mirror of
https://github.com/hedge-dev/XenonRecomp.git
synced 2025-11-04 06:47:09 +00:00
Some vector load & store instructions.
This commit is contained in:
Skyth
@@ -138,6 +138,7 @@ int main()
|
|||||||
println("PPC_FUNC void sub_{:X}(PPCContext& __restrict ctx, uint8_t* base) {{", fn.base);
|
println("PPC_FUNC void sub_{:X}(PPCContext& __restrict ctx, uint8_t* base) {{", fn.base);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
println("\t__assume((reinterpret_cast<size_t>(base) & 0xFFFFFFFF) == 0);");
|
||||||
println("\tPPCRegister temp;");
|
println("\tPPCRegister temp;");
|
||||||
println("\tuint32_t ea;\n");
|
println("\tuint32_t ea;\n");
|
||||||
|
|
||||||
@@ -714,19 +715,46 @@ int main()
|
|||||||
case PPC_INST_LVEWX128:
|
case PPC_INST_LVEWX128:
|
||||||
case PPC_INST_LVX:
|
case PPC_INST_LVX:
|
||||||
case PPC_INST_LVX128:
|
case PPC_INST_LVX128:
|
||||||
// TODO: endian swap
|
// NOTE: for endian swapping, we reverse the whole vector instead of individual elements.
|
||||||
print("\t_mm_store_ps(ctx.v{}.f32, _mm_load_ps(reinterpret_cast<float*>(base + ", insn.operands[0]);
|
// this is accounted for in every instruction (eg. dp3 sums yzw instead of xyz)
|
||||||
|
print("\t_mm_store_si128((__m128i*)ctx.v{}.u8, _mm_shuffle_epi8(_mm_load_si128((__m128i*)(base + ((", insn.operands[0]);
|
||||||
if (insn.operands[1] != 0)
|
if (insn.operands[1] != 0)
|
||||||
print("ctx.r{}.u32 + ", insn.operands[1]);
|
print("ctx.r{}.u32 + ", insn.operands[1]);
|
||||||
println("ctx.r{}.u32)));", insn.operands[2]);
|
println("ctx.r{}.u32) & ~0xF))), _mm_load_si128((__m128i*)VectorMaskL)));", insn.operands[2]);
|
||||||
break;
|
break;
|
||||||
|
|
||||||
case PPC_INST_LVLX:
|
case PPC_INST_LVLX:
|
||||||
case PPC_INST_LVLX128:
|
case PPC_INST_LVLX128:
|
||||||
|
print("\ttemp.u32 = ");
|
||||||
|
if (insn.operands[1] != 0)
|
||||||
|
print("ctx.r{}.u32 + ", insn.operands[1]);
|
||||||
|
println("ctx.r{}.u32;", insn.operands[2]);
|
||||||
|
println("\t_mm_store_si128((__m128i*)ctx.v{}.u8, _mm_shuffle_epi8(_mm_load_si128((__m128i*)(base + (temp.u32 & ~0xF))), _mm_load_si128((__m128i*)&VectorMaskL[(temp.u32 & 0xF) * 16])));", insn.operands[0]);
|
||||||
|
break;
|
||||||
|
|
||||||
case PPC_INST_LVRX:
|
case PPC_INST_LVRX:
|
||||||
case PPC_INST_LVRX128:
|
case PPC_INST_LVRX128:
|
||||||
|
print("\ttemp.u32 = ");
|
||||||
|
if (insn.operands[1] != 0)
|
||||||
|
print("ctx.r{}.u32 + ", insn.operands[1]);
|
||||||
|
println("ctx.r{}.u32;", insn.operands[2]);
|
||||||
|
println("\t_mm_store_si128((__m128i*)ctx.v{}.u8, temp.u32 & 0xF ? _mm_shuffle_epi8(_mm_load_si128((__m128i*)(base + (temp.u32 & ~0xF))), _mm_load_si128((__m128i*)&VectorMaskR[(temp.u32 & 0xF) * 16])) : _mm_setzero_si128());", insn.operands[0]);
|
||||||
|
break;
|
||||||
|
|
||||||
case PPC_INST_LVSL:
|
case PPC_INST_LVSL:
|
||||||
|
print("\ttemp.u32 = ");
|
||||||
|
if (insn.operands[1] != 0)
|
||||||
|
print("ctx.r{}.u32 + ", insn.operands[1]);
|
||||||
|
println("ctx.r{}.u32;", insn.operands[2]);
|
||||||
|
println("\t_mm_store_si128((__m128i*)ctx.v{}.u8, _mm_load_si128((__m128i*)&VectorShiftTableL[(temp.u32 & 0xF) * 16]));", insn.operands[0]);
|
||||||
|
break;
|
||||||
|
|
||||||
case PPC_INST_LVSR:
|
case PPC_INST_LVSR:
|
||||||
|
print("\ttemp.u32 = ");
|
||||||
|
if (insn.operands[1] != 0)
|
||||||
|
print("ctx.r{}.u32 + ", insn.operands[1]);
|
||||||
|
println("ctx.r{}.u32;", insn.operands[2]);
|
||||||
|
println("\t_mm_store_si128((__m128i*)ctx.v{}.u8, _mm_load_si128((__m128i*)&VectorShiftTableR[(temp.u32 & 0xF) * 16]));", insn.operands[0]);
|
||||||
break;
|
break;
|
||||||
|
|
||||||
case PPC_INST_LWA:
|
case PPC_INST_LWA:
|
||||||
@@ -1115,8 +1143,14 @@ int main()
|
|||||||
case PPC_INST_STVLX128:
|
case PPC_INST_STVLX128:
|
||||||
case PPC_INST_STVRX:
|
case PPC_INST_STVRX:
|
||||||
case PPC_INST_STVRX128:
|
case PPC_INST_STVRX128:
|
||||||
|
break;
|
||||||
|
|
||||||
case PPC_INST_STVX:
|
case PPC_INST_STVX:
|
||||||
case PPC_INST_STVX128:
|
case PPC_INST_STVX128:
|
||||||
|
print("\t_mm_store_si128((__m128i*)(base + ((");
|
||||||
|
if (insn.operands[1] != 0)
|
||||||
|
print("ctx.r{}.u32 + ", insn.operands[1]);
|
||||||
|
println("ctx.r{}.u32) & ~0xF)), _mm_shuffle_epi8(_mm_load_si128((__m128i*)ctx.v{}.u8), _mm_load_si128((__m128i*)VectorMaskL)));", insn.operands[2], insn.operands[0]);
|
||||||
break;
|
break;
|
||||||
|
|
||||||
case PPC_INST_STW:
|
case PPC_INST_STW:
|
||||||
|
|||||||
@@ -11,6 +11,7 @@
|
|||||||
#define _byteswap_ulong __builtin_bswap32
|
#define _byteswap_ulong __builtin_bswap32
|
||||||
#define _byteswap_uint64 __builtin_bswap64
|
#define _byteswap_uint64 __builtin_bswap64
|
||||||
#define isnan __builtin_isnan
|
#define isnan __builtin_isnan
|
||||||
|
#define __assume __builtin_assume
|
||||||
#define PPC_FUNC __attribute__((weak,noinline))
|
#define PPC_FUNC __attribute__((weak,noinline))
|
||||||
#else
|
#else
|
||||||
#include <intrin.h>
|
#include <intrin.h>
|
||||||
@@ -342,6 +343,86 @@ struct PPCContext
|
|||||||
};
|
};
|
||||||
};
|
};
|
||||||
|
|
||||||
|
inline uint8_t VectorMaskL[] =
|
||||||
|
{
|
||||||
|
0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00,
|
||||||
|
0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01,
|
||||||
|
0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02,
|
||||||
|
0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D, 0x0C,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E, 0x0D,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F, 0x0E,
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0F,
|
||||||
|
};
|
||||||
|
|
||||||
|
inline uint8_t VectorMaskR[] =
|
||||||
|
{
|
||||||
|
0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF, 0xFF,
|
||||||
|
0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF, 0xFF,
|
||||||
|
0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00, 0xFF,
|
||||||
|
};
|
||||||
|
|
||||||
|
inline uint8_t VectorShiftTableL[] =
|
||||||
|
{
|
||||||
|
0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01, 0x00,
|
||||||
|
0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01,
|
||||||
|
0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02,
|
||||||
|
0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03,
|
||||||
|
0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04,
|
||||||
|
0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05,
|
||||||
|
0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06,
|
||||||
|
0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07,
|
||||||
|
0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08,
|
||||||
|
0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09,
|
||||||
|
0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A,
|
||||||
|
0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B,
|
||||||
|
0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C,
|
||||||
|
0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D,
|
||||||
|
0x1D, 0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E,
|
||||||
|
0x1E, 0x1D, 0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F,
|
||||||
|
};
|
||||||
|
|
||||||
|
inline uint8_t VectorShiftTableR[] =
|
||||||
|
{
|
||||||
|
0x1F, 0x1E, 0x1D, 0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10,
|
||||||
|
0x1E, 0x1D, 0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F,
|
||||||
|
0x1D, 0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E,
|
||||||
|
0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D,
|
||||||
|
0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C,
|
||||||
|
0x1A, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B,
|
||||||
|
0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A,
|
||||||
|
0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09,
|
||||||
|
0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08,
|
||||||
|
0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07,
|
||||||
|
0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06,
|
||||||
|
0x14, 0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05,
|
||||||
|
0x13, 0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04,
|
||||||
|
0x12, 0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03,
|
||||||
|
0x11, 0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02,
|
||||||
|
0x10, 0x0F, 0x0E, 0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04, 0x03, 0x02, 0x01,
|
||||||
|
};
|
||||||
|
|
||||||
inline __m128i _mm_adds_epu32(__m128i a, __m128i b)
|
inline __m128i _mm_adds_epu32(__m128i a, __m128i b)
|
||||||
{
|
{
|
||||||
return _mm_add_epi32(_mm_min_epu32(a, _mm_xor_si128(b, _mm_cmpeq_epi32(b, b))), b);
|
return _mm_add_epi32(_mm_min_epu32(a, _mm_xor_si128(b, _mm_cmpeq_epi32(b, b))), b);
|
||||||
|
|||||||
Reference in New Issue
Block a user