Re: [FFmpeg-devel] [PATCH] avcodec/mips: Improve avc put mc 11, 31, 13 and 33 msa functions

2017-10-25 Thread Michael Niedermayer 
On Tue, Oct 24, 2017 at 11:21:21AM +, Manojkumar Bhosale wrote:
> LGTM

will apply

thanks

[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Rewriting code that is poorly written but fully understood is good.
Rewriting code that one doesnt understand is a sign that one is less smart
then the original author, trying to rewrite it will not make it better.


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] avcodec/mips: Improve avc put mc 11, 31, 13 and 33 msa functions

2017-10-24 Thread Manojkumar Bhosale 
LGTM

-Original Message-
From: ffmpeg-devel [mailto:ffmpeg-devel-boun...@ffmpeg.org] On Behalf Of 
kaustubh.ra...@imgtec.com
Sent: Tuesday, October 24, 2017 12:39 PM
To: ffmpeg-devel@ffmpeg.org
Cc: Kaustubh Raste
Subject: [FFmpeg-devel] [PATCH] avcodec/mips: Improve avc put mc 11, 31, 13 and 
33 msa functions

From: Kaustubh Raste <kaustubh.ra...@imgtec.com>

Remove loops and unroll as block sizes are known.

Signed-off-by: Kaustubh Raste <kaustubh.ra...@imgtec.com>
---
 libavcodec/mips/h264qpel_msa.c |  400 
 1 file changed, 240 insertions(+), 160 deletions(-)

diff --git a/libavcodec/mips/h264qpel_msa.c b/libavcodec/mips/h264qpel_msa.c 
index f11fce8..fcccb98 100644
--- a/libavcodec/mips/h264qpel_msa.c
+++ b/libavcodec/mips/h264qpel_msa.c
@@ -171,23 +171,27 @@ static const uint8_t luma_mask_arr[16 * 8] = {
 out0_m; \
 } )
 
-static void avc_luma_hv_qrt_4w_msa(const uint8_t *src_x, const uint8_t *src_y,
-   int32_t src_stride, uint8_t *dst,
-   int32_t dst_stride, int32_t height)
+static void avc_luma_hv_qrt_4x4_msa(const uint8_t *src_x, const uint8_t *src_y,
+uint8_t *dst, int32_t stride)
 {
-uint32_t loop_cnt;
-v16i8 src_hz0, src_hz1, src_hz2, src_hz3;
-v16i8 src_vt0, src_vt1, src_vt2, src_vt3, src_vt4;
-v16i8 src_vt5, src_vt6, src_vt7, src_vt8;
-v16i8 mask0, mask1, mask2;
-v8i16 hz_out0, hz_out1, vert_out0, vert_out1;
-v8i16 out0, out1;
+const int16_t filt_const0 = 0xfb01;
+const int16_t filt_const1 = 0x1414;
+const int16_t filt_const2 = 0x1fb;
 v16u8 out;
+v16i8 src_hz0, src_hz1, src_hz2, src_hz3, src_vt7, src_vt8;
+v16i8 src_vt0, src_vt1, src_vt2, src_vt3, src_vt4, src_vt5, src_vt6;
+v16i8 src_vt10_r, src_vt32_r, src_vt54_r, src_vt76_r;
+v16i8 mask0, mask1, mask2, filt0, filt1, filt2;
+v8i16 hz_out0, hz_out1, vt_out0, vt_out1, out0, out1;
+
+filt0 = (v16i8) __msa_fill_h(filt_const0);
+filt1 = (v16i8) __msa_fill_h(filt_const1);
+filt2 = (v16i8) __msa_fill_h(filt_const2);
 
 LD_SB3(_mask_arr[48], 16, mask0, mask1, mask2);
 
-LD_SB5(src_y, src_stride, src_vt0, src_vt1, src_vt2, src_vt3, src_vt4);
-src_y += (5 * src_stride);
+LD_SB5(src_y, stride, src_vt0, src_vt1, src_vt2, src_vt3, src_vt4);
+src_y += (5 * stride);
 
 src_vt0 = (v16i8) __msa_insve_w((v4i32) src_vt0, 1, (v4i32) src_vt1);
 src_vt1 = (v16i8) __msa_insve_w((v4i32) src_vt1, 1, (v4i32) src_vt2); @@ 
-196,149 +200,237 @@ static void avc_luma_hv_qrt_4w_msa(const uint8_t *src_x, 
const uint8_t *src_y,
 
 XORI_B4_128_SB(src_vt0, src_vt1, src_vt2, src_vt3);
 
