# HG changeset patch # User michael # Date 1188059325 0 # Node ID fadffa1e5aefdc4de739bdda4c8ee70df3632f1e # Parent 642588a60570f1e4fdc76e1c92ad12da9d4fb3ad prevent one overflow in the first vertical lifting step diff -r 642588a60570 -r fadffa1e5aef i386/snowdsp_mmx.c --- a/i386/snowdsp_mmx.c Sat Aug 25 15:20:56 2007 +0000 +++ b/i386/snowdsp_mmx.c Sat Aug 25 16:28:45 2007 +0000 @@ -448,16 +448,16 @@ snow_vertical_compose_sse2_load(REG_S,"xmm0","xmm2","xmm4","xmm6") snow_vertical_compose_sse2_add(REG_a,"xmm0","xmm2","xmm4","xmm6") snow_vertical_compose_sse2_move("xmm0","xmm2","xmm4","xmm6","xmm1","xmm3","xmm5","xmm7") - snow_vertical_compose_sse2_r2r_add("xmm0","xmm2","xmm4","xmm6","xmm0","xmm2","xmm4","xmm6") + snow_vertical_compose_sse2_sra("1","xmm0","xmm2","xmm4","xmm6") snow_vertical_compose_sse2_r2r_add("xmm1","xmm3","xmm5","xmm7","xmm0","xmm2","xmm4","xmm6") "pcmpeqd %%xmm1, %%xmm1 \n\t" "psllw $15, %%xmm1 \n\t" - "psrlw $13, %%xmm1 \n\t" + "psrlw $14, %%xmm1 \n\t" "mov %5, %%"REG_a" \n\t" snow_vertical_compose_sse2_r2r_add("xmm1","xmm1","xmm1","xmm1","xmm0","xmm2","xmm4","xmm6") - snow_vertical_compose_sse2_sra("3","xmm0","xmm2","xmm4","xmm6") + snow_vertical_compose_sse2_sra("2","xmm0","xmm2","xmm4","xmm6") snow_vertical_compose_sse2_load(REG_a,"xmm1","xmm3","xmm5","xmm7") snow_vertical_compose_sse2_sub("xmm0","xmm2","xmm4","xmm6","xmm1","xmm3","xmm5","xmm7") snow_vertical_compose_sse2_store(REG_a,"xmm1","xmm3","xmm5","xmm7") @@ -550,16 +550,16 @@ snow_vertical_compose_mmx_load(REG_S,"mm0","mm2","mm4","mm6") snow_vertical_compose_mmx_add(REG_a,"mm0","mm2","mm4","mm6") snow_vertical_compose_mmx_move("mm0","mm2","mm4","mm6","mm1","mm3","mm5","mm7") - snow_vertical_compose_mmx_r2r_add("mm0","mm2","mm4","mm6","mm0","mm2","mm4","mm6") + snow_vertical_compose_mmx_sra("1","mm0","mm2","mm4","mm6") snow_vertical_compose_mmx_r2r_add("mm1","mm3","mm5","mm7","mm0","mm2","mm4","mm6") "pcmpeqw %%mm1, %%mm1 \n\t" "psllw $15, %%mm1 \n\t" - "psrlw $13, %%mm1 \n\t" + "psrlw $14, %%mm1 \n\t" "mov %5, %%"REG_a" \n\t" snow_vertical_compose_mmx_r2r_add("mm1","mm1","mm1","mm1","mm0","mm2","mm4","mm6") - snow_vertical_compose_mmx_sra("3","mm0","mm2","mm4","mm6") + snow_vertical_compose_mmx_sra("2","mm0","mm2","mm4","mm6") snow_vertical_compose_mmx_load(REG_a,"mm1","mm3","mm5","mm7") snow_vertical_compose_mmx_sub("mm0","mm2","mm4","mm6","mm1","mm3","mm5","mm7") snow_vertical_compose_mmx_store(REG_a,"mm1","mm3","mm5","mm7")