Skip to content

F
ffmpeg.wasm-core
Commit d7117138 authored by Nedeljko Babic's avatar Nedeljko Babic Committed by Michael Niedermayer
Browse files
Options

mips: ac3 downmix updated to the new data layout. 

Signed-off-by: 's avatarNedeljko Babic <nbabic@mips.com>
Signed-off-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parent 2dbc84b1
Hide whitespace changes
Inline Side-by-side
Showing with 127 additions and 87 deletions
libavcodec/mips/ac3dsp_mips.c
View file @ d7117138
...@@ -26,7 +26,8 @@ ...@@ -26,7 +26,8 @@
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE. * SUCH DAMAGE.
* *
* Author: Branimir Vasic (bvasic@mips.com) * Authors: Branimir Vasic (bvasic@mips.com)
* Nedeljko Babic (nbabic@mips.com)
* *
* Various AC-3 DSP Utils optimized for MIPS * Various AC-3 DSP Utils optimized for MIPS
* *
...@@ -198,7 +199,7 @@ static void ac3_update_bap_counts_mips(uint16_t mant_cnt[16], uint8_t *bap, ...@@ -198,7 +199,7 @@ static void ac3_update_bap_counts_mips(uint16_t mant_cnt[16], uint8_t *bap,
} }
#endif #endif
#if HAVE_MIPSFPU #if HAVE_MIPSFPU && HAVE_MIPS32R2
static void float_to_fixed24_mips(int32_t *dst, const float *src, unsigned int len) static void float_to_fixed24_mips(int32_t *dst, const float *src, unsigned int len)
{ {
const float scale = 1 << 24; const float scale = 1 << 24;
...@@ -266,93 +267,132 @@ static void float_to_fixed24_mips(int32_t *dst, const float *src, unsigned int l ...@@ -266,93 +267,132 @@ static void float_to_fixed24_mips(int32_t *dst, const float *src, unsigned int l
} while (len > 0); } while (len > 0);
} }
static void ac3_downmix_mips(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len) static void ac3_downmix_mips(float **samples, float (*matrix)[2],
int out_ch, int in_ch, int len)
{ {
int i, j; int i, j, i1, i2, i3;
float v0, v1, v2, v3; float v0, v1, v2, v3;
float v4, v5, v6, v7; float v4, v5, v6, v7;
float samples0, samples1, samples2, samples3, matrix_j, matrix_j2; float samples0, samples1, samples2, samples3, matrix_j, matrix_j2;
float *samples_p,*matrix_p; float *samples_p,*matrix_p, **samples_x, **samples_end, **samples_sw;
if (out_ch == 2) {
for (i = 0; i < len; i += 4) { __asm__ volatile(
v0 = v1 = v2 = v3 = 0.0f; ".set push \n\t"
v4 = v5 = v6 = v7 = 0.0f; ".set noreorder \n\t"
samples_p = &samples[0][i];
matrix_p = &matrix[0][0]; "li %[i1], 2 \n\t"
__asm__ volatile ( "sll %[len], 2 \n\t"
"move %[j], $zero \n\t" "move %[i], $zero \n\t"
"1: \n\t" "sll %[j], %[in_ch], 2 \n\t"
"lwc1 %[matrix_j], 0(%[matrix_p]) \n\t"
"lwc1 %[matrix_j2], 4(%[matrix_p]) \n\t" "bne %[out_ch], %[i1], 3f \n\t" // if (out_ch == 2)
"lwc1 %[samples0], 0(%[samples_p]) \n\t" " li %[i2], 1 \n\t"
"lwc1 %[samples1], 4(%[samples_p]) \n\t"
"lwc1 %[samples2], 8(%[samples_p]) \n\t" "2: \n\t" // start of the for loop (for (i = 0; i < len; i+=4))
"lwc1 %[samples3], 12(%[samples_p]) \n\t" "move %[matrix_p], %[matrix] \n\t"
"addiu %[matrix_p], 8 \n\t" "move %[samples_x], %[samples] \n\t"
"madd.s %[v0], %[v0], %[samples0], %[matrix_j] \n\t" "mtc1 $zero, %[v0] \n\t"
"madd.s %[v1], %[v1], %[samples1], %[matrix_j] \n\t" "mtc1 $zero, %[v1] \n\t"
"madd.s %[v2], %[v2], %[samples2], %[matrix_j] \n\t" "mtc1 $zero, %[v2] \n\t"
