search for: vmull_s32

Hi all, I'm sending patch "Optimize silk_warped_autocorrelation_FIX() for ARM NEON" in an separate email. It is based on Tim’s aarch64v8 branch https://git.xiph.org/?p=users/tterribe/opus.git;a=shortlog;h=refs/heads/aarch64v8 Thanks for your comments. Linfeng

I rebased my previous 3 patches to the current master with minor changes. Patches 1 to 3 replace all my previous submitted patches. Patches 4 and 5 are new. Thanks, Linfeng Zhang

...), 14 ); S[ 1 ] = silk_SMLAWB( S[ 1 ], out32_Q14, A1_U_Q28 ); S[ 1 ] = silk_SMLAWB( S[ 1 ], B_Q28[ 2 ], inval ); A_Q28 is split to 2 14-bit (or 16-bit, whatever) integers, to make the multiplication operation within 32-bits. NEON can do 32-bit x 32-bit = 64-bit using 'int64x2_t vmull_s32(int32x2_t a, int32x2_t b)', and it could possibly be faster and less rounding/shifting errors than above C code. But it may increase difficulties for other CPUs not supporting 32-bit multiplication. Thanks, Linfeng -------------- next part -------------- An HTML attachment was scrubbed... URL:...

...q_s32(coef32 + 8); > + int32x4_t coef3 = vld1q_s32(coef32 + 12); > + > + int32x4_t a0 = vld1q_s32(buf32 - 15); > + int32x4_t a1 = vld1q_s32(buf32 - 11); > + int32x4_t a2 = vld1q_s32(buf32 - 7); > + int32x4_t a3 = vld1q_s32(buf32 - 3); > + > + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); > + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); > + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); > + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); > + int64x2_...

On Tue, Apr 25, 2017 at 10:31 PM, Jean-Marc Valin <jmvalin at jmvalin.ca> wrote: > > > A_Q28 is split to 2 14-bit (or 16-bit, whatever) integers, to make the > > multiplication operation within 32-bits. NEON can do 32-bit x 32-bit = > > 64-bit using 'int64x2_t vmull_s32(int32x2_t a, int32x2_t b)', and it > > could possibly be faster and less rounding/shifting errors than above C > > code. But it may increase difficulties for other CPUs not supporting > > 32-bit multiplication. > > OK, so I'm not totally opposed to that, but it incre...

...n at jmvalin.ca <mailto:jmvalin at jmvalin.ca>> wrote: > > > > A_Q28 is split to 2 14-bit (or 16-bit, whatever) integers, to make the > > multiplication operation within 32-bits. NEON can do 32-bit x 32-bit = > > 64-bit using 'int64x2_t vmull_s32(int32x2_t a, int32x2_t b)', and it > > could possibly be faster and less rounding/shifting errors than above C > > code. But it may increase difficulties for other CPUs not supporting > > 32-bit multiplication. > > OK, so I'm not tota...

...%. It would just be good to check that the 0.8% indeed comes from Neon as opposed to just unrolling the channels. > A_Q28 is split to 2 14-bit (or 16-bit, whatever) integers, to make the > multiplication operation within 32-bits. NEON can do 32-bit x 32-bit = > 64-bit using 'int64x2_t vmull_s32(int32x2_t a, int32x2_t b)', and it > could possibly be faster and less rounding/shifting errors than above C > code. But it may increase difficulties for other CPUs not supporting > 32-bit multiplication. OK, so I'm not totally opposed to that, but it increases the testing/mainten...

...int32x4_t coef3 = vld1q_s32(coef32 + 12); >> + >> + int32x4_t a0 = vld1q_s32(buf32 - 15); >> + int32x4_t a1 = vld1q_s32(buf32 - 11); >> + int32x4_t a2 = vld1q_s32(buf32 - 7); >> + int32x4_t a3 = vld1q_s32(buf32 - 3); >> + >> + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); >> + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); >> + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); >> + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); >&g...

..., a01, 3); // data0[0] data1[0] ...[2] + int32x4_t a1 = vld1q_s32(data1 + 3); // data1[3] ... [6] + + int16x8_t coef16 = vld1q_s16(coef); + int32x4_t coef0 = vmovl_s16(vget_low_s16(coef16)); + int32x4_t coef1 = vmovl_s16(vget_high_s16(coef16)); + + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); + + int64x1_t...

