On Wed, Oct 7, 2009 at 1:13 PM, Nishanth Menon nm@ti.com wrote:
Remove the predefined static initialization and generate the map dynamically to reduce code size.
This patch benefits were pointed out by Peter: http://www.nabble.com/forum/Permalink.jtp?root=25518020&post=25523748&am...
Signed-off-by: Nishanth Menon nm@ti.com Cc: Peter Tyser ptyser@xes-inc.com Cc: Sandeep Paulraj s-paulraj@ti.com Cc: Tom Rix tom.rix@windriver.com
common/dlmalloc.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)
diff --git a/common/dlmalloc.c b/common/dlmalloc.c index 241db8c..25e5314 100644 --- a/common/dlmalloc.c +++ b/common/dlmalloc.c @@ -1474,6 +1474,7 @@ typedef struct malloc_chunk* mbinptr;
#define IAV(i) bin_at(i), bin_at(i)
+#ifdef CONFIG_X86 static mbinptr av_[NAV * 2 + 2] = { 0, 0, IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7), @@ -1493,6 +1494,9 @@ static mbinptr av_[NAV * 2 + 2] = { IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119), IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127) }; +#else +static mbinptr av_[NAV * 2 + 2]; +#endif
Is there any reason why X86 is treated differently? I know the previous patch to dlmalloc.c did, but in the end it really didn't need to and ended up breaking the x86 build anyway (my fault for not getting time to test it before it went mainline)
Feel free to remove this #ifdef unless you are 100% sure it IS needed. Mainline x86 is broken anyway - if this change introduces another break, I will submit a consolidated fix soon
void malloc_bin_reloc (void) { @@ -1527,6 +1531,11 @@ void *sbrk(ptrdiff_t increment) */ void mem_malloc_init(ulong start, ulong size) {
u8 i;av_[0] = av_[1] = 0;for (i = 0; i < 128; i++)av_[2 + i * 2] = av_[2 + i * 2 + 1] = bin_at(i);mem_malloc_start = start; mem_malloc_end = start + size; mem_malloc_brk = start;-- 1.6.0.4
Regards,
Graeme