[coreboot] [HEADS UP] Time stamps for AGESA boards now measured relatively to start and not ramstage

[Paul Menzel]

Dear coreboot folks,

thanks to the latest time stamp work and especially thanks to Aaron
Durbin’s latest fix up commit bd1499d3 (timestamps: don't drop ramstage
timestamps with EARLY_CBMEM_INIT) [1], the ramstage time stamps on
AGESA boards – I only tested with the ASRock E350M1 – seem to be
measured relative to the system start and not relatively to start of
ramstage.

Before time stamps were measured relatively to ramstage.

$ more asrock/e350m1/4.0-10267-g72dd909/2015-07-10T00:11:47Z/coreboot_timestamps.txt
12 entries total:

10:start of ramstage                                 0
  30:device enumeration                                8 (8)
  40:device configuration                              94,880 (94,872)
  50:device enable                                     98,697 (3,816)
  60:device initialization                             108,662 (9,965)
  70:device setup done                                 122,998 (14,335)
  75:cbmem post                                        123,369 (371)
  80:write tables                                      123,374 (4)
  90:load payload                                      128,020 (4,646)
  15:starting LZMA decompress (ignore for x86)         128,275 (254)
  16:finished LZMA decompress (ignore for x86)         146,142 (17,866)
  99:selfboot jump                                     146,163 (21)

With the latest changes we they are measured relatively to system start.

$ more asrock/e350m1/4.0-10270-gbd1499d/2015-07-10T13:23:53Z/coreboot_timestamps.txt
12 entries total:

10:start of ramstage                                 385,974
  30:device enumeration                                385,982 (8)
  40:device configuration                              480,233 (94,250)
  50:device enable                                     484,088 (3,855)
  60:device initialization                             494,049 (9,960)
  70:device setup done                                 508,368 (14,318)
  75:cbmem post                                        508,736 (368)
  80:write tables                                      508,741 (4)
  90:load payload                                      513,320 (4,579)
  15:starting LZMA decompress (ignore for x86)         513,574 (253)
  16:finished LZMA decompress (ignore for x86)         531,423 (17,848)
  99:selfboot jump                                     531,445 (21)

So 386 ms for romstage with AGESA is still an acceptable number. This
confirms, which was already known. AGESA does a pretty good job.

Time stamps from romstage are still not preserved though, which is a
well known limitation though.

Thanks,

Paul

[1] http://review.coreboot.org/10880