Werner Almesberger
|
71cdeb80df
|
m1/perf/favg: tool to extract values from ./tabulate -h and to calculate averages
|
2011-09-24 20:34:09 -03:00 |
|
Werner Almesberger
|
3da0074896
|
m1/perf/sched.c: if register allocation fails, return an error instead of aborting
|
2011-09-24 20:33:09 -03:00 |
|
Werner Almesberger
|
bb8c1ce741
|
m1/perf/sched.c: tried and rejected more accurate LCPF distance heuristics
They turned out to produce the same or even slightly worse results.
|
2011-09-24 20:20:04 -03:00 |
|
Werner Almesberger
|
9db83ae72c
|
m1/perf/sched.c: removed "assert" hack (since we now use regular libc)
|
2011-09-24 09:38:55 -03:00 |
|
Werner Almesberger
|
b2e2364231
|
m1/perf/Makefile: use libfpvm/x86-linux/Makefile
|
2011-09-24 09:37:49 -03:00 |
|
Werner Almesberger
|
af8935f5be
|
m1/perf/tabular: print total run time at the end
|
2011-09-23 22:19:36 -03:00 |
|
Werner Almesberger
|
e2c133e056
|
m1/perf/Makefile: use -fno-builtin, use profiling libc, set -DPRINTF_FLOAT
|
2011-09-23 22:18:31 -03:00 |
|
Werner Almesberger
|
0f80dcf118
|
m1/perf/sched.c: don't try to unblock references of next setter twice (more)
Also changed:
- added commented-out #define DEBUG
- work-around for abort() being a no-op in libbase
|
2011-09-23 22:15:11 -03:00 |
|
Werner Almesberger
|
69cfc407d5
|
m1/perf/tabulate: added HTML mode with highlighting (-h)
|
2011-09-22 03:23:15 -03:00 |
|
Werner Almesberger
|
c5df76a7b9
|
m1/perf/: calculate effective register allocation and how it in comparison table
|
2011-09-22 02:07:42 -03:00 |
|
Werner Almesberger
|
b3796c855f
|
m1/perf/eval.pl: process both programs (frame and vertex) and dump all assigned-to fixed registers
|
2011-09-22 01:44:05 -03:00 |
|
Werner Almesberger
|
403917a034
|
m1/perf/try: disable profiling
|
2011-09-21 18:00:26 -03:00 |
|
Werner Almesberger
|
2bc4c33814
|
m1/perf/sched.c (schedule): reduced conditionally compiled code size for optimizer
|
2011-09-21 18:00:26 -03:00 |
|
Werner Almesberger
|
e98f08e857
|
m1/perf/sched.c: light code cleanup
|
2011-09-21 18:00:26 -03:00 |
|
Werner Almesberger
|
b177acd632
|
m1/perf/sched.c (foreach): improved type safety
|
2011-09-21 17:50:24 -03:00 |
|
Werner Almesberger
|
23c7bd6ce1
|
m1/perf/: "spotless" target that also removes the "main" executable
|
2011-09-21 17:48:18 -03:00 |
|
Werner Almesberger
|
a19a196785
|
m1/perf/main.c: exit immediately on compilation error
|
2011-09-21 17:46:25 -03:00 |
|
Werner Almesberger
|
8f82a0e8d4
|
m1/perf/sched.c: revamped to handle static registers correctly as well
|
2011-09-21 17:44:52 -03:00 |
|
Werner Almesberger
|
1ad7db70f1
|
m1/perf/sched.c: slight cleanup (no functional changes)
|
2011-09-19 11:32:29 -03:00 |
|
Werner Almesberger
|
e34e8ebb16
|
m1/perf/sched.c: fixed compilation with DEBUG enabled
|
2011-09-18 19:14:13 -03:00 |
|
Werner Almesberger
|
60cf6f5e9a
|
m1/perf/sched.c: added asserts and list poisoning
|
2011-09-18 19:11:17 -03:00 |
|
Werner Almesberger
|
16b4a9a55b
|
m1/perf/tabulate: some small improvements and corrections
|
2011-09-18 12:36:09 -03:00 |
|
Werner Almesberger
|
3727324a2c
|
m1/perf/: scripts for bulk test runs and tabulating the results
|
2011-09-18 11:06:35 -03:00 |
|
Werner Almesberger
|
c02c02903d
|
m1/perf/: a bunch of bug fixes all over the place
|
2011-09-18 10:14:12 -03:00 |
|
Werner Almesberger
|
5dc1c6de36
|
m1/perf/: added LCPF scheduler
|
2011-09-18 07:45:17 -03:00 |
|
Werner Almesberger
|
abd7de7265
|
m1/perf/: added instrumentation for monitoring register allocation
|
2011-09-18 06:56:00 -03:00 |
|
Werner Almesberger
|
b4abaffa5c
|
m1/perf/: allocate scheduler context (about 0.5 MB) dynamically
|
2011-09-18 06:18:25 -03:00 |
|
Werner Almesberger
|
c009a13f5c
|
m1/perf/: cleaned up sched.c a little; added TODO
|
2011-09-18 01:28:02 -03:00 |
|
Werner Almesberger
|
64e2a3c6b7
|
m1/perf/: scheduler cleanup; output comparison tool
|
2011-09-18 00:51:17 -03:00 |
|
Werner Almesberger
|
9a5a22eda5
|
m1/perf/: main.c was missing. Added on-going work.
|
2011-09-17 23:02:34 -03:00 |
|
Werner Almesberger
|
001ca49cc6
|
m1/perf/: build the patch compiler with gcc and with profiling
|
2011-09-17 03:24:08 -03:00 |
|
Werner Almesberger
|
cc2b07ba16
|
m1/jtag-boot/: added "standby" (by Cristian Paul Penaranda Rojas)
|
2011-09-12 00:17:23 -03:00 |
|
Werner Almesberger
|
47f13281fe
|
m1/jtag-boot/mkboot: eradicate traces of time travel (reported by Michael Walle)
|
2011-09-08 19:07:55 -03:00 |
|
Werner Almesberger
|
757f19cd7b
|
moved m1/torture/ to more specific m1rc3/norruption/
|
2011-09-07 04:23:32 -03:00 |
|
Werner Almesberger
|
29677f4707
|
m1/torture/LOG: activity log (with various build instructions)
|
2011-09-07 04:20:08 -03:00 |
|
Werner Almesberger
|
bcd12fe644
|
m1/jtag-boot/: moved explanation from mkboot to README and added more details
|
2011-09-06 18:54:39 -03:00 |
|
Werner Almesberger
|
f3b289413d
|
m1/jtag-boot/mkboot: added comments with pointers to information sources
|
2011-09-06 12:29:31 -03:00 |
|
Werner Almesberger
|
96a5dd3b60
|
m1/torture/: quick and dirty power-cycling test loop
|
2011-09-06 12:09:56 -03:00 |
|
Werner Almesberger
|
2c896d860d
|
m1/jtag-boot/Makefile: run UrJTAG in quiet mode
|
2011-09-06 11:59:09 -03:00 |
|
Werner Almesberger
|
2ef6c7e13f
|
m1/jtag-boot/boot.bit: removed generated file, oops
|
2011-09-06 11:38:54 -03:00 |
|
Werner Almesberger
|
87aa1e7345
|
m1/jtag-boot/: quick hack to boot an M1 with the "regular" bitstream (from NOR)
|
2011-09-06 11:37:01 -03:00 |
|