Add Support for TOSHIBA TMPM066 board #4840

ghost · 2017-08-01T07:32:53Z

Description

Add mbed support for TOSHIBA's TMPM066 board.
DAPLink PR: ARMmbed/DAPLink#295

Status

READY

Migrations

NO

Test Status

Pass: Mbed SDK automated test suite and greentea with ARM, GCC_ARM and IAR toolchains.

0xc0170 · 2017-08-02T13:13:45Z

@ganesh-ramachandran Hello, thanks for your contribution. It would be good to clean the git history as the last 3 commits are identical. The first two also but I assume they are adding different bits.

we recommend our contributors follow Chris Beam’s seven rules of great commit messages to keep the commit history clear. We find the commit.template feature particularly helpful.

Pass: Mbed SDK automated test suite and greentea with ARM, GCC_ARM and IAR toolchains.

Would it be possible to share them here (paste or attach txt file with the test results) .

ghost · 2017-08-03T11:29:35Z

@0xc0170 Hello, Thank you for your inputs. Cleaned the git history. Please find below the test results attached.

Pass: Mbed SDK automated test suite and greentea with ARM, GCC_ARM and IAR toolchains.

Would it be possible to share them here (paste or attach txt file with the test results) .

M066_SingleAutomated_TestResults.txt
M066_GreenteaAutomated_TestResults.txt

0xc0170 · 2017-08-03T16:14:25Z

targets/TARGET_TOSHIBA/mbed_rtx.h

+#define INITIAL_SP                        (0x20004000UL)
+#endif
+
+#ifdef TOOLCHAIN_GCC_ARM


why is this main stack dependant on toolchain?

GCC tool chain usually generates bigger code than other tool chains. Because of this few of the test cases need more memory on the Heap. That's why, we have reduced mbed main stack size to 2KB from 4KB for our target.

I believe it should not be tool dependent. neither the library should change this. This is an application specific, therefore I would recommend to remove it

cc @c1728p9 @bulislaw

Removed the main stack updated for GCC tool chain. However, on removal two test suites of greentea fails due to out of memory. Please find attached the log.

M066_GCC_ARM_GreenTea_fail.txt

I cant spot from the logs why it timeouts ? What is the reason for the failure? not enough memory? We got targets that are quite low with the memory, and haven't seen this errors.

What is the memory for this target - flash/ram, how much is left for these 2 tests?

Not enough memory is the message from the log - "Operator new[] out of memory" - Line numbers 191 & 289 from the attached log "M066_GCC_ARM_GreenTea_MainStack_4K.txt". Did not face this memory issue with ARM and IAR.

M066 Memory : RAM - 16K and Flash - 128K.

Attached below the GCC_ARM logs with 4K and 2K main stack size.
M066_GCC_ARM_GreenTea_MainStack_2K.txt
M066_GCC_ARM_GreenTea_MainStack_4K.txt

Can you create a separate issue for this? If 16K is available, it should be sufficient for this test. we got targets with the same numbers that do not cause the issues.

Separate issue #4901 created

0xc0170 · 2017-08-03T16:14:58Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/us_ticker.c

+static volatile uint32_t acc_us_ticker = 0;
+
+// 16Bb high timer counter
+static uint32_t us_ticker_16h = 0;


should be volatile, as its acc_us_ticker

0xc0170 · 2017-08-03T16:15:17Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/us_ticker.c

+    NVIC_EnableIRQ(INT16A0_IRQn);
+    // Match counter set to max value
+    TSB_T16A0->RG = TMR16A_100US;
+    // TSB_T16A0->CP = 0x00;


please remove any dead code

0xc0170 · 2017-08-03T16:16:42Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/serial_api.c

+    pinmap_pinout(tx, PinMap_UART_TX);
+}
+
+void serial_set_flow_control(serial_t *obj, FlowControl type, PinName rxflow, PinName txflow)


This can be removed, and SERIAL_FC disabled

0xc0170 · 2017-08-03T16:29:27Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/i2c_api.c

+    obj->myi2c.PrescalerClkDiv = clk.prsck;
+
+    I2C_Init(obj->i2c, &obj->myi2c);
+    NVIC_DisableIRQ(obj->IRQn);


why is this disabling IRQ (related to slave functionality that operates with IRQs) ?

Currently, the I2C implementation is based on polling. Hence disabled IRQ.

