Repository navigation

firmware: module code in PSRAM, a 41% faster blit, a crash record, and a console that outlives a hung pattern - #466

Merged

engmung merged 18 commits into

mainfrom

dev

Oct 1, 2026

Commits on Oct 1, 2026
- firmware: a module's code runs from PSRAM
  
  engmung
  and
  claude
  committed
  88f9e5a
- firmware: a cut-off form, a quoted name, a global constructor and day 25
  
  engmung
  and
  claude
  committed
  3464e07
- firmware: a host test replays cut-off requests through the vendored parser
  
  engmung
  and
  claude
  committed
  e677991
- firmware: blit runs as three short passes, 635 -> 426 instructions a column pair
  
  engmung
  and
  claude
  committed
  a33d53b
- firmware: blit skips the gamma/WB LUTs when they are the identity
  
  engmung
  and
  claude
  committed
  ef9cc14
- firmware: a plain POST to an upload route, and a 9,000-character boundary, no longer reboot the panel
  
  engmung
  and
  claude
  committed
  3fa2baf
- firmware: module code in PSRAM is aligned by hand, not by the allocator
  
  engmung
  and
  claude
  committed
  617f058
- firmware: the blit's speed is measured, its flip condition is a formula, and the loop task's stack is in status
  
  engmung
  and
  claude
  committed
  20c9af2
- docs: the changelog says what the blit change measured
  
  engmung
  and
  claude
  committed
  81fa5e8
- firmware: a crashed board says where - the core dump's summary and a pattern breadcrumb in status
  
  engmung
  and
  claude
  committed
  ed6863a
- firmware: the crash record costs 232 bytes of internal RAM, not 8
  
  engmung
  and
  claude
  committed
  8466b49
- firmware: the crash breadcrumb records where a module's code is fetched from, and lives in DRAM
  
  engmung
  and
  claude
  committed
  164dd09
- firmware: a host check boots the crash record through every reset, breadcrumb and dump
  
  engmung
  and
  claude
  committed
  5cdb683
- docs: the crash record was checked on a board
  
  engmung
  and
  claude
  committed
  040b729
- firmware: a hung render loop no longer takes the console with it
  
  engmung
  and
  claude
  committed
  43e4ae5
- firmware: the status page says when the panel is frozen
  
  engmung
  and
  claude
  committed
  9dc5afb
- firmware: "not answering", not "frozen" - what the loop's stall time can and cannot say
  
  engmung
  and
  claude
  committed
  5551ebd
- firmware: footprint re-pinned for the crash record and the bounded loop sync
  
  engmung
  and
  claude
  committed
  df2106f

Navigation Menu

Repository navigation

firmware: module code in PSRAM, a 41% faster blit, a crash record, and a console that outlives a hung pattern - #466

firmware: module code in PSRAM, a 41% faster blit, a crash record, and a console that outlives a hung pattern#466
engmung merged 18 commits into
mainfrom
dev

Commits on Oct 1, 2026

firmware: a module's code runs from PSRAM

firmware: a cut-off form, a quoted name, a global constructor and day 25

firmware: a host test replays cut-off requests through the vendored parser

firmware: blit runs as three short passes, 635 -> 426 instructions a column pair

firmware: blit skips the gamma/WB LUTs when they are the identity

firmware: a plain POST to an upload route, and a 9,000-character boundary, no longer reboot the panel

firmware: module code in PSRAM is aligned by hand, not by the allocator

firmware: the blit's speed is measured, its flip condition is a formula, and the loop task's stack is in status

docs: the changelog says what the blit change measured

firmware: a crashed board says where - the core dump's summary and a pattern breadcrumb in status

firmware: the crash record costs 232 bytes of internal RAM, not 8

firmware: the crash breadcrumb records where a module's code is fetched from, and lives in DRAM

firmware: a host check boots the crash record through every reset, breadcrumb and dump

docs: the crash record was checked on a board

firmware: a hung render loop no longer takes the console with it

firmware: the status page says when the panel is frozen

firmware: "not answering", not "frozen" - what the loop's stall time can and cannot say

firmware: footprint re-pinned for the crash record and the bounded loop sync