"madd.s %[v3], %[v3], %[samples3], %[matrix_j] \n\t" "mtc1 $zero, %[v3] \n\t"
"madd.s %[v4], %[v4], %[samples0], %[matrix_j2]\n\t" "mtc1 $zero, %[v4] \n\t"
"madd.s %[v5], %[v5], %[samples1], %[matrix_j2]\n\t" "mtc1 $zero, %[v5] \n\t"
"madd.s %[v6], %[v6], %[samples2], %[matrix_j2]\n\t" "mtc1 $zero, %[v6] \n\t"
"madd.s %[v7], %[v7], %[samples3], %[matrix_j2]\n\t" "mtc1 $zero, %[v7] \n\t"
"addiu %[j], 1 \n\t" "addiu %[i1], %[i], 4 \n\t"
"addiu %[samples_p], 1024 \n\t" "addiu %[i2], %[i], 8 \n\t"
"bne %[j], %[in_ch], 1b \n\t" "lw %[samples_p], 0(%[samples_x]) \n\t"
:[samples0]"=&f"(samples0), [samples1]"=&f"(samples1), "addiu %[i3], %[i], 12 \n\t"
[samples2]"=&f"(samples2), [samples3]"=&f"(samples3), "addu %[samples_end], %[samples_x], %[j] \n\t"
[samples_p]"+r"(samples_p), [matrix_j]"=&f"(matrix_j), "move %[samples_sw], %[samples_p] \n\t"
[matrix_p]"+r"(matrix_p), [v0]"+f"(v0), [v1]"+f"(v1),
[v2]"+f"(v2), [v3]"+f"(v3), [v4]"+f"(v4), [v5]"+f"(v5), "1: \n\t" // start of the inner for loop (for (j = 0; j < in_ch; j++))
[v6]"+f"(v6), [v7]"+f"(v7),[j]"=&r"(j), [matrix_j2]"=&f"(matrix_j2) "lwc1 %[matrix_j], 0(%[matrix_p]) \n\t"
:[in_ch]"r"(in_ch) "lwc1 %[matrix_j2], 4(%[matrix_p]) \n\t"
:"memory" "lwxc1 %[samples0], %[i](%[samples_p]) \n\t"
); "lwxc1 %[samples1], %[i1](%[samples_p]) \n\t"
samples[0][i ] = v0; "lwxc1 %[samples2], %[i2](%[samples_p]) \n\t"
samples[0][i+1] = v1; "lwxc1 %[samples3], %[i3](%[samples_p]) \n\t"
samples[0][i+2] = v2; "addiu %[matrix_p], 8 \n\t"
samples[0][i+3] = v3; "addiu %[samples_x], 4 \n\t"
samples[1][i ] = v4; "madd.s %[v0], %[v0], %[samples0], %[matrix_j] \n\t"
samples[1][i+1] = v5; "madd.s %[v1], %[v1], %[samples1], %[matrix_j] \n\t"
samples[1][i+2] = v6; "madd.s %[v2], %[v2], %[samples2], %[matrix_j] \n\t"
samples[1][i+3] = v7; "madd.s %[v3], %[v3], %[samples3], %[matrix_j] \n\t"
} "madd.s %[v4], %[v4], %[samples0], %[matrix_j2]\n\t"
} else if (out_ch == 1) { "madd.s %[v5], %[v5], %[samples1], %[matrix_j2]\n\t"
for (i = 0; i < len; i += 4) { "madd.s %[v6], %[v6], %[samples2], %[matrix_j2]\n\t"
v0 = v1 = v2 = v3 = 0.0f; "madd.s %[v7], %[v7], %[samples3], %[matrix_j2]\n\t"
samples_p = &samples[0][i]; "bne %[samples_x], %[samples_end], 1b \n\t"
matrix_p = &matrix[0][0]; " lw %[samples_p], 0(%[samples_x]) \n\t"
__asm__ volatile (
"move %[j], $zero \n\t" "lw %[samples_p], 4(%[samples]) \n\t"
"1: \n\t" "swxc1 %[v0], %[i](%[samples_sw]) \n\t"
"lwc1 %[matrix_j], 0(%[matrix_p]) \n\t" "swxc1 %[v1], %[i1](%[samples_sw]) \n\t"
"lwc1 %[samples0], 0(%[samples_p]) \n\t" "swxc1 %[v2], %[i2](%[samples_sw]) \n\t"
"lwc1 %[samples1], 4(%[samples_p]) \n\t" "swxc1 %[v3], %[i3](%[samples_sw]) \n\t"
"lwc1 %[samples2], 8(%[samples_p]) \n\t" "swxc1 %[v4], %[i](%[samples_p]) \n\t"
"lwc1 %[samples3], 12(%[samples_p]) \n\t" "addiu %[i], 16 \n\t"
"addiu %[matrix_p], 8 \n\t" "swxc1 %[v5], %[i1](%[samples_p]) \n\t"
"madd.s %[v0], %[v0], %[samples0], %[matrix_j] \n\t" "swxc1 %[v6], %[i2](%[samples_p]) \n\t"
"madd.s %[v1], %[v1], %[samples1], %[matrix_j] \n\t" "bne %[i], %[len], 2b \n\t"
"madd.s %[v2], %[v2], %[samples2], %[matrix_j] \n\t" " swxc1 %[v7], %[i3](%[samples_p]) \n\t"
"madd.s %[v3], %[v3], %[samples3], %[matrix_j] \n\t"
"addiu %[j], 1 \n\t" "3: \n\t"
