1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
|
{ gccStdenv, lib, pkgs,
git, openssl, autoconf, coreutils,
src, version, git-version,
stampYmd ? 0, stampHms ? 0,
gambit-support,
optimizationSetting ? "-O1",
gambit-params ? pkgs.gambit-support.stable-params,
rev ? git-version }:
# Note that according to a benchmark run by Marc Feeley on May 2018,
# clang is 10x (with default settings) to 15% (with -O2) slower than GCC at compiling
# Gambit output, producing code that is 3x slower. IIRC the benchmarks from Gambit@30,
# the numbers were still heavily in favor of GCC in October 2019.
# Thus we use GCC over clang, even on macOS.
#
# Also note that I (fare) just ran benchmarks from https://github.com/ecraven/r7rs-benchmarks
# with Gambit 4.9.3 with -O1 vs -O2 vs -Os on Feb 2020. Which wins depends on the benchmark.
# The fight is unclear between -O1 and -O2, where -O1 wins more often, by up to 17%,
# but sometimes -O2 wins, once by up to 43%, so that overall -O2 is 5% faster.
# However, -Os seems more consistent in winning slightly against both -O1 and -O2,
# and is overall 15% faster than -O2. As for compile times, -O1 is fastest,
# -Os is about 29%-33% slower than -O1, while -O2 is about 40%-50% slower than -O1.
#
# Overall, -Os seems like the best choice, but I care more about compile-time,
# so I stick with -O1 (in the defaults above), which is also the default for Gambit.
gccStdenv.mkDerivation rec {
pname = "gambit";
inherit src version git-version;
bootstrap = gambit-support.gambit-bootstrap;
passthru = {
inherit src version git-version rev stampYmd stampHms optimizationSetting openssl;
};
nativeBuildInputs = [ git autoconf ];
# TODO: if/when we can get all the library packages we depend on to have static versions,
# we could use something like (makeStaticLibraries openssl) to enable creation
# of statically linked binaries by gsc.
buildInputs = [ openssl ];
# TODO: patch gambit's source so it has the full path to sed, grep, fgrep? Is there more?
# Or wrap relevant programs to add a suitable PATH ?
#runtimeDeps = [ gnused gnugrep ];
configureFlags = [
"--enable-targets=${gambit-params.targets}"
"--enable-single-host"
"--enable-c-opt=${optimizationSetting}"
"--enable-c-opt-rts=-O2"
"--enable-gcc-opts"
"--enable-trust-c-tco"
"--enable-shared"
"--enable-absolute-shared-libs" # Yes, NixOS will want an absolute path, and fix it.
"--enable-openssl"
"--enable-dynamic-clib"
#"--enable-default-compile-options='(compactness 9)'" # Make life easier on the JS backend
"--enable-default-runtime-options=${gambit-params.defaultRuntimeOptions}"
# "--enable-rtlib-debug" # used by Geiser, but only on recent-enough gambit, and messes js runtime
# "--enable-debug" # Nope: enables plenty of good stuff, but also the costly console.log
# "--enable-multiple-versions" # Nope, NixOS already does version multiplexing
# "--enable-guide"
# "--enable-track-scheme"
# "--enable-high-res-timing"
# "--enable-max-processors=4"
# "--enable-multiple-vms"
# "--enable-dynamic-tls"
# "--enable-multiple-threaded-vms" # when SMP branch is merged in
# "--enable-thread-system=posix" # default when --enable-multiple-vms is on.
# "--enable-profile"
# "--enable-coverage"
# "--enable-inline-jumps"
# "--enable-char-size=1" # default is 4
# "--enable-march=native" # Nope, makes it not work on machines older than the builder
] ++ gambit-params.extraOptions
# TODO: pick an appropriate architecture to optimize on on x86-64?
# https://gcc.gnu.org/onlinedocs/gcc-4.8.4/gcc/i386-and-x86-64-Options.html#i386-and-x86-64-Options
# ++ lib.optional pkgs.stdenv.hostPlatform.isx86_64 "--enable-march=core-avx2"
# Do not enable poll on darwin due to https://github.com/gambit/gambit/issues/498
++ lib.optional (!gccStdenv.hostPlatform.isDarwin) "--enable-poll";
configurePhase = ''
export CC=${gccStdenv.cc}/bin/${gccStdenv.cc.targetPrefix}gcc \
CXX=${gccStdenv.cc}/bin/${gccStdenv.cc.targetPrefix}g++ \
CPP=${gccStdenv.cc}/bin/${gccStdenv.cc.targetPrefix}cpp \
CXXCPP=${gccStdenv.cc}/bin/${gccStdenv.cc.targetPrefix}cpp \
LD=${gccStdenv.cc}/bin/${gccStdenv.cc.targetPrefix}ld \
XMKMF=${coreutils}/bin/false
unset CFLAGS LDFLAGS LIBS CPPFLAGS CXXFLAGS
${gambit-params.fixStamp git-version stampYmd stampHms}
./configure --prefix=$out/gambit ${builtins.concatStringsSep " " configureFlags}
# OS-specific paths are hardcoded in ./configure
substituteInPlace config.status \
${lib.optionalString (gccStdenv.hostPlatform.isDarwin && !gambit-params.stable)
''--replace "/usr/local/opt/openssl@1.1" "${lib.getLib openssl}"''} \
--replace "/usr/local/opt/openssl" "${lib.getLib openssl}"
./config.status
'';
buildPhase = ''
# The MAKEFLAGS setting is a workaround for https://github.com/gambit/gambit/issues/833
export MAKEFLAGS="--output-sync=recurse"
echo "Make bootstrap compiler, from release bootstrap"
mkdir -p boot
cp -rp ${bootstrap}/gambit/. boot/.
chmod -R u+w boot
cd boot
cp ../gsc/makefile.in ../gsc/*.scm gsc/
echo > include/stamp.h # No stamp needed for the bootstrap compiler
./configure
for i in lib gsi gsc ; do (cd $i ; make -j$NIX_BUILD_CORES) ; done
cd ..
cp boot/gsc/gsc gsc-boot
echo "Now use the bootstrap compiler to build the real thing!"
make -j$NIX_BUILD_CORES from-scratch
${lib.optionalString gambit-params.modules "make -j$NIX_BUILD_CORES modules"}
'';
postInstall = ''
mkdir $out/bin
cd $out/bin
ln -s ../gambit/bin/* .
'';
doCheck = true;
dontStrip = true;
meta = gambit-support.meta;
}
|