0xc0170 · 2017-08-03T16:30:05Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/i2c_api.c

+
+#endif
+
+void INTI2C0_IRQHandler(void)


it's recommended to use nvic set /get vectors to register own handlers, than using weak symbols.

Why are these empty?

0xc0170 · 2017-08-03T16:38:17Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/gpio_irq_api.c

+    obj->irq_id = pinmap_peripheral(pin, PinMap_GPIO_IRQ);
+
+    // Disable interrupt by CPU
+    __set_PRIMASK(1);


please use critical section API

0xc0170 · 2017-08-03T16:38:43Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/gpio_irq_api.c

+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include "string.h"


why is string header included here?

0xc0170 · 2017-08-03T16:39:08Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/gpio_api.c

+    }
+}
+
+int  gpio_read (gpio_t *obj)


one less space around gpio_read

ghost · 2017-08-04T12:54:40Z

@0xc0170 Hello, Thanks for your review comments. We have addressed the comments and committed the changes. Also checked for additional format changes.

0xc0170 · 2017-08-04T14:13:19Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/sleep.c

+    // Enter stop1 mode
+    __WFI();
+    // Switch over from IHOSC to EHOSC
+    external_losc_enable();


Out of curiosity, how long it takes to switch from IHOSC to EHOSC ? What is thte stop1 power consumptions and wake up time?

Switching from Internal Oscillator to External Oscillator takes as WarmUpTime (CG_WUODR_INT).
600mW is the peak power consumption at T = 85degC.
When using IOSC 10MHz,<WUP[15:0]> = 0x2700
Warm-up time = (0x2700 + 0x10) × (1/10 MHz)
= 0x2710 × 100ns
= 10000 × 100ns
= 1.0 ms
When using EOSC 16MHz,<WUPT[15:0]> = 0x3E70
Warm-up time = (0x3E70 + 0x10) × (1/16 MHz)
= 0x3E80 × 62.5 ns
= 16000 × 62.5 ns
= 1.0 ms

0xc0170 · 2017-08-04T14:32:01Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/analogin_api.c

+    // Set pin function as ADC
+    pinmap_pinout(pin, PinMap_ADC);
+    // Software reset ADC
+    ADC_SWReset();


What does this reset do? Can it conflict with some already configured ADC pins?

A software reset will initializes all the registers again (even pre-configured ADC settings), except clock register. This is fixed to avoid conflict with some already configured ADC pins and committed the changes.

0xc0170 · 2017-08-07T13:12:03Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/us_ticker.c

+    if (!us_ticker_inited) {
+        us_ticker_init();
+    }
+    ret_val = (((acc_us_ticker << 16) + us_ticker_16h) * 100);


How this read() works? There is not hw counter used, only sw counters? When accessing this variable, shouldn't you protect it via at least disabling ticker interrupt (while this is accessing this variable, interrupt might happen and plus check again if there is pending, reread those sw counters to be up to date) ?

How much does this ticker ticks (interrupt invoked)? I am reading this as it ticks every 1us (16b counter) ?

Assume a single tick is 100us, the acc_us_ticker (lower 16bits) will overflow in 119.3 hours after initializing us_ticker. Is it the intended behavior?

@0xc0170, Thank you for your input. Yes, we have used only sw counter. Now we have handled to check again the variable, if interrupt happens while accessing. Ticker ticks for every 100 us.

@soramame21, Thank you for your input. Overflow has been handled now using 32 bit value.

0xc0170 · 2017-08-09T13:45:16Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/us_ticker.c

@@ -98,5 +110,5 @@ void us_ticker_disable_interrupt(void)

 void us_ticker_clear_interrupt(void)
 {
-    //no flags to clear
+    NVIC_ClearPendingIRQ(INT16A1_IRQn);


I believe this should be part of fire interrupt now, clear can stay empty as it was?

Yes, thank you. Made empty as it was.

0xc0170 · 2017-08-09T13:46:03Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/us_ticker.c

    // Stops and clear count operation
    TSB_T16A1->RUN = TMR16A_STOP;
    TSB_T16A1->CR = TMR16A_SYSCK;
    // Set the compare register
    delta = (int)(timestamp - us_ticker_read());
+    if (delta < 0) {


This should not happen (it is handled in the upper layer, if it is, then fire interrupt is invoked, not set interrupt), will be documented in tickers API once we do update soon. Plus even if this would handle, should use set pending bit, rather than invoking handler directly (might cause stack overflow)

Thank you for the update. Removed the check.

theotherjimmy · 2017-08-10T14:17:24Z

@0xc0170 Could you review again?

theotherjimmy · 2017-08-14T16:32:01Z

/morph test

theotherjimmy · 2017-08-14T16:32:09Z

/morph export-build

mbed-bot · 2017-08-14T18:35:08Z

Result: FAILURE

Your command has finished executing! Here's what you wrote!

