#include "mini.h"
#include "ir-emit.h"
+#include "mono/utils/bsearch.h"
+#include <mono/metadata/abi-details.h>
/*
General notes on SIMD intrinsics
TODO add stuff to man pages
TODO document this under /docs
TODO make passing a xmm as argument not cause it to be LDADDR'ed (introduce an OP_XPUSH)
-TODO revamp the .ctor sequence as it looks very fragile, maybe use a var just like iconv_to_r8_raw. (or just pinst sse ops)
+TODO revamp the .ctor sequence as it looks very fragile, maybe use a var just like move_i4_to_f. (or just pinst sse ops)
TODO figure out what's wrong with OP_STOREX_MEMBASE_REG and OP_STOREX_MEMBASE (the 2nd is for imm operands)
TODO maybe add SSE3 emulation on top of SSE2, or just implement the corresponding functions using SSE2 intrinsics.
TODO pass simd arguments in registers or, at least, add SSE support for pushing large (>=16) valuetypes
without a OP_LDADDR.
*/
-#ifdef MONO_ARCH_SIMD_INTRINSICS
+#if defined (MONO_ARCH_SIMD_INTRINSICS)
+
+#if defined (DISABLE_JIT)
+
+void
+mono_simd_intrinsics_init (void)
+{
+}
+
+#else
//#define IS_DEBUG_ON(cfg) (0)
typedef struct {
guint16 name;
guint16 opcode;
+ guint8 simd_version_flags;
guint8 simd_emit_mode : 4;
- guint8 simd_version : 4;
- guint8 flags;
+ guint8 flags : 4;
} SimdIntrinsc;
static const SimdIntrinsc vector4f_intrinsics[] = {
- { SN_ctor, OP_EXPAND_R4, SIMD_EMIT_CTOR },
- { SN_AddSub, OP_ADDSUBPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE3 },
- { SN_AndNot, OP_ANDNPS, SIMD_EMIT_BINARY },
- { SN_CompareEqual, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_CompareLessEqual, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_LE },
- { SN_CompareLessThan, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_LT },
- { SN_CompareNotEqual, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_CompareNotLessEqual, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_NLE },
- { SN_CompareNotLessThan, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_NLT },
- { SN_CompareOrdered, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_ORD },
- { SN_CompareUnordered, OP_COMPPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_UNORD },
- { SN_DuplicateHigh, OP_DUPPS_HIGH, SIMD_EMIT_UNARY, SIMD_VERSION_SSE3 },
- { SN_DuplicateLow, OP_DUPPS_LOW, SIMD_EMIT_UNARY, SIMD_VERSION_SSE3 },
- { SN_HorizontalAdd, OP_HADDPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE3 },
- { SN_HorizontalSub, OP_HSUBPS, SIMD_EMIT_BINARY, SIMD_VERSION_SSE3 },
- { SN_InterleaveHigh, OP_UNPACK_HIGHPS, SIMD_EMIT_BINARY },
- { SN_InterleaveLow, OP_UNPACK_LOWPS, SIMD_EMIT_BINARY },
- { SN_InvSqrt, OP_RSQRTPS, SIMD_EMIT_UNARY },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_Max, OP_MAXPS, SIMD_EMIT_BINARY },
- { SN_Min, OP_MINPS, SIMD_EMIT_BINARY },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_Reciprocal, OP_RCPPS, SIMD_EMIT_UNARY },
- { SN_Shuffle, OP_PSHUFLED, SIMD_EMIT_SHUFFLE },
- { SN_Sqrt, OP_SQRTPS, SIMD_EMIT_UNARY },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_StoreNonTemporal, OP_STOREX_NTA_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_get_W, 3, SIMD_EMIT_GETTER },
- { SN_get_X, 0, SIMD_EMIT_GETTER },
- { SN_get_Y, 1, SIMD_EMIT_GETTER },
- { SN_get_Z, 2, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_ADDPS, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_ANDPS, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_ORPS, SIMD_EMIT_BINARY },
- { SN_op_Division, OP_DIVPS, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_COMPPS, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_XORPS, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_COMPPS, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_Multiply, OP_MULPS, SIMD_EMIT_BINARY },
- { SN_op_Subtraction, OP_SUBPS, SIMD_EMIT_BINARY },
- { SN_set_W, 3, SIMD_EMIT_SETTER },
- { SN_set_X, 0, SIMD_EMIT_SETTER },
- { SN_set_Y, 1, SIMD_EMIT_SETTER },
- { SN_set_Z, 2, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_R4, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_AddSub, OP_ADDSUBPS, SIMD_VERSION_SSE3, SIMD_EMIT_BINARY},
+ { SN_AndNot, OP_ANDNPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY},
