I've tried to keep the changes isolated to a single commit, so feel free to push back on any changes, they can be reverted individually pretty easily. The changes are:
Have the early-boot dummies for string-printing EFP functions still update A0 to point to the end of the string, as documented.
Remove the INIT_SSP constant that's not used anywhere anymore. (The initial SSP was previously set to be below the stage2 loading location instead, so this value wasn't valid anymore.)
Tweak to remove a magic number shift for C_DEVICE_SIZE. (This one isn't particularly important, actually less efficient, and shouldn't apply unless the CHAR_DEVICE structure is grown in the future.)
(2 commits) Move some memory layout constants out from code and into linker scripts. (The main significant one is RAM_TOP, since it was using an ifd on REVISION1X which ignored the actual ROM size and memory layout.
Minor change that addresses a comment about it being inefficient to use sz-strings in ANSI code printing, this replaces the single-character sequences with single-character prints.
There were two copies of serial.asm, one for each stage. The version in stage1 unused except for RECVCHAR, which I folded into trap14.asm (along with the other functions), since that's where the other EFP wrappers live as FW_<...>.
Fix one of the linker script changes from above only making it into the 1M version.
Fix a Makefile issue that I noticed while preparing the above commit. When printing the final ROM size message, HUGEROM was only checked against false. If it was empty (use default for this board revision), it would print that the firmware was built for 1M ROMs. This was wrong in the case of REVISION1X=true without a HUGEROM set.
0xTJ
commented
8 months ago
I've tried to keep the changes isolated to a single commit, so feel free to push back on any changes, they can be reverted individually pretty easily. The changes are:
A0to point to the end of the string, as documented.C_DEVICE_SIZE. (This one isn't particularly important, actually less efficient, and shouldn't apply unless theCHAR_DEVICEstructure is grown in the future.)RAM_TOP, since it was using anifdon REVISION1X which ignored the actual ROM size and memory layout.serial.asm, one for each stage. The version in stage1 unused except forRECVCHAR, which I folded intotrap14.asm(along with the other functions), since that's where the other EFP wrappers live asFW_<...>.Makefileissue that I noticed while preparing the above commit. When printing the final ROM size message,HUGEROMwas only checked againstfalse. If it was empty (use default for this board revision), it would print that the firmware was built for 1M ROMs. This was wrong in the case ofREVISION1X=truewithout aHUGEROMset.