Macintosh-Tools/macemu
1
0
Fork 0
mirror of https://github.com/kanjitalk755/macemu.git synced 2024-10-02 11:54:46 +00:00
Code Issues Projects Releases Wiki Activity

Add raw_emit_nop_filler() with more efficient no-op fillers stolen from

Browse Source 
GNU binutils 2.12.90.0.15. Speed bump is marginal (less than 6%). Make it
default though, that's conditionalized by tune_nop_fillers constant.
This commit is contained in:
gbeauche 2002-10-12 16:27:13 +00:00
parent 7914fdc677
commit aa6b264d21
2 changed files with 77 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

62
BasiliskII/src/uae_cpu/compiler/codegen_x86.cpp
View File

@ -2057,6 +2057,68 @@ static __inline__ void raw_nop(void)
emit_byte(0x90);
}
static __inline__ void raw_emit_nop_filler(int nbytes)
{
/* Source: GNU Binutils 2.12.90.0.15 */
/* Various efficient no-op patterns for aligning code labels.
Note: Don't try to assemble the instructions in the comments.
0L and 0w are not legal. */
static const uae_u8 f32_1[] =
{0x90}; /* nop */
static const uae_u8 f32_2[] =
{0x89,0xf6}; /* movl %esi,%esi */
static const uae_u8 f32_3[] =
{0x8d,0x76,0x00}; /* leal 0(%esi),%esi */
static const uae_u8 f32_4[] =
{0x8d,0x74,0x26,0x00}; /* leal 0(%esi,1),%esi */
static const uae_u8 f32_5[] =
{0x90, /* nop */
0x8d,0x74,0x26,0x00}; /* leal 0(%esi,1),%esi */
static const uae_u8 f32_6[] =
{0x8d,0xb6,0x00,0x00,0x00,0x00}; /* leal 0L(%esi),%esi */
static const uae_u8 f32_7[] =
{0x8d,0xb4,0x26,0x00,0x00,0x00,0x00}; /* leal 0L(%esi,1),%esi */
static const uae_u8 f32_8[] =
{0x90, /* nop */
0x8d,0xb4,0x26,0x00,0x00,0x00,0x00}; /* leal 0L(%esi,1),%esi */
static const uae_u8 f32_9[] =
{0x89,0xf6, /* movl %esi,%esi */
0x8d,0xbc,0x27,0x00,0x00,0x00,0x00}; /* leal 0L(%edi,1),%edi */
static const uae_u8 f32_10[] =
{0x8d,0x76,0x00, /* leal 0(%esi),%esi */
0x8d,0xbc,0x27,0x00,0x00,0x00,0x00}; /* leal 0L(%edi,1),%edi */
static const uae_u8 f32_11[] =
{0x8d,0x74,0x26,0x00, /* leal 0(%esi,1),%esi */
0x8d,0xbc,0x27,0x00,0x00,0x00,0x00}; /* leal 0L(%edi,1),%edi */
static const uae_u8 f32_12[] =
{0x8d,0xb6,0x00,0x00,0x00,0x00, /* leal 0L(%esi),%esi */
0x8d,0xbf,0x00,0x00,0x00,0x00}; /* leal 0L(%edi),%edi */
static const uae_u8 f32_13[] =
{0x8d,0xb6,0x00,0x00,0x00,0x00, /* leal 0L(%esi),%esi */
0x8d,0xbc,0x27,0x00,0x00,0x00,0x00}; /* leal 0L(%edi,1),%edi */
static const uae_u8 f32_14[] =
{0x8d,0xb4,0x26,0x00,0x00,0x00,0x00, /* leal 0L(%esi,1),%esi */
0x8d,0xbc,0x27,0x00,0x00,0x00,0x00}; /* leal 0L(%edi,1),%edi */
static const uae_u8 f32_15[] =
{0xeb,0x0d,0x90,0x90,0x90,0x90,0x90, /* jmp .+15; lotsa nops */
0x90,0x90,0x90,0x90,0x90,0x90,0x90,0x90};
static const uae_u8 f32_16[] =
{0xeb,0x0d,0x90,0x90,0x90,0x90,0x90, /* jmp .+15; lotsa nops */
0x90,0x90,0x90,0x90,0x90,0x90,0x90,0x90,0x90};
static const uae_u8 *const f32_patt[] = {
f32_1, f32_2, f32_3, f32_4, f32_5, f32_6, f32_7, f32_8,
f32_9, f32_10, f32_11, f32_12, f32_13, f32_14, f32_15
};
int nloops = nbytes / 16;
while (nloops-- > 0)
emit_block(f32_16, sizeof(f32_16));
nbytes %= 16;
if (nbytes)
emit_block(f32_patt[nbytes - 1], nbytes);
}
/*************************************************************************
* Flag handling, to and fro UAE flag register *

19
BasiliskII/src/uae_cpu/compiler/compemu_support.cpp
View File

@ -128,7 +128,8 @@ static bool lazy_flush = true; // Flag: lazy translation cache invalidation
static bool avoid_fpu = true; // Flag: compile FPU instructions ?
static bool have_cmov = false; // target has CMOV instructions ?
static bool have_rat_stall = true; // target has partial register stalls ?
static bool tune_alignment = true; // Tune code alignments for running CPU ?
const bool tune_alignment = true; // Tune code alignments for running CPU ?
const bool tune_nop_fillers = true; // Tune no-op fillers for architecture
static int align_loops = 32; // Align the start of loops
static int align_jumps = 32; // Align the start of jumps
static int zero_fd = -1;
@ -698,6 +699,12 @@ static __inline__ void emit_long(uae_u32 x)
target+=4;
}
static __inline__ void emit_block(const uae_u8 *block, uae_u32 blocklen)
{
memcpy((uae_u8 *)target,block,blocklen);
target+=blocklen;
}
static __inline__ uae_u32 reverse32(uae_u32 v)
{
#if 1
@ -4965,9 +4972,13 @@ void freescratch(void)
static void align_target(uae_u32 a)
{
/* Fill with NOPs --- makes debugging with gdb easier */
while ((uae_u32)target&(a-1))
*target++=0x90;
if (tune_nop_fillers)
raw_emit_nop_filler(a - (((uae_u32)target) & (a - 1)));
else {
/* Fill with NOPs --- makes debugging with gdb easier */
while ((uae_u32)target&(a-1))
*target++=0x90;
}
}
static __inline__ int isinrom(uintptr addr)