libavcodec/ppc: remove broken AltiVec acceleration for lossless audio

It fails sporadically on big- or little-endian POWER/PPC, as well as
triggers the undefined-behaviour & address scanners in GCC.

libavcodec/bit_depth_template.c

@@ -43,11 +43,13 @@
 #   undef FUNCC
 #   undef av_clip_pixel
 #   undef PIXEL_SPLAT_X4
+#   undef PIXELSIZE
 #else
 #   define AVCODEC_BIT_DEPTH_TEMPLATE_C
 #endif
 
 #if BIT_DEPTH > 8
+#   define PIXELSIZE 16
 #   define pixel  uint16_t
 #   define pixel2 uint32_t
 #   define pixel4 uint64_t
@@ -76,6 +78,7 @@
 #   define av_clip_pixel(a) av_clip_uintp2(a, BIT_DEPTH)
 #   define CLIP(a)          av_clip_uintp2(a, BIT_DEPTH)
 #else
+#   define PIXELSIZE 8
 #   define pixel  uint8_t
 #   define pixel2 uint16_t
 #   define pixel4 uint32_t
@@ -100,6 +103,7 @@
 #define FUNC2(a, b, c)  FUNC3(a, b, c)
 #define FUNC(a)  FUNC2(a, BIT_DEPTH,)
 #define FUNCC(a) FUNC2(a, BIT_DEPTH, _c)
+#define FUNCC2(a) FUNC2(a, PIXELSIZE, _c)
 #define FUNC4(a, b, c)  a ## _int ## b ## _ ## c ## bit
 #define FUNC5(a, b, c)  FUNC4(a, b, c)
 #define FUNC6(a)  FUNC5(a, IN_IDCT_DEPTH, BIT_DEPTH)

libavcodec/h264dsp.c

@@ -69,14 +69,19 @@ av_cold void ff_h264dsp_init(H264DSPContext *c, const int bit_depth,
 #undef FUNC
 #define FUNC(a, depth) a ## _ ## depth ## _c
 
-#define ADDPX_DSP(depth) \
+#define SET_PIXSIZE_FUNCS(depth) \
+    c->h264_luma_dc_dequant_idct= FUNC(ff_h264_luma_dc_dequant_idct, depth);\
+    if (chroma_format_idc <= 1)\
+        c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma_dc_dequant_idct, depth);\
+    else\
+        c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma422_dc_dequant_idct, depth);\
     c->h264_add_pixels4_clear = FUNC(ff_h264_add_pixels4, depth);\
     c->h264_add_pixels8_clear = FUNC(ff_h264_add_pixels8, depth)
 
     if (bit_depth > 8 && bit_depth <= 16) {
-        ADDPX_DSP(16);
+        SET_PIXSIZE_FUNCS(16);
     } else {
-        ADDPX_DSP(8);
+        SET_PIXSIZE_FUNCS(8);
     }
 
 #define H264_DSP(depth) \
@@ -91,11 +96,6 @@ av_cold void ff_h264dsp_init(H264DSPContext *c, const int bit_depth,
     else\
         c->h264_idct_add8  = FUNC(ff_h264_idct_add8_422, depth);\
     c->h264_idct_add16intra= FUNC(ff_h264_idct_add16intra, depth);\
-    c->h264_luma_dc_dequant_idct= FUNC(ff_h264_luma_dc_dequant_idct, depth);\
-    if (chroma_format_idc <= 1)\
-        c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma_dc_dequant_idct, depth);\
-    else\
-        c->h264_chroma_dc_dequant_idct= FUNC(ff_h264_chroma422_dc_dequant_idct, depth);\
 \
     c->weight_h264_pixels_tab[0]= FUNC(weight_h264_pixels16, depth);\
     c->weight_h264_pixels_tab[1]= FUNC(weight_h264_pixels8, depth);\

libavcodec/h264idct.h

@@ -31,9 +31,6 @@ void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffse
 void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[5 * 8]);\
 void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 * 8]);\
 void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, int16_t *block, int stride, const uint8_t nnzc[15 * 8]);\
-void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(int16_t *output, int16_t *input, int qmul);\
-void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(int16_t *block, int qmul);\
-void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(int16_t *block, int qmul);
 
