Because N8 CPU doesn't have floating point unit,
so we implement an extra floating point engine
(single-precision addition, subtraction, multiplication,
and division) into it8320 to improve performance of
floating point operation.
To make CPU's instruction compatible, we use register (DLMB)
to switch ALU (Arithmetic Logic Unit). eg:
Instruction 'ADD45' adds the contents of two registers then
writes the result to the source register.
But if we switch ALU to floating point operation mode,
this instruction will do a floating-point addition instead.
For the other FPU that we don't support as far,
we have to use soft float library routines of nds32.
Signed-off-by: Dino Li <dino.li@ite.com.tw>
BRANCH=none
BUG=none
TEST=add the following console command and test different
scenarios by changing variable a and b.
#define PRINTF_FLOAT(x) ((int)((x) * 1000.0f))
static int it83xx_fpu_test(int argc, char **argv)
{
volatile float a = 1.23f;
volatile float b = 4.56f;
volatile float c;
c = a + b;
ccprintf("__addsf3: (%d)\n", PRINTF_FLOAT(c));
c = a - b;
ccprintf("__subsf3: (%d)\n", PRINTF_FLOAT(c));
c = a * b;
ccprintf("__mulsf3: (%d)\n", PRINTF_FLOAT(c));
c = a / b;
ccprintf("__divsf3: (%d)\n", PRINTF_FLOAT(c));
return EC_SUCCESS;
}
DECLARE_CONSOLE_COMMAND(fpu, it83xx_fpu_test, "", "");
Change-Id: I4fc1c08d8c2376156bec9f098491187675c4a88f
Reviewed-on: https://chromium-review.googlesource.com/427640
Commit-Ready: Dino Li <Dino.Li@ite.com.tw>
Tested-by: Dino Li <Dino.Li@ite.com.tw>
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>
We use the latest nde32 official toolchain to verify the bug of
GP-related instruction was fixed (we can remove "-mno-gp-direct" flag to
save code space), and code optimization.
We got some error on official toolchain, so we fix them as well:
- "break" instead of "trap".
N8 CPU does not support "trap" instruction to generate an unconditional
trap exception, but no error occurred while compiling by using current
GCC toolchain (this will trigger a reserved instruction exception,
so "ASSERT()" still work).
We use "break" to generate a exception in this change.
- "li" instead of "la".
To fix the error: "Error: la must use with symbol".
Also fix:
- The "_bss_start" has to be word-aligned because we use "lwi" instruction
to load a word from the memory into the general register.
BRANCH=none
BUG=none
TEST=1. console command "crash assert".
2. check registers settings: f02030h, f0203eh, and f02044h.
Change-Id: I33404a1d60eeebfa135bf43d3a7d5e73ab35c678
Reviewed-on: https://chromium-review.googlesource.com/422608
Commit-Ready: Dino Li <Dino.Li@ite.com.tw>
Tested-by: Dino Li <Dino.Li@ite.com.tw>
Reviewed-by: Randall Spangler <rspangler@chromium.org>
We have a limitation for EC interrupt vector registers.
System may read incorrect interrupt number in ISR so we need to add
a workaround to prevent it.
The following is a example that got incorrect interrupt number:
1. REG IVCTx = 0x10. (no interrupt pending)
2. EC INT6 interrupt occurs (IVCTx = 0x16) and jump to ISR.
3. Read interrupt vector register to determine interrupt number.
4. Higher priority interrupt of same interrupt group occurs
(for example: INT134, IVCTx = 0x96) while the system is reading the
interrupt vector register for EC INT6, we may end up with an incorrect
interrupt number between 0x16 and 0x96.
Signed-off-by: Dino Li <dino.li@ite.com.tw>
BRANCH=none
BUG=none
TEST=1. EC interrupts work normally: WUI (GPIO interrupt), timer, uart,
LPC, I2C, and PECI.
2. Console command 'taskinfo'.
Change-Id: I54e61f417ad506eb3b4cd5d0652f64eed9a28a17
Reviewed-on: https://chromium-review.googlesource.com/322097
Commit-Ready: Dino Li <dino.li@ite.com.tw>
Tested-by: Dino Li <dino.li@ite.com.tw>
Reviewed-by: Randall Spangler <rspangler@chromium.org>
[symptom]
The 'vboot_hash_start()' always return busy error and variable 'in_progress'
got a strange value(should be 0 or 1).
'start_irq_handler()' causes scratchpad overflow in first context switch.
It must be called after SP switch to system stack in ISR.