+ { SN_CompareEqual, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_EQ },
+ { SN_CompareLessEqual, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_LE },
+ { SN_CompareLessThan, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_LT },
+ { SN_CompareNotEqual, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_NEQ },
+ { SN_CompareNotLessEqual, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_NLE },
+ { SN_CompareNotLessThan, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_NLT },
+ { SN_CompareOrdered, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_ORD },
+ { SN_CompareUnordered, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_UNORD },
+ { SN_ConvertToDouble, OP_CVTPS2PD, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_ConvertToInt, OP_CVTPS2DQ, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_ConvertToIntTruncated, OP_CVTTPS2DQ, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_DuplicateHigh, OP_DUPPS_HIGH, SIMD_VERSION_SSE3, SIMD_EMIT_UNARY },
+ { SN_DuplicateLow, OP_DUPPS_LOW, SIMD_VERSION_SSE3, SIMD_EMIT_UNARY },
+ { SN_HorizontalAdd, OP_HADDPS, SIMD_VERSION_SSE3, SIMD_EMIT_BINARY },
+ { SN_HorizontalSub, OP_HSUBPS, SIMD_VERSION_SSE3, SIMD_EMIT_BINARY },
+ { SN_InterleaveHigh, OP_UNPACK_HIGHPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_InterleaveLow, OP_UNPACK_LOWPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_InvSqrt, OP_RSQRTPS, SIMD_VERSION_SSE1, SIMD_EMIT_UNARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_Max, OP_MAXPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_Min, OP_MINPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_Reciprocal, OP_RCPPS, SIMD_VERSION_SSE1, SIMD_EMIT_UNARY },
+ { SN_Shuffle, OP_PSHUFLED, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_Sqrt, OP_SQRTPS, SIMD_VERSION_SSE1, SIMD_EMIT_UNARY },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_StoreNonTemporal, OP_STOREX_NTA_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_get_W, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_Z, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_ADDPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_ANDPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_ORPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Division, OP_DIVPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_XORPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_COMPPS, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_Multiply, OP_MULPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Subtraction, OP_SUBPS, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_W, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Z, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER }
};
static const SimdIntrinsc vector2d_intrinsics[] = {
- { SN_ctor, OP_EXPAND_R8, SIMD_EMIT_CTOR },
- { SN_AddSub, OP_ADDSUBPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE3 },
- { SN_AndNot, OP_ANDNPD, SIMD_EMIT_BINARY },
- { SN_CompareEqual, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_CompareLessEqual, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_LE },
- { SN_CompareLessThan, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_LT },
- { SN_CompareNotEqual, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_CompareNotLessEqual, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_NLE },
- { SN_CompareNotLessThan, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_NLT },
- { SN_CompareOrdered, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_ORD },
- { SN_CompareUnordered, OP_COMPPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1, SIMD_COMP_UNORD },
- { SN_Duplicate, OP_DUPPD, SIMD_EMIT_UNARY, SIMD_VERSION_SSE3 },
- { SN_HorizontalAdd, OP_HADDPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE3 },
- { SN_HorizontalSub, OP_HSUBPD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE3 },
- { SN_InterleaveHigh, OP_UNPACK_HIGHPD, SIMD_EMIT_BINARY },
- { SN_InterleaveLow, OP_UNPACK_LOWPD, SIMD_EMIT_BINARY },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_Max, OP_MAXPD, SIMD_EMIT_BINARY },
