- updated to 1.4

[packages/xavs2.git] / xavs2-x86.patch

--- xavs2-1.4/source/common/x86/quant8.asm.orig	2023-04-09 09:12:17.196476871 +0200
+++ xavs2-1.4/source/common/x86/quant8.asm	2023-04-09 09:13:53.645954359 +0200
@@ -48,9 +48,15 @@ cextern pd_n32768
 INIT_XMM sse4\r
 cglobal quant, 2,3,8\r
 ;{\r
+%if ARCH_X86_64\r
     movq        m4, r2mp              ; m4[0] = scale\r
     movq        m5, r3mp              ; m5[0] = shift\r
     movq        m6, r4mp              ; m6[0] = add\r
+%else\r
+    movd        m4, r2mp              ; m4[0] = scale\r
+    movd        m5, r3mp              ; m5[0] = shift\r
+    movd        m6, r4mp              ; m6[0] = add\r
+%endif\r
     mov         r2, r1                ; r2    = i_coef\r
     shr         r1, 3                 ; r1    = i_coef/8\r
     pxor        m7, m7                ; m7 <-- num_non_zero = 0\r
@@ -86,9 +92,15 @@ cglobal quant, 2,3,8
                                       ;\r
     packuswb    m7, m7                ;\r
     psadbw      m7, m0                ;\r
+%if ARCH_X86_64\r
     movifnidn  rax, r2                ; eax <-- i_coef\r
     movq        r1, m7                ;\r
     sub        rax, r1                ; return value: num_non_zero\r
+%else\r
+    movifnidn  eax, r2                ; eax <-- i_coef\r
+    movd        r1, m7                ;\r
+    sub        eax, r1                ; return value: num_non_zero\r
+%endif\r
     RET                               ; return\r
 ;}\r
 \r
@@ -103,13 +103,22 @@ INIT_XMM sse4
 cglobal dequant, 2,4,7\r
 ;{\r
     mov         r3, r3mp              ; r3  <-- shift\r
+%if ARCH_X86_64\r
     movq        m4, r2mp              ; m4[0] = scale\r
     movq        m6, r3                ; m6[0] = shift\r
+%else\r
+    movd        m4, r2mp              ; m4[0] = scale\r
+    movd        m6, r3                ; m6[0] = shift\r
+%endif\r
     dec         r3                    ; r3d <-- shift - 1\r
     xor         r2, r2                ; r2 <-- 0\r
     shr         r1, 4                 ; r1    = i_coef/16\r
     bts         r2, r3                ; r2 <-- add = 1 < (shift - 1)\r
+%if ARCH_X86_64\r
     movq        m5, r2                ; m5[0] = add\r
+%else\r
+    movd        m5, r2                ; m5[0] = add\r
+%endif\r
     pshufd      m4, m4, 0             ; m4[3210] = scale\r
     pshufd      m5, m5, 0             ; m5[3210] = add\r
                                       ;\r