x86: replace explicit REP_RETs with RETs

From x86inc:
> On AMD cpus <=K10, an ordinary ret is slow if it immediately follows either
> a branch or a branch target. So switch to a 2-byte form of ret in that case.
> We can automatically detect "follows a branch", but not a branch target.
> (SSSE3 is a sufficient condition to know that your cpu doesn't have this problem.)

x86inc can automatically determine whether to use REP_RET rather than
REP in most of these cases, so impact is minimal. Additionally, a few
REP_RETs were used unnecessary, despite the return being nowhere near a
branch.

The only CPUs affected were AMD K10s, made between 2007 and 2011, 16
years ago and 12 years ago, respectively.

In the future, everyone involved with x86inc should consider dropping
REP_RETs altogether.

libavcodec/x86/aacpsdsp.asm

@@ -49,7 +49,7 @@ align 16
     add  dstq, mmsize
     add    nq, mmsize*2
     jl .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM sse
@@ -83,7 +83,7 @@ align 16
     add   src2q, mmsize
     add      nq, mmsize*2
     jl .loop
-    REP_RET
+    RET
 
 ;***********************************************************************
 ;void ff_ps_stereo_interpolate_sse3(float (*l)[2], float (*r)[2],
@@ -116,7 +116,7 @@ align 16
     movhps [rq+nq], m2
     add      nq, 8
     jl .loop
-    REP_RET
+    RET
 
 ;***************************************************************************
 ;void ps_stereo_interpolate_ipdopd_sse3(float (*l)[2], float (*r)[2],
@@ -164,7 +164,7 @@ align 16
     movhps [rq+nq], m2
     add      nq, 8
     jl .loop
-    REP_RET
+    RET
 
 ;**********************************************************
 ;void ps_hybrid_analysis_ileave_sse(float out[2][38][64],
@@ -484,7 +484,7 @@ align 16
     add    outq, strideq
     add      nq, 64
     jl .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM sse

libavcodec/x86/ac3dsp.asm

@@ -60,7 +60,7 @@ cglobal ac3_exponent_min, 3, 4, 2, exp, reuse_blks, expn, offset
     sub        expnq, mmsize
     jg .nextexp
 .end:
-    REP_RET
+    RET
 %endmacro
 
 %define LOOP_ALIGN ALIGN 16
@@ -126,7 +126,7 @@ cglobal float_to_fixed24, 3, 3, 9, dst, src, len
     sub      lenq, 16
 %endif
     ja .loop
-    REP_RET
+    RET
 
 ;------------------------------------------------------------------------------
 ; int ff_ac3_compute_mantissa_size(uint16_t mant_cnt[6][16])
@@ -220,7 +220,7 @@ cglobal ac3_extract_exponents, 3, 3, 4, exp, coef, len
 
     add     lenq, 4
     jl .loop
-    REP_RET
+    RET
 %endmacro
 
 %if HAVE_SSE2_EXTERNAL

libavcodec/x86/alacdsp.asm

@@ -100,7 +100,7 @@ align 16
 
     add     lenq, mmsize*2
     jl .loop
-    REP_RET
+    RET
 
 %if ARCH_X86_64
 cglobal alac_append_extra_bits_mono, 2, 5, 3, buf, exbuf, exbits, ch, len
@@ -130,4 +130,4 @@ align 16
 
     add     lenq, mmsize*2
     jl .loop
-    REP_RET
+    RET

libavcodec/x86/audiodsp.asm

@@ -123,7 +123,7 @@ cglobal vector_clip_int32%5, 5,5,%1, dst, src, min, max, len
     add     dstq, mmsize*4*(%2+%3)
     sub     lend, mmsize*(%2+%3)
     jg .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM sse2

libavcodec/x86/dirac_dwt.asm

@@ -75,7 +75,7 @@ cglobal vertical_compose53iL0_%1, 4,4,1, b0, b1, b2, width
     COMPOSE_53iL0 m0, m1, [b2q+2*widthq], m2
     mova    [b1q+2*widthq], m0
     jg      .loop
-    REP_RET
+    RET
 
 ; void vertical_compose_dirac53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
 ;                                  int width)
@@ -93,7 +93,7 @@ cglobal vertical_compose_dirac53iH0_%1, 4,4,1, b0, b1, b2, width
     paddw   m0, [b1q+2*widthq]
     mova    [b1q+2*widthq], m0
     jg      .loop
-    REP_RET
+    RET
 