-for (loop_cnt = (height >> 2); loop_cnt--;) {
-LD_SB4(src_x, src_stride, src_hz0, src_hz1, src_hz2, src_hz3);
-src_x += (4 * src_stride);
-
-XORI_B4_128_SB(src_hz0, src_hz1, src_hz2, src_hz3);
-
-hz_out0 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src_hz0,
-  src_hz1, mask0,
-  mask1, mask2);
-hz_out1 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src_hz2,
-  src_hz3, mask0,
-  mask1, mask2);
-
-SRARI_H2_SH(hz_out0, hz_out1, 5);
-SAT_SH2_SH(hz_out0, hz_out1, 7);
-
-LD_SB4(src_y, src_stride, src_vt5, src_vt6, src_vt7, src_vt8);
-src_y += (4 * src_stride);
-
-src_vt4 = (v16i8) __msa_insve_w((v4i32) src_vt4, 1, (v4i32) src_vt5);
-src_vt5 = (v16i8) __msa_insve_w((v4i32) src_vt5, 1, (v4i32) src_vt6);
-src_vt6 = (v16i8) __msa_insve_w((v4i32) src_vt6, 1, (v4i32) src_vt7);
-src_vt7 = (v16i8) __msa_insve_w((v4i32) src_vt7, 1, (v4i32) src_vt8);
-
-XORI_B4_128_SB(src_vt4, src_vt5, src_vt6, src_vt7);
+LD_SB4(src_x, stride, src_hz0, src_hz1, src_hz2, src_hz3);
+XORI_B4_128_SB(src_hz0, src_hz1, src_hz2, src_hz3);
+hz_out0 = AVC_HORZ_FILTER_SH(src_hz0, src_hz1, mask0, mask1, mask2);
+hz_out1 = AVC_HORZ_FILTER_SH(src_hz2, src_hz3, mask0, mask1, 
+ mask2);
 
-/* filter calc */
-vert_out0 = AVC_CALC_DPADD_B_6PIX_2COEFF_R_SH(src_vt0, src_vt1,
-  src_vt2, src_vt3,
-  src_vt4, src_vt5);
-vert_out1 = AVC_CALC_DPADD_B_6PIX_2COEFF_R_SH(src_vt2, src_vt3,
-  src_vt4, src_vt5,
-  src_vt6, src_vt7);
+SRARI_H2_SH(hz_out0, hz_out1, 5);
+SAT_SH2_SH(hz_out0, hz_out1, 7);
 
-SRARI_H2_SH(vert_out0, vert_out1, 5);
-SAT_SH2_SH(vert_out0, vert_out1, 7);
+LD_SB4(src_y, stride

[FFmpeg-devel] [PATCH] avcodec/mips: Improve avc put mc 11, 31, 13 and 33 msa functions

2017-10-24 Thread kaustubh.raste 
From: Kaustubh Raste 

Remove loops and unroll as block sizes are known.

Signed-off-by: Kaustubh Raste 
---
 libavcodec/mips/h264qpel_msa.c |  400 
 1 file changed, 240 insertions(+), 160 deletions(-)

diff --git a/libavcodec/mips/h264qpel_msa.c b/libavcodec/mips/h264qpel_msa.c
index f11fce8..fcccb98 100644
--- a/libavcodec/mips/h264qpel_msa.c
+++ b/libavcodec/mips/h264qpel_msa.c
@@ -171,23 +171,27 @@ static const uint8_t luma_mask_arr[16 * 8] = {
 out0_m; \
 } )
 
-static void avc_luma_hv_qrt_4w_msa(const uint8_t *src_x, const uint8_t *src_y,
-   int32_t src_stride, uint8_t *dst,
-   int32_t dst_stride, int32_t height)
+static void avc_luma_hv_qrt_4x4_msa(const uint8_t *src_x, const uint8_t *src_y,
+uint8_t *dst, int32_t stride)
 {
-uint32_t loop_cnt;
-v16i8 src_hz0, src_hz1, src_hz2, src_hz3;
-v16i8 src_vt0, src_vt1, src_vt2, src_vt3, src_vt4;
-v16i8 src_vt5, src_vt6, src_vt7, src_vt8;
-v16i8 mask0, mask1, mask2;
-v8i16 hz_out0, hz_out1, vert_out0, vert_out1;
-v8i16 out0, out1;
+const int16_t filt_const0 = 0xfb01;
+const int16_t filt_const1 = 0x1414;
+const int16_t filt_const2 = 0x1fb;
 v16u8 out;
+v16i8 src_hz0, src_hz1, src_hz2, src_hz3, src_vt7, src_vt8;
+v16i8 src_vt0, src_vt1, src_vt2, src_vt3, src_vt4, src_vt5, src_vt6;
+v16i8 src_vt10_r, src_vt32_r, src_vt54_r, src_vt76_r;
+v16i8 mask0, mask1, mask2, filt0, filt1, filt2;
+v8i16 hz_out0, hz_out1, vt_out0, vt_out1, out0, out1;
+
+filt0 = (v16i8) __msa_fill_h(filt_const0);
+filt1 = (v16i8) __msa_fill_h(filt_const1);
+filt2 = (v16i8) __msa_fill_h(filt_const2);
 
 LD_SB3(_mask_arr[48], 16, mask0, mask1, mask2);
 
-LD_SB5(src_y, src_stride, src_vt0, src_vt1, src_vt2, src_vt3, src_vt4);
-src_y += (5 * src_stride);
+LD_SB5(src_y, stride, src_vt0, src_vt1, src_vt2, src_vt3, src_vt4);
+src_y += (5 * stride);
 
 src_vt0 = (v16i8) __msa_insve_w((v4i32) src_vt0, 1, (v4i32) src_vt1);
 src_vt1 = (v16i8) __msa_insve_w((v4i32) src_vt1, 1, (v4i32) src_vt2);
@@ -196,149 +200,237 @@ static void avc_luma_hv_qrt_4w_msa(const uint8_t 
*src_x, const uint8_t *src_y,
 
 XORI_B4_128_SB(src_vt0, src_vt1, src_vt2, src_vt3);
 
-for (loop_cnt = (height >> 2); loop_cnt--;) {
-LD_SB4(src_x, src_stride, src_hz0, src_hz1, src_hz2, src_hz3);
-src_x += (4 * src_stride);
-
-XORI_B4_128_SB(src_hz0, src_hz1, src_hz2, src_hz3);
-
-hz_out0 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src_hz0,
-  src_hz1, mask0,
-  mask1, mask2);
-hz_out1 = AVC_XOR_VSHF_B_AND_APPLY_6TAP_HORIZ_FILT_SH(src_hz2,
-  src_hz3, mask0,
-  mask1, mask2);
-
-SRARI_H2_SH(hz_out0, hz_out1, 5);
-SAT_SH2_SH(hz_out0, hz_out1, 7);
-
-LD_SB4(src_y, src_stride, src_vt5, src_vt6, src_vt7, src_vt8);
-src_y += (4 * src_stride);
-
-src_vt4 = (v16i8) __msa_insve_w((v4i32) src_vt4, 1, (v4i32) src_vt5);
-src_vt5 = (v16i8) __msa_insve_w((v4i32) src_vt5, 1, (v4i32) src_vt6);
-src_vt6 = (v16i8) __msa_insve_w((v4i32) src_vt6, 1, (v4i32) src_vt7);
-src_vt7 = (v16i8) __msa_insve_w((v4i32) src_vt7, 1, (v4i32) src_vt8);
-
-XORI_B4_128_SB(src_vt4, src_vt5, src_vt6, src_vt7);
+LD_SB4(src_x, stride, src_hz0, src_hz1, src_hz2, src_hz3);
+XORI_B4_128_SB(src_hz0, src_hz1, src_hz2, src_hz3);
+hz_out0 = AVC_HORZ_FILTER_SH(src_hz0, src_hz1, mask0, mask1, mask2);
+hz_out1 = AVC_HORZ_FILTER_SH(src_hz2, src_hz3, mask0, mask1, mask2);
 
-/* filter calc */
-vert_out0 = AVC_CALC_DPADD_B_6PIX_2COEFF_R_SH(src_vt0, src_vt1,
-  src_vt2, src_vt3,
-  src_vt4, src_vt5);
-vert_out1 = AVC_CALC_DPADD_B_6PIX_2COEFF_R_SH(src_vt2, src_vt3,
-  src_vt4, src_vt5,
-  src_vt6, src_vt7);
+SRARI_H2_SH(hz_out0, hz_out1, 5);
+SAT_SH2_SH(hz_out0, hz_out1, 7);
 
-SRARI_H2_SH(vert_out0, vert_out1, 5);
-SAT_SH2_SH(vert_out0, vert_out1, 7);
+LD_SB4(src_y, stride, src_vt5, src_vt6, src_vt7, src_vt8);
 
-out0 = __msa_srari_h((hz_out0 + vert_out0), 1);
-out1 = __msa_srari_h((hz_out1 + vert_out1), 1);
+src_vt4 = (v16i8) __msa_insve_w((v4i32) src_vt4, 1, (v4i32) src_vt5);
+src_vt5 = (v16i8) __msa_insve_w((v4i32) src_vt5, 1, (v4i32) src_vt6);
+src_vt6 = (v16i8)