x64 always has MMX, MMXEXT, SSE and SSE2 and this means that some functions for MMX, MMXEXT, SSE and 3dnow are always overridden by other functions (unless one e.g. explicitly disables SSE2). So given that the only systems which benefit from these functions are truely ancient 32bit x86s they are removed.
Unified · Split
| @@ -45,7 +45,6 @@ SECTION .text · 44 unmodified lines | ||
| 45 | 45 | jnz .%1_y_loop |
| 46 | 46 | %endmacro |
| 47 | 47 | |
| 48 | −%macro vvar_fn 0 | |
| 49 | 48 | ; .----. <- zero |
| 50 | 49 | ; | | <- top is copied from first line in body of source |
| 51 | 50 | ; |----| <- start_y |
| @@ -53,6 +52,7 @@ SECTION .text · 1 unmodified line | ||
| 53 | 52 | ; |----| <- end_y |
| 54 | 53 | ; | | <- bottom is copied from last line in body of source |
| 55 | 54 | ; '----' <- bh |
| 55 | +INIT_XMM sse | |
| 56 | 56 | %if ARCH_X86_64 |
| 57 | 57 | cglobal emu_edge_vvar, 7, 8, 1, dst, dst_stride, src, src_stride, \ |
| 58 | 58 | start_y, end_y, bh, w |
| @@ -81,15 +81,6 @@ cglobal emu_edge_vvar, 1, 6, 1, dst, src, start_y, end_y, bh, w · 22 unmodified lines | ||
| 81 | 81 | V_COPY_ROW bottom, bhq ; v_copy_row(bottom, bh) |
| 82 | 82 | .end: ; } |
| 83 | 83 | RET |
| 84 | −%endmacro | |
| 85 | − | |
| 86 | −%if ARCH_X86_32 | |
| 87 | −INIT_MMX mmx | |
| 88 | −vvar_fn | |
| 89 | −%endif | |
| 90 | − | |
| 91 | −INIT_XMM sse | |
| 92 | −vvar_fn | |
| 93 | 84 | |
| 94 | 85 | %macro hvar_fn 0 |
| 95 | 86 | cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w |
| @@ -105,11 +96,7 @@ cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w · 9 unmodified lines | ||
| 105 | 96 | imul wd, 0x01010101 ; w *= 0x01010101 |
| 106 | 97 | movd m0, wd |
| 107 | 98 | mov wq, n_wordsq ; initialize w |
| 108 | −%if cpuflag(sse2) | |
| 109 | 99 | pshufd m0, m0, q0000 ; splat |
| 110 | −%else ; mmx | |
| 111 | − punpckldq m0, m0 ; splat | |
| 112 | −%endif ; mmx/sse | |
| 113 | 100 | %endif ; avx2 |
| 114 | 101 | .x_loop: ; do { |
| 115 | 102 | movu [dstq+wq*2], m0 ; write($reg, $mmsize) |
| @@ -123,11 +110,6 @@ cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w · 7 unmodified lines | ||
| 123 | 110 | RET |
| 124 | 111 | %endmacro |
| 125 | 112 | |
| 126 | −%if ARCH_X86_32 | |
| 127 | −INIT_MMX mmx | |
| 128 | −hvar_fn | |
| 129 | −%endif | |
| 130 | − | |
| 131 | 113 | INIT_XMM sse2 |
| 132 | 114 | hvar_fn |
| 133 | 115 | |
| @@ -338,9 +320,6 @@ cglobal emu_edge_vfix %+ %%n, 1, 5, 1, dst, src, start_y, end_y, bh · 204 unmodified lines | ||
| 338 | 320 | |
| 339 | 321 | INIT_MMX mmx |
| 340 | 322 | VERTICAL_EXTEND 1, 15 |
| 341 | −%if ARCH_X86_32 | |
| 342 | −VERTICAL_EXTEND 16, 22 | |
| 343 | −%endif | |
| 344 | 323 | |
| 345 | 324 | INIT_XMM sse |
| 346 | 325 | VERTICAL_EXTEND 16, 22 |
| @@ -438,9 +417,6 @@ cglobal emu_edge_hfix %+ %%n, 4, 5, 1, dst, dst_stride, start_x, bh, val · 91 unmodified lines | ||
| 438 | 417 | |
| 439 | 418 | INIT_MMX mmx |
| 440 | 419 | H_EXTEND 2, 14 |
| 441 | −%if ARCH_X86_32 | |
| 442 | −H_EXTEND 16, 22 | |
| 443 | −%endif | |
| 444 | 420 | |
| 445 | 421 | INIT_XMM sse2 |
| 446 | 422 | H_EXTEND 16, 22 |
| @@ -450,19 +426,11 @@ INIT_XMM avx2 · 3 unmodified lines | ||
