dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-12-27 22:28:22 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

MIPS: MIPS32r1: Optimization of function WebPRescalerExportRow. [2/6]

Browse Source 
Change-Id: I420e646deefac5df9a8dcec3b02da8dff02b1167
This commit is contained in:
Djordje Pesut 2013-12-25 11:46:37 +01:00 committed by James Zern
parent 9882b2f953
commit 88230854e3
1 changed files with 54 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
src/utils/rescaler.c
View File

@ -92,14 +92,66 @@ void WebPRescalerImportRow(WebPRescaler* const wrk,
uint8_t* WebPRescalerExportRow(WebPRescaler* const wrk) { uint8_t* WebPRescalerExportRow(WebPRescaler* const wrk) {
if (wrk->y_accum <= 0) { if (wrk->y_accum <= 0) {
int x_out; int x_out = 0;
uint8_t* const dst = wrk->dst; uint8_t* const dst = wrk->dst;
int32_t* const irow = wrk->irow; int32_t* const irow = wrk->irow;
const int32_t* const frow = wrk->frow; const int32_t* const frow = wrk->frow;
const int yscale = wrk->fy_scale * (-wrk->y_accum); const int yscale = wrk->fy_scale * (-wrk->y_accum);
const int x_out_max = wrk->dst_width * wrk->num_channels; const int x_out_max = wrk->dst_width * wrk->num_channels;
for (x_out = 0; x_out < x_out_max; ++x_out) { #if defined(__mips__)
// if wrk->fxy_scale can fit into 32 bits use optimized code,
// otherwise use C code
if ((wrk->fxy_scale >> 32) == 0) {
int temp0, temp1, temp3, temp4, temp5, temp6, temp7, loop_end;
const int temp2 = (int)(wrk->fxy_scale);
const int temp8 = x_out_max << 2;
uint8_t* dst_t = (uint8_t*)dst;
int32_t* irow_t = (int32_t*)irow;
const int32_t* frow_t = (const int32_t*)frow;
__asm__ volatile(
"addiu %[temp6], $zero, -256 \n\t"
"addiu %[temp7], $zero, 255 \n\t"
"li %[temp3], 0x10000 \n\t"
"li %[temp4], 0x8000 \n\t"
"addu %[loop_end], %[frow_t], %[temp8] \n\t"
"1: \n\t"
"lw %[temp0], 0(%[frow_t]) \n\t"
"mult %[temp3], %[temp4] \n\t"
"addiu %[frow_t], %[frow_t], 4 \n\t"
"sll %[temp0], %[temp0], 2 \n\t"
"madd %[temp0], %[yscale] \n\t"
"mfhi %[temp1] \n\t"
"lw %[temp0], 0(%[irow_t]) \n\t"
"addiu %[dst_t], %[dst_t], 1 \n\t"
"addiu %[irow_t], %[irow_t], 4 \n\t"
"subu %[temp0], %[temp0], %[temp1] \n\t"
"mult %[temp3], %[temp4] \n\t"
"sll %[temp0], %[temp0], 2 \n\t"
"madd %[temp0], %[temp2] \n\t"
"mfhi %[temp5] \n\t"
"sw %[temp1], -4(%[irow_t]) \n\t"
"and %[temp0], %[temp5], %[temp6] \n\t"
"slti %[temp1], %[temp5], 0 \n\t"
"beqz %[temp0], 2f \n\t"
"xor %[temp5], %[temp5], %[temp5] \n\t"
"movz %[temp5], %[temp7], %[temp1] \n\t"
"2: \n\t"
"sb %[temp5], -1(%[dst_t]) \n\t"
"bne %[frow_t], %[loop_end], 1b \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp3]"=&r"(temp3),
[temp4]"=&r"(temp4), [temp5]"=&r"(temp5), [temp6]"=&r"(temp6),
[temp7]"=&r"(temp7), [frow_t]"+r"(frow_t), [irow_t]"+r"(irow_t),
[dst_t]"+r"(dst_t), [loop_end]"=&r"(loop_end)
: [temp2]"r"(temp2), [yscale]"r"(yscale), [temp8]"r"(temp8)
: "memory", "hi", "lo"
);
x_out = x_out_max;
}
#endif // __mips__
for (; x_out < x_out_max; ++x_out) {
const int frac = (int)MULT_FIX(frow[x_out], yscale); const int frac = (int)MULT_FIX(frow[x_out], yscale);
const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale); const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
dst[x_out] = (!(v & ~0xff)) ? v : (v < 0) ? 0 : 255; dst[x_out] = (!(v & ~0xff)) ? v : (v < 0) ? 0 : 255;