"addiu %[samples_p], 1024 \n\t" "bne %[out_ch], %[i2], 6f \n\t" // if (out_ch == 1)
"bne %[j], %[in_ch], 1b \n\t" " nop \n\t"
:[samples0]"=&f"(samples0), [samples1]"=&f"(samples1),
[samples2]"=&f"(samples2), [samples3]"=&f"(samples3), "5: \n\t" // start of the outer for loop (for (i = 0; i < len; i+=4))
[samples_p]"+r"(samples_p), [matrix_j]"=&f"(matrix_j), "move %[matrix_p], %[matrix] \n\t"
[matrix_p]"+r"(matrix_p), [v0]"+f"(v0), [v1]"+f"(v1), "move %[samples_x], %[samples] \n\t"
[v2]"+f"(v2), [v3]"+f"(v3), [j]"=&r"(j) "mtc1 $zero, %[v0] \n\t"
:[in_ch]"r"(in_ch) "mtc1 $zero, %[v1] \n\t"
:"memory" "mtc1 $zero, %[v2] \n\t"
); "mtc1 $zero, %[v3] \n\t"
samples[0][i ] = v0; "addiu %[i1], %[i], 4 \n\t"
samples[0][i+1] = v1; "addiu %[i2], %[i], 8 \n\t"
samples[0][i+2] = v2; "lw %[samples_p], 0(%[samples_x]) \n\t"
samples[0][i+3] = v3; "addiu %[i3], %[i], 12 \n\t"
} "addu %[samples_end], %[samples_x], %[j] \n\t"
} "move %[samples_sw], %[samples_p] \n\t"
"4: \n\t" // start of the inner for loop (for (j = 0; j < in_ch; j++))
"lwc1 %[matrix_j], 0(%[matrix_p]) \n\t"
"lwxc1 %[samples0], %[i](%[samples_p]) \n\t"
"lwxc1 %[samples1], %[i1](%[samples_p]) \n\t"
"lwxc1 %[samples2], %[i2](%[samples_p]) \n\t"
"lwxc1 %[samples3], %[i3](%[samples_p]) \n\t"
"addiu %[matrix_p], 8 \n\t"
"addiu %[samples_x], 4 \n\t"
"madd.s %[v0], %[v0], %[samples0], %[matrix_j] \n\t"
"madd.s %[v1], %[v1], %[samples1], %[matrix_j] \n\t"
"madd.s %[v2], %[v2], %[samples2], %[matrix_j] \n\t"
"madd.s %[v3], %[v3], %[samples3], %[matrix_j] \n\t"
"bne %[samples_x], %[samples_end], 4b \n\t"
" lw %[samples_p], 0(%[samples_x]) \n\t"
"swxc1 %[v0], %[i](%[samples_sw]) \n\t"
"addiu %[i], 16 \n\t"
"swxc1 %[v1], %[i1](%[samples_sw]) \n\t"
"swxc1 %[v2], %[i2](%[samples_sw]) \n\t"
"bne %[i], %[len], 5b \n\t"
" swxc1 %[v3], %[i3](%[samples_sw]) \n\t"
"6: \n\t"
".set pop"
:[samples_p]"=&r"(samples_p), [matrix_j]"=&f"(matrix_j), [matrix_j2]"=&f"(matrix_j2),
[samples0]"=&f"(samples0), [samples1]"=&f"(samples1),
[samples2]"=&f"(samples2), [samples3]"=&f"(samples3),
[v0]"=&f"(v0), [v1]"=&f"(v1), [v2]"=&f"(v2), [v3]"=&f"(v3),
[v4]"=&f"(v4), [v5]"=&f"(v5), [v6]"=&f"(v6), [v7]"=&f"(v7),
[samples_x]"=&r"(samples_x), [matrix_p]"=&r"(matrix_p),
[samples_end]"=&r"(samples_end), [samples_sw]"=&r"(samples_sw),
[i1]"=&r"(i1), [i2]"=&r"(i2), [i3]"=&r"(i3), [i]"=&r"(i),
[j]"=&r"(j), [len]"+r"(len)
:[samples]"r"(samples), [matrix]"r"(matrix),
[in_ch]"r"(in_ch), [out_ch]"r"(out_ch)
:"memory"
);
} }
#endif #endif
#endif /* HAVE_INLINE_ASM */ #endif /* HAVE_INLINE_ASM */
...@@ -363,9 +403,9 @@ void ff_ac3dsp_init_mips(AC3DSPContext *c, int bit_exact) { ...@@ -363,9 +403,9 @@ void ff_ac3dsp_init_mips(AC3DSPContext *c, int bit_exact) {
c->bit_alloc_calc_bap = ac3_bit_alloc_calc_bap_mips; c->bit_alloc_calc_bap = ac3_bit_alloc_calc_bap_mips;
c->update_bap_counts = ac3_update_bap_counts_mips; c->update_bap_counts = ac3_update_bap_counts_mips;
#endif #endif
#if HAVE_MIPSFPU #if HAVE_MIPSFPU && HAVE_MIPS32R2
c->float_to_fixed24 = float_to_fixed24_mips; c->float_to_fixed24 = float_to_fixed24_mips;
// c->downmix = ac3_downmix_mips; c->downmix = ac3_downmix_mips;
#endif #endif
#endif #endif
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment