Use SRAM2 32Kbytes on STM32L475 / L476 and L486 devices #6172

adustm · 2018-02-22T16:58:01Z

Description

STM32L475 STM32L476 and STM32L486 devices have 32kbytes of SRAM2 that is used only for interrupt vector at the moment.
With this PR, the stack is now located in SRAM2.
This gives more space in SRAM1 for variables and heap.
It is mandatory for heavy applications like IOT connections

This has already been pushed and merged for IAR toolchain. ARM Toolchain will come soon

Note that the mbed-os code for GCC_ARM is made for heap and stack in the same RAM area. I've rewritten __wrap__sbrk (code already available in mbed-os) and used a new define in targets.json : TWO_RAM_REGIONS

@ARMmbed/team-avnetsilica Please pay attention. I have modified your GCC_ARM linker file and added TWO_REGION_RAM in targets.json for SILICA_SENSOR_NODE. It compiles and link, but I can't test it.
If you don't want to benefit from it, you will have to remove this define and copy the previous GCC linker file in your own directory, instead of using ST' one.
(according to the use cases you are playing with this platform, I guess you would be happy to gain 32kbytes... up to you to decide)

@MarceloSalazar I'm sure you would like to use that...

kind regards
Armelle

… stack in SRAM2 (after the interrupt vector)

…use of TWO_RAM_REGIONS define

… STM32L486 devices

MarceloSalazar · 2018-02-22T17:29:04Z

@0xc0170 please let's get it in asap! ;-)

bcostm · 2018-02-23T07:04:24Z

targets/TARGET_STM/mbed_rtx.h

-#if (defined(TARGET_STM32F051R8) ||\
+#if (defined(TARGET_STM32L475VG))
+/* only GCC_ARM and IAR toolchain have the stack on SRAM2 */
+#if (defined(TOOLCHAIN_GCC_ARM) || defined(TOOLCHAIN_GCC_CR) || defined(__IAR_SYSTEMS_ICC__ ))


I would expect to have also && defined(TWO_RAM_REGIONS) on this line ?

bcostm · 2018-02-23T07:10:05Z

targets/TARGET_STM/TARGET_STM32L4/l4_retarget.c

+  ******************************************************************************  
+  */ 
+#if defined(TWO_RAM_REGIONS)
+#if defined(TOOLCHAIN_GCC_ARM) || defined(TOOLCHAIN_GCC_CR)


Just a a remark, in Cube we use __CC_ARM for ARM, __GNUC__ for GCC and __ICCARM__ for IAR. I don't know which macros are the best to use ?

Just a a remark, in Cube we use __CC_ARM for ARM, GNUC for GCC and ICCARM for IAR. I don't know which macros are the best to use ?

+1, we should use those that come from the toolchains (be careful that GNUC is defined for ARMCC in mbed as we define gnu extension there)

Ok, I will (but I took those from platform/mbed_retarget.cpp, so both defines exist in mbed...)
The multiple use of GNUC was a problem for me, that's why I did not use it.

adustm · 2018-02-23T09:45:26Z

@0xc0170 @bcostm I applied the requested changes. Thanks for the review

0xc0170 · 2018-02-23T10:48:16Z

platform/mbed_retarget.cpp

@@ -1053,12 +1053,13 @@ extern "C" uint32_t  __HeapLimit;
 extern "C" int errno;

 // Dynamic memory allocation related syscall.
-#if defined(TARGET_NUVOTON)
+#if (defined(TARGET_NUVOTON) || defined(TWO_RAM_REGIONS))


cleaner would be nuvoton would also define two ram regions so we would have just one config and no target here

0xc0170 you are right.
I suggest we create an issue in mbed-os so that NUVOTON can use the same define once this PR is merged. What do you think ?

