electronic_eel has quit [Ping timeout: 272 seconds]
electronic_eel has joined #amaranth-lang
<cr1901>
Did some optimizations on icache... down to 105 LUTs, 11 DFFs, and 4 BRAMs for direct-mapped. And also deployed using LiteX. SPI flash sequential reads are 7x times faster*, SPI flash random reads 2x faster. Basic memtest now completes 5-5.5x faster.
<whitequark[cis]>
niiiiiice
<whitequark[cis]>
good job
<cr1901>
Tyvm :D
<cr1901>
* This is not a real result. I forgot to bring out logic to disable caching on non-insn fetches to SPI flash region. Though maybe I should keep it and document "only use for read-only mem" or something...
* cr1901
crosses out "icache" and writes "dual instruction/read-only memory cache"