Use cached region for benchmarks, added some optimization opts.

CMakeLists.txt

@@ -26,14 +26,16 @@ set(TARGET_INCLUDES
 )
 
 set(TARGET_DEFS
-    "ITERATIONS=65536"
+    "ITERATIONS=2000UL"
     "PERFORMANCE_RUN=1"
 )
 
 set(TARGET_FLAGS_HARDWARE "-m4-300 -ml")
 
-set(CMAKE_C_FLAGS_DEBUG          "-DDEBUG -g -O2")
+set(TARGET_MANUAL_OPTS "-fauto-inc-dec -fbranch-count-reg -fcombine-stack-adjustments -fcompare-elim -fcprop-registers -fdce -fdefer-pop -fdelayed-branch -fdse -fforward-propagate")
-set(CMAKE_CXX_FLAGS_DEBUG        "-DDEBUG -g -O2")
+
+set(CMAKE_C_FLAGS_DEBUG          "-DDEBUG -g -O0 ${TARGET_MANUAL_OPTS}")
+set(CMAKE_CXX_FLAGS_DEBUG        "-DDEBUG -g -O0 ${TARGET_MANUAL_OPTS}")
 set(CMAKE_EXE_LINKER_FLAGS_DEBUG "")
 
 set(CMAKE_C_FLAGS_RELEASE          "-DNDEBUG -O2 -flto")
@@ -42,7 +44,7 @@ set(CMAKE_EXE_LINKER_FLAGS_RELEASE "-flto")
 
 set(CMAKE_C_FLAGS          "${TARGET_FLAGS_HARDWARE} -Wall -ffunction-sections -fdata-sections")
 set(CMAKE_CXX_FLAGS        "${TARGET_FLAGS_HARDWARE} -Wall -ffunction-sections -fdata-sections")
-set(CMAKE_EXE_LINKER_FLAGS "${TARGET_FLAGS_HARDWARE} -Wall -lc -lm -nostartfiles -Wl,--print-memory-usage -Wl,--gc-sections")
+set(CMAKE_EXE_LINKER_FLAGS "${TARGET_FLAGS_HARDWARE} -specs=nosys.specs -Wall -lc -lm -nostartfiles -Wl,--print-memory-usage -Wl,--gc-sections")
 
 add_compile_definitions(${TARGET_DEFS})
 include_directories(${TARGET_INCLUDES})

include/core_portme.h

@@ -78,7 +78,7 @@ typedef uint32_t CORE_TICKS;
 #endif
 #endif
 #ifndef COMPILER_FLAGS
-#define COMPILER_FLAGS "-O2"
+#define COMPILER_FLAGS "-O0"
 #endif
 #ifndef MEM_LOCATION
 #define MEM_LOCATION "DDR2 LMI"
@@ -98,7 +98,7 @@ typedef int32_t  ee_s32;
 typedef float    ee_f32;
 typedef uint8_t  ee_u8;
 typedef uint32_t ee_u32;
-typedef ee_u32   ee_ptr_int;
+typedef ee_s32   ee_ptr_int;
 typedef size_t   ee_size_t;
 /* align_mem :
         This macro is used to align an offset to point to a 32b value. It is
@@ -167,7 +167,7 @@ typedef size_t   ee_size_t;
    greater then 1.
 */
 #ifndef MAIN_HAS_NOARGC
-#define MAIN_HAS_NOARGC 0
+#define MAIN_HAS_NOARGC 1
 #endif
 
 /* Configuration : MAIN_HAS_NORETURN

lib/coremark

@@ -1 +1 @@
-Subproject commit cfa9ab377835911f23d9b0831c7be302ed1f58de
+Subproject commit eefc986ebd3452d6adde22eafaff3e5c859f29e4

src/core_portme.c

@@ -51,7 +51,7 @@ volatile ee_s32 seed5_volatile = 0;
    time.h and windows.h definitions included.
 */
 
-#define EE_TICKS_PER_SEC 97656
+#define EE_TICKS_PER_SEC (100000000.0 / 1024.0)
 
 /** Define Host specific (POSIX), or target specific global time variables. */
 static uint32_t start_time_val, stop_time_val;
@@ -86,8 +86,6 @@ void start_time(void) {
 void stop_time(void) {
     TMU->TSTR &= ~TMU_TSTR_STR0_Msk; /* Stop counter */
     stop_time_val = TMU->TCNT0;
-
-    printf("Stop time: 0x%09lx\r\n", stop_time_val);
 }
 /* Function : get_time
         Return an abstract "ticks" number that signifies time on the system.
@@ -99,9 +97,7 @@ void stop_time(void) {
    controlled by <TIMER_RES_DIVIDER>
 */
 CORE_TICKS get_time(void) {
-    CORE_TICKS elapsed = (CORE_TICKS)(start_time_val - stop_time_val);
+    return start_time_val - stop_time_val;
-    printf("Get time: 0x%08lx\r\n", elapsed);
-    return elapsed;
 }
 /* Function : time_in_secs
         Convert the value returned by get_time to seconds.
@@ -142,8 +138,6 @@ void portable_init(core_portable *p, int *argc, char *argv[]) {
 
     uart_init();
 
-    ee_printf("Portable initialized\r\n");
-
     if (sizeof(ee_ptr_int) != sizeof(ee_u8 *)) {
         ee_printf(
             "ERROR! Please define ee_ptr_int to a type that holds a "
@@ -162,7 +156,14 @@ void portable_fini(core_portable *p) {
 }
 
 void _putchar(char ch) {
-    while (CONSOLE_ASC->STA & 1 << 9U) {
+    if (ch == '\n') {
+        while (CONSOLE_ASC->STA & (1 << 9U)) {
+            /**/
+        }
+
+        CONSOLE_ASC->TX_BUF = '\r';
+    }
+    while (CONSOLE_ASC->STA & (1 << 9U)) {
         // wait for TX FIFO slot.
     }
     CONSOLE_ASC->TX_BUF = ch;

startup_stx7105.S

@@ -63,10 +63,11 @@ _main_entry:
     jsr @r0
     or r0, r0
 
-    mov r0, r4
+    mov.l _exit_loop_k, r0
-    mov.l _exit_k, r0
+_exit_loop:
-    jsr @r0
+    sleep
-    or r0, r0
+    jmp @r0
+    nop
 
     .balign 4
 /* libc FPU routine */
@@ -88,8 +89,8 @@ _end_k:
 /* Function pointers */
 _main_k:
     .long _main /* Same address as main */
-_exit_k:
+_exit_loop_k:
-    .long _exit
+    .long _exit_loop
 _exc_base_k:
     .long _exc_base
 _exc_imask_k:

stx7105.ld

@@ -7,8 +7,8 @@ ENTRY(_start)
 /* We don't use 29-bit mode since PMB and LMI initialization has to be done anyway. */
 
 MEMORY {
-    EMI  (rx) : ORIGIN = 0x80000000, LENGTH = 0x01000000 /* LMI virtual address: 0x8000_0000 */
+    EMI  (rx) : ORIGIN = 0x90000000, LENGTH = 0x01000000 /* LMI virtual address: 0x8000_0000 */
-    LMI (rwx) : ORIGIN = 0x81000000, LENGTH = 0x0F000000 /* LMI virtual address: 0x8100_0000 */
+    LMI (rwx) : ORIGIN = 0x91000000, LENGTH = 0x0F000000 /* LMI virtual address: 0x8100_0000 */
 }
 
 SECTIONS {