This simplifies allocating outside of the VM because the VM doesn't
take up all remaining memory by default.
On ESP we delegate to the IDF for allocations. For all other ports,
we use TLSF to manage an outer "port" heap. The IDF uses TLSF
internally and we use their fork for the other ports.
This also removes the dynamic C stack sizing. It wasn't often used
and is not possible with a fixed outer heap.
Fixes#8512. Fixes#7334.
micropython puts the pointer-ness into the typedef; we can put the
const-ness there too.
this reduces the delta to micropython; for instance, emitinlinextensa
and emitinlinethumb now match upstream.
.. and switch makerfabs tft7 over to it as a test.
We have our existing way of "reserving" PSRAM for esp-idf (we actually
control it all but add back the "reserved" part). However, this does
not work with off the shelf esp_lcd, which only will allocate a
framebuffer in PSRAM if CONFIG_SPIRAM_USE_CAPS_ALLOC (or CONFIG_SPIRAM_USE_ALLOC)
is defined, not if CONFIG_SPIRAM_USE_MEMMAP is.
This new way is possibly compatible with more esp-idf code, but it complicates
CircuitPython's initial startup since nothing until port_heap_init is
permitted to use the CP heap or supervisor allocator. In practice this
seems to be OK today.
Right now this doesn't change the setting across all boards with PSRAM and so
it does not revert esp-idf to its prior state. Instead, what I'm thinking is
that we can do it during or just after the IDF5 update when sdkconfig files
will be getting an overhaul anyway.
.. the primary user of which will be the camera, since the framebuffers
must be allocated via esp-idf allocation function and never from the
gc heap.
A board can have a default value, and the value can also be set in the
/.env file using the key CIRCUITPY_RESERVED_PSRAM with the value being
the reserved size in bytes.
Co-authored-by: Dan Halbert <halbert@adafruit.com>