- { SN_Min, OP_MINPD, SIMD_EMIT_BINARY },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_get_X, 0, SIMD_EMIT_GETTER_QWORD },
- { SN_get_Y, 1, SIMD_EMIT_GETTER_QWORD },
- { SN_op_Addition, OP_ADDPD, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_ANDPD, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_ORPD, SIMD_EMIT_BINARY },
- { SN_op_Division, OP_DIVPD, SIMD_EMIT_BINARY },
- { SN_op_ExclusiveOr, OP_XORPD, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Multiply, OP_MULPD, SIMD_EMIT_BINARY },
- { SN_op_Subtraction, OP_SUBPD, SIMD_EMIT_BINARY },
- { SN_set_X, 0, SIMD_EMIT_SETTER },
- { SN_set_Y, 1, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_R8, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_AddSub, OP_ADDSUBPD, SIMD_VERSION_SSE3, SIMD_EMIT_BINARY,},
+ { SN_AndNot, OP_ANDNPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareEqual, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_EQ },
+ { SN_CompareLessEqual, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_LE },
+ { SN_CompareLessThan, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_LT },
+ { SN_CompareNotEqual, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_NEQ },
+ { SN_CompareNotLessEqual, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_NLE },
+ { SN_CompareNotLessThan, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_NLT },
+ { SN_CompareOrdered, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_ORD },
+ { SN_CompareUnordered, OP_COMPPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_COMP_UNORD },
+ { SN_ConvertToFloat, OP_CVTPD2PS, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_ConvertToInt, OP_CVTPD2DQ, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_ConvertToIntTruncated, OP_CVTTPD2DQ, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_Duplicate, OP_DUPPD, SIMD_VERSION_SSE3, SIMD_EMIT_UNARY },
+ { SN_HorizontalAdd, OP_HADDPD, SIMD_VERSION_SSE3, SIMD_EMIT_BINARY },
+ { SN_HorizontalSub, OP_HSUBPD, SIMD_VERSION_SSE3, SIMD_EMIT_BINARY },
+ { SN_InterleaveHigh, OP_UNPACK_HIGHPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_InterleaveLow, OP_UNPACK_LOWPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_Max, OP_MAXPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_Min, OP_MINPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_Shuffle, OP_SHUFPD, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_Sqrt, OP_SQRTPD, SIMD_VERSION_SSE1, SIMD_EMIT_UNARY },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_get_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER_QWORD },
+ { SN_get_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER_QWORD },
+ { SN_op_Addition, OP_ADDPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_ANDPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_ORPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Division, OP_DIVPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_ExclusiveOr, OP_XORPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Multiply, OP_MULPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Subtraction, OP_SUBPD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector2ul_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I8, SIMD_EMIT_CTOR },
- { SN_CompareEqual, OP_PCMPEQQ, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_UnpackHigh, OP_UNPACK_HIGHQ, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWQ, SIMD_EMIT_BINARY },
- { SN_get_X, 0, SIMD_EMIT_GETTER_QWORD },
- { SN_get_Y, 1, SIMD_EMIT_GETTER_QWORD },
- { SN_op_Addition, OP_PADDQ, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_LeftShift, OP_PSHLQ, SIMD_EMIT_SHIFT },
- { SN_op_Multiply, OP_PMULQ, SIMD_EMIT_BINARY },
- { SN_op_RightShift, OP_PSHRQ, SIMD_EMIT_SHIFT },
- { SN_op_Subtraction, OP_PSUBQ, SIMD_EMIT_BINARY },
- { SN_set_X, 0, SIMD_EMIT_SETTER },
- { SN_set_Y, 1, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I8, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_CompareEqual, OP_PCMPEQQ, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_Shuffle, OP_SHUFPD, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_UnpackHigh, OP_UNPACK_HIGHQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER_QWORD },
+ { SN_get_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER_QWORD },
+ { SN_op_Addition, OP_PADDQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1 },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_LeftShift, OP_PSHLQ, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Multiply, OP_PMULQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_RightShift, OP_PSHRQ, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Subtraction, OP_PSUBQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector2l_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I8, SIMD_EMIT_CTOR },
- { SN_CompareEqual, OP_PCMPEQQ, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_CompareGreaterThan, OP_PCMPGTQ, SIMD_EMIT_BINARY, SIMD_VERSION_SSE42 },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_LogicalRightShift, OP_PSHRQ, SIMD_EMIT_SHIFT },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_UnpackHigh, OP_UNPACK_HIGHQ, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWQ, SIMD_EMIT_BINARY },
- { SN_get_X, 0, SIMD_EMIT_GETTER_QWORD },
- { SN_get_Y, 1, SIMD_EMIT_GETTER_QWORD },
- { SN_op_Addition, OP_PADDQ, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_LeftShift, OP_PSHLQ, SIMD_EMIT_SHIFT },
- { SN_op_Multiply, OP_PMULQ, SIMD_EMIT_BINARY },
- { SN_op_Subtraction, OP_PSUBQ, SIMD_EMIT_BINARY },
- { SN_set_X, 0, SIMD_EMIT_SETTER },
- { SN_set_Y, 1, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I8, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_CompareEqual, OP_PCMPEQQ, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_CompareGreaterThan, OP_PCMPGTQ, SIMD_VERSION_SSE42, SIMD_EMIT_BINARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_LogicalRightShift, OP_PSHRQ, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_Shuffle, OP_SHUFPD, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_UnpackHigh, OP_UNPACK_HIGHQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER_QWORD },
+ { SN_get_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER_QWORD },
+ { SN_op_Addition, OP_PADDQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_LeftShift, OP_PSHLQ, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Multiply, OP_PMULQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Subtraction, OP_PSUBQ, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector4ui_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I4, SIMD_EMIT_CTOR },
- { SN_ArithmeticRightShift, OP_PSARD, SIMD_EMIT_SHIFT },
- { SN_CompareEqual, OP_PCMPEQD, SIMD_EMIT_BINARY },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_Max, OP_PMAXD_UN, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_Min, OP_PMIND_UN, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_Shuffle, OP_PSHUFLED, SIMD_EMIT_SHUFFLE },
- { SN_SignedPackWithSignedSaturation, OP_PACKD, SIMD_EMIT_BINARY },
- { SN_SignedPackWithUnsignedSaturation, OP_PACKD_UN, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_UnpackHigh, OP_UNPACK_HIGHD, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWD, SIMD_EMIT_BINARY },
- { SN_get_W, 3, SIMD_EMIT_GETTER },
- { SN_get_X, 0, SIMD_EMIT_GETTER },
- { SN_get_Y, 1, SIMD_EMIT_GETTER },
- { SN_get_Z, 2, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_PADDD, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_PCMPEQD, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_PCMPEQD, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_LeftShift, OP_PSHLD, SIMD_EMIT_SHIFT },
- { SN_op_Multiply, OP_PMULD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_op_RightShift, OP_PSHRD, SIMD_EMIT_SHIFT },
- { SN_op_Subtraction, OP_PSUBD, SIMD_EMIT_BINARY },
- { SN_set_W, 3, SIMD_EMIT_SETTER },
- { SN_set_X, 0, SIMD_EMIT_SETTER },
- { SN_set_Y, 1, SIMD_EMIT_SETTER },
- { SN_set_Z, 2, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I4, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_ArithmeticRightShift, OP_PSARD, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_CompareEqual, OP_PCMPEQD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_Max, OP_PMAXD_UN, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_Min, OP_PMIND_UN, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_Shuffle, OP_PSHUFLED, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_SignedPackWithSignedSaturation, OP_PACKD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_SignedPackWithUnsignedSaturation, OP_PACKD_UN, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_UnpackHigh, OP_UNPACK_HIGHD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_W, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_Z, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_PADDD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_PCMPEQD, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_PCMPEQD, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_LeftShift, OP_PSHLD, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Multiply, OP_PMULD, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_op_RightShift, OP_PSHRD, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Subtraction, OP_PSUBD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_W, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Z, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector4i_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I4, SIMD_EMIT_CTOR },
- { SN_CompareEqual, OP_PCMPEQD, SIMD_EMIT_BINARY },
- { SN_CompareGreaterThan, OP_PCMPGTD, SIMD_EMIT_BINARY },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_LogicalRightShift, OP_PSHRD, SIMD_EMIT_SHIFT },
- { SN_Max, OP_PMAXD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_Min, OP_PMIND, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_PackWithSignedSaturation, OP_PACKD, SIMD_EMIT_BINARY },
- { SN_PackWithUnsignedSaturation, OP_PACKD_UN, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_Shuffle, OP_PSHUFLED, SIMD_EMIT_SHUFFLE },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_UnpackHigh, OP_UNPACK_HIGHD, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWD, SIMD_EMIT_BINARY },
- { SN_get_W, 3, SIMD_EMIT_GETTER },
- { SN_get_X, 0, SIMD_EMIT_GETTER },
- { SN_get_Y, 1, SIMD_EMIT_GETTER },
- { SN_get_Z, 2, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_PADDD, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_PCMPEQD, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_PCMPEQD, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_LeftShift, OP_PSHLD, SIMD_EMIT_SHIFT },
- { SN_op_Multiply, OP_PMULD, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_op_RightShift, OP_PSARD, SIMD_EMIT_SHIFT },
- { SN_op_Subtraction, OP_PSUBD, SIMD_EMIT_BINARY },
- { SN_set_W, 3, SIMD_EMIT_SETTER },
- { SN_set_X, 0, SIMD_EMIT_SETTER },
- { SN_set_Y, 1, SIMD_EMIT_SETTER },
- { SN_set_Z, 2, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I4, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_CompareEqual, OP_PCMPEQD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareGreaterThan, OP_PCMPGTD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_ConvertToDouble, OP_CVTDQ2PD, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_ConvertToFloat, OP_CVTDQ2PS, SIMD_VERSION_SSE2, SIMD_EMIT_UNARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_LogicalRightShift, OP_PSHRD, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_Max, OP_PMAXD, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_Min, OP_PMIND, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_PackWithSignedSaturation, OP_PACKD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PackWithUnsignedSaturation, OP_PACKD_UN, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_Shuffle, OP_PSHUFLED, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_UnpackHigh, OP_UNPACK_HIGHD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_W, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_Z, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_PADDD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_PCMPEQD, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_PCMPEQD, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_LeftShift, OP_PSHLD, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Multiply, OP_PMULD, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_op_RightShift, OP_PSARD, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Subtraction, OP_PSUBD, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_W, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_X, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Y, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_Z, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector8us_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I2, SIMD_EMIT_CTOR },