Yes, can be done separately, would be nice o t have a tracking issue so we do not forget, thanks

0xc0170 · 2018-02-23T10:51:16Z

targets/targets.json

@@ -1552,7 +1552,7 @@
            }
        },
        "detect_code": ["0765"],
-        "macros_add": ["USBHOST_OTHER"],
+        "macros_add": ["USBHOST_OTHER", "TWO_RAM_REGIONS"],


I assume this is fine here but instead of macro, if this is a config that would result in MBED_CONF_MULTIPLE_RAM_REGIONS (if there are 3 , would it matter? if yes, then two is fine in the name) ?

@0xc0170 the 'TWO' means that heap and stack are located in 2 separate ram regions.
It won't matter if you have 3 ram regions on the MCU. Only 2 will be concerned by this change.

adustm · 2018-02-23T11:16:47Z

Hello, I can see 'continuous-integration/jenkins/pr-head — This commit cannot be built'
Could you explain me what is failing ?

Kind regards

0xc0170 · 2018-02-23T13:43:57Z

Hello, I can see 'continuous-integration/jenkins/pr-head — This commit cannot be built'

This is the error from jenkins: mbed_retarget.cpp@1069,21: conflicting declaration 'uint32_t __end__'

adustm · 2018-02-23T14:18:10Z

This is the error from jenkins: mbed_retarget.cpp@1069,21: conflicting declaration 'uint32_t end'

What target and toolchain, please ? Do you have any idea why it does not fail the same for Nuvoton platform (PFM_M453 or PFM_NUC472) ?

adustm · 2018-02-23T16:21:05Z

Hi @0xc0170
I will fix the jenkins issue, but be aware that Mbed already contains the conflicting declarations

/c/.../mbed-os  (l4_linkergcc)$ git grep __end__ | grep uint32_t
features/FEATURE_UVISOR/source/rtx/unsupported_malloc.c:extern uint32_t __end__[];      /* __heap_start */
features/FEATURE_UVISOR/source/rtx/unsupported_malloc.c:                          (uint32_t) __end__);
platform/mbed_retarget.cpp:extern "C" uint32_t __end__;
rtos/TARGET_CORTEX/mbed_boot.c:        extern uint32_t         __end__[];
targets/TARGET_NUVOTON/mbed_rtx.h:    extern uint32_t               __end__[];
targets/TARGET_RENESAS/mbed_rtx.h:    extern uint32_t               __end__;
targets/TARGET_STM/mbed_rtx.h:    extern uint32_t               __end__[];];

0xc0170 · 2018-02-26T10:02:02Z

/morph build

mbed-ci · 2018-02-26T10:41:24Z

Build : SUCCESS

Build number : 1254
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/6172/

Triggering tests

/morph test
/morph uvisor-test
/morph export-build
/morph mbed2-build

mbed-ci · 2018-02-26T13:35:59Z

Test : SUCCESS

Build number : 1051
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/6172/1051

mbed-ci · 2018-02-26T13:41:06Z

Exporter Build : SUCCESS

Build number : 918
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/6172/

0xc0170 · 2018-02-26T14:31:02Z

targets/TARGET_STM/mbed_rtx.h

+     defined(TARGET_STM32L476VG) ||\
+     defined(TARGET_STM32L486RG))
+/* only GCC_ARM and IAR toolchains have the stack on SRAM2 */
+#if (((defined(__GNUC__) && !defined(__CC_ARM)) ||\


Why not for ARMCC? only those 2 toolchains are here supported?

Hello @0xc0170
IAR was done in #5844
This PR is for GCC_ARM (required for Cloud example).
ARM will come soon.

