commit e5fb99c18c8f83364e7826c7a722ca8369ae3de4
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
64/x86_64/amd/k8/k8_64: Added '-mfpmath=sse' and others
Switching to sse is said to be considerably faster and more stable,
but may break existing code.
Note that the default math unit (-mfpmath) is 387 for ia32.
Switching to sse is said to be considerably faster and more stable,
but may break existing code.
commit f95be7933a08ee5856df4950266074bb122ab353
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
32/ia32/intel/em64t/core2_32: Add '-mfpmath=sse' and others
Note that the default math unit (-mfpmath) is 387 for ia32.
Switching to sse is said to be considerably faster and more stable,
but may break existing code.
Note that the default math unit (-mfpmath) is 387 for ia32. Switching
to sse is said to be considerably faster and more stable, but may
break existing code.
Note that the default math unit (-mfpmath) is 387 for ia32. Switching
to sse is said to be considerably faster and more stable, but may break
existing code.
commit 73dd1bb32c9d8c1963c391b658968ae6486f4964
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
32/ia32/generic/x86_64/x86_32: Re-add specific flags
Note that the default math unit (-mfpmath) is 387 for ia32. Switching
to sse is said to be considerably faster and more stable, but may break
existing code.
commit c3c0de02e5aea37548c10b01b1e080b28f671fc1
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
64/x86_64/native-x86_64-64: Native architecture selection for x86_64
For GCC 4.2+, use '-march=native -mtune=native', or all but GCC2,
use '-march=x86_64", and for GCC2 use '-march=i486"
march=i486 was chosen as this is the lowest the ISO goes and the
lowest required for NPTL in glibc.
commit 9c7e7944f4d132d18242b6bbba4fddb5ed7cf31f
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
32/ia32/native-x86_64-32: Native architecture selection for x86/x86_64
For GCC 4.2+, use '-march=native -mtune=native', for everything else,
use '-march=i486".
march=i486 was chosen as this is the lowest the ISO goes and the lowest
required for NPTL in glibc.
commit 52fdf10756a800b5285de006f5857de4b9ee39eb
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
32/ia32/intel/em64t/nocona_32: Added '-mmmx -mfpmath=sse -msse -msse2
-msse3' for GCC 4.3
Note that the default math unit (-mfpmath) is 387 for ia32. Switching to
sse is said to be considerably faster and more stable, but may break
existing code.
Note that the default math unit (-mfpmath) is 387 for ia32. Switching to
sse is said to be considerably faster and more stable, but may break
existing code.
commit 5b71e78ae116c6544e10fe8600caf9ab228575d3
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
.mailmap: Add eric AT sandall.us map
commit e31466200b60bad999a47b529479410f6d524502
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
.mailmap: Add mail mapping file
commit 2d9db7c5f066d330b94d5c324dc6a8835de1c79d
Author: Eric Sandall <sandalle AT sourcemage.org>
Commit: Eric Sandall <sandalle AT sourcemage.org>
Make core2_{32,64} future proof (e.g. for gcc 4.4), thanks Jaka
This does mean that if you are using a compiler version which we do not
check for, that Sorcery will use the GCC 4.3+ -march=core2 flag and not
fail back to -march=nocona (which your non-detected compiler version
probably doesn't have anyways).
This moves the Core 2 support from nocona (lower class processor to the
Core 2) to its own file. GCC 4.3.x+ supports -march=core2.
commit 201eca962d3faf4e0e413ae926f6e29ea9d0bd30
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
This -m32 is necessary!
commit 5c4b903551b64d7000fcd7eba3e2e5a93561aaab
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
Merging the minmal archspecs with the main ones (omitting unneeded flags)
and clearing unneeded -m64 or -m32 flags.
Some drive-by fixing like enhancing flags for gcc-3.3 since it can handle
most modern -marchs now.
Noted the Core and Core 2 processors in prescott and nocona archs.
commit 78bd82efedfe174882a4add6238050d20b2ff448
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
being straight and applying the 64 bit tidyness here too:
- no -m64; the HOST should trigger build a 64 bit compiler, that one
doesn't need -m64
- not using 32 bit march on 64 bit
- but not yet deleting gcc2 entry, despite of realizing that the HOST
setting would mess up there in addition of it gcc2 being useless in pure 64
bit
commit 585186f6cc76070c14132fdf93f0d4d567e0c066
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
applied suggestions from bug 10198; removing possibly counterproductive
flags (march=athlon-xp; forced m64) for k8_64 ...
commit 29eb8e882282c78b201d6227d333c47283392f19
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
host is powerpc64-unknown-linux-gnu, see bug 8649
I wonder how well a change of hostspec works for a running system...
commit ea549a6732c37bb76747baa51f28bec05a7b1143
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
changing the host to powerpc-unknown-linux-gnu to match the scheme used
elsewhere, from bug 8649
commit 40b72dd4ac54d10da843d25ba205083142b79d84
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
adding Pentium M archspecs, difference between minimal and normal is just
presence of -mfpmath=sse (the only one of the usual -m flags that has an
effect together with -march)
commit d14c6da7b7745b6b5bd4b53d07014432d0b9ff37
Author: Thomas Orgis <sobukus AT sourcemage.org>
Commit: Thomas Orgis <sobukus AT sourcemage.org>
commit 215f2a91b45719bd60dbfe9db481c61ffd7ca964
Author: root <root AT adams.agnld.uni-potsdam.de>
Commit: root <root AT adams.agnld.uni-potsdam.de>
less alpha in dir names (now vendor dir dec), more in file names (alpha-
prefix)
commit 0083a15248825f2291838fc9b351fed9a6356dcd
Author: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
Commit: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
add prescott archspec #13032
thanks go to Elisamuel Resto
commit 7794c7110184de53d40e1fbfd368ca36caffb51a
Author: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
Commit: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
forgot to make the alpha archspecs executable
commit 9de395819a73e70357a679ed940b928e821d9817
Author: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
Commit: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
fixed #12715 - new archspecs: alpha
I modelled the directory structure according to the rest, but
half of it has no rational base. If someone has a better layout in mind
...
this one is for sobokus :)
commit 49ffc7c0573ceecb8a2638900e5414f536877103
Author: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>
Commit: Jaka Kranjc <lynxlynxlynx AT sourcemage.org>