/morph test

mbed-bot · 2017-08-14T18:35:27Z

Result: FAILURE

Your command has finished executing! Here's what you wrote!

/morph export-build

theotherjimmy · 2017-08-14T18:40:41Z

@ganesh-ramachandran It looks like you have conflicts in targets/targets.json. Could you rebase to resolve those conflicts? We can continue testing afterwards.

ghost · 2017-08-16T08:32:17Z

@theotherjimmy Rebased with master and resolved the conflict.

theotherjimmy · 2017-08-16T14:41:33Z

Rebased with master and resolved the conflict.

Thanks! But I'm questioning if it was a merge or a rebase. You have a commit, Resolved merge conflict, that indicates that you may have done a merge (the default action of git pull). Generally, a rebase to resolve a conflict will not create a commit, and it will change the dates of all of the commits you rebased to be after the date of my "could you rebase" comment above.

Is the aforementioned commit a merge commit? Could you do git fetch armmbed then git rebase armmbed/master (assuming your remote for this repo is armmbed to remove that merge commit ?

SERIAL_FC disabled, critical section API and Format changes updated

ADC Reset conflict with already configured ADC pins is fixed

Functions 'serial_break_set' & 'serial_break_clear' wrongly removed during 'SERIAL_FC disabled, critical section API Updation' commit. Due to this 'mbed compile' command for GCC_ARM tool fails to compile.

ghost · 2017-08-17T07:04:09Z

@theotherjimmy Yes, the previous one was merge commit. Got it about rebase. Now we have rebased.

0xc0170 · 2017-08-17T10:25:41Z

targets/TARGET_TOSHIBA/TARGET_TMPM066/serial_api.c

@@ -239,3 +239,11 @@ void serial_pinout_tx(PinName tx)
 {
    pinmap_pinout(tx, PinMap_UART_TX);
 }
+
+void serial_break_set(serial_t *obj)


Can you implement these 2 functions? A question is why these 2 are empty

@0xc0170 Line control register is not available in M066 controller. Due to this these 2 functions are not implemented and hence they are kept empty.

0xc0170 · 2017-08-17T10:27:08Z

/morph test

studavekar · 2017-08-17T21:02:03Z

/morph export-build

mbed-bot · 2017-08-18T06:26:27Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph export-build

Output

mbed Build Number: 128

All exports and builds passed!

0xc0170 · 2017-08-21T14:01:08Z

/morph test

theotherjimmy · 2017-08-21T15:28:53Z

/morph test

mbed-bot · 2017-08-21T21:16:31Z

Result: SUCCESS

Your command has finished executing! Here's what you wrote!

/morph test

Output

mbed Build Number: 1063

All builds and test passed!

mbed OS 5.5.6 release We are pleased to announce the [mbed OS 5.5.6 release](https://github.com/ARMmbed/mbed-os/releases/tag/mbed-os-5.5.6) is now available. This release includes ... Known Issues The following list of known issues apply to this release: Contents Ports for Upcoming Targets [4608](ARMmbed#4608) Support Nuvoton's new target NUMAKER_PFM_M487 [4840](ARMmbed#4840) Add Support for TOSHIBA TMPM066 board Fixes and Changes [4801](ARMmbed#4801) STM32 CAN: Fix issue with speed function calculation [4808](ARMmbed#4808) Make HAL & US tickers idle safe [4812](ARMmbed#4812) Use DSPI SDK driver API's in SPI HAL driver [4832](ARMmbed#4832) NUC472/M453: Fix several startup and hal bugs [4842](ARMmbed#4842) Add call to DAC_Enable as this is no longer done as part [4849](ARMmbed#4849) Allow using of malloc() for reserving the Nanostack's heap. [4850](ARMmbed#4850) Add list of defines to vscode exporter [4863](ARMmbed#4863) Optimize memory usage of wifi scan for REALTEK_RTW8195AM [4869](ARMmbed#4869) HAL LPCs SPI: Fix mask bits for SPI clock rate [4873](ARMmbed#4873) Fix Cortex-A cache file [4878](ARMmbed#4878) STM32 : Separate internal ADC channels with new pinmap [4392](ARMmbed#4392) Enhance memap, and configure depth level [4895](ARMmbed#4895) Turn on doxygen for DEVICE_* features [4817](ARMmbed#4817) Move RTX error handlers into RTX handler file [4902](ARMmbed#4902) Using CMSIS/RTX Exclusive access macro [4923](ARMmbed#4923) fix export static_files to zip [4844](ARMmbed#4844) bd: Add ProfilingBlockDevice for measuring higher-level applications [4896](ARMmbed#4896) target BLUEPILL_F103C8 compile fix [4921](ARMmbed#4921) Update gcc-arm-embedded PPA in Travis [4926](ARMmbed#4926) STM32L053x8: Refactor NUCLEO_L053R8 and DISCO_L053C8 targets [4831](ARMmbed#4831) Remove excessive use of printf/scanf in mbed_fdopen/_open [4922](ARMmbed#4922) bug fix: xdot clock config [4935](ARMmbed#4935) STM32: fix F410RB vectors size [4940](ARMmbed#4940) Update mbed-coap to version 4.0.9 [4941](ARMmbed#4941) Update of MemoryPool.h header file. You can fetch this release from the [mbed-os GitHub](https://github.com/ARMmbed/mbed-os) repository, using the tag "mbed-os-5.5.6". Please feel free to ask any questions or provide feedback on this release [on the forum](https://forums.mbed.com/), or to contact us at [[email protected]](mailto:[email protected]).