..., a01, 3); // data0[0] data1[0] ...[2] + int32x4_t a1 = vld1q_s32(data1 + 3); // data1[3] ... [6] + + int16x8_t coef16 = vld1q_s16(coef); + int32x4_t coef0 = vmovl_s16(vget_low_s16(coef16)); + int32x4_t coef1 = vmovl_s16(vget_high_s16(coef16)); + + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); + + int64x1_t...

...17 at 10:31 PM, Jean-Marc Valin <jmvalin at jmvalin.ca> > wrote: > >> >> > A_Q28 is split to 2 14-bit (or 16-bit, whatever) integers, to make the >> > multiplication operation within 32-bits. NEON can do 32-bit x 32-bit = >> > 64-bit using 'int64x2_t vmull_s32(int32x2_t a, int32x2_t b)', and it >> > could possibly be faster and less rounding/shifting errors than above C >> > code. But it may increase difficulties for other CPUs not supporting >> > 32-bit multiplication. >> >> OK, so I'm not totally opposed t...

Hi Jean-Marc, Tested on my chromebook, when stride (channel) == 1, the optimization has no gain compared with C function. When stride (channel) == 2, the optimization is 1.2%-1.8% faster (1.6% at Complexity 8) compared with C function. Please let me know and I can remove the optimization of stride 1 case. If it's allowed to skip the split of A_Q28 and replace by 32-bit multiplication

...n at jmvalin.ca>> wrote: > > > > > > > A_Q28 is split to 2 14-bit (or 16-bit, whatever) integers, to > make the > > > multiplication operation within 32-bits. NEON can do 32-bit x > 32-bit = > > > 64-bit using 'int64x2_t vmull_s32(int32x2_t a, int32x2_t b)', > and it > > > could possibly be faster and less rounding/shifting errors > than above C > > > code. But it may increase difficulties for other CPUs not > supporting > > > 32-bit multiplication. > > &g...

As promised, here's a re-send of all my Aarch64 patches, following comments by John Ridges. Note that they actually affect more than just Aarch64 -- other than the ones specifically guarded by AARCH64_NEON defines, the Neon intrinsics all also apply on armv7; and the OPUS_FAST_INT64 patches apply on any 64-bit machine. The patches should largely be independent and independently useful, other

...coef32 + 4); + int32x4_t coef2 = vld1q_s32(coef32 + 8); + int32x4_t coef3 = vld1q_s32(coef32 + 12); + + int32x4_t a0 = vld1q_s32(buf32 - 15); + int32x4_t a1 = vld1q_s32(buf32 - 11); + int32x4_t a2 = vld1q_s32(buf32 - 7); + int32x4_t a3 = vld1q_s32(buf32 - 3); + + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); + int64x2_t b4 = vmlal_s32(b3,...

...coef32 + 4); + int32x4_t coef2 = vld1q_s32(coef32 + 8); + int32x4_t coef3 = vld1q_s32(coef32 + 12); + + int32x4_t a0 = vld1q_s32(buf32 - 15); + int32x4_t a1 = vld1q_s32(buf32 - 11); + int32x4_t a2 = vld1q_s32(buf32 - 7); + int32x4_t a3 = vld1q_s32(buf32 - 3); + + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); + int64x2_t b4 = vmlal_s32(b3,...

This sequence of patches provides arm64 support for Opus. Tested on iOS, Android, and Ubuntu 14.04. The patch sequence was written on top of Viswanath Puttagunta's Ne10 patches, but all but the second ("Reorganize pitch_arm.h") should, I think, apply independently of it. It does depends on my previous intrinsics configury reorganization, however. Comments welcome. With this and

Here are my aarch64 patches rebased to the current tip of Opus master. They're largely the same as my previous patch set, with the addition of the final one (the Neon fixed-point implementation of xcorr_kernel). This replaces Viswanath's Neon fixed-point celt_pitch_xcorr, since xcorr_kernel is used in celt_fir and celt_iir as well. These have been tested for correctness under qemu

Following Tim's comments, here are my reworked patches for the Neon intrinsic function patches of of my Aarch64 patchset, i.e. replacing patches 5-8 of the v2 series. Patches 1-4 and 9-18 of the old series still apply unmodified. The one new (as opposed to changed) patch is the first one in this series, to add named constants for the ARM architecture variants. There are also some minor code