I finally got CONFIG_OVMS_HW_SPIMEM_AGGRESSIVE working properly. It seems we need to include sdkconfig.h to access these CONFIG_* variables (if we don’t include other esp libraries that do that already). Other than that change, the rest was easy.

For my running config (a car, a v2 server connected, and wifi), I originally had this:

OVMS# module memory Free 8-bit 87484/230708, 32-bit 7296/11532, SPIRAM 4079552/4194252

Now, with CONFIG_OVMS_HW_SPIMEM_AGGRESSIVE working properly, I get this:

OVMS# module memory Free 8-bit 104640/230700, 32-bit 7296/11532, SPIRAM 4062400/4194252

OVMS# test spiram Metrics (0x3f802024) are in SPI RAM

With that config option enabled, all the C++ allocations are by default in SPI RAM, and we gain about 17KB of conventional RAM.

A quick ‘fgrep -r alloc main components’ shows about 1,000 uses of calloc/malloc in our system. I’ve got a large number of them now done, but probably a few more still to do. For example, changing wolf ssh/ssl to use it means most of the incremental allocations are from SPI RAM. Duktape would be a good win (but non-trivial, I think).

Even with bluetooth compiled in, and running, I now have:

OVMS# module memory Free 8-bit 76476/230696, 32-bit 7296/11532, SPIRAM 4033960/4194252

OVMS# module tasks Number of Tasks = 21 Stack: Now Max Total Heap 32-bit SPIRAM 3FFCFF98 1 Blk esp_timer 388 628 4096 23928 644 44632 3FFD1C38 2 Blk eventTask 436 1876 4608 0 0 0 3FFD4158 3 Blk OVMS Events 436 3124 8192 73960 0 22128 3FFD56C0 4 Blk OVMS CanRx 428 476 1024 0 0 0 3FFBC3D4 5 Blk ipc0 400 480 1024 7776 0 0 3FFBC9D4 6 Blk ipc1 400 496 1024 12 0 0 3FFBD5E8 9 Rdy IDLE 372 548 1024 0 0 0 3FFBD768 10 Rdy IDLE 372 548 1024 0 0 0 3FFBD8E8 11 Blk Tmr Svc 344 888 3072 88 0 0 3FFBCCAC 16 Blk tiT 496 1504 3072 696 0 884 3FFB7A2C 17 Blk OVMS SIMCOM 464 512 4096 0 0 0 3FFB7D58 18 Blk wifi 432 2384 4096 360 0 1576 3FFDAD44 19 Blk OVMS Vehicle 460 508 6144 0 0 0 3FFDF718 20 Rdy OVMS Console 740 2612 6144 9072 3488 24044 3FFDFBC0 21 Blk mdns 456 1384 4096 108 0 4 3FFE37BC 22 Blk OVMS NetMan 728 2552 8192 104 0 1136 3FFE79B0 23 Blk btController 444 1980 4096 88 0 0 3FFE8CA0 24 Blk Btc_task 412 2572 3584 14924 0 1688 3FFE9EEC 25 Blk hciHostT 408 744 2560 0 0 0 3FFEAFF0 26 Blk hciH4T 412 604 2560 0 0 0 3FFEC854 27 Blk btuT 416 2016 4608 2760 0 23996

That seems very workable.

Regards, Mark

Begin forwarded message:

From: GitHub <noreply@github.com> Subject: [openvehicles/Open-Vehicle-Monitoring-System-3] c2d3f7: Add a test spiram command to show whether SPI RAM ... Date: 11 June 2018 at 12:51:21 PM HKT

Branch: refs/heads/master Home: https://github.com/openvehicles/Open-Vehicle-Monitoring-System-3 Commit: c2d3f7f7ae8655d9d2809e15ad6f989a3459c8df https://github.com/openvehicles/Open-Vehicle-Monitoring-System-3/commit/c2d3... Author: Mark Webb-Johnson <mark@webb-johnson.net> Date: 2018-06-11 (Mon, 11 Jun 2018)

Changed paths: M vehicle/OVMS.V3/main/test_framework.cpp

Log Message:

Add a test spiram command to show whether SPI RAM is being globally used or not

Commit: c1f96e71dd51c44f9aeff9cf5940982b939a8b24 https://github.com/openvehicles/Open-Vehicle-Monitoring-System-3/commit/c1f9... Author: Mark Webb-Johnson <mark@webb-johnson.net> Date: 2018-06-11 (Mon, 11 Jun 2018)

Changed paths: M vehicle/OVMS.V3/main/ovms.cpp M vehicle/OVMS.V3/main/ovms.h

Log Message:

Fix CONFIG_OVMS_HW_SPIMEM_AGGRESSIVE