Arthur Heymans has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/37198 )
Change subject: nb/amd/agesa: select ROMSTAGE_CACHED_CBMEM ......................................................................
Patch Set 7:
Patch Set 7:
On the ASRock E350M1:
4.12-374-g0013623b7c:
27 entries total: 0:1st timestamp 7,931 11:start of bootblock 14,414 (6,482) 12:end of bootblock 15,624 (1,210) 13:starting to load romstage 16,214 (589) 14:finished loading romstage 16,301 (87) 1:start of romstage 16,605 (303) 2:before ram initialization 129,739 (113,134) 3:after ram initialization 411,357 (281,617) 4:end of romstage 426,279 (14,921) 100:start of postcar 426,418 (138) 101:end of postcar 426,418 (0) 8:starting to load ramstage 426,528 (109) 15:starting LZMA decompress (ignore for x86) 426,564 (36) 16:finished LZMA decompress (ignore for x86) 518,200 (91,636) 9:finished loading ramstage 518,247 (46) 10:start of ramstage 518,269 (21) 30:device enumeration 522,196 (3,926) 40:device configuration 659,462 (137,266) 50:device enable 672,386 (12,924) 60:device initialization 672,859 (473) 70:device setup done 687,333 (14,474) 75:cbmem post 932,112 (244,778) 80:write tables 932,113 (1) 85:finalize chips 973,568 (41,454) 90:load payload 973,572 (4) 15:starting LZMA decompress (ignore for x86) 974,055 (483) 16:finished LZMA decompress (ignore for x86) 1,027,445 (53,389) 99:selfboot jump 1,027,502 (57) Total Time: 1,019,558
Your change-set rebased on origin/master (`cpu/x86/cache.h` added for `clflush_supported()`):
28 entries total: 0:1st timestamp 7,886 11:start of bootblock 13,556 (5,670) 12:end of bootblock 14,705 (1,148) 13:starting to load romstage 15,285 (580) 14:finished loading romstage 15,371 (86) 1:start of romstage 15,652 (280) 2:before ram initialization 129,582 (113,930) 3:after ram initialization 414,538 (284,955) 17:starting LZ4 decompress (ignore for x86) 421,713 (7,175) 18:finished LZ4 decompress (ignore for x86) 427,201 (5,487) 100:start of postcar 428,672 (1,471) 101:end of postcar 428,672 (0) 8:starting to load ramstage 428,783 (110) 15:starting LZMA decompress (ignore for x86) 428,819 (36) 16:finished LZMA decompress (ignore for x86) 520,412 (91,593) 9:finished loading ramstage 520,459 (46) 10:start of ramstage 520,480 (21) 30:device enumeration 524,407 (3,927) 40:device configuration 661,704 (137,297) 50:device enable 674,636 (12,931) 60:device initialization 675,105 (468) 70:device setup done 689,625 (14,520) 75:cbmem post 953,917 (264,291) 80:write tables 953,918 (1) 85:finalize chips 995,266 (41,348) 90:load payload 995,271 (4) 15:starting LZMA decompress (ignore for x86) 995,755 (483) 16:finished LZMA decompress (ignore for x86) 1,049,208 (53,453) 99:selfboot jump 1,049,265 (57) Total Time: 1,041,368
Looks like this improves loading postcar stage with ~1ms. Thanks for testing.