Commit 2149f214 authored by Jingning Han's avatar Jingning Han Committed by Gerrit Code Review
Browse files

Merge "Reduce required xmm number by one in block_error_fp"

parents fff4654d f2cf3c06
......@@ -78,7 +78,7 @@ cglobal block_error, 3, 3, 8, uqc, dqc, size, ssz
; intptr_t block_size)
cglobal block_error_fp, 3, 3, 8, uqc, dqc, size
cglobal block_error_fp, 3, 3, 6, uqc, dqc, size
pxor m4, m4 ; sse accumulator
pxor m5, m5 ; dedicated zero register
lea uqcq, [uqcq+sizeq*2]
......@@ -96,13 +96,13 @@ cglobal block_error_fp, 3, 3, 8, uqc, dqc, size
pmaddwd m0, m0
pmaddwd m1, m1
; accumulate in 64bit
punpckldq m7, m0, m5
punpckldq m3, m0, m5
punpckhdq m0, m5
paddq m4, m7
punpckldq m7, m1, m5
paddq m4, m3
punpckldq m3, m1, m5
paddq m4, m0
punpckhdq m1, m5
paddq m4, m7
paddq m4, m3
paddq m4, m1
add sizeq, mmsize
jl .loop
Supports Markdown
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