 H264_IDCT( 8)
 H264_IDCT( 9)
@@ -41,4 +38,12 @@ H264_IDCT(10)
 H264_IDCT(12)
 H264_IDCT(14)
 
+#define H264_IDCT2(pixsize) \
+void ff_h264_luma_dc_dequant_idct_ ## pixsize ## _c(int16_t *output, int16_t *input, int qmul);\
+void ff_h264_chroma422_dc_dequant_idct_ ## pixsize ## _c(int16_t *block, int qmul);\
+void ff_h264_chroma_dc_dequant_idct_ ## pixsize ## _c(int16_t *block, int qmul);
+
+H264_IDCT2( 8)
+H264_IDCT2(16)
+
 #endif /* AVCODEC_H264IDCT_H */

libavcodec/h264idct_template.c

@@ -244,11 +244,13 @@ void FUNCC(ff_h264_idct_add8_422)(uint8_t **dest, const int *block_offset, int16
     }
 }
 
+#if BIT_DEPTH == 8 || BIT_DEPTH == 9
 /**
  * IDCT transforms the 16 dc values and dequantizes them.
  * @param qmul quantization parameter
  */
-void FUNCC(ff_h264_luma_dc_dequant_idct)(int16_t *_output, int16_t *_input, int qmul){
+void FUNCC2(ff_h264_luma_dc_dequant_idct)(int16_t *_output, int16_t *_input, int qmul)
+{
 #define stride 16
     int i;
     int temp[16];
@@ -283,7 +285,8 @@ void FUNCC(ff_h264_luma_dc_dequant_idct)(int16_t *_output, int16_t *_input, int
 #undef stride
 }
 
-void FUNCC(ff_h264_chroma422_dc_dequant_idct)(int16_t *_block, int qmul){
+void FUNCC2(ff_h264_chroma422_dc_dequant_idct)(int16_t *_block, int qmul)
+{
     const int stride= 16*2;
     const int xStride= 16;
     int i;
@@ -310,7 +313,8 @@ void FUNCC(ff_h264_chroma422_dc_dequant_idct)(int16_t *_block, int qmul){
     }
 }
 
-void FUNCC(ff_h264_chroma_dc_dequant_idct)(int16_t *_block, int qmul){
+void FUNCC2(ff_h264_chroma_dc_dequant_idct)(int16_t *_block, int qmul)
+{
     const int stride= 16*2;
     const int xStride= 16;
     SUINT a,b,c,d,e;
@@ -331,3 +335,4 @@ void FUNCC(ff_h264_chroma_dc_dequant_idct)(int16_t *_block, int qmul){
     block[stride*1 + xStride*0]= (int)((a-c)*qmul) >> 7;
     block[stride*1 + xStride*1]= (int)((e-b)*qmul) >> 7;
 }
+#endif

libavcodec/lossless_audiodsp.c

@@ -61,8 +61,6 @@ av_cold void ff_llauddsp_init(LLAudDSPContext *c)
 
 #if ARCH_ARM
     ff_llauddsp_init_arm(c);
-#elif ARCH_PPC
-    ff_llauddsp_init_ppc(c);
 #elif ARCH_RISCV
     ff_llauddsp_init_riscv(c);
 #elif ARCH_X86

libavcodec/lossless_audiodsp.h

@@ -45,7 +45,6 @@ typedef struct LLAudDSPContext {
 
 void ff_llauddsp_init(LLAudDSPContext *c);
 void ff_llauddsp_init_arm(LLAudDSPContext *c);
-void ff_llauddsp_init_ppc(LLAudDSPContext *c);
 void ff_llauddsp_init_riscv(LLAudDSPContext *c);
 void ff_llauddsp_init_x86(LLAudDSPContext *c);
 

libavcodec/ppc/Makefile

@@ -21,7 +21,6 @@ OBJS-$(CONFIG_VP8DSP)                  += ppc/vp8dsp_altivec.o
 
 # decoders/encoders
 OBJS-$(CONFIG_HEVC_DECODER)            += ppc/hevcdsp.o
-OBJS-$(CONFIG_LLAUDDSP)                += ppc/lossless_audiodsp_altivec.o
 OBJS-$(CONFIG_MPEG4_DECODER)           += ppc/mpeg4videodsp.o
 OBJS-$(CONFIG_SVQ1_ENCODER)            += ppc/svq1enc_altivec.o
 OBJS-$(CONFIG_VORBIS_DECODER)          += ppc/vorbisdsp_altivec.o

