kingkevin
/
espressif_tinyusb
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

clean up _ff_push_n

This commit is contained in:
hathach 2021-04-07 12:52:57 +07:00
parent d6737fb5fe
commit 9042e973d3
1 changed files with 41 additions and 50 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

91
src/common/tusb_fifo.c
View File

@ -168,18 +168,20 @@ static inline void _ff_push(tu_fifo_t* f, void const * data, uint16_t wRel)
// send n items to FIFO WITHOUT updating write pointer // send n items to FIFO WITHOUT updating write pointer
static void _ff_push_n(tu_fifo_t* f, void const * data, uint16_t n, uint16_t wRel, tu_fifo_copy_mode_t copy_mode) static void _ff_push_n(tu_fifo_t* f, void const * data, uint16_t n, uint16_t wRel, tu_fifo_copy_mode_t copy_mode)
{ {
uint16_t const nLin = f->depth - wRel;
uint16_t const nWrap = n - nLin; // only used if wrapped
switch (copy_mode) switch (copy_mode)
{ {
case TU_FIFO_COPY_INC: case TU_FIFO_COPY_INC:
if(n <= f->depth-wRel) if(n <= nLin)
{ {
// Linear mode only // Linear only
memcpy(f->buffer + (wRel * f->item_size), data, n*f->item_size); memcpy(f->buffer + (wRel * f->item_size), data, n*f->item_size);
} }
else else
{ {
// Wrap around // Wrap around
uint16_t nLin = f->depth - wRel;
// Write data to linear part of buffer // Write data to linear part of buffer
memcpy(f->buffer + (wRel * f->item_size), data, nLin*f->item_size); memcpy(f->buffer + (wRel * f->item_size), data, nLin*f->item_size);
@ -189,61 +191,51 @@ static void _ff_push_n(tu_fifo_t* f, void const * data, uint16_t n, uint16_t wRe
} }
break; break;
case TU_FIFO_COPY_CST_FULL_WORDS: // Intended for hardware buffers from which it can be read word by word only case TU_FIFO_COPY_CST_FULL_WORDS:
if(n <= f->depth-wRel) // Intended for hardware buffers from which it can be read word by word only
if(n <= nLin)
{ {
// Linear mode only // Linear only
_ff_push_const_addr(f->buffer + (wRel * f->item_size), data, n*f->item_size); _ff_push_const_addr(f->buffer + (wRel * f->item_size), data, n*f->item_size);
} }
else else
{ {
// Wrap around case // Wrap around case
uint16_t nLin = (f->depth - wRel) * f->item_size; uint16_t nLin_bytes = nLin * f->item_size;
uint16_t nWrap = (n - nLin) * f->item_size; uint16_t nWrap_bytes = nWrap * f->item_size;
// Optimize for fast word copies uint8_t* dst = f->buffer + (wRel * f->item_size);
typedef struct{
uint32_t val;
} __attribute((__packed__)) unaligned_uint32_t;
unaligned_uint32_t* dst = (unaligned_uint32_t*)(f->buffer + (wRel * f->item_size)); // Write full words to linear part of buffer
uint16_t nLin_4n_bytes = nLin_bytes & 0xFFFC;
_ff_push_const_addr(dst, data, nLin_4n_bytes);
dst += nLin_4n_bytes;
// There could be odd 1-3 bytes before the wrap-around boundary
volatile uint32_t * rx_fifo = (volatile uint32_t *) data; volatile uint32_t * rx_fifo = (volatile uint32_t *) data;
uint8_t rem = nLin_bytes & 0x03;
// Write full words of linear part to buffer
uint16_t full_words = nLin >> 2;
while(full_words--)
{
dst->val = *rx_fifo;
dst++;
}
uint8_t * dst_u8;