 ; void vertical_compose_dd97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
 ;                               IDWTELEM *b3, IDWTELEM *b4, int width)
@@ -110,7 +110,7 @@ cglobal vertical_compose_dd97iH0_%1, 6,6,5, b0, b1, b2, b3, b4, width
     COMPOSE_DD97iH0 [b2q+2*widthq], [b3q+2*widthq], [b4q+2*widthq]
     mova    [b2q+2*widthq], m1
     jg      .loop
-    REP_RET
+    RET
 
 ; void vertical_compose_dd137iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
 ;                                IDWTELEM *b3, IDWTELEM *b4, int width)
@@ -139,7 +139,7 @@ cglobal vertical_compose_dd137iL0_%1, 6,6,6, b0, b1, b2, b3, b4, width
     psubw   m5, m1
     mova    [b2q+2*widthq], m5
     jg      .loop
-    REP_RET
+    RET
 
 ; void vertical_compose_haar(IDWTELEM *b0, IDWTELEM *b1, int width)
 cglobal vertical_compose_haar_%1, 3,4,3, b0, b1, width
@@ -159,7 +159,7 @@ cglobal vertical_compose_haar_%1, 3,4,3, b0, b1, width
     paddw   m2, m0
     mova    [b1q+2*widthq], m2
     jg      .loop
-    REP_RET
+    RET
 %endmacro
 
 ; extend the left and right edges of the tmp array by %1 and %2 respectively
@@ -225,7 +225,7 @@ cglobal horizontal_compose_haar%2i_%1, 3,6,4, b, tmp, w, x, w2, b_w2
     cmp     xq, w2q
     jl      .highpass_loop
 .end:
-    REP_RET
+    RET
 %endmacro
 
 
@@ -290,7 +290,7 @@ cglobal horizontal_compose_dd97i_ssse3, 3,6,8, b, tmp, w, x, w2, b_w2
     cmp     xd, w2d
     jl      .highpass_loop
 .end:
-    REP_RET
+    RET
 
 
 INIT_XMM

libavcodec/x86/fft.asm

@@ -475,7 +475,7 @@ cglobal fft_calc, 2,5,8
     mov     r0, r1
     mov     r1, r3
     FFT_DISPATCH _interleave %+ SUFFIX, r1
-    REP_RET
+    RET
 
 %endif
 
@@ -510,7 +510,7 @@ cglobal fft_calc, 2,5,8
     add      r2, mmsize*2
     jl       .loop
 .end:
-    REP_RET
+    RET
 
 cglobal fft_permute, 2,7,1
     mov     r4,  [r0 + FFTContext.revtab]
@@ -543,7 +543,7 @@ cglobal fft_permute, 2,7,1
     movaps  [r1 + r2 + 16], xmm1
     add     r2, 32
     jl      .loopcopy
-    REP_RET
+    RET
 
 INIT_XMM sse
 cglobal imdct_calc, 3,5,3
@@ -583,7 +583,7 @@ cglobal imdct_calc, 3,5,3
     sub     r3, mmsize
     add     r2, mmsize
     jl      .loop
-    REP_RET
+    RET
 
 %ifdef PIC
 %define SECTION_REL - $$

libavcodec/x86/flacdsp.asm

@@ -79,7 +79,7 @@ ALIGN 16
     movd   [decodedq+4], m1
     jg .loop_sample
 .ret:
-    REP_RET
+    RET
 %endmacro
 
 %if HAVE_XOP_EXTERNAL
@@ -133,7 +133,7 @@ align 16
     mova [outq + lenq], m%2
     add      lenq, 16
     jl .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM sse2
@@ -177,7 +177,7 @@ align 16
     add      outq, mmsize*2
     sub      lend, mmsize/4
     jg .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM sse2
@@ -302,7 +302,7 @@ align 16
     add      outq, mmsize*REPCOUNT
     sub      lend, mmsize/4
     jg .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM ssse3

libavcodec/x86/h264_chromamc.asm

@@ -112,7 +112,7 @@ cglobal %1_%2_chroma_mc8%3, 6, 7 + extra_regs, 0
     jne .at_least_one_non_zero
     ; mx == 0 AND my == 0 - no filter needed
     mv0_pixels_mc8
-    REP_RET
+    RET
 
 .at_least_one_non_zero:
 %ifidn %2, rv40
@@ -192,7 +192,7 @@ cglobal %1_%2_chroma_mc8%3, 6, 7 + extra_regs, 0
     add           r1, r2
     dec           r3d
     jne .next1drow
-    REP_RET
+    RET
 
