ppc: itx pwr9
3 unresolved threads
3 unresolved threads
It is on top of the loopfilter work since also this rely on pwr9-only instructions
-
4x4 -
4x8 -
8x4 -
8x8 -
16x4 -
4x16
Tested with the argon samples (thank you @unlord for pointing me to it)
Edited by Luca Barbato
Merge request reports
Activity
Filter activity
- Automatically resolved by Luca Barbato
- Resolved by Luca Barbato
- Automatically resolved by Luca Barbato
- Resolved by Luca Barbato
- Automatically resolved by Luca Barbato
added performance label
added 28 commits
-
71f69261...2d2c6c65 - 25 commits from branch
videolan:master
- a57194bf - ppc: itx 4x4 pwr9
- 35637738 - ppc: itx 4x8 and 8x4 pwr9
- 89cce85e - ppc: itx 8x8 pwr9
Toggle commit list-
71f69261...2d2c6c65 - 25 commits from branch
added 16 commits
-
89cce85e...37155c11 - 12 commits from branch
videolan:master
- c5ec0f83 - ppc: itx 4x4 pwr9
- 9c667961 - ppc: itx 4x8 and 8x4 pwr9
- 6618c007 - ppc: itx 8x8 pwr9
- 9907e1ce - ppc: Remove high bitdepth macros from the 8bit-only code
Toggle commit list-
89cce85e...37155c11 - 12 commits from branch
- Automatically resolved by Luca Barbato
- src/ppc/itx_tmpl.c 0 → 100644
923 STORE_4(dst + 4 * stride, stride, e, f, g, hh) 924 } 925 926 927 #define inv_txfm_fn4x8(type1, type2) \ 928 void dav1d_inv_txfm_add_##type1##_##type2##_4x8_8bpc_pwr9(uint8_t *dst, const ptrdiff_t stride, \ 929 int16_t *const coeff, const int eob) \ 930 { \ 931 i16x8 v = vec_splats((int16_t)(2896*8)); \ 932 LOAD_SCALE_COEFF_4x8(coeff, v) \ 933 type1##_4_in(c0, c1, c2, c3, c01, c23) \ 934 type1##_4_in(c4, c5, c6, c7, c45, c67) \ 935 memset(coeff, 0, sizeof(*coeff) * 4 * 8); \ 936 TRANSPOSE4_I32(c0, c1, c2, c3); \ 937 TRANSPOSE4_I32(c4, c5, c6, c7); \ 938 type2##_8_out(c0, c1, c2, c3, c4, c5, c6, c7, c01, c23, c45, c67) \ - Automatically resolved by Luca Barbato
65 assign_itx16_fn( , 8, 8, pwr9); 66 assign_itx16_fn(R, 8, 16, pwr9); 67 assign_itx2_fn (R, 8, 32, pwr9); 68 assign_itx16_fn(R, 16, 4, pwr9); 69 assign_itx16_fn(R, 16, 8, pwr9); 70 assign_itx12_fn( , 16, 16, pwr9); 71 assign_itx2_fn (R, 16, 32, pwr9); 72 assign_itx1_fn (R, 16, 64, pwr9); 73 assign_itx2_fn (R, 32, 8, pwr9); 74 assign_itx2_fn (R, 32, 16, pwr9); 75 assign_itx2_fn ( , 32, 32, pwr9); 76 assign_itx1_fn (R, 32, 64, pwr9); 77 assign_itx1_fn (R, 64, 16, pwr9); 78 assign_itx1_fn (R, 64, 32, pwr9); 79 assign_itx1_fn ( , 64, 64, pwr9); 80 */ added 7 commits
-
9907e1ce...3623543c - 3 commits from branch
videolan:master
- 669c6c43 - ppc: itx 4x4 pwr9
- 91b109a3 - ppc: itx 4x8 and 8x4 pwr9
- aa260dc6 - ppc: itx 8x8 pwr9
- 6cf54de5 - ppc: Remove high bitdepth macros from the 8bit-only code
Toggle commit list-
9907e1ce...3623543c - 3 commits from branch
added 2 commits
changed milestone to %1.5.0
added 21 commits
Toggle commit list- Automatically resolved by Luca Barbato
mentioned in merge request !1723 (merged)
added 45 commits
Toggle commit list- Automatically resolved by Luca Barbato
- src/ppc/itx_tmpl.c 0 → 100644
19 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR 20 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 22 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 23 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 25 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include "src/ppc/dav1d_types.h" 29 #include "src/ppc/itx.h" 30 #include "src/ppc/utils.h" 31 32 #include "src/itx_1d.h" 33 34 #if BITDEPTH == 8 added 9 commits
-
ccbca2ea...21235966 - 2 commits from branch
videolan:master
- 33b9d514 - ppc: itx 4x4 pwr9
- da51b123 - ppc: itx 4x8 and 8x4 pwr9
- b1d847be - ppc: itx 8x8 pwr9
- 19e122ee - ppc: Remove high bitdepth macros from the 8bit-only code
- 0bf331a1 - ppc: itx 4x16 pwr9
- 75d3ad14 - ppc: itx 16x4 pwr9
- 8d9b1e26 - ppc: Factor out dc_only itx
Toggle commit list-
ccbca2ea...21235966 - 2 commits from branch