From 59effb34124ffe82d8dbed7caf8b56abfade6e43 Mon Sep 17 00:00:00 2001 From: Nekotekina Date: Thu, 2 Apr 2015 18:22:43 +0300 Subject: [PATCH] PPU: some instructions replaced --- rpcs3/Emu/Cell/PPUInterpreter.cpp | 229 +++++++----------------------- 1 file changed, 55 insertions(+), 174 deletions(-) diff --git a/rpcs3/Emu/Cell/PPUInterpreter.cpp b/rpcs3/Emu/Cell/PPUInterpreter.cpp index 887266218f..5546f95e41 100644 --- a/rpcs3/Emu/Cell/PPUInterpreter.cpp +++ b/rpcs3/Emu/Cell/PPUInterpreter.cpp @@ -516,100 +516,50 @@ void ppu_interpreter::VMLADDUHM(PPUThread& CPU, ppu_opcode_t op) void ppu_interpreter::VMRGHB(PPUThread& CPU, ppu_opcode_t op) { - u128 VA = CPU.VPR[op.va]; - u128 VB = CPU.VPR[op.vb]; - for (uint h = 0; h < 8; h++) - { - CPU.VPR[op.vd]._u8[15 - h * 2] = VA._u8[15 - h]; - CPU.VPR[op.vd]._u8[15 - h * 2 - 1] = VB._u8[15 - h]; - } + CPU.VPR[op.vd].vi = _mm_unpackhi_epi8(CPU.VPR[op.vb].vi, CPU.VPR[op.va].vi); } void ppu_interpreter::VMRGHH(PPUThread& CPU, ppu_opcode_t op) { - u128 VA = CPU.VPR[op.va]; - u128 VB = CPU.VPR[op.vb]; - for (uint w = 0; w < 4; w++) - { - CPU.VPR[op.vd]._u16[7 - w * 2] = VA._u16[7 - w]; - CPU.VPR[op.vd]._u16[7 - w * 2 - 1] = VB._u16[7 - w]; - } + CPU.VPR[op.vd].vi = _mm_unpackhi_epi16(CPU.VPR[op.vb].vi, CPU.VPR[op.va].vi); } void ppu_interpreter::VMRGHW(PPUThread& CPU, ppu_opcode_t op) { - u128 VA = CPU.VPR[op.va]; - u128 VB = CPU.VPR[op.vb]; - for (uint d = 0; d < 2; d++) - { - CPU.VPR[op.vd]._u32[3 - d * 2] = VA._u32[3 - d]; - CPU.VPR[op.vd]._u32[3 - d * 2 - 1] = VB._u32[3 - d]; - } + CPU.VPR[op.vd].vi = _mm_unpackhi_epi32(CPU.VPR[op.vb].vi, CPU.VPR[op.va].vi); } void ppu_interpreter::VMRGLB(PPUThread& CPU, ppu_opcode_t op) { - u128 VA = CPU.VPR[op.va]; - u128 VB = CPU.VPR[op.vb]; - for (uint h = 0; h < 8; h++) - { - CPU.VPR[op.vd]._u8[15 - h * 2] = VA._u8[7 - h]; - CPU.VPR[op.vd]._u8[15 - h * 2 - 1] = VB._u8[7 - h]; - } + CPU.VPR[op.vd].vi = _mm_unpacklo_epi8(CPU.VPR[op.vb].vi, CPU.VPR[op.va].vi); } void ppu_interpreter::VMRGLH(PPUThread& CPU, ppu_opcode_t op) { - u128 VA = CPU.VPR[op.va]; - u128 VB = CPU.VPR[op.vb]; - for (uint w = 0; w < 4; w++) - { - CPU.VPR[op.vd]._u16[7 - w * 2] = VA._u16[3 - w]; - CPU.VPR[op.vd]._u16[7 - w * 2 - 1] = VB._u16[3 - w]; - } + CPU.VPR[op.vd].vi = _mm_unpacklo_epi16(CPU.VPR[op.vb].vi, CPU.VPR[op.va].vi); } void ppu_interpreter::VMRGLW(PPUThread& CPU, ppu_opcode_t op) { - u128 VA = CPU.VPR[op.va]; - u128 VB = CPU.VPR[op.vb]; - for (uint d = 0; d < 2; d++) - { - CPU.VPR[op.vd]._u32[3 - d * 2] = VA._u32[1 - d]; - CPU.VPR[op.vd]._u32[3 - d * 2 - 1] = VB._u32[1 - d]; - } + CPU.VPR[op.vd].vi = _mm_unpacklo_epi32(CPU.VPR[op.vb].vi, CPU.VPR[op.va].vi); } void