- { SN_AddWithSaturation, OP_PADDW_SAT_UN, SIMD_EMIT_BINARY },
- { SN_ArithmeticRightShift, OP_PSARW, SIMD_EMIT_SHIFT },
- { SN_Average, OP_PAVGW_UN, SIMD_EMIT_BINARY },
- { SN_CompareEqual, OP_PCMPEQW, SIMD_EMIT_BINARY },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_Max, OP_PMAXW_UN, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_Min, OP_PMINW_UN, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_MultiplyStoreHigh, OP_PMULW_HIGH_UN, SIMD_EMIT_BINARY },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_ShuffleHigh, OP_PSHUFLEW_HIGH, SIMD_EMIT_SHUFFLE },
- { SN_ShuffleLow, OP_PSHUFLEW_LOW, SIMD_EMIT_SHUFFLE },
- { SN_SignedPackWithSignedSaturation, OP_PACKW, SIMD_EMIT_BINARY },
- { SN_SignedPackWithUnsignedSaturation, OP_PACKW_UN, SIMD_EMIT_BINARY },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_SubtractWithSaturation, OP_PSUBW_SAT_UN, SIMD_EMIT_BINARY },
- { SN_UnpackHigh, OP_UNPACK_HIGHW, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWW, SIMD_EMIT_BINARY },
- { SN_get_V0, 0, SIMD_EMIT_GETTER },
- { SN_get_V1, 1, SIMD_EMIT_GETTER },
- { SN_get_V2, 2, SIMD_EMIT_GETTER },
- { SN_get_V3, 3, SIMD_EMIT_GETTER },
- { SN_get_V4, 4, SIMD_EMIT_GETTER },
- { SN_get_V5, 5, SIMD_EMIT_GETTER },
- { SN_get_V6, 6, SIMD_EMIT_GETTER },
- { SN_get_V7, 7, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_PADDW, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_PCMPEQW, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_PCMPEQW, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_LeftShift, OP_PSHLW, SIMD_EMIT_SHIFT },
- { SN_op_Multiply, OP_PMULW, SIMD_EMIT_BINARY },
- { SN_op_RightShift, OP_PSHRW, SIMD_EMIT_SHIFT },
- { SN_op_Subtraction, OP_PSUBW, SIMD_EMIT_BINARY },
- { SN_set_V0, 0, SIMD_EMIT_SETTER },
- { SN_set_V1, 1, SIMD_EMIT_SETTER },
- { SN_set_V2, 2, SIMD_EMIT_SETTER },
- { SN_set_V3, 3, SIMD_EMIT_SETTER },
- { SN_set_V4, 4, SIMD_EMIT_SETTER },
- { SN_set_V5, 5, SIMD_EMIT_SETTER },
- { SN_set_V6, 6, SIMD_EMIT_SETTER },
- { SN_set_V7, 7, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I2, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_AddWithSaturation, OP_PADDW_SAT_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_ArithmeticRightShift, OP_PSARW, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_Average, OP_PAVGW_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareEqual, OP_PCMPEQW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY, SIMD_VERSION_SSE1 },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_Max, OP_PMAXW_UN, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_Min, OP_PMINW_UN, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_MultiplyStoreHigh, OP_PMULW_HIGH_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_ShuffleHigh, OP_PSHUFLEW_HIGH, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_ShuffleLow, OP_PSHUFLEW_LOW, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_SignedPackWithSignedSaturation, OP_PACKW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_SignedPackWithUnsignedSaturation, OP_PACKW_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_SubtractWithSaturation, OP_PSUBW_SAT_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackHigh, OP_UNPACK_HIGHW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_PADDW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_PCMPEQW, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_PCMPEQW, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_LeftShift, OP_PSHLW, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Multiply, OP_PMULW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_RightShift, OP_PSHRW, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Subtraction, OP_PSUBW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector8s_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I2, SIMD_EMIT_CTOR },
