History

x86/sbrdsp: zero extend m_max in apply_noise_main

James Almer committed 9 years ago

Commit bcbe9e41 parent 44028541 file · +14 −14Browse files

Unified · Split

libavcodec/x86/sbrdsp.asm+14−14
@@ -378,24 +378,24 @@ cglobal sbr_hf_apply_noise_3, 5,5+NREGS+UNIX64,8, Y,s_m,q_filt,noise,kx,m_max · 377 unmodified lines
378apply_noise_main:378apply_noise_main:
379%if ARCH_X86_64 == 0 || WIN64379%if ARCH_X86_64 == 0 || WIN64
380 mov kxd, m_maxm380 mov kxd, m_maxm
381%define count kxq381 DEFINE_ARGS Y, s_m, q_filt, noise, count
382%else382%else
383%define count m_maxq383 DEFINE_ARGS Y, s_m, q_filt, noise, kx, count
384%endif384%endif
385 movsxdifnidn noiseq, noised385 movsxdifnidn noiseq, noised
386 dec noiseq386 dec noiseq
387 shl count, 2387 shl countd, 2
388%ifdef PIC388%ifdef PIC
389 lea NOISE_TABLE, [sbr_noise_table]389 lea NOISE_TABLE, [sbr_noise_table]
390%endif390%endif
391 lea Yq, [Yq + 2*count]391 lea Yq, [Yq + 2*countq]
392 add s_mq, count392 add s_mq, countq
393 add q_filtq, count393 add q_filtq, countq
394 shl noiseq, 3394 shl noiseq, 3
395 pxor m5, m5395 pxor m5, m5
396 neg count396 neg countq
397.loop:397.loop:
398 mova m1, [q_filtq + count]398 mova m1, [q_filtq + countq]
399 movu m3, [noiseq + NOISE_TABLE + 1*mmsize]399 movu m3, [noiseq + NOISE_TABLE + 1*mmsize]
400 movu m4, [noiseq + NOISE_TABLE + 2*mmsize]400 movu m4, [noiseq + NOISE_TABLE + 2*mmsize]
401 add noiseq, 2*mmsize401 add noiseq, 2*mmsize
@@ -404,7 +404,7 @@ apply_noise_main: · 2 unmodified lines
404 punpckldq m1, m1404 punpckldq m1, m1
405 mulps m1, m3 ; m2 = q_filt[m] * ff_sbr_noise_table[noise]405 mulps m1, m3 ; m2 = q_filt[m] * ff_sbr_noise_table[noise]
406 mulps m2, m4 ; m2 = q_filt[m] * ff_sbr_noise_table[noise]406 mulps m2, m4 ; m2 = q_filt[m] * ff_sbr_noise_table[noise]
407 mova m3, [s_mq + count]407 mova m3, [s_mq + countq]
408 ; TODO: replace by a vpermd in AVX2408 ; TODO: replace by a vpermd in AVX2
409 punpckhdq m4, m3, m3409 punpckhdq m4, m3, m3
410 punpckldq m3, m3410 punpckldq m3, m3
@@ -414,15 +414,15 @@ apply_noise_main: · 3 unmodified lines
414 mulps m4, m0 ; s_m[m] * phi_sign414 mulps m4, m0 ; s_m[m] * phi_sign
415 pand m1, m6415 pand m1, m6
416 pand m2, m7416 pand m2, m7
417 movu m6, [Yq + 2*count]417 movu m6, [Yq + 2*countq]
418 movu m7, [Yq + 2*count + mmsize]418 movu m7, [Yq + 2*countq + mmsize]
419 addps m3, m1419 addps m3, m1
420 addps m4, m2420 addps m4, m2
421 addps m6, m3421 addps m6, m3
422 addps m7, m4422 addps m7, m4
423 movu [Yq + 2*count], m6423 movu [Yq + 2*countq], m6
424 movu [Yq + 2*count + mmsize], m7424 movu [Yq + 2*countq + mmsize], m7
425 add count, mmsize425 add countq, mmsize
426 jl .loop426 jl .loop
427 RET427 RET
428 428

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t