I assume will be created as a new PR that will address ARMCC support for this. I would like to avoid having different behavior between toolchains

We will do it .

adustm added 5 commits February 22, 2018 17:36

Call to specific _wrap_sbrk to support 2 region model for heap and stack

0efc876

Change STM32L475/76/86 GCC_ARM linker files to have HEAP in SRAM1 and…

02b2b01

… stack in SRAM2 (after the interrupt vector)

Add support of separate memories for heap and stack region swith the …

f551255

…use of TWO_RAM_REGIONS define

Add defined for GCC_ARM to pass heap and stack tests

815be94

Use the TWO_RAM_REGIONS in targets.json for STM32L475 / STM32L476 and…

9865744

… STM32L486 devices

cmonr added the needs: review label Feb 22, 2018

cmonr requested a review from MarceloSalazar February 22, 2018 19:12

bcostm reviewed Feb 23, 2018

View reviewed changes

0xc0170 added needs: work and removed needs: review labels Feb 23, 2018

adustm added 2 commits February 23, 2018 10:29

Use official toolchain defines

6795325

Add initial_sp value for STM32L476 and STM32L486 devices

28a43b5

0xc0170 reviewed Feb 23, 2018

View reviewed changes

Allow jenkins script to pass

d1e6e8f

LMESTM approved these changes Feb 23, 2018

View reviewed changes

bcostm approved these changes Feb 24, 2018

View reviewed changes

MarceloSalazar approved these changes Feb 26, 2018

View reviewed changes

0xc0170 removed the needs: work label Feb 26, 2018

0xc0170 added the needs: CI label Feb 26, 2018

0xc0170 reviewed Feb 26, 2018

View reviewed changes

0xc0170 approved these changes Feb 26, 2018

View reviewed changes

cmonr added ready for merge and removed needs: CI labels Feb 27, 2018

cmonr merged commit 153fa05 into ARMmbed:master Feb 27, 2018

cmonr added release-version: 5.7.7 and removed ready for merge labels Feb 27, 2018

adustm deleted the l4_linkergcc branch March 1, 2018 08:37

jeromecoutant mentioned this pull request Mar 12, 2018

Pretty bars feature optimisation with dual RAM targets #6334

Closed

0xc0170 mentioned this pull request Mar 13, 2018

Failure to flash for MTB_ADV_WISE_1570 target #6252

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use SRAM2 32Kbytes on STM32L475 / L476 and L486 devices #6172

Use SRAM2 32Kbytes on STM32L475 / L476 and L486 devices #6172

adustm commented Feb 22, 2018

MarceloSalazar commented Feb 22, 2018

bcostm Feb 23, 2018

adustm Feb 23, 2018

bcostm Feb 23, 2018

0xc0170 Feb 23, 2018 •

edited

Loading

adustm Feb 23, 2018

adustm commented Feb 23, 2018

0xc0170 Feb 23, 2018

adustm Feb 23, 2018

0xc0170 Feb 23, 2018

0xc0170 Feb 23, 2018

adustm Feb 23, 2018

adustm commented Feb 23, 2018

0xc0170 commented Feb 23, 2018

adustm commented Feb 23, 2018 •

edited

Loading

adustm commented Feb 23, 2018 •

edited

Loading

0xc0170 commented Feb 26, 2018

mbed-ci commented Feb 26, 2018

mbed-ci commented Feb 26, 2018

mbed-ci commented Feb 26, 2018

0xc0170 Feb 26, 2018

adustm Feb 26, 2018

0xc0170 Feb 26, 2018

adustm Feb 26, 2018

Use SRAM2 32Kbytes on STM32L475 / L476 and L486 devices #6172

Use SRAM2 32Kbytes on STM32L475 / L476 and L486 devices #6172

Conversation

adustm commented Feb 22, 2018

Description

MarceloSalazar commented Feb 22, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

0xc0170 Feb 23, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

adustm commented Feb 23, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

adustm commented Feb 23, 2018

0xc0170 commented Feb 23, 2018

adustm commented Feb 23, 2018 • edited Loading

adustm commented Feb 23, 2018 • edited Loading

0xc0170 commented Feb 26, 2018

mbed-ci commented Feb 26, 2018

Triggering tests

mbed-ci commented Feb 26, 2018

mbed-ci commented Feb 26, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

0xc0170 Feb 23, 2018 •

edited

Loading

adustm commented Feb 23, 2018 •

edited

Loading

adustm commented Feb 23, 2018 •

edited

Loading