ppu_interpreter::VMSUMMBM(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - s32 result = 0; - - for (uint b = 0; b < 4; b++) - { - result += CPU.VPR[op.va]._s8[w * 4 + b] * CPU.VPR[op.vb]._u8[w * 4 + b]; - } - - result += CPU.VPR[op.vc]._s32[w]; - CPU.VPR[op.vd]._s32[w] = result; - } + const auto a = CPU.VPR[op.va].vi; // signed bytes + const auto b = CPU.VPR[op.vb].vi; // unsigned bytes + const auto ah = _mm_srai_epi16(a, 8); + const auto bh = _mm_srli_epi16(b, 8); + const auto al = _mm_srai_epi16(_mm_srli_epi16(a, 8), 8); + const auto bl = _mm_and_si128(b, _mm_set1_epi16(0x00ff)); + const auto sh = _mm_madd_epi16(ah, bh); + const auto sl = _mm_madd_epi16(al, bl); + CPU.VPR[op.vd].vi = _mm_add_epi32(_mm_add_epi32(CPU.VPR[op.vc].vi, sh), sl); } void ppu_interpreter::VMSUMSHM(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - s32 result = 0; - - for (uint h = 0; h < 2; h++) - { - result += CPU.VPR[op.va]._s16[w * 2 + h] * CPU.VPR[op.vb]._s16[w * 2 + h]; - } - - result += CPU.VPR[op.vc]._s32[w]; - CPU.VPR[op.vd]._s32[w] = result; - } + CPU.VPR[op.vd].vi = _mm_add_epi32(_mm_madd_epi16(CPU.VPR[op.va].vi, CPU.VPR[op.vb].vi), CPU.VPR[op.vc].vi); } void ppu_interpreter::VMSUMSHS(PPUThread& CPU, ppu_opcode_t op) @@ -643,34 +593,27 @@ void ppu_interpreter::VMSUMSHS(PPUThread& CPU, ppu_opcode_t op) void ppu_interpreter::VMSUMUBM(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - u32 result = 0; - - for (uint b = 0; b < 4; b++) - { - result += (u32)CPU.VPR[op.va]._u8[w * 4 + b] * (u32)CPU.VPR[op.vb]._u8[w * 4 + b]; - } - - result += CPU.VPR[op.vc]._u32[w]; - CPU.VPR[op.vd]._u32[w] = result; - } + const auto a = CPU.VPR[op.va].vi; + const auto b = CPU.VPR[op.vb].vi; + const auto mask = _mm_set1_epi16(0x00ff); + const auto ah = _mm_srli_epi16(a, 8); + const auto al = _mm_and_si128(a, mask); + const auto bh = _mm_srli_epi16(b, 8); + const auto bl = _mm_and_si128(b, mask); + const auto sh = _mm_madd_epi16(ah, bh); + const auto sl = _mm_madd_epi16(al, bl); + CPU.VPR[op.vd].vi = _mm_add_epi32(_mm_add_epi32(CPU.VPR[op.vc].vi, sh), sl); } void ppu_interpreter::VMSUMUHM(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - u32 result = 0; - - for (uint h = 0; h < 2; h++) - { - result += (u32)CPU.VPR[op.va]._u16[w * 2 + h] * (u32)CPU.VPR[op.vb]._u16[w * 2 + h]; - } - - result += CPU.VPR[op.vc]._u32[w]; - CPU.VPR[op.vd]._u32[w] = result; - } + const auto a = CPU.VPR[op.va].vi; + const auto b = CPU.VPR[op.vb].vi; + const auto ml = _mm_mullo_epi16(a, b); // low results + const auto mh = _mm_mulhi_epu16(a, b); // high results + const auto ls = _mm_add_epi32(_mm_srli_epi32(ml, 16), _mm_and_si128(ml, _mm_set1_epi32(0x0000ffff))); + const auto hs = _mm_add_epi32(_mm_slli_epi32(mh, 16), _mm_and_si128(mh, _mm_set1_epi32(0xffff0000))); + CPU.VPR[op.vd].vi = _mm_add_epi32(_mm_add_epi32(CPU.VPR[op.vc].vi, ls), hs); } void ppu_interpreter::VMSUMUHS(PPUThread& CPU, ppu_opcode_t op) @@ -700,66 +643,52 @@ void ppu_interpreter::VMSUMUHS(PPUThread& CPU, ppu_opcode_t op) void ppu_interpreter::VMULESB(PPUThread& CPU, ppu_opcode_t op) { - for (uint h = 0; h < 8; h++) - { - CPU.VPR[op.vd]._s16[h] = (s16)CPU.VPR[op.va]._s8[h * 2 + 1] * (s16)CPU.VPR[op.vb]._s8[h * 2 + 1]; - } + CPU.VPR[op.vd].vi = _mm_mullo_epi16(_mm_srai_epi16(CPU.VPR[op.va].vi, 8), _mm_srai_epi16(CPU.VPR[op.vb].vi, 8)); } void ppu_interpreter::VMULESH(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - CPU.VPR[op.vd]._s32[w] = (s32)CPU.VPR[op.va]._s16[w * 2 + 1] * (s32)CPU.VPR[op.vb]._s16[w * 2 + 1]; - } + CPU.VPR[op.vd].vi = _mm_madd_epi16(_mm_srli_epi16(CPU.VPR[op.va].vi, 16), _mm_srli_epi16(CPU.VPR[op.vb].vi, 16)); } void ppu_interpreter::VMULEUB(PPUThread& CPU, ppu_opcode_t op) { - for (uint h = 0; h < 8; h++) - { - CPU.VPR[op.vd]._u16[h] = (u16)CPU.VPR[op.va]._u8[h * 2 + 1] * (u16)CPU.VPR[op.vb]._u8[h * 2 + 1]; - } + CPU.VPR[op.vd].vi = _mm_mullo_epi16(_mm_srli_epi16(CPU.VPR[op.va].vi, 8), _mm_srli_epi16(CPU.VPR[op.vb].vi, 8)); } void ppu_interpreter::VMULEUH(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - CPU.VPR[op.vd]._u32[w] = (u32)CPU.VPR[op.va]._u16[w * 2 + 1] * (u32)CPU.VPR[op.vb]._u16[w * 2 + 1]; - } + const auto a = CPU.VPR[op.va].vi; + const auto b = CPU.VPR[op.vb].vi; + const auto ml = _mm_mullo_epi16(a, b); + const auto mh = _mm_mulhi_epu16(a, b); + CPU.VPR[op.vd].vi = _mm_or_si128(_mm_srli_epi32(ml, 16), _mm_and_si128(mh, _mm_set1_epi32(0xffff0000))); } void ppu_interpreter::VMULOSB(PPUThread& CPU, ppu_opcode_t op) { - for (uint h = 0; h < 8; h++) - { - CPU.VPR[op.vd]._s16[h] = (s16)CPU.VPR[op.va]._s8[h * 2] * (s16)CPU.VPR[op.vb]._s8[h * 2]; - } + CPU.VPR[op.vd].vi = _mm_mullo_epi16(_mm_srai_epi16(_mm_slli_epi16(CPU.VPR[op.va].vi, 8), 8), _mm_srai_epi16(_mm_slli_epi16(CPU.VPR[op.vb].vi, 8), 8)); } void ppu_interpreter::VMULOSH(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - CPU.VPR[op.vd]._s32[w] = (s32)CPU.VPR[op.va]._s16[w * 2] * (s32)CPU.VPR[op.vb]._s16[w * 2]; - } + const auto mask = _mm_set1_epi32(0x0000ffff); + CPU.VPR[op.vd].vi = _mm_madd_epi16(_mm_and_si128(CPU.VPR[op.va].vi, mask), _mm_and_si128(CPU.VPR[op.vb].vi, mask)); } void ppu_interpreter::VMULOUB(PPUThread& CPU, ppu_opcode_t op) { - for (uint h = 0; h < 8; h++) - { - CPU.VPR[op.vd]._u16[h] = (u16)CPU.VPR[op.va]._u8[h * 2] * (u16)CPU.VPR[op.vb]._u8[h * 2]; - } + const auto mask = _mm_set1_epi16(0x00ff); + CPU.VPR[op.vd].vi = _mm_mullo_epi16(_mm_and_si128(CPU.VPR[op.va].vi, mask), _mm_and_si128(CPU.VPR[op.vb].vi, mask)); } void ppu_interpreter::VMULOUH(PPUThread& CPU, ppu_opcode_t op) { - for (uint w = 0; w < 4; w++) - { - CPU.VPR[op.vd]._u32[w] = (u32)CPU.VPR[op.va]._u16[w * 2] * (u32)CPU.VPR[op.vb]._u16[w * 2]; - } + const auto a = CPU.VPR[op.va].vi; + const auto b = CPU.VPR[op.vb].vi; + const auto ml = _mm_mullo_epi16(a, b); + const auto mh = _mm_mulhi_epu16(a, b); + CPU.VPR[op.vd].vi = _mm_or_si128(_mm_slli_epi32(mh, 16), _mm_and_si128(ml, _mm_set1_epi32(0x0000ffff))); } void ppu_interpreter::VNMSUBFP(PPUThread& CPU, ppu_opcode_t op) @@ -1305,40 +1234,12 @@ void ppu_interpreter::VSUBFP(PPUThread& CPU, ppu_opcode_t op) void ppu_interpreter::VSUBSBS(PPUThread& CPU, ppu_opcode_t op) { - for (uint b = 0; b < 16; b++) - { - s16 result = (s16)CPU.VPR[op.va]._s8[b] - (s16)CPU.VPR[op.vb]._s8[b]; - - if (result < INT8_MIN) - { - CPU.VPR[op.vd]._s8[b] = INT8_MIN; - } - else if (result > INT8_MAX) - { - CPU.VPR[op.vd]._s8[b] = INT8_MAX; - } - else - CPU.VPR[op.vd]._s8[b] = (s8)result; - } + CPU.VPR[op.vd].vi = _mm_subs_epi8(CPU.VPR[op.va].vi, CPU.VPR[op.vb].vi); } void ppu_interpreter::VSUBSHS(PPUThread& CPU, ppu_opcode_t op) { - for (uint h = 0; h < 8; h++) - { - s32 result = (s32)CPU.VPR[op.va]._s16[h] - (s32)CPU.VPR[op.vb]._s16[h]; - - if (result < INT16_MIN) - { - CPU.VPR[op.vd]._s16[h] = (s16)INT16_MIN; - } - else if (result > INT16_MAX) - { - CPU.VPR[op.vd]._s16[h] = (s16)INT16_MAX; - } - else - CPU.VPR[op.vd]._s16[h] = (s16)result; - } + CPU.VPR[op.vd].vi = _mm_subs_epi16(CPU.VPR[op.va].vi, CPU.VPR[op.vb].vi); } void ppu_interpreter::VSUBSWS(PPUThread& CPU, ppu_opcode_t op) @@ -1367,17 +1268,7 @@ void ppu_interpreter::VSUBUBM(PPUThread& CPU, ppu_opcode_t op) void ppu_interpreter::VSUBUBS(PPUThread& CPU, ppu_opcode_t op) { - for (uint b = 0; b < 16; b++) - { - s16 result = (s16)CPU.VPR[op.va]._u8[b] - (s16)CPU.VPR[op.vb]._u8[b]; - - if (result < 0) - { - CPU.VPR[op.vd]._u8[b] = 0; - } - else - CPU.VPR[op.vd]._u8[b] = (u8)result; - } + CPU.VPR[op.vd].vi = _mm_subs_epu8(CPU.VPR[op.va].vi, CPU.VPR[op.vb].vi); } void ppu_interpreter::VSUBUHM(PPUThread& CPU, ppu_opcode_t op) @@ -1387,17 +1278,7 @@ void ppu_interpreter::VSUBUHM(PPUThread& CPU, ppu_opcode_t op) void ppu_interpreter::VSUBUHS(PPUThread& CPU, ppu_opcode_t op) { - for (uint h = 0; h < 8; h++) - { - s32 result = (s32)CPU.VPR[op.va]._u16[h] - (s32)CPU.VPR[op.vb]._u16[h]; - - if (result < 0) - { - CPU.VPR[op.vd]._u16[h] = 0; - } - else - CPU.VPR[op.vd]._u16[h] = (u16)result; - } + CPU.VPR[op.vd].vi = _mm_subs_epu16(CPU.VPR[op.va].vi, CPU.VPR[op.vb].vi); } void ppu_interpreter::VSUBUWM(PPUThread& CPU, ppu_opcode_t op)