uint8_t rem = nLin & 0x03;
// Handle wrap around
if (rem > 0) if (rem > 0)
{ {
dst_u8 = (uint8_t *)dst; uint8_t remrem = tu_min16(nWrap_bytes, 4-rem);
uint8_t remrem = tu_min16(nWrap, 4-rem); nWrap_bytes -= remrem;
nWrap -= remrem;
uint32_t tmp = *rx_fifo; uint32_t tmp32 = *rx_fifo;
uint8_t * src_u8 = ((uint8_t *) &tmp); uint8_t * src_u8 = ((uint8_t *) &tmp32);
while(rem--)
{ // Write 1-3 bytes before wrapped boundary
*dst_u8++ = *src_u8++; while(rem--) *dst++ = *src_u8++;
}
dst_u8 = f->buffer; // Read more bytes to beginning to complete a word
while(remrem--) dst = f->buffer;
{ while(remrem--) *dst++ = *src_u8++;
*dst_u8++ = *src_u8++;
}
} }
else else
{ {
dst_u8 = f->buffer; dst = f->buffer; // wrap around to beginning
} }
// Final part // Write data wrapped part
if (nWrap > 0) _ff_push_const_addr(dst_u8, data, nWrap); if (nWrap_bytes > 0) _ff_push_const_addr(dst, data, nWrap_bytes);
} }
break; break;
} }
@ -284,7 +276,7 @@ static void _ff_pull_n(tu_fifo_t* f, void* p_buffer, uint16_t n, uint16_t rRel,
case TU_FIFO_COPY_CST_FULL_WORDS: case TU_FIFO_COPY_CST_FULL_WORDS:
if ( n <= nLin ) if ( n <= nLin )
{ {
// Linear mode only // Linear only
_ff_pull_const_addr(p_buffer, f->buffer + (rRel * f->item_size), n*f->item_size); _ff_pull_const_addr(p_buffer, f->buffer + (rRel * f->item_size), n*f->item_size);
} }
else else
@ -296,13 +288,12 @@ static void _ff_pull_n(tu_fifo_t* f, void* p_buffer, uint16_t n, uint16_t rRel,
uint8_t* src = f->buffer + (rRel * f->item_size); uint8_t* src = f->buffer + (rRel * f->item_size);
// Read data from linear part of buffer // Read full words from linear part of buffer
uint16_t nLin_4n_bytes = nLin_bytes & 0xFFFC; uint16_t nLin_4n_bytes = nLin_bytes & 0xFFFC;
_ff_pull_const_addr(p_buffer, src, nLin_4n_bytes); _ff_pull_const_addr(p_buffer, src, nLin_4n_bytes);
src += nLin_4n_bytes; src += nLin_4n_bytes;
// There could be odd 1-3 bytes before the wrap-around boundary // There could be odd 1-3 bytes before the wrap-around boundary
// Handle wrap around - do it manually as these are only 4 bytes and its faster without memcpy
volatile uint32_t * tx_fifo = (volatile uint32_t *) p_buffer; volatile uint32_t * tx_fifo = (volatile uint32_t *) p_buffer;
uint8_t rem = nLin_bytes & 0x03; uint8_t rem = nLin_bytes & 0x03;
if (rem > 0) if (rem > 0)
@ -310,17 +301,17 @@ static void _ff_pull_n(tu_fifo_t* f, void* p_buffer, uint16_t n, uint16_t rRel,
uint8_t remrem = tu_min16(nWrap_bytes, 4-rem); uint8_t remrem = tu_min16(nWrap_bytes, 4-rem);
nWrap_bytes -= remrem; nWrap_bytes -= remrem;
uint32_t tmp; uint32_t tmp32;
uint8_t * dst_u8 = (uint8_t *)&tmp; uint8_t * dst_u8 = (uint8_t *)&tmp32;
// Get 1-3 bytes before wrapped boundary // Read 1-3 bytes before wrapped boundary
while(rem--) *dst_u8++ = *src++; while(rem--) *dst_u8++ = *src++;
// Get more bytes from beginning to form a complete word // Read more bytes from beginning to complete a word
src = f->buffer; src = f->buffer;
while(remrem--) *dst_u8++ = *src++; while(remrem--) *dst_u8++ = *src++;
*tx_fifo = tmp; *tx_fifo = tmp32;
} }
else else
{ {