~sandmann/x86info - Unnamed repository; edit this file 'description' to name the repository.

Age	Commit message (Collapse)	Author	Files	Lines
2011-02-22	move MHz display to mhz.c	Dave Jones	1	-0/+12

2009-12-11	Only calculate MHz on the first cpu, unless --all-cpus is passed.	Dave Jones	1	-0/+7

2009-10-26	Do the rdtsc cleanup how I meant to do it the first time.	Dave Jones	1	-3/+1

2009-10-26	Revert "cleanup the rdtsc type usage a little."	Dave Jones	2	-4/+6
	This reverts commit 3aa21b5d657644980e5acd1ae87461b46bb24790.
2009-10-26	handle rdtsc wrapping.	Dave Jones	1	-1/+6

2009-10-26	cleanup the rdtsc type usage a little.	Dave Jones	2	-6/+4

2008-12-24	Fix up warnings. Add -Werror to catch new ones.	Dave Jones	1	-1/+1
	The lsmsr ones are kind of brute-force fixes because I'm lazy right now. We should possibly just use unsigned long longs everywhere, instead of all those u_int64_t's.
2008-08-29	Only do CPU binding once per cpu.	Dave Jones	2	-4/+0

2008-07-02	Factor out cpu binding to a common function.	Dave Jones	2	-22/+2

2008-07-02	Make the benchmarks code work again	Dave Jones	3	-28/+34

2008-07-02	Improve MHz estimation.	Dave Jones	1	-23/+14
	The fatal flaw was that rdtsc was wrapping too soon as we were only using 32bits of accuracy. Extending this to 64bit increases the resolution of the timing dramatically. We can also do away with all the other timing code as a bonus.
2008-07-02	The estimate MHz routine is utterly busted.	Dave Jones	1	-13/+35
	At the least it should be binding to a CPU. Also, the usleep doesn't make the CPU busy, so replace it with a loop and wake up on an alarm after 3 seconds. It's still wildly inaccurate. Not sure what to do yet, but this is no more broken than it already was.
2008-07-02	* Vastly rework the core cpuid handling on uniprocessor.	Dave Jones	1	-1/+1
	- No more nasty pages and pages of assembly - x86-64 'just works' * cpuid takes 32 bit registers. always. wtf were we passing around longs? * implement cpuid4() * start using cpuid4 to determine number of cores.
2008-07-01	ridiculous amount of whitespace cleanups	Dave Jones	1	-2/+2

2008-07-01	Remove CVS tags from headers, remove pointless crap, and dead urls.	Dave Jones	1	-2/+0
	general cleanup
2007-06-06	Solaris support.	John Levon	1	-3/+8

2006-05-21	Clean up warnings from sparse	Dave Jones	2	-4/+3

2005-08-27	Remove support for non Linux OS's.	Dave Jones	1	-39/+0
	The ifdefs really were ugly, and I've not heard a report from a single user of this code since it was merged.
2003-06-09	* bench/MHz.c: Rounding up to nearest 50MHz if MHz is within a 15MHz	davej	1	-1/+8
	discrepancy of the nearest 50. Crap, needs improving.
2003-05-02	* identify.c: Display MHz as GHz if >1000. Always do MHz calculation,	davej	1	-8/+7
	even if no --MHz is specified * Mhz.c: Store results in cpu struct. Use integer instead of fp.
2003-03-14	* bench/bench.h: improve timing routine a little. still needs work.	davej	2	-8/+12
	* bench/benchmarks.c: remove lea (eax), 1(eax) test for now.
2003-02-07	benchmark bswap	davej	1	-2/+3

2003-01-28	Remove dud protos	davej	1	-3/+0

2003-01-27	Drop broken benchmarks for now	davej	1	-3/+0

2003-01-27	Add nehemiah	davej	1	-0/+4

2003-01-23	couple of extra benchmarks	davej	1	-1/+3

2003-01-18	Various cleanups and addition of locked add benchmark	davej	4	-61/+31

2002-12-22	* bench/cpuid.c: Benchmark cpuid.	davej	4	-1/+35
	* bench/int80h.c: Report time in int not float.
2002-12-18	split the int 0x80 and new sysenter stuff.	davej	4	-13/+40
	add --bench
2002-12-18	new directory for benchmark type things	davej	2	-0/+135