minicube.net .
Fork of mbed-src file paths change. LPC1114FN28 use only.
Fork of
mbed-src
by 
mbed official
Information
この情報は2013/10/28時点での解決方法です。
現在はmbed-src、標準ライブラリで問題なくコンパイルが可能です。
・使う物
LPC1114FN28
mbed SDK
LPC1114FN28でmbed-SDKのLibraryを使うとCompile出来ない。（2013/10/28）
パスが通ってないだけのようなのでファイルを以下に移動する。
| mbed-src\targets\cmsis\TARGET_NXP\TARGET_LPC11XX_11CXX\ mbed-src\targets\cmsis\TARGET_NXP\TARGET_LPC11XX_11CXX\TARGET_LPC11XX\ |
にあるファイルをすべて
| mbed-src\targets\cmsis\TARGET_NXP\ |
へ移動
mbed-src\targets\cmsis\TARGET_NXP\TARGET_LPC11XX_11CXX\にある
| TOOLCHAIN_ARM_MICRO |
をフォルダごと
| mbed-src\targets\cmsis\TARGET_NXP\ |
へ移動
| mbed-src\targets\hal\TARGET_NXP\TARGET_LPC11XX_11CXX\ mbed-src\targets\hal\TARGET_NXP\TARGET_LPC11XX_11CXX\TARGET_LPC11XX\ |
にあるファイルをすべて
| mbed-src\targets\hal\TARGET_NXP\ |
へ移動
移動後は以下のような構成になると思います。
※不要なファイルは削除してあります。
 |
ファイルの移動が面倒なので以下に本家からフォークしたライブラリを置いておきます。
Import librarymbed-src-LPC1114FN28
Fork of mbed-src file paths change. LPC1114FN28 use only.
Last commit 27 Oct 2013 by 
minicube.net .
エラーが出力される場合
"TOOLCHAIN_ARM_MICRO"が無いとエラーになる。
| Error: Undefined symbol _initial_sp (referred from entry2.o). Error: Undefined symbol _heap_base (referred from malloc.o). Error: Undefined symbol _heap_limit (referred from malloc.o). |
LPC1114FN28はMicrolibを使ってCompileされるため上記のエラーになるようです。
Diff: targets/cmsis/TARGET_NXP/cmsis_nvic.c
- Revision:
- 43:b3acfef78949
- Parent:
- 30:91c1d09ada54
--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/targets/cmsis/TARGET_NXP/cmsis_nvic.c Sun Oct 27 20:12:31 2013 +0000
@@ -0,0 +1,57 @@
+/* mbed Microcontroller Library - cmsis_nvic for LPC11U24
+ * Copyright (c) 2011 ARM Limited. All rights reserved.
+ *
+ * CMSIS-style functionality to support dynamic vectors
+ */
+
+#include "cmsis_nvic.h"
+
+/* In the M0, there is no VTOR. In the LPC range such as the LPC11U,
+ * whilst the vector table may only be something like 48 entries (192 bytes, 0xC0),
+ * the SYSMEMREMAP register actually remaps the memory from 0x10000000-0x100001FF
+ * to adress 0x0-0x1FF. In this case, RAM can be addressed at both 0x10000000 and 0x0
+ *
+ * If we just copy the vectors to RAM and switch the SYSMEMMAP, any accesses to FLASH
+ * above the vector table before 0x200 will actually go to RAM. So we need to provide
+ * a solution where the compiler gets the right results based on the memory map
+ *
+ * Option 1 - We allocate and copy 0x200 of RAM rather than just the table
+ * - const data and instructions before 0x200 will be copied to and fetched/exec from RAM
+ * - RAM overhead: 0x200 - 0xC0 = 320 bytes, FLASH overhead: 0
+ *
+ * Option 2 - We pad the flash to 0x200 to ensure the compiler doesn't allocate anything there
+ * - No flash accesses will go to ram, as there will be nothing there
+ * - RAM only needs to be allocated for the vectors, as all other ram addresses are normal
+ * - RAM overhead: 0, FLASH overhead: 320 bytes
+ *
+ * Option 2 is the one to go for, as RAM is the most valuable resource
+ */
+
+#define NVIC_RAM_VECTOR_ADDRESS (0x10000000) // Vectors positioned at start of RAM
+
+void NVIC_SetVector(IRQn_Type IRQn, uint32_t vector) {
+ int i;
+ // Space for dynamic vectors, initialised to allocate in R/W
+ static volatile uint32_t* vectors = (uint32_t*)NVIC_RAM_VECTOR_ADDRESS;
+
+ // Copy and switch to dynamic vectors if first time called
+ if((LPC_SYSCON->SYSMEMREMAP & 0x3) != 0x1) {
+ uint32_t *old_vectors = (uint32_t *)0; // FLASH vectors are at 0x0
+ for(i = 0; i < NVIC_NUM_VECTORS; i++) {
+ vectors[i] = old_vectors[i];
+ }
+ LPC_SYSCON->SYSMEMREMAP = 0x1; // Remaps 0x0-0x1FF FLASH block to RAM block
+ }
+
+ // Set the vector
+ vectors[IRQn + 16] = vector;
+}
+
+uint32_t NVIC_GetVector(IRQn_Type IRQn) {
+ // We can always read vectors at 0x0, as the addresses are remapped
+ uint32_t *vectors = (uint32_t*)0;
+
+ // Return the vector
+ return vectors[IRQn + 16];
+}
+