- treat floating point constants as single precision
(this removes all dependencies for double)
- use internal.ldf linker script
(puts libc functions in local memory)
speedup is immense. running code from external memory
is extremely slow, especially when all cores fight for it.
# target toolchain
ECC = e-gcc
EOC = e-objcopy
-ECFLAGS = -Os -std=c99 -falign-loops=8 -falign-functions=8 -Wall
-ELFLAGS = -T$(EPIPHANY_HOME)/bsps/current/fast.ldf -le-lib
+ECFLAGS = -Os -std=c99 -falign-loops=8 -falign-functions=8 -Wall -fsingle-precision-constant -ffast-math
+ELFLAGS = -T$(EPIPHANY_HOME)/bsps/current/internal.ldf -le-lib
EOFLAGS = -R .shared_dram
# host application