0xc0170 added scope: new-target needs: review labels Aug 1, 2017

0xc0170 requested changes Aug 3, 2017

View reviewed changes

0xc0170 added needs: work and removed needs: review labels Aug 3, 2017

0xc0170 reviewed Aug 4, 2017

View reviewed changes

0xc0170 reviewed Aug 7, 2017

View reviewed changes

0xc0170 reviewed Aug 9, 2017

View reviewed changes

theotherjimmy added needs: review and removed needs: work labels Aug 10, 2017

0xc0170 approved these changes Aug 14, 2017

View reviewed changes

ghost mentioned this pull request Aug 14, 2017

Greentea fails for GCC_ARM in TMPM066 #4901

Closed

theotherjimmy added needs: CI and removed needs: review labels Aug 14, 2017

Ganesh Ramachandran added 5 commits August 17, 2017 10:54

Added Support for Toshiba TMPM066

bd8fa4e

SERIAL_FC disabled, critical section API Updation

624620b

SERIAL_FC disabled, critical section API and Format changes updated

ADC Reset Conflict Fixed

0a781c4

ADC Reset conflict with already configured ADC pins is fixed

Interrupt and Overflow check for us_ticker

9dca3fe

Clear interrupt and delta check removed

48ee6a0

Ganesh Ramachandran added 2 commits August 17, 2017 10:54

Main stack dependant on GCC toolchain removed

552fecc

Issue with mbed compile for GCC_ARM tool resolved

f79216c

Functions 'serial_break_set' & 'serial_break_clear' wrongly removed during 'SERIAL_FC disabled, critical section API Updation' commit. Due to this 'mbed compile' command for GCC_ARM tool fails to compile.

0xc0170 reviewed Aug 17, 2017

View reviewed changes

0xc0170 added ready for merge and removed needs: CI labels Aug 18, 2017

0xc0170 added needs: CI and removed ready for merge labels Aug 21, 2017

0xc0170 added ready for merge and removed needs: CI labels Aug 22, 2017

theotherjimmy merged commit 0a063a0 into ARMmbed:master Aug 24, 2017

theotherjimmy added release-version: 5.5.6 and removed ready for merge labels Aug 24, 2017

Add Support for TOSHIBA TMPM066 board #4840

Add Support for TOSHIBA TMPM066 board #4840

Uh oh!

Conversation

ghost commented Aug 1, 2017

Description

Status

Migrations

Test Status

Uh oh!

0xc0170 commented Aug 2, 2017

Uh oh!

ghost commented Aug 3, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ghost Aug 11, 2017 • edited by ghost Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ghost commented Aug 4, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

0xc0170 Aug 7, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

0xc0170 Aug 9, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

theotherjimmy commented Aug 10, 2017

Uh oh!

ghost Aug 11, 2017 •

edited by ghost

Loading

0xc0170 Aug 7, 2017 •

edited

Loading

0xc0170 Aug 9, 2017 •

edited

Loading

0xc0170 Aug 17, 2017 •

edited

Loading