ref: 97e6cc619834cc6d9eccfd5d402e6808e0eb4dbb
parent: 16eba81f6955adf8b60fd7a71b775ee004f970ba
author: Scott LaVarnway <slavarnway@google.com>
date: Mon Nov 23 05:24:09 EST 2015
VPX: Removed unnecessary pmulhrsw in IDCT32X32_34 and fixed macro name. Change-Id: I306b98a2b4ec80b130ae80290b4cd9c7a5363311
--- a/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
+++ b/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
@@ -368,7 +368,7 @@
%define idx30 16 * 6
%define idx31 16 * 7
-%macro IDCT32X32_34x 4
+%macro IDCT32X32_34 4
; FROM idct32x32_add_neon.asm
;
; Instead of doing the transforms stage by stage, it is done by loading
@@ -571,9 +571,7 @@
; BLOCK D STAGE 4 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
mova m0, [rsp + transposed_in + 16 * 0]
mova m10, [pw_11585x2]
- mova m7, m0
pmulhrsw m0, m10 ; stp1_1
- pmulhrsw m7, m10 ; stp1_0
mova m14, m11 ; stp1_4
mova m13, m12 ; stp1_7
@@ -587,6 +585,7 @@
BUTTERFLY_4X 13, 14, 11585, 11585, m8, 9, 10 ; stp1_5, stp1_6
SWAP 13, 14
%endif
+ mova m7, m0 ; stp1_0 = stp1_1
mova m4, m0 ; stp1_1
mova m2, m7 ; stp1_0
@@ -752,7 +751,7 @@
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
- IDCT32X32_34x 16*0, 16*32, 16*64, 16*96
+ IDCT32X32_34 16*0, 16*32, 16*64, 16*96
lea stp, [stp + 16 * 8]
mov r6, 4
lea stp, [rsp + pass_one_start]
@@ -774,7 +773,7 @@
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
- IDCT32X32_34x 16*0, 16*8, 16*16, 16*24
+ IDCT32X32_34 16*0, 16*8, 16*16, 16*24
lea stp, [stp + 16 * 32]
add r9, 16 * 32