- { SN_AddWithSaturation, OP_PADDW_SAT, SIMD_EMIT_BINARY },
- { SN_CompareEqual, OP_PCMPEQW, SIMD_EMIT_BINARY },
- { SN_CompareGreaterThan, OP_PCMPGTW, SIMD_EMIT_BINARY },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_LogicalRightShift, OP_PSHRW, SIMD_EMIT_SHIFT },
- { SN_Max, OP_PMAXW, SIMD_EMIT_BINARY },
- { SN_Min, OP_PMINW, SIMD_EMIT_BINARY },
- { SN_MultiplyStoreHigh, OP_PMULW_HIGH, SIMD_EMIT_BINARY },
- { SN_PackWithSignedSaturation, OP_PACKW, SIMD_EMIT_BINARY },
- { SN_PackWithUnsignedSaturation, OP_PACKW_UN, SIMD_EMIT_BINARY },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_ShuffleHigh, OP_PSHUFLEW_HIGH, SIMD_EMIT_SHUFFLE },
- { SN_ShuffleLow, OP_PSHUFLEW_LOW, SIMD_EMIT_SHUFFLE },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_SubtractWithSaturation, OP_PSUBW_SAT_UN, SIMD_EMIT_BINARY },
- { SN_UnpackHigh, OP_UNPACK_HIGHW, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWW, SIMD_EMIT_BINARY },
- { SN_get_V0, 0, SIMD_EMIT_GETTER },
- { SN_get_V1, 1, SIMD_EMIT_GETTER },
- { SN_get_V2, 2, SIMD_EMIT_GETTER },
- { SN_get_V3, 3, SIMD_EMIT_GETTER },
- { SN_get_V4, 4, SIMD_EMIT_GETTER },
- { SN_get_V5, 5, SIMD_EMIT_GETTER },
- { SN_get_V6, 6, SIMD_EMIT_GETTER },
- { SN_get_V7, 7, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_PADDW, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_PCMPEQW, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_PCMPEQW, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_LeftShift, OP_PSHLW, SIMD_EMIT_SHIFT },
- { SN_op_Multiply, OP_PMULW, SIMD_EMIT_BINARY },
- { SN_op_RightShift, OP_PSARW, SIMD_EMIT_SHIFT },
- { SN_op_Subtraction, OP_PSUBW, SIMD_EMIT_BINARY },
- { SN_set_V0, 0, SIMD_EMIT_SETTER },
- { SN_set_V1, 1, SIMD_EMIT_SETTER },
- { SN_set_V2, 2, SIMD_EMIT_SETTER },
- { SN_set_V3, 3, SIMD_EMIT_SETTER },
- { SN_set_V4, 4, SIMD_EMIT_SETTER },
- { SN_set_V5, 5, SIMD_EMIT_SETTER },
- { SN_set_V6, 6, SIMD_EMIT_SETTER },
- { SN_set_V7, 7, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I2, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_AddWithSaturation, OP_PADDW_SAT, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareEqual, OP_PCMPEQW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareGreaterThan, OP_PCMPGTW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_LogicalRightShift, OP_PSHRW, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_Max, OP_PMAXW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_Min, OP_PMINW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_MultiplyStoreHigh, OP_PMULW_HIGH, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PackWithSignedSaturation, OP_PACKW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PackWithUnsignedSaturation, OP_PACKW_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_ShuffleHigh, OP_PSHUFLEW_HIGH, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_ShuffleLow, OP_PSHUFLEW_LOW, SIMD_VERSION_SSE1, SIMD_EMIT_SHUFFLE },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_SubtractWithSaturation, OP_PSUBW_SAT_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackHigh, OP_UNPACK_HIGHW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_PADDW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_PCMPEQW, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_PCMPEQW, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_LeftShift, OP_PSHLW, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Multiply, OP_PMULW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_RightShift, OP_PSARW, SIMD_VERSION_SSE1, SIMD_EMIT_SHIFT },
+ { SN_op_Subtraction, OP_PSUBW, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static const SimdIntrinsc vector16b_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I1, SIMD_EMIT_CTOR },
- { SN_AddWithSaturation, OP_PADDB_SAT_UN, SIMD_EMIT_BINARY },
- { SN_Average, OP_PAVGB_UN, SIMD_EMIT_BINARY },
- { SN_CompareEqual, OP_PCMPEQB, SIMD_EMIT_BINARY },
- { SN_ExtractByteMask, 0, SIