETSTART;\
 					SETEND;\
 				}\
-				BLUR_PIXEL(START, LOOP, DISTANCE, SUFFIX);\
+				BLUR_PIXEL(TYPE, START, LOOP, DISTANCE);\
 				BLUR_PIXEL_EPILOGUE(DIR);\
 			}\
 		}\
 		ejoin_jobs(is_master, children);\
 	} while (0)
 
-#define PROCESS(TYPE, SUFFIX)\
+#define PROCESS(TYPE)\
 	do {\
 		typedef TYPE pixel_t[4];\
 		\
@@ -226,8 +222,7 @@ static size_t spread = 0;
 			     x2end = spread + 1 > colour->width - x1 ? colour->width : x1 + spread + 1,\
 			     i2 = y1 * colour->width + x2start,\
 			     x2 = x2start; x2 < x2end; (x2++, i2++),\
-			     (ssize_t)x1 - (ssize_t)x2,\
-			     SUFFIX);\
+			     (ssize_t)x1 - (ssize_t)x2);\
 		if (horizontal && vertical)\
 			memcpy(clr, img, colour->frame_size);\
 		if (vertical)\
@@ -236,8 +231,7 @@ static size_t spread = 0;
 			     y2end = spread + 1 > colour->height - y1 ? colour->height : y1 + spread + 1,\
 			     i2 = y2start * colour->width + x1,\
 			     y2 = y2start; y2 < y2end; (y2++, i2 += colour->width),\
-			     (ssize_t)y1 - (ssize_t)y2,\
-			     SUFFIX);\
+			     (ssize_t)y1 - (ssize_t)y2);\
 		\
 		start = 0, end = colour->height;\
 		is_master = efork_jobs(&start, &end, jobs, &children);\
@@ -257,7 +251,7 @@ static size_t spread = 0;
 		i1 = start * colour->width;\
 		for (y1 = start; y1 < end; y1++) {\
 			for (x1 = 0; x1 < colour->width; x1++, i1++) {\
-				if (!img[i1][3])\
+				if (img[i1][3] != 0)\
 					continue;\
 				img[i1][0] /= img[i1][3];\
 				img[i1][1] /= img[i1][3];\
@@ -282,14 +276,14 @@ static void
 process_xyza(char *restrict output, char *restrict cbuf, char *restrict sbuf,
 	     struct stream *colour, struct stream *sigma)
 {
-	PROCESS(double,);
+	PROCESS(double);
 }
 
 static void
 process_xyzaf(char *restrict output, char *restrict cbuf, char *restrict sbuf,
 	     struct stream *colour, struct stream *sigma)
 {
-	PROCESS(float, f);
+	PROCESS(float);
 }
 
 int
diff --git a/src/blind-invert-luma.c b/src/blind-invert-luma.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-iw] mask-stream")
 
diff --git a/src/blind-kernel.c b/src/blind-kernel.c
@@ -1,13 +1,14 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
-#include <math.h>
-#include <string.h>
+#if defined(__GNUC__) && !defined(__clang__)
+# pragma GCC diagnostic push
+# pragma GCC diagnostic ignored "-Wunsafe-loop-optimizations"
+#endif
 
 USAGE("[-xyza] kernel [parameter] ...")
 
-#define SUBUSAGE(FORMAT)          "Usage: %s [-xyza] " FORMAT, argv0
+#define SUBUSAGE(FORMAT)          "usage: %s [-xyza] " FORMAT, argv0
 #define STRCASEEQ3(A, B1, B2, B3) (!strcasecmp(A, B1) || !strcasecmp(A, B2) || !strcasecmp(A, B3))
 
 #define LIST_KERNELS\
@@ -110,6 +111,7 @@ kernel_sharpen(int argc, char *argv[], size_t *rows, size_t *cols, double **free
 	*rows = *cols = 3;
 
 #define argv0 arg
+	(void) arg;
 	argc++, argv--;
 	ARGBEGIN {
 	case 'i':
@@ -126,7 +128,6 @@ kernel_sharpen(int argc, char *argv[], size_t *rows, size_t *cols, double **free
 usage:
 	eprintf(SUBUSAGE("'sharpen' [-i]"));
 	return NULL;
-	(void) arg;
 }
 
 static const double *
@@ -156,7 +157,7 @@ kernel_gaussian(int argc, char *argv[], size_t *rows, size_t *cols, double **fre
 	if (argc != 1)
 		goto usage;
 
-	sigma = etof_arg("standard-deviation", argv[0]);
+	sigma = etolf_arg("standard-deviation", argv[0]);
 
 	if (!spread)
 		spread = (size_t)(sigma * 3.0 + 0.5);
@@ -171,7 +172,7 @@ kernel_gaussian(int argc, char *argv[], size_t *rows, size_t *cols, double **fre
 	k = 1.0 / -k;
 
 	for (x = 0; x <= spread; x++) {
-		value = spread - x;
+		value = (double)(spread - x);
 		value *= value * k;
 		value = exp(value) * c;
 		for (y = 0; y < *rows; y++) {
@@ -181,7 +182,7 @@ kernel_gaussian(int argc, char *argv[], size_t *rows, size_t *cols, double **fre
 	}
 
 	for (y = 0; y <= spread; y++) {
-		value = spread - y;
+		value = (double)(spread - y);
 		value *= value * k;
 		value = exp(value) * c;
 		for (x = 0; x < *cols; x++) {
diff --git a/src/blind-make-kernel.c b/src/blind-make-kernel.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <ctype.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-d denominator] ... [-nxyza] [-- value ...] ...")
 
@@ -58,7 +54,7 @@ static double *
 read_matrix_stdin(size_t *rows, size_t *cols)
 {
 	char *line = NULL, *p, *q;
-	size_t size = 0, col;
+	size_t size = 0, col = 0;
 	double *kernel = NULL;
 	ssize_t len;
 	*rows = *cols = 0;
diff --git a/src/blind-next-frame.c b/src/blind-next-frame.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-f frames] width height pixel-format ...")
 
diff --git a/src/blind-read-head.c b/src/blind-read-head.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <ctype.h>
+#include "common.h"
 
 USAGE("")
 
@@ -50,5 +47,4 @@ main(int argc, char *argv[])
 	return 0;
 bad_format:
 	eprintf("<stdin>: file format not supported\n");
-	return 0;
 }
diff --git a/src/blind-repeat.c b/src/blind-repeat.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("(count | 'inf') file")
 
@@ -14,8 +11,8 @@ static int
 repeat_regular_file(void)
 {
 	while (inf || count--) {
-		fadvise_sequential(stream.fd, stream.headlen, 0);
-		elseek(stream.fd, stream.headlen, SEEK_SET, stream.file);
+		fadvise_sequential(stream.fd, (off_t)(stream.headlen), 0);
+		elseek(stream.fd, (off_t)(stream.headlen), SEEK_SET, stream.file);
 		if (esend_stream(&stream, STDOUT_FILENO, NULL))
 			return -1;
 	}
diff --git a/src/blind-reverse.c b/src/blind-reverse.c
@@ -1,6 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
 USAGE("[-i] file")
 
@@ -9,15 +8,15 @@ to_stdout(struct stream *stream)
 {
 	size_t ptr, end, n;
 	char buf[BUFSIZ];
-	ssize_t r;
 
 	while (stream->frames--) {
 		ptr = stream->frames * stream->frame_size + stream->headlen;
 		end = ptr + stream->frame_size;
 		while (ptr < end) {
-			if (!(r = epread(stream->fd, buf, MIN(sizeof(buf), end - ptr), ptr, stream->file)))
+			if (!(n = epread(stream->fd, buf, MIN(sizeof(buf), end - ptr),
+					 (off_t)ptr, stream->file)))
 				eprintf("%s: file is shorter than expected\n", stream->file);
-			ptr += n = (size_t)r;
+			ptr += n;
 			ewriteall(STDOUT_FILENO, buf, n, "<stdout>");
 		}
 	}
@@ -57,8 +56,8 @@ in_place(struct stream *stream)
 	bufb = emalloc(stream->frame_size);
 
 	for (f = 0; f < stream->frames >> 1; f++) {
-		pa = f        * stream->frame_size + stream->headlen;
-		pb = (fm - f) * stream->frame_size + stream->headlen;
+		pa = (off_t)f        * (off_t)(stream->frame_size) + (off_t)(stream->headlen);
+		pb = (off_t)(fm - f) * (off_t)(stream->frame_size) + (off_t)(stream->headlen);
 
 		epread_frame(stream, bufa, pa);
 		epread_frame(stream, bufb, pb);
diff --git a/src/blind-rewrite-head.c b/src/blind-rewrite-head.c
@@ -1,10 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <sys/mman.h>
-#include <sys/stat.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-h] file [(frames | 'auto') [(width | 'same') (height | 'same') [format | 'same']]]")
 
@@ -40,7 +35,7 @@ rewrite(struct stream *stream, int frames_auto)
 		eprintf("%s: video is too long\n", stream->file);
 
 	if ((size_t)headlen > stream->headlen)
-		if (ftruncate(stream->fd, length + headlen))
+		if (ftruncate(stream->fd, (off_t)length + (off_t)headlen))
 			eprintf("ftruncate %s:", stream->file);
 
 	data = mmap(0, length + (size_t)headlen, PROT_READ | PROT_WRITE, MAP_SHARED, stream->fd, 0);
@@ -52,7 +47,7 @@ rewrite(struct stream *stream, int frames_auto)
 	munmap(data, length + (size_t)headlen);
 
 	if ((size_t)headlen < stream->headlen)
-		if (ftruncate(stream->fd, length + headlen))
+		if (ftruncate(stream->fd, (off_t)length + (off_t)headlen))
 			eprintf("ftruncate %s:", stream->file);
 }
 
diff --git a/src/blind-set-alpha.c b/src/blind-set-alpha.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-i] alpha-stream")
 
diff --git a/src/blind-set-luma.c b/src/blind-set-luma.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("luma-stream")
 
diff --git a/src/blind-set-saturation.c b/src/blind-set-saturation.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-w] saturation-stream")
 
diff --git a/src/blind-single-colour.c b/src/blind-single-colour.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-f frames | -f 'inf'] [-F pixel-format] -w width -h height (X Y Z | Y) [alpha]")
 
diff --git a/src/blind-skip-pattern.c b/src/blind-skip-pattern.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("(skipped-frames | +included-frames) ...")
 
diff --git a/src/blind-split.c b/src/blind-split.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <alloca.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-L] (file (end-point | 'end')) ...")
 
diff --git a/src/blind-stack.c b/src/blind-stack.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-b] bottom-stream ... top-stream")
 
@@ -22,7 +19,7 @@ USAGE("[-b] bottom-stream ... top-stream")
 				z2 = ((TYPE *)(streams[j].buf + i))[2];\
 				a2 = ((TYPE *)(streams[j].buf + i))[3];\
 				if (BLEND)\
-					a2 /= j + 1;\
+					a2 /= (TYPE)(j + 1);\
 				x1 = x1 * a1 * (1 - a2) + x2 * a2;\
 				y1 = y1 * a1 * (1 - a2) + y2 * a2;\
 				z1 = z1 * a1 * (1 - a2) + z2 * a2;\
diff --git a/src/blind-temporal-mean.c b/src/blind-temporal-mean.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <math.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-g | -h | -l power | -p power]")
 /* TODO add -w weight-stream */
@@ -21,23 +17,23 @@ typedef void (*process_func)(struct stream *stream, void *buffer, void *image, s
  * X-parameter 7: pre-finalise assignments
  * X-parameter 8: subcell finalisation
  */
-#define LIST_MEANS(TYPE, SUFFIX)\
+#define LIST_MEANS(TYPE)\
 	/* [default] arithmetic mean */\
 	X(ARITHMETIC, arithmetic, 1, COPY_FRAME,, *img1 += *buf,\
 	  a = (TYPE)1.0 / (TYPE)frame, *img1 *= a)\
 	/* geometric mean */\
 	X(GEOMETRIC, geometric, 1, COPY_FRAME,, *img1 *= *buf,\
-	  a = (TYPE)1.0 / (TYPE)frame, *img1 = nnpow##SUFFIX(*img1, a))\
+	  a = (TYPE)1.0 / (TYPE)frame, *img1 = nnpow(*img1, a))\
 	/* harmonic mean */\
 	X(HARMONIC, harmonic, 1, ZERO_AND_PROCESS_FRAME,, *img1 += (TYPE)1 / *buf,\
 	  a = (TYPE)frame, *img1 = a / *img1)\
 	/* lehmer mean */\
 	X(LEHMER, lehmer, 2, ZERO_AND_PROCESS_FRAME, (a = (TYPE)power, b = a - (TYPE)1),\
-	  (*img1 += nnpow##SUFFIX(*buf, a), *img2 += nnpow##SUFFIX(*buf, b)),, *img1 /= *img2)\
+	  (*img1 += nnpow(*buf, a), *img2 += nnpow(*buf, b)),, *img1 /= *img2)\
 	/* power mean (Hölder mean) (m = 2 for root square mean; m = 3 for cubic mean) */\
 	X(POWER, power, 1, ZERO_AND_PROCESS_FRAME, a = (TYPE)power,\
-	  *img1 += nnpow##SUFFIX(*buf, a), (a = (TYPE)1 / (TYPE)frame, b = (TYPE)(1.0 / power)),\
-	  *img1 = a * nnpow##SUFFIX(*img1, b))
+	  *img1 += nnpow(*buf, a), (a = (TYPE)1 / (TYPE)frame, b = (TYPE)(1.0 / power)),\
+	  *img1 = a * nnpow(*img1, b))
 
 enum first_frame_action {
 	COPY_FRAME,
@@ -46,28 +42,12 @@ enum first_frame_action {
 };
 
 #define X(V, ...) V,
-enum method { LIST_MEANS(,) };
+enum method { LIST_MEANS() };
 #undef X
 
 static double power;
 
-static inline double
-nnpow(double a, double b)
-{
-	int neg = a < 0;
-	a = pow(neg ? -a : a, b);
-	return neg ? -a : a;
-}
-
-static inline float
-nnpowf(float a, float b)
-{
-	int neg = a < 0;
-	a = powf(neg ? -a : a, b);
-	return neg ? -a : a;
-}
-
-#define MAKE_PROCESS(PIXFMT, TYPE, SUFFIX,\
+#define MAKE_PROCESS(PIXFMT, TYPE,\
 		     _1, NAME, _3, _4, PRE_PROCESS, PROCESS_SUBCELL, PRE_FINALISE, FINALISE_SUBCELL)\
 	static void\
 	process_##PIXFMT##_##NAME(struct stream *stream, void *buffer, void *image, size_t frame)\
@@ -86,22 +66,22 @@ nnpowf(float a, float b)
 				for (x = 0; x < stream->width; x++, img1++, img2++, buf++)\
 					PROCESS_SUBCELL;\
 		}\
-		(void) img2, (void) a, (void) b;\
+		(void) img2, (void) a, (void) b, (void) frame;\
 	}
-#define X(...) MAKE_PROCESS(xyza, double,, __VA_ARGS__)
-LIST_MEANS(double,)
+#define X(...) MAKE_PROCESS(xyza, double, __VA_ARGS__)
+LIST_MEANS(double)
 #undef X
-#define X(...) MAKE_PROCESS(xyzaf, float, f, __VA_ARGS__)
-LIST_MEANS(float, f)
+#define X(...) MAKE_PROCESS(xyzaf, float, __VA_ARGS__)
+LIST_MEANS(float)
 #undef X
 #undef MAKE_PROCESS
 
 #define X(ID, NAME, ...) [ID] = process_xyza_##NAME,
-static const process_func process_functions_xyza[] = { LIST_MEANS(,) };
+static const process_func process_functions_xyza[] = { LIST_MEANS() };
 #undef X
 
 #define X(ID, NAME, ...) [ID] = process_xyzaf_##NAME,
-static const process_func process_functions_xyzaf[] = { LIST_MEANS(,) };
+static const process_func process_functions_xyzaf[] = { LIST_MEANS() };
 #undef X
 
 int
@@ -142,7 +122,7 @@ main(int argc, char *argv[])
 		first_frame_action = FIRST_FRAME_ACTION;\
 		break;
 	switch (method) {
-	LIST_MEANS(,)
+	LIST_MEANS()
 	default:
 		abort();
 	}
diff --git a/src/blind-time-blur.c b/src/blind-time-blur.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
-#include <unistd.h>
+#include "common.h"
 
 USAGE("alpha-stream")
 
diff --git a/src/blind-to-image.c b/src/blind-to-image.c
@@ -1,9 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <arpa/inet.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-d depth | -f]")
 
@@ -13,7 +9,7 @@ static int alpha_warning_triggered = 0;
 static unsigned long long int max;
 static int bytes;
 
-#define WRITE_PIXEL(TYPE, SUFFIX)\
+#define WRITE_PIXEL(TYPE)\
 	do {\
 		unsigned long long int colours[4];\
 		unsigned char buf[4 * 8];\
@@ -38,10 +34,10 @@ static int bytes;
 			A = A < 0 ? 0 : 1;\
 		}\
 		\
-		colours[0] = srgb_encode##SUFFIX(R) * max + (TYPE)0.5;\
-		colours[1] = srgb_encode##SUFFIX(G) * max + (TYPE)0.5;\
-		colours[2] = srgb_encode##SUFFIX(B) * max + (TYPE)0.5;\
-		colours[3] = A * max + (TYPE)0.5;\
+		colours[0] = (unsigned long long int)(srgb_encode(R) * (TYPE)max + (TYPE)0.5);\
+		colours[1] = (unsigned long long int)(srgb_encode(G) * (TYPE)max + (TYPE)0.5);\
+		colours[2] = (unsigned long long int)(srgb_encode(B) * (TYPE)max + (TYPE)0.5);\
+		colours[3] = (unsigned long long int)(A * (TYPE)max + (TYPE)0.5);\
 		\
 		for (i = k = 0; i < 4; i++, k += bytes) {\
 			for (j = 0; j < bytes; j++) {\
@@ -50,7 +46,7 @@ static int bytes;
 			}\
 		}\
 		\
-		ewriteall(STDOUT_FILENO, buf, k, "<stdout>");\
+		ewriteall(STDOUT_FILENO, buf, (size_t)k, "<stdout>");\
 	} while (0)
 
 #define PROCESS(TYPE, SUFFIX)\
@@ -71,15 +67,15 @@ static int bytes;
 				}\
 			}\
 			\
-			ciexyz_to_srgb##SUFFIX(X, Y, Z, &R, &G, &B);\
+			ciexyz_to_srgb(X, Y, Z, &R, &G, &B);\
 			write_pixel##SUFFIX(R, G, B, A);\
 		}\
 	} while (0)
 
-static void write_pixel(double R, double G, double B, double A) {WRITE_PIXEL(double,);}
-static void write_pixel_f(float R, float G, float B, float A) {WRITE_PIXEL(float, _f);}
+static void write_pixel_d(double R, double G, double B, double A) {WRITE_PIXEL(double);}
+static void write_pixel_f(float  R, float  G, float  B, float  A) {WRITE_PIXEL(float);}
 
-static void process_xyza (struct stream *stream, size_t n) {PROCESS(double,);}
+static void process_xyza (struct stream *stream, size_t n) {PROCESS(double, _d);}
 static void process_xyzaf(struct stream *stream, size_t n) {PROCESS(float, _f);}
 
 int
@@ -117,7 +113,8 @@ main(int argc, char *argv[])
 		eprintf("pixel format %s is not supported, try xyza\n", stream.pixfmt);
 
 	if (farbfeld) {
-		uint32_t width = stream.width, height = stream.height;
+		uint32_t width  = (uint32_t)(stream.width);
+		uint32_t height = (uint32_t)(stream.height);
 		if (stream.width > UINT32_MAX)
 			eprintf("%s: frame is too wide\n", stream.file);
 		if (stream.height > UINT32_MAX)
diff --git a/src/blind-to-portable.c b/src/blind-to-portable.c
@@ -1,17 +1,18 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
-#include <alloca.h>
-#include <math.h>
-#include <string.h>
+/* Disable warnings in <math.h> */
+#if defined(__clang__)
+# pragma clang diagnostic ignored "-Wdouble-promotion"
+# pragma clang diagnostic ignored "-Wconversion"
+#endif
 
 USAGE("[-s]")
 
 #define USING_BINARY32 0
 #define USING_BINARY64 0
 
-#define CONV(ITYPE, OTYPE, SOTYPE, EXPONENT, HA2EXPONENT, FRACTION, SUFFIX)\
+#define CONV(ITYPE, OTYPE, SOTYPE, EXPONENT, HA2EXPONENT, FRACTION)\
 	do {\
 		static int cache_i = 0;\
 		static ITYPE cache_in[] = {0, 0, 0, 0};\
@@ -24,22 +25,23 @@ USAGE("[-s]")
 			cache_i &= 3;\
 			return ret;\
 		}\
-		signb = signbit(host);\
+		cache_in[cache_i] = host;\
+		signb = (OTYPE)signbit(host);\
 		u = signb ? -host : host;\
-		if (isnan(host) || !isfinite(host)) {\
-			ret = ((((OTYPE)1 << EXPONENT) - 1) << FRACTION) | !isinf(host);\
+		if (g_isnan(host) || g_isinf(host)) {\
+			ret = ((((OTYPE)1 << EXPONENT) - (OTYPE)1) << FRACTION) | (OTYPE)!g_isinf(host);\
 		} else if (u == (ITYPE)0.0) {\
 			ret = 0;\
 		} else {\
-			dexponent = log2##SUFFIX(u);\
+			dexponent = log2(u);\
 			exponent = (SOTYPE)dexponent;\
-			if (u == pow##SUFFIX(2.0, (ITYPE)exponent)) {\
+			if (u == pow((ITYPE)2.0, (ITYPE)exponent)) {\
 				exponent += HA2EXPONENT;\
 				fraction = 0;\
 			} else {\
 				/* TODO subnormals are a bit rounded off */\
-				u *= pow##SUFFIX(2.0, (ITYPE)(FRACTION + 1 - exponent));\
-				fraction = u;\
+				u *= pow((ITYPE)2.0, (ITYPE)(FRACTION + 1 - exponent));\
+				fraction = (OTYPE)u;\
 				while (fraction >= (OTYPE)2 << FRACTION) {\
 					fraction >>= 1;\
 					exponent += 1;\
@@ -51,11 +53,11 @@ USAGE("[-s]")
 				/* TODO subnormal result */\
 				exponent = 0;\
 				fraction = 0;\
-			} else if (exponent >= ((SOTYPE)1 << EXPONENT) - (SOTYPE)1) { \
-				exponent = ((SOTYPE)1 << EXPONENT) - (SOTYPE)1;\
+			} else if (exponent >= ((SOTYPE)1 << EXPONENT) - 1) { \
+				exponent = ((SOTYPE)1 << EXPONENT) - 1;\
 				fraction = 0;\
 			}\
-			ret = (exponent << FRACTION) + fraction;\
+			ret = ((OTYPE)exponent << FRACTION) + fraction;\
 		}\
 		ret |= signb << (FRACTION + EXPONENT);\
 		cache_out[cache_i++] = ret;\
@@ -89,8 +91,8 @@ USAGE("[-s]")
 			eprintf("%s: incomplete frame\n", stream->file);\
 	} while (0)
 
-static uint64_t conv_double(double host) {CONV(double, uint64_t, int64_t, 11, 1023, 52,);}
-static uint32_t conv_float (float host)  {CONV(float, uint32_t, int32_t, 8, 127, 23, f);}
+static uint64_t conv_double(double host) {CONV(double, uint64_t, int64_t, 11, 1023, 52);}
+static uint32_t conv_float (float  host) {CONV(float, uint32_t, int32_t, 8, 127, 23);}
 
 static void process_xyza (struct stream *stream, int strict) {PROCESS(double, uint64_t, 64);}
 static void process_xyzaf(struct stream *stream, int strict) {PROCESS(float, uint32_t, 32);}
diff --git a/src/blind-to-text.c b/src/blind-to-text.c
@@ -1,6 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
 #include <string.h>
 
diff --git a/src/blind-to-video.c b/src/blind-to-video.c
@@ -1,18 +1,12 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <signal.h>
-#include <stdio.h>
-#include <stdint.h>
-#include <string.h>
+#include "common.h"
 
 USAGE("[-d] frame-rate ffmpeg-arguments ...")
 
 static int draft = 0;
 static int fd;
 
-#define PROCESS(TYPE, SUFFIX)\
+#define PROCESS(TYPE)\
 	do {\
 		char *buf = stream->buf;\
 		TYPE *pixel, r, g, b;\
@@ -25,14 +19,14 @@ static int fd;
 		if (draft) {\
 			for (ptr = 0; ptr < n; ptr += 4 * sizeof(TYPE)) {\
 				pixel = (TYPE *)(buf + ptr);\
-				ciexyz_to_scaled_yuv##SUFFIX(pixel[0], pixel[1], pixel[2], &r, &g, &b);\
+				ciexyz_to_scaled_yuv(pixel[0], pixel[1], pixel[2], &r, &g, &b);\
 				y = (long int)r +  16L * 256L;\
 				u = (long int)g + 128L * 256L;\
 				v = (long int)b + 128L * 256L;\
 				*pixels++ = 0xFFFFU;\
-				*pixels++ = htole16((uint16_t)CLIP(0, y, 0xFFFFL));\
-				*pixels++ = htole16((uint16_t)CLIP(0, u, 0xFFFFL));\
-				*pixels++ = htole16((uint16_t)CLIP(0, v, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, y, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, u, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, v, 0xFFFFL));\
 				if (pixels == end)\
 					ewriteall(fd, pixels = pixbuf, sizeof(pixbuf), "<subprocess>");\
 			}\
@@ -40,18 +34,18 @@ static int fd;
 			for (ptr = 0; ptr < n; ptr += 4 * sizeof(TYPE)) {\
 				pixel = (TYPE *)(buf + ptr);\
 				a = (long int)(pixel[3] * 0xFFFFL);\
-				ciexyz_to_srgb##SUFFIX(pixel[0], pixel[1], pixel[2], &r, &g, &b);\
-				r = srgb_encode##SUFFIX(r);\
-				g = srgb_encode##SUFFIX(g);\
-				b = srgb_encode##SUFFIX(b);\
-				srgb_to_yuv##SUFFIX(r, g, b, pixel + 0, pixel + 1, pixel + 2);\
+				ciexyz_to_srgb(pixel[0], pixel[1], pixel[2], &r, &g, &b);\
+				r = srgb_encode(r);\
+				g = srgb_encode(g);\
+				b = srgb_encode(b);\
+				srgb_to_yuv(r, g, b, pixel + 0, pixel + 1, pixel + 2);\
 				y = (long int)(pixel[0] * 0xFFFFL) +  16L * 256L;\
 				u = (long int)(pixel[1] * 0xFFFFL) + 128L * 256L;\
 				v = (long int)(pixel[2] * 0xFFFFL) + 128L * 256L;\
-				*pixels++ = htole16((uint16_t)CLIP(0, a, 0xFFFFL));\
-				*pixels++ = htole16((uint16_t)CLIP(0, y, 0xFFFFL));\
-				*pixels++ = htole16((uint16_t)CLIP(0, u, 0xFFFFL));\
-				*pixels++ = htole16((uint16_t)CLIP(0, v, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, a, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, y, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, u, 0xFFFFL));\
+				*pixels++ = htole((uint16_t)CLIP(0, v, 0xFFFFL));\
 				if (pixels == end)\
 					ewriteall(fd, pixels = pixbuf, sizeof(pixbuf), "<subprocess>");\
 			}\
@@ -59,8 +53,8 @@ static int fd;
 		ewriteall(fd, pixbuf, (size_t)(pixels - pixbuf) * sizeof(*pixels), "<subprocess>");\
 	} while (0)
 
-static void process_xyza (struct stream *stream, size_t n) {PROCESS(double,);}
-static void process_xyzaf(struct stream *stream, size_t n) {PROCESS(float, _f);}
+static void process_xyza (struct stream *stream, size_t n) {PROCESS(double);}
+static void process_xyzaf(struct stream *stream, size_t n) {PROCESS(float);}
 
 int
 main(int argc, char *argv[])
diff --git a/src/blind-translate.c b/src/blind-translate.c
@@ -1,8 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <string.h>
+#include "common.h"
 
 USAGE("[-wp] translation-stream")
 
@@ -59,7 +56,6 @@ process_frame(struct stream *stream, char *buf, size_t above, size_t below, size
 	return 1;
 eof:
 	eprintf("%s: file is shorter than expected\n", stream->file);
-	return 0;
 }
 
 static void
diff --git a/src/blind-transpose.c b/src/blind-transpose.c
@@ -1,6 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
+#include "common.h"
 
 USAGE("")
 
diff --git a/src/blind-write-head.c b/src/blind-write-head.c
@@ -1,5 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "util.h"
+#include "common.h"
 
 USAGE("parameters ...")
 
diff --git a/src/common.h b/src/common.h
@@ -0,0 +1,37 @@
+/* See LICENSE file for copyright and license details. */
+#if defined(__clang__)
+# pragma clang diagnostic ignored "-Wdisabled-macro-expansion"
+# pragma clang diagnostic ignored "-Wcomma"
+# pragma clang diagnostic ignored "-Wcast-align"
+# pragma clang diagnostic ignored "-Wassign-enum"
+# pragma clang diagnostic ignored "-Wfloat-equal"
+# pragma clang diagnostic ignored "-Wformat-nonliteral"
+# pragma clang diagnostic ignored "-Wcovered-switch-default"
+#elif defined(__GNUC__)
+# pragma GCC diagnostic ignored "-Wfloat-equal"
+#endif
+
+#include "stream.h"
+#include "util.h"
+#include "video-math.h"
+
+#include <arpa/inet.h>
+#if defined(HAVE_EPOLL)
+# include <sys/epoll.h>
+#endif
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <sys/wait.h>
+#include <alloca.h>
+#include <ctype.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <limits.h>
+#include <signal.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <strings.h>
+#include <unistd.h>
diff --git a/src/stream.c b/src/stream.c
@@ -1,14 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "stream.h"
-#include "util.h"
-
-#include <sys/mman.h>
-#include <sys/stat.h>
-#include <errno.h>
-#include <inttypes.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
+#include "common.h"
 
 static inline int
 get_dimension(int status, size_t *out, const char *s, const char *fname, const char *dim)
diff --git a/src/util.c b/src/util.c
@@ -1,18 +1,5 @@
 /* See LICENSE file for copyright and license details. */
-#include "util.h"
-
-#include <sys/wait.h>
-#include <ctype.h>
-#include <errno.h>
-#include <fcntl.h>
-#include <limits.h>
-#include <signal.h>
-#include <stdarg.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <strings.h>
-#include <unistd.h>
+#include "common.h"
 
 char *argv0;
 
@@ -139,17 +126,17 @@ readall(int fd, void *buf, size_t n)
 }
 
 int
-pwriteall(int fd, void *buf, size_t n, size_t ptr)
+pwriteall(int fd, void *buf, size_t n, off_t ptr)
 {
 	char *buffer = buf;
 	ssize_t r;
 	while (n) {
-		r = pwrite(fd, buffer, n, ptr);
+		r = pwrite(fd, buffer, n, (off_t)ptr);
 		if (r < 0)
 			return -1;
 		buffer += (size_t)r;
 		n -= (size_t)r;
-		ptr += (size_t)r;
+		ptr += (off_t)r;
 	}
 	return 0;
 }
@@ -171,7 +158,7 @@ getfile(int fd, void *buffer, size_t *restrict ptr, size_t *restrict size)
 {
 	char *restrict *restrict buf = buffer;
 	void *new;
-	size_t r;
+	ssize_t r;
 
 	for (;;) {
 		if (*ptr == *size) {
diff --git a/src/util.h b/src/util.h
@@ -1,6 +1,12 @@
 /* See LICENSE file for copyright and license details. */
 #include "arg.h"
 
+#if defined(__GNUC__) || defined(__clang__)
+# define ATTRIBUTE_NORETURN __attribute__((noreturn))
+#else
+# define ATTRIBUTE_NORETURN
+#endif
+
 #define ELEMENTSOF(ARRAY) (sizeof(ARRAY) / sizeof(*(ARRAY)))
 #define MIN(A, B)         ((A) < (B) ? (A) : (B))
 #define MAX(A, B)         ((A) > (B) ? (A) : (B))
@@ -9,8 +15,9 @@
 #define INTSTRLEN(TYPE)   ((sizeof(TYPE) == 1 ? 3 : (5 * sizeof(TYPE) / 2)) + ((TYPE)-1 < 1))
 
 #define USAGE(SYNOPSIS)\
+	ATTRIBUTE_NORETURN\
 	static void usage(void)\
-	{ eprintf("usage: %s%s%s\n", argv0, SYNOPSIS ? " " : "", SYNOPSIS); }
+	{ eprintf("usage: %s%s%s\n", argv0, *SYNOPSIS ? " " : "", SYNOPSIS); }
 
 #include "util/eprintf.h"
 #include "util/efflush.h"
diff --git a/src/util/colour.h b/src/util/colour.h
@@ -7,210 +7,118 @@
 #define D65_XYZ_X (D65_XYY_X / D65_XYY_Y)
 #define D65_XYZ_Z (1 / D65_XYY_Y - 1 - D65_XYZ_X)
 
-static inline double
-srgb_encode(double t)
-{
-	double sign = 1;
-	if (t < 0) {
-		t = -t;
-		sign = -1;
+#define SRGB_ENCODE(TYPE, NAME_SUFFIX, MATH_SUFFIX)\
+	static inline TYPE\
+	srgb_encode##NAME_SUFFIX(TYPE t)\
+	{\
+		TYPE sign = 1;\
+		if (t < 0) {\
+			t = -t;\
+			sign = -1;\
+		}\
+		t = t <= (TYPE)0.0031306684425217108\
+			? (TYPE)12.92 * t\
+			: (TYPE)1.055 * pow##MATH_SUFFIX(t, (TYPE)1.0 / (TYPE)2.4) - (TYPE)0.055;\
+		return t * sign;\
 	}
-	t = t <= 0.0031306684425217108
-		? 12.92 * t
-		: 1.055 * pow(t, 1 / 2.4) - 0.055;
-	return t * sign;
-}
-
-static inline float
-srgb_encode_f(float t)
-{
-	float sign = 1;
-	if (t < 0) {
-		t = -t;
-		sign = -1;
-	}
-	t = t <= (float)0.0031306684425217108
-		? (float)12.92 * t
-		: (float)1.055 * powf(t, 1 / (float)2.4) - (float)0.055;
-	return t * sign;
-}
-
-static inline double
-srgb_decode(double t)
-{
-	double sign = 1;
-	if (t < 0) {
-		t = -t;
-		sign = -1;
+SRGB_ENCODE(double, _d,)
+SRGB_ENCODE(float, _f, f)
+#undef SRGB_ENCODE
+
+#define SRGB_DECODE(TYPE, NAME_SUFFIX, MATH_SUFFIX)\
+	static inline TYPE\
+	srgb_decode##NAME_SUFFIX(TYPE t)\
+	{\
+		TYPE sign = 1;\
+		if (t < 0) {\
+			t = -t;\
+			sign = -1;\
+		}\
+		t = t <= (TYPE)0.0031306684425217108 * (TYPE)12.92\
+			? t / (TYPE)12.92\
+			: pow##MATH_SUFFIX((t + (TYPE)0.055) / (TYPE)1.055, (TYPE)2.4);\
+		return t * sign;\
 	}
-	t = t <= 0.0031306684425217108 * 12.92
-		? t / 12.92
-		: pow((t + 0.055) / 1.055, 2.4);
-	return t * sign;
-}
-
-static inline float
-srgb_decode_f(float t)
-{
-	float sign = 1;
-	if (t < 0) {
-		t = -t;
-		sign = -1;
+SRGB_DECODE(double, _d,)
+SRGB_DECODE(float, _f, f)
+#undef SRGB_DECODE
+
+
+#define MATRIX_MULTIPLY_FUNCTION(FUNCTION, TYPE, R1C1, R1C2, R1C3, R2C1, R2C2, R2C3, R3C1, R3C2, R3C3)\
+	static inline void\
+	FUNCTION(TYPE ia, TYPE ib, TYPE ic, TYPE *oa, TYPE *ob, TYPE *oc)\
+	{\
+		*oa = (TYPE)(R1C1) * ia + (TYPE)(R1C2) * ib + (TYPE)(R1C3) * ic;\
+		*ob = (TYPE)(R2C1) * ia + (TYPE)(R2C2) * ib + (TYPE)(R2C3) * ic;\
+		*oc = (TYPE)(R3C1) * ia + (TYPE)(R3C2) * ib + (TYPE)(R3C3) * ic;\
 	}
-	t = t <= (float)0.0031306684425217108 * (float)12.92
-		? t / (float)12.92
-		: powf((t + (float)0.055) / (float)1.055, (float)2.4);
-	return t * sign;
-}
-
-static inline void
-yuv_to_srgb(double y, double u, double v, double *r, double *g, double *b)
-{
-#define MULTIPLY(CY, CU, CV)  ((CY) * y + (CU) * u + (CV) * v)
-	*r = MULTIPLY(1,  0.00028328010485821202317155420580263580632163211703,  1.14070449590558520291949662350816652178764343261719);
-	*g = MULTIPLY(1, -0.39630886669497211727275498560629785060882568359375, -0.58107364288228224857846271333983168005943298339844);
-	*b = MULTIPLY(1,  2.03990003507541306504435851820744574069976806640625,  0.00017179031692307700847528739718228507626918144524);
-#undef MULTIPLY
-}
-
-static inline void
-yuv_to_srgb_f(float y, float u, float v, float *r, float *g, float *b)
-{
-#define MULTIPLY(CY, CU, CV)  ((float)(CY) * y + (float)(CU) * u + (float)(CV) * v)
-	*r = MULTIPLY(1,  0.00028328010485821202317155420580263580632163211703,  1.14070449590558520291949662350816652178764343261719);
-	*g = MULTIPLY(1, -0.39630886669497211727275498560629785060882568359375, -0.58107364288228224857846271333983168005943298339844);
-	*b = MULTIPLY(1,  2.03990003507541306504435851820744574069976806640625,  0.00017179031692307700847528739718228507626918144524);
-#undef MULTIPLY
-}
-
-static inline void
-srgb_to_yuv(double r, double g, double b, double *y, double *u, double *v)
-{
-#define MULTIPLY(CR, CG, CB) ((CR) * r + (CG) * g + (CB) * b)
-	*y = MULTIPLY(0.299, 0.587, 0.114);
-	*u = MULTIPLY(-0.14662756598240470062854967636667424812912940979004,
-		      -0.28771586836102963635752871596196200698614120483398,
-		       0.43434343434343436474165400795754976570606231689453);
-	*v = MULTIPLY( 0.61456892577224520035628074765554629266262054443359,
-		      -0.51452282157676354490405401520547457039356231689453,
-		      -0.10004610419548178035231700278018251992762088775635);
-#undef MULTIPLY
-}
-
-static inline void
-srgb_to_yuv_f(float r, float g, float b, float *y, float *u, float *v)
-{
-#define MULTIPLY(CR, CG, CB) ((float)(CR) * r + (float)(CG) * g + (float)(CB) * b)
-	*y = MULTIPLY(0.299, 0.587, 0.114);
-	*u = MULTIPLY(-0.14662756598240470062854967636667424812912940979004,
-		      -0.28771586836102963635752871596196200698614120483398,
-		       0.43434343434343436474165400795754976570606231689453);
-	*v = MULTIPLY( 0.61456892577224520035628074765554629266262054443359,
-		      -0.51452282157676354490405401520547457039356231689453,
-		      -0.10004610419548178035231700278018251992762088775635);
-#undef MULTIPLY
-}
-
-static inline void
-ciexyz_to_srgb(double x, double y, double z, double *r, double *g, double *b)
-{
-#define MULTIPLY(CX, CY, CZ)  ((CX) * x + (CY) * y + (CZ) * z)
-	*r = MULTIPLY(3.240446254647737500675930277794, -1.537134761820080575134284117667, -0.498530193022728718155178739835);
-	*g = MULTIPLY(-0.969266606244679751469561779231, 1.876011959788370209167851498933, 0.041556042214430065351304932619);
-	*b = MULTIPLY(0.055643503564352832235773149705, -0.204026179735960239147729566866, 1.057226567722703292062647051353);
-#undef MULTIPLY
-}
-
-static inline void
-ciexyz_to_srgb_f(float x, float y, float z, float *r, float *g, float *b)
-{
-#define MULTIPLY(CX, CY, CZ)  ((float)(CX) * x + (float)(CY) * y + (float)(CZ) * z)
-	*r = MULTIPLY(3.240446254647737500675930277794, -1.537134761820080575134284117667, -0.498530193022728718155178739835);
-	*g = MULTIPLY(-0.969266606244679751469561779231, 1.876011959788370209167851498933, 0.041556042214430065351304932619);
-	*b = MULTIPLY(0.055643503564352832235773149705, -0.204026179735960239147729566866, 1.057226567722703292062647051353);
-#undef MULTIPLY
-}
-
-static inline void
-srgb_to_ciexyz(double r, double g, double b, double *x, double *y, double *z)
-{
-#define MULTIPLY(CR, CG, CB) ((CR) * r + (CG) * g + (CB) * b)
-	*x = MULTIPLY(0.412457445582367576708548995157, 0.357575865245515878143578447634, 0.180437247826399665973085006954);
-	*y = MULTIPLY(0.212673370378408277403536885686, 0.715151730491031756287156895269, 0.072174899130559869164791564344);
-	*z = MULTIPLY(0.019333942761673460208893260415, 0.119191955081838593666354597644, 0.950302838552371742508739771438);
-#undef MULTIPLY
-}
-
-static inline void
-srgb_to_ciexyz_f(float r, float g, float b, float *x, float *y, float *z)
-{
-#define MULTIPLY(CR, CG, CB) ((float)(CR) * r + (float)(CG) * g + (float)(CB) * b)
-	*x = MULTIPLY(0.412457445582367576708548995157, 0.357575865245515878143578447634, 0.180437247826399665973085006954);
-	*y = MULTIPLY(0.212673370378408277403536885686, 0.715151730491031756287156895269, 0.072174899130559869164791564344);
-	*z = MULTIPLY(0.019333942761673460208893260415, 0.119191955081838593666354597644, 0.950302838552371742508739771438);
-#undef MULTIPLY
-}
-
-static inline void
-scaled_yuv_to_ciexyz(double y, double u, double v, double *xp, double *yp, double *zp)
-{
-#define MULTIPLY(CY, CU, CV) ((CY) * y + (CU) * u + (CV) * v)
-	*xp = MULTIPLY( 0.00001450325106667098632156481796684488472237717360,
-		        0.00000345586790639342739093228633329157872822179343,
-		        0.00000400923398630552893485111398685916128670214675);
-	*yp = MULTIPLY( 0.00001525902189669641837040624243737596543724066578,
-		       -0.00000207722814409390653614547427030512238843584782,
-		       -0.00000263898607692305410302407824019166326934282552);
-	*zp = MULTIPLY( 0.00001661446153041708825425643025752719950105529279,
-		        0.00002885925752619118069149627137104374696718878113,
-			-0.00000071781086875769179526501342566979779746816348);
-#undef MULTIPLY
-}
-
-static inline void
-scaled_yuv_to_ciexyz_f(float y, float u, float v, float *xp, float *yp, float *zp)
-{
-#define MULTIPLY(CY, CU, CV) ((float)(CY) * y + (float)(CU) * u + (float)(CV) * v)
-	*xp = MULTIPLY( 0.00001450325106667098632156481796684488472237717360,
-		        0.00000345586790639342739093228633329157872822179343,
-		        0.00000400923398630552893485111398685916128670214675);
-	*yp = MULTIPLY( 0.00001525902189669641837040624243737596543724066578,
-		       -0.00000207722814409390653614547427030512238843584782,
-		       -0.00000263898607692305410302407824019166326934282552);
-	*zp = MULTIPLY( 0.00001661446153041708825425643025752719950105529279,
-		        0.00002885925752619118069149627137104374696718878113,
-			-0.00000071781086875769179526501342566979779746816348);
-#undef MULTIPLY
-}
-
-static inline void
-ciexyz_to_scaled_yuv(double x, double y, double z, double *yp, double *up, double *vp)
-{
-#define MULTIPLY(CX, CY, CZ) ((CX) * x + (CY) * y + (CZ) * z)
-	*yp = MULTIPLY(  26625.38231027395886485464870929718017578125,
-		         40524.0090949436053051613271236419677734375,
-		          -271.5313105642117079696618020534515380859375);
-	*up = MULTIPLY( -11278.3751445417292416095733642578125,
-		        -26409.91773157499847002327442169189453125,
-		         34100.5706543184860493056476116180419921875);
-	*vp = MULTIPLY( 162829.60100012840121053159236907958984375,
-		       -123829.313212639070115983486175537109375,
-			-28411.65702312920984695665538311004638671875);
-#undef MULTIPLY
-}
-
-static inline void
-ciexyz_to_scaled_yuv_f(float x, float y, float z, float *yp, float *up, float *vp)
-{
-#define MULTIPLY(CX, CY, CZ) ((float)(CX) * x + (float)(CY) * y + (float)(CZ) * z)
-	*yp = MULTIPLY(  26625.38231027395886485464870929718017578125,
-		         40524.0090949436053051613271236419677734375,
-		          -271.5313105642117079696618020534515380859375);
-	*up = MULTIPLY( -11278.3751445417292416095733642578125,
-		        -26409.91773157499847002327442169189453125,
-		         34100.5706543184860493056476116180419921875);
-	*vp = MULTIPLY( 162829.60100012840121053159236907958984375,
-		       -123829.313212639070115983486175537109375,
-			-28411.65702312920984695665538311004638671875);
-#undef MULTIPLY
-}
+#define MATRIX_MULTIPLY_FUNCTIONS(FUNCTION_BASE, ...)\
+	MATRIX_MULTIPLY_FUNCTION(FUNCTION_BASE##_d, double, __VA_ARGS__)\
+	MATRIX_MULTIPLY_FUNCTION(FUNCTION_BASE##_f, float,  __VA_ARGS__)
+
+MATRIX_MULTIPLY_FUNCTIONS(yuv_to_srgb,
+			  1,
+			  0.00028328010485821202317155420580263580632163211703,
+			  1.14070449590558520291949662350816652178764343261719,
+			  1,
+			  -0.39630886669497211727275498560629785060882568359375,
+			  -0.58107364288228224857846271333983168005943298339844,
+			  1,
+			  2.03990003507541306504435851820744574069976806640625,
+			  0.00017179031692307700847528739718228507626918144524)
+
+MATRIX_MULTIPLY_FUNCTIONS(srgb_to_yuv,
+			  0.299, 0.587, 0.114,
+			  -0.14662756598240470062854967636667424812912940979004,
+			  -0.28771586836102963635752871596196200698614120483398,
+			   0.43434343434343436474165400795754976570606231689453,
+			   0.61456892577224520035628074765554629266262054443359,
+			  -0.51452282157676354490405401520547457039356231689453,
+			  -0.10004610419548178035231700278018251992762088775635)
+
+MATRIX_MULTIPLY_FUNCTIONS(ciexyz_to_srgb,
+			   3.240446254647737500675930277794,
+			  -1.537134761820080575134284117667,
+			  -0.498530193022728718155178739835,
+			  -0.969266606244679751469561779231,
+			   1.876011959788370209167851498933,
+			   0.041556042214430065351304932619,
+			   0.055643503564352832235773149705,
+			  -0.204026179735960239147729566866,
+			   1.057226567722703292062647051353)
+
+MATRIX_MULTIPLY_FUNCTIONS(srgb_to_ciexyz,
+			  0.412457445582367576708548995157,
+			  0.357575865245515878143578447634,
+			  0.180437247826399665973085006954,
+			  0.212673370378408277403536885686,
+			  0.715151730491031756287156895269,
+			  0.072174899130559869164791564344,
+			  0.019333942761673460208893260415,
+			  0.119191955081838593666354597644,
+			  0.950302838552371742508739771438)
+
+MATRIX_MULTIPLY_FUNCTIONS(scaled_yuv_to_ciexyz,
+			   0.00001450325106667098632156481796684488472237717360,
+			   0.00000345586790639342739093228633329157872822179343,
+			   0.00000400923398630552893485111398685916128670214675,
+			   0.00001525902189669641837040624243737596543724066578,
+			  -0.00000207722814409390653614547427030512238843584782,
+			  -0.00000263898607692305410302407824019166326934282552,
+			   0.00001661446153041708825425643025752719950105529279,
+			   0.00002885925752619118069149627137104374696718878113,
+			  -0.00000071781086875769179526501342566979779746816348)
+
+MATRIX_MULTIPLY_FUNCTIONS(ciexyz_to_scaled_yuv,
+			    26625.38231027395886485464870929718017578125,
+			    40524.0090949436053051613271236419677734375,
+			     -271.5313105642117079696618020534515380859375,
+			   -11278.3751445417292416095733642578125,
+			   -26409.91773157499847002327442169189453125,
+			    34100.5706543184860493056476116180419921875,
+			   162829.60100012840121053159236907958984375,
+			  -123829.313212639070115983486175537109375,
+			   -28411.65702312920984695665538311004638671875)
+
+#undef MATRIX_MULTIPLY_FUNCTIONS
+#undef MATRIX_MULTIPLY_FUNCTION
diff --git a/src/util/io.h b/src/util/io.h
@@ -40,10 +40,10 @@ enreadall(int status, int fd, void *buf, size_t n, const char *fname)
 	return (size_t)r;
 }
 