NOTE:
The scratchpad is still also need more size even if
'start_irq_handler()' is using system stack. following is detail:
1. uint32_t scratchpad[19] 0x81d34 ~ 0x81d7f
[__task_start:]
2. /* put the dummy stack pointer at the top of the stack in scratchpad */
addi $sp, $r3, 4 * 18
-> SP 0x81d7c
3. syscall push return address (-4)
-> SP 0x81d78
[ISR:]
4. push r15, fp, lp, and sp (-0x10)
-> SP 0x81d68
5. push r0 ~ r5 (-0x18)
-> SP 0x81d50
[__switch_task:]
6. /* save ipsw, ipc, r6, r7, r8, r9, r10 on the current process stack */ (-0x1C)
-> SP 0x81d34
Signed-off-by: Dino Li <dino.li@ite.com.tw>
BRANCH=none
BUG=none
TEST=1. hash done.
2. console command 'taskinfo'.
3. the scratchpad does not overflow after first context switch.
Change-Id: If5d89ff5c945a777010492fcfb54bf41f434ed69
Reviewed-on: https://chromium-review.googlesource.com/317468
Commit-Ready: Dino Li <dino.li@ite.com.tw>
Tested-by: Dino Li <dino.li@ite.com.tw>
Reviewed-by: Randall Spangler <rspangler@chromium.org>
[chip config]
1. No hardware specific udelay().
2. Enable watchdog.
[watchdog]
3. Watchdog period is "CONFIG_WATCHDOG_PERIOD_MS" of config.h.
4. Watchdog auxiliary timer period is "CONFIG_AUX_TIMER_PERIOD_MS".
[task and irq]
5. Write 1 to clear interrupt pending status, no |.
6. A global variable for store interrupt number of software interrupt.
[uart]
7. Always reset UART module before config it.
[hwtimer]
8. Use more external timers for HW timer module.
[task]
9. Fix task profiling.
Signed-off-by: Dino Li <dino.li@ite.com.tw>
BRANCH=none
BUG=none
TEST=[watchdog]
1. console "waitms 1100", only pre-watchdog warning message.
2. console "waitms 1600", warning message and watchdog reset.
[hwtimer]
3. console commands "gettime", "timerinfo", and "forcetime".
4. enable hook debug and there is no delayed by more than 10%
warning message over 48 hours.
5. There is no watchdog reset too.
[task]
6. console 'taskinfo'
Task Ready Name Events Time (s) StkUsed
0 R << idle >> 00000000 32.927724 308/512
1 HOOKS 00000000 0.034267 372/768
2 R CONSOLE 00000000 0.116763 468/768
3 HOSTCMD 00000000 0.000641 372/512
4 KEYPROTO 00000000 0.000042 212/512
5 KEYSCAN 00000000 0.000908 356/512
IRQ counts by type:
38 2932
155 1
158 261
160 67
Service calls: 87
Total exceptions: 3348
Task switches: 167
Task switching started: 0.001999 s
Time in tasks: 33.282819 s
Time in exceptions: 0.164717 s
Change-Id: I234085cec231cd855d2a5e639ea1b0966c61d796
Reviewed-on: https://chromium-review.googlesource.com/296939
Commit-Ready: Dino Li <dino.li@ite.com.tw>
Tested-by: Dino Li <dino.li@ite.com.tw>
Reviewed-by: Randall Spangler <rspangler@chromium.org>
When there is an interrupt event, N8 CPU will save PSW register to IPSW
register and clear GIE then jump to interrupt service routine. N8 will
restore PSW from IPSW after "iret" instruction (the above are purely
hardware mechanism).
Nested interrupt will occur if we set GIE again in interrupt context.
symptom:
power button pressed while LID open -> exception or unknown reset.
Signed-off-by: Dino Li <dino.li@ite.com.tw>
BRANCH=none
BUG=none
TEST=1. Manually pressed power button x200.
2. Console "eflash" erase and write eflash OK.
Change-Id: Ic04a23d473ebc6417dffea814a27583cb8d63a1f
Reviewed-on: https://chromium-review.googlesource.com/289437
Reviewed-by: Randall Spangler <rspangler@chromium.org>
Commit-Queue: Dino Li <dino.li@ite.com.tw>
Tested-by: Dino Li <dino.li@ite.com.tw>
Some section(entry point, interrupt vector, e-flash signature
and so on) of linker script file are not linked.
The start address of e-flash signature should always at 00000080h.
Default firmware treats VCC logic high to prevent pin 11 logic low
but use following functions.
(EC2I, KBC, SWUC, PMC, CIR, SSPI, UART, BRAM, and PECI)
Signed-off-by: Dino Li <dino.li@ite.com.tw>
BRANCH=none
BUG=none
TEST=Firmware can startup on IT8380 emulation board.
Change-Id: I9860ac5b99dcc6e9e00dbc9d1e79a141237b7789
Reviewed-on: https://chromium-review.googlesource.com/190008
Reviewed-by: Alec Berg <alecaberg@chromium.org>
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>
Tested-by: Dino Li <dino.li@ite.com.tw>
Commit-Queue: Dino Li <dino.li@ite.com.tw>