 .both_non_zero: ; general case, bilinear
     movd          m4, r4d         ; x
@@ -365,7 +365,7 @@ cglobal %1_%2_chroma_mc4, 6, 6 + extra_regs, 0
     add           r0, r2
     sub          r3d, 2
     jnz .next2rows
-    REP_RET
+    RET
 %endmacro
 
 %macro chroma_mc2_mmx_func 2
@@ -407,7 +407,7 @@ cglobal %1_%2_chroma_mc2, 6, 7, 0
     add           r0, r2
     sub          r3d, 1
     jnz .nextrow
-    REP_RET
+    RET
 %endmacro
 
 %define rnd_1d_h264 pw_4
@@ -453,7 +453,7 @@ cglobal %1_%2_chroma_mc8%3, 6, 7, 8
     jne .at_least_one_non_zero
     ; mx == 0 AND my == 0 - no filter needed
     mv0_pixels_mc8
-    REP_RET
+    RET
 
 .at_least_one_non_zero:
     test         r5d, r5d
@@ -514,7 +514,7 @@ cglobal %1_%2_chroma_mc8%3, 6, 7, 8
     sub          r3d, 2
     lea           r0, [r0+r2*2]
     jg .next2rows
-    REP_RET
+    RET
 
 .my_is_zero:
     mov          r5d, r4d
@@ -551,7 +551,7 @@ cglobal %1_%2_chroma_mc8%3, 6, 7, 8
     lea           r0, [r0+r2*2]
     lea           r1, [r1+r2*2]
     jg .next2xrows
-    REP_RET
+    RET
 
 .mx_is_zero:
     mov          r4d, r5d
@@ -588,7 +588,7 @@ cglobal %1_%2_chroma_mc8%3, 6, 7, 8
     sub          r3d, 2
     lea           r0, [r0+r2*2]
     jg .next2yrows
-    REP_RET
+    RET
 %endmacro
 
 %macro chroma_mc4_ssse3_func 2
@@ -638,7 +638,7 @@ cglobal %1_%2_chroma_mc4, 6, 7, 0
     sub          r3d, 2
     lea           r0, [r0+r2*2]
     jg .next2rows
-    REP_RET
+    RET
 %endmacro
 
 %define CHROMAMC_AVG NOTHING

libavcodec/x86/h264_chromamc_10bit.asm

@@ -67,7 +67,7 @@ cglobal %1_h264_chroma_mc8_10, 6,7,8
     jne .at_least_one_non_zero
     ; mx == 0 AND my == 0 - no filter needed
     MV0_PIXELS_MC8
-    REP_RET
+    RET
 
 .at_least_one_non_zero:
     mov          r6d, 2
@@ -102,7 +102,7 @@ cglobal %1_h264_chroma_mc8_10, 6,7,8
     add           r1, r2
     dec           r3d
     jne .next1drow
-    REP_RET
+    RET
 
 .xy_interpolation: ; general case, bilinear
     movd          m4, r4m         ; x
@@ -144,7 +144,7 @@ cglobal %1_h264_chroma_mc8_10, 6,7,8
     add           r0, r2
     dec          r3d
     jne .next2drow
-    REP_RET
+    RET
 %endmacro
 
 ;-----------------------------------------------------------------------------
@@ -194,7 +194,7 @@ cglobal %1_h264_chroma_mc4_10, 6,6,7
     MC4_OP m6, m0
     sub   r3d, 2
     jnz .next2rows
-    REP_RET
+    RET
 %endmacro
 
 ;-----------------------------------------------------------------------------
@@ -234,7 +234,7 @@ cglobal %1_h264_chroma_mc2_10, 6,7
     add           r0, r2
     dec          r3d
     jnz .nextrow
-    REP_RET
+    RET
 %endmacro
 
 %macro NOTHING 2-3

libavcodec/x86/h264_deblock_10bit.asm

@@ -372,7 +372,7 @@ cglobal deblock_v_luma_10, 5,5,15
     add         r4, 2
     dec         r3
     jg .loop
-    REP_RET
+    RET
 
 cglobal deblock_h_luma_10, 5,7,15
     shl        r2d, 2
@@ -411,7 +411,7 @@ cglobal deblock_h_luma_10, 5,7,15
     lea         r5, [r5+r1*8]
     dec         r6
     jg .loop
-    REP_RET
+    RET
 %endmacro
 
 INIT_XMM sse2
@@ -648,7 +648,7 @@ cglobal deblock_v_luma_intra_10, 4,7,16
     add     r4, mmsize
     dec     r6
     jg .loop
-    REP_RET
+    RET
 
 ;-----------------------------------------------------------------------------
 ; void ff_deblock_h_luma_intra_10(uint16_t *pix, int stride, int alpha,