libavcodec/ppc/lossless_audiodsp_altivec.c

@@ -1,91 +0,0 @@
-/*
- * Copyright (c) 2007 Luca Barbato <lu_zero@gentoo.org>
- *
- * This file is part of FFmpeg.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "config.h"
-
-#include "libavutil/attributes.h"
-#include "libavutil/cpu.h"
-#include "libavutil/ppc/cpu.h"
-#include "libavutil/ppc/util_altivec.h"
-
-#include "libavcodec/lossless_audiodsp.h"
-
-#if HAVE_BIGENDIAN
-#define GET_T(tt0,tt1,src,a,b){       \
-        a = vec_ld(16, src);          \
-        tt0 = vec_perm(b, a, align);  \
-        b = vec_ld(32, src);          \
-        tt1 = vec_perm(a, b, align);  \
- }
-#else
-#define GET_T(tt0,tt1,src,a,b){       \
-        tt0 = vec_vsx_ld(0, src);     \
-        tt1 = vec_vsx_ld(16, src);    \
- }
-#endif
-
-#if HAVE_ALTIVEC
-static int32_t scalarproduct_and_madd_int16_altivec(int16_t *v1,
-                                                    const int16_t *v2,
-                                                    const int16_t *v3,
-                                                    int order, int mul)
-{
-    LOAD_ZERO;
-    vec_s16 *pv1 = (vec_s16 *) v1;
-    register vec_s16 muls = { mul, mul, mul, mul, mul, mul, mul, mul };
-    register vec_s16 t0, t1, i0, i1, i4, i2, i3;
-    register vec_s32 res = zero_s32v;
-#if HAVE_BIGENDIAN
-    register vec_u8 align = vec_lvsl(0, v2);
-    i2 = vec_ld(0, v2);
-    i3 = vec_ld(0, v3);
-#endif
-    int32_t ires;
-
-    order >>= 4;
-    do {
-        GET_T(t0,t1,v2,i1,i2);
-        i0     = pv1[0];
-        i1     = pv1[1];
-        res    = vec_msum(t0, i0, res);
-        res    = vec_msum(t1, i1, res);
-        GET_T(t0,t1,v3,i4,i3);
-        pv1[0] = vec_mladd(t0, muls, i0);
-        pv1[1] = vec_mladd(t1, muls, i1);
-        pv1   += 2;
-        v2    += 16;
-        v3    += 16;
-    } while (--order);
-    res = vec_splat(vec_sums(res, zero_s32v), 3);
-    vec_ste(res, 0, &ires);
-
-    return ires;
-}
-#endif /* HAVE_ALTIVEC */
-
-av_cold void ff_llauddsp_init_ppc(LLAudDSPContext *c)
-{
-#if HAVE_ALTIVEC
-    if (!PPC_ALTIVEC(av_get_cpu_flags()))
-        return;
-
-    c->scalarproduct_and_madd_int16 = scalarproduct_and_madd_int16_altivec;
-#endif /* HAVE_ALTIVEC */
-}

libavutil/hwcontext_vulkan.c

@@ -1772,6 +1772,11 @@ static int vulkan_device_create_internal(AVHWDeviceContext *ctx,
     dev_info.pNext = p->feats.device.pNext;
     dev_info.pEnabledFeatures = &p->feats.device.features;
 
+    /* Limit queues to a given number if needed */
+    opt_d = av_dict_get(opts, "limit_queues", NULL, 0);
+    if (opt_d)
+        p->limit_queues = strtol(opt_d->value, NULL, 10);
+
     /* Setup enabled queue families */
     if ((err = setup_queue_families(ctx, &dev_info)))
         goto end;
@@ -1799,11 +1804,6 @@ static int vulkan_device_create_internal(AVHWDeviceContext *ctx,
     if (opt_d)
         p->use_linear_images = strtol(opt_d->value, NULL, 10);
 
-    /* Limit queues to a given number if needed */
-    opt_d = av_dict_get(opts, "limit_queues", NULL, 0);
-    if (opt_d)
-        p->limit_queues = strtol(opt_d->value, NULL, 10);
-
     /* The disable_multiplane argument takes precedent over the option */
     p->disable_multiplane = disable_multiplane;
     if (!p->disable_multiplane) {