| 450 | 426 | H_EXTEND 8, 22 |
| 451 | 427 | %endif |
| 452 | 428 | |
| 453 | −%macro PREFETCH_FN 1 | |
| 429 | +INIT_MMX mmxext | |
| 454 | 430 | cglobal prefetch, 3, 3, 0, buf, stride, h |
| 455 | 431 | .loop: |
| 456 | − %1 [bufq] | |
| 432 | + prefetcht0 [bufq] | |
| 457 | 433 | add bufq, strideq |
| 458 | 434 | dec hd |
| 459 | 435 | jg .loop |
| 460 | 436 | REP_RET |
| 461 | −%endmacro | |
| 462 | − | |
| 463 | −INIT_MMX mmxext | |
| 464 | −PREFETCH_FN prefetcht0 | |
| 465 | −%if ARCH_X86_32 | |
| 466 | −INIT_MMX 3dnow | |
| 467 | −PREFETCH_FN prefetch | |
| 468 | −%endif | |
| @@ -52,26 +52,6 @@ extern emu_edge_vfix_func ff_emu_edge_vfix12_mmx; · 51 unmodified lines | ||
| 52 | 52 | extern emu_edge_vfix_func ff_emu_edge_vfix13_mmx; |
| 53 | 53 | extern emu_edge_vfix_func ff_emu_edge_vfix14_mmx; |
| 54 | 54 | extern emu_edge_vfix_func ff_emu_edge_vfix15_mmx; |
| 55 | −extern emu_edge_vfix_func ff_emu_edge_vfix16_mmx; | |
| 56 | −extern emu_edge_vfix_func ff_emu_edge_vfix17_mmx; | |
| 57 | −extern emu_edge_vfix_func ff_emu_edge_vfix18_mmx; | |
| 58 | −extern emu_edge_vfix_func ff_emu_edge_vfix19_mmx; | |
| 59 | −extern emu_edge_vfix_func ff_emu_edge_vfix20_mmx; | |
| 60 | −extern emu_edge_vfix_func ff_emu_edge_vfix21_mmx; | |
| 61 | −extern emu_edge_vfix_func ff_emu_edge_vfix22_mmx; | |
| 62 | −#if ARCH_X86_32 | |
| 63 | −static emu_edge_vfix_func * const vfixtbl_mmx[22] = { | |
| 64 | − &ff_emu_edge_vfix1_mmx, &ff_emu_edge_vfix2_mmx, &ff_emu_edge_vfix3_mmx, | |
| 65 | − &ff_emu_edge_vfix4_mmx, &ff_emu_edge_vfix5_mmx, &ff_emu_edge_vfix6_mmx, | |
| 66 | − &ff_emu_edge_vfix7_mmx, &ff_emu_edge_vfix8_mmx, &ff_emu_edge_vfix9_mmx, | |
| 67 | − &ff_emu_edge_vfix10_mmx, &ff_emu_edge_vfix11_mmx, &ff_emu_edge_vfix12_mmx, | |
| 68 | − &ff_emu_edge_vfix13_mmx, &ff_emu_edge_vfix14_mmx, &ff_emu_edge_vfix15_mmx, | |
| 69 | − &ff_emu_edge_vfix16_mmx, &ff_emu_edge_vfix17_mmx, &ff_emu_edge_vfix18_mmx, | |
| 70 | − &ff_emu_edge_vfix19_mmx, &ff_emu_edge_vfix20_mmx, &ff_emu_edge_vfix21_mmx, | |
| 71 | − &ff_emu_edge_vfix22_mmx | |
| 72 | −}; | |
| 73 | −#endif | |
| 74 | −extern emu_edge_vvar_func ff_emu_edge_vvar_mmx; | |
| 75 | 55 | extern emu_edge_vfix_func ff_emu_edge_vfix16_sse; |
| 76 | 56 | extern emu_edge_vfix_func ff_emu_edge_vfix17_sse; |
| 77 | 57 | extern emu_edge_vfix_func ff_emu_edge_vfix18_sse; |
| @@ -103,19 +83,6 @@ extern emu_edge_hfix_func ff_emu_edge_hfix8_mmx; · 25 unmodified lines | ||
| 103 | 83 | extern emu_edge_hfix_func ff_emu_edge_hfix10_mmx; |
| 104 | 84 | extern emu_edge_hfix_func ff_emu_edge_hfix12_mmx; |
| 105 | 85 | extern emu_edge_hfix_func ff_emu_edge_hfix14_mmx; |
| 106 | −extern emu_edge_hfix_func ff_emu_edge_hfix16_mmx; | |
| 107 | −extern emu_edge_hfix_func ff_emu_edge_hfix18_mmx; | |
| 108 | −extern emu_edge_hfix_func ff_emu_edge_hfix20_mmx; | |
| 109 | −extern emu_edge_hfix_func ff_emu_edge_hfix22_mmx; | |
| 110 | −#if ARCH_X86_32 | |
| 111 | −static emu_edge_hfix_func * const hfixtbl_mmx[11] = { | |
| 112 | − ff_emu_edge_hfix2_mmx, ff_emu_edge_hfix4_mmx, ff_emu_edge_hfix6_mmx, | |
| 113 | − ff_emu_edge_hfix8_mmx, ff_emu_edge_hfix10_mmx, ff_emu_edge_hfix12_mmx, | |
| 114 | − ff_emu_edge_hfix14_mmx, ff_emu_edge_hfix16_mmx, ff_emu_edge_hfix18_mmx, | |
| 115 | − ff_emu_edge_hfix20_mmx, ff_emu_edge_hfix22_mmx | |
| 116 | −}; | |
| 117 | −#endif | |
| 118 | −extern emu_edge_hvar_func ff_emu_edge_hvar_mmx; | |
| 119 | 86 | extern emu_edge_hfix_func ff_emu_edge_hfix16_sse2; |
| 120 | 87 | extern emu_edge_hfix_func ff_emu_edge_hfix18_sse2; |
| 121 | 88 | extern emu_edge_hfix_func ff_emu_edge_hfix20_sse2; |
| @@ -221,30 +188,6 @@ static av_always_inline void emulated_edge_mc(uint8_t *dst, const uint8_t *src, · 99 unmodified lines | ||
| 221 | 188 | } |
| 222 | 189 | } |
| 223 | 190 | |
| 224 | −#if ARCH_X86_32 | |
| 225 | −static av_noinline void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src, | |
| 226 | − ptrdiff_t buf_stride, | |
| 227 | − ptrdiff_t src_stride, | |
| 228 | − int block_w, int block_h, | |
| 229 | − int src_x, int src_y, int w, int h) | |
| 230 | −{ | |
| 231 | − emulated_edge_mc(buf, src, buf_stride, src_stride, block_w, block_h, | |
| 232 | − src_x, src_y, w, h, vfixtbl_mmx, &ff_emu_edge_vvar_mmx, | |
| 233 | − hfixtbl_mmx, &ff_emu_edge_hvar_mmx); | |
| 234 | −} | |
| 235 | − | |
| 236 | −static av_noinline void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src, | |
| 237 | − ptrdiff_t buf_stride, | |
| 238 | − ptrdiff_t src_stride, | |
| 239 | − int block_w, int block_h, | |
| 240 | − int src_x, int src_y, int w, int h) | |
| 241 | −{ | |
| 242 | − emulated_edge_mc(buf, src, buf_stride, src_stride, block_w, block_h, | |
| 243 | − src_x, src_y, w, h, vfixtbl_sse, &ff_emu_edge_vvar_sse, | |
| 244 | − hfixtbl_mmx, &ff_emu_edge_hvar_mmx); | |
| 245 | −} | |
| 246 | −#endif | |
| 247 | − | |
| 248 | 191 | static av_noinline void emulated_edge_mc_sse2(uint8_t *buf, const uint8_t *src, |
| 249 | 192 | ptrdiff_t buf_stride, |
| 250 | 193 | ptrdiff_t src_stride, |
| @@ -273,29 +216,15 @@ static av_noinline void emulated_edge_mc_avx2(uint8_t *buf, const uint8_t *src, · 22 unmodified lines | ||
| 273 | 216 | #endif /* HAVE_X86ASM */ |
| 274 | 217 | |
| 275 | 218 | void ff_prefetch_mmxext(uint8_t *buf, ptrdiff_t stride, int h); |
| 276 | −void ff_prefetch_3dnow(uint8_t *buf, ptrdiff_t stride, int h); | |
| 277 | 219 | |
| 278 | 220 | av_cold void ff_videodsp_init_x86(VideoDSPContext *ctx, int bpc) |
| 279 | 221 | { |
| 280 | 222 | #if HAVE_X86ASM |
| 281 | 223 | int cpu_flags = av_get_cpu_flags(); |
| 282 | 224 | |
| 283 | −#if ARCH_X86_32 | |
| 284 | − if (EXTERNAL_MMX(cpu_flags) && bpc <= 8) { | |
| 285 | − ctx->emulated_edge_mc = emulated_edge_mc_mmx; | |
| 286 | − } | |
| 287 | − if (EXTERNAL_AMD3DNOW(cpu_flags)) { | |
| 288 | − ctx->prefetch = ff_prefetch_3dnow; | |
| 289 | − } | |
| 290 | −#endif /* ARCH_X86_32 */ | |
| 291 | 225 | if (EXTERNAL_MMXEXT(cpu_flags)) { |
| 292 | 226 | ctx->prefetch = ff_prefetch_mmxext; |
| 293 | 227 | } |
| 294 | −#if ARCH_X86_32 | |
| 295 | − if (EXTERNAL_SSE(cpu_flags) && bpc <= 8) { | |
| 296 | − ctx->emulated_edge_mc = emulated_edge_mc_sse; | |
| 297 | − } | |
| 298 | −#endif /* ARCH_X86_32 */ | |
| 299 | 228 | if (EXTERNAL_SSE2(cpu_flags) && bpc <= 8) { |
| 300 | 229 | ctx->emulated_edge_mc = emulated_edge_mc_sse2; |
| 301 | 230 | } |