-int pwriteall(int fd, void *buf, size_t n, size_t ptr);
+int pwriteall(int fd, void *buf, size_t n, off_t ptr);
 
 static inline void
-enpwriteall(int status, int fd, void *buf, size_t n, size_t ptr, const char *fname)
+enpwriteall(int status, int fd, void *buf, size_t n, off_t ptr, const char *fname)
 {
 	if (pwriteall(fd, buf, n, ptr))
 		enprintf(status, "pwrite %s:", fname);
diff --git a/src/video-math.h b/src/video-math.h
@@ -0,0 +1,90 @@
+/* See LICENSE file for copyright and license details. */
+#include <math.h>
+
+static inline double
+nnpow(double a, double b)
+{
+	int neg = a < 0;
+	a = pow(neg ? -a : a, b);
+	return neg ? -a : a;
+}
+
+static inline float
+nnpowf(float a, float b)
+{
+	int neg = a < 0;
+	a = powf(neg ? -a : a, b);
+	return neg ? -a : a;
+}
+
+#define GENERIC(TYPE, FUNC, ...)\
+	TYPE:           FUNC(__VA_ARGS__),\
+	TYPE *:         FUNC(__VA_ARGS__),\
+	TYPE **:        FUNC(__VA_ARGS__),\
+	TYPE ***:       FUNC(__VA_ARGS__),\
+	const TYPE:     FUNC(__VA_ARGS__),\
+	const TYPE *:   FUNC(__VA_ARGS__),\
+	const TYPE **:  FUNC(__VA_ARGS__),\
+	const TYPE ***: FUNC(__VA_ARGS__)
+
+#define MATH_GENERIC_1(FUNC, A)       (_Generic((A),\
+						GENERIC(double, FUNC,    A),\
+						GENERIC(float,  FUNC##f, A)))
+
+#define MATH_GENERIC_N(FUNC, A, ...)  (_Generic((A),\
+						GENERIC(double, FUNC,    A, __VA_ARGS__),\
+						GENERIC(float,  FUNC##f, A, __VA_ARGS__)))
+
+#define BLIND_GENERIC_1(FUNC, A)      (_Generic((A),\
+						GENERIC(double, FUNC##_d, A),\
+						GENERIC(float,  FUNC##_f, A)))
+
+#define BLIND_GENERIC_N(FUNC, A, ...) (_Generic((A),\
+						GENERIC(double, FUNC##_d, A, __VA_ARGS__), \
+						GENERIC(float,  FUNC##_f, A, __VA_ARGS__)))
+
+#define pow(...)         MATH_GENERIC_N(pow,      __VA_ARGS__)
+#define log2(...)        MATH_GENERIC_1(log2,     __VA_ARGS__)
+#define abs(...)         MATH_GENERIC_1(fabs,     __VA_ARGS__)
+#define sqrt(...)        MATH_GENERIC_1(sqrt,     __VA_ARGS__)
+#define exp(...)         MATH_GENERIC_1(exp,      __VA_ARGS__)
+#define g_isnan(...)     MATH_GENERIC_1(isnan,    __VA_ARGS__)
+#define g_isinf(...)     MATH_GENERIC_1(isinf,    __VA_ARGS__)
+#define g_isfinite(...)  MATH_GENERIC_1(isfinite, __VA_ARGS__)
+#define nnpow(...)       MATH_GENERIC_N(nnpow,    __VA_ARGS__)
+
+#define srgb_encode(...) BLIND_GENERIC_1(srgb_encode, __VA_ARGS__)
+#define srgb_decode(...) BLIND_GENERIC_1(srgb_decode, __VA_ARGS__)
+
+#define yuv_to_srgb(a, b, c, d, e, f)\
+	BLIND_GENERIC_N(yuv_to_srgb, (a), (b), (c), (void *)(d), (void *)(e), (void *)(f))
+#define srgb_to_yuv(a, b, c, d, e, f)\
+	BLIND_GENERIC_N(srgb_to_yuv, (a), (b), (c), (void *)(d), (void *)(e), (void *)(f))
+#define ciexyz_to_srgb(a, b, c, d, e, f)\
+	BLIND_GENERIC_N(ciexyz_to_srgb, (a), (b), (c), (void *)(d), (void *)(e), (void *)(f))
+#define srgb_to_ciexyz(a, b, c, d, e, f)\
+	BLIND_GENERIC_N(srgb_to_ciexyz, (a), (b), (c), (void *)(d), (void *)(e), (void *)(f))
+#define scaled_yuv_to_ciexyz(a, b, c, d, e, f)\
+	BLIND_GENERIC_N(scaled_yuv_to_ciexyz, (a), (b), (c), (void *)(d), (void *)(e), (void *)(f))
+#define ciexyz_to_scaled_yuv(a, b, c, d, e, f)\
+	BLIND_GENERIC_N(ciexyz_to_scaled_yuv, (a), (b), (c), (void *)(d), (void *)(e), (void *)(f))
+
+#define htole(A) (_Generic((A),\
+			   uint8_t: (A),\
+			    int8_t: (uint8_t)(A),\
+			   uint16_t: htole16(A),\
+			    int16_t: (uint16_t)htole16((uint16_t)(A)),\
+			   uint32_t: htole32(A),\
+			    int32_t: (uint32_t)htole32((uint32_t)(A)),\
+			   uint64_t: htole64(A),\
+			    int64_t: (uint64_t)htole64((uint64_t)(A))))
+
+#define letoh(A) (_Generic((A),\
+			   uint8_t: (A),\
+			    int8_t: (uint8_t)(A),\
+			   uint16_t: le16toh(A),\
+			    int16_t: (uint16_t)le16toh((uint16_t)(A)),\
+			   uint32_t: le32toh(A),\
+			    int32_t: (uint32_t)le32toh((uint32_t)(A)),\
+			   uint64_t: le64toh(A),\
+			    int64_t: (uint64_t)le64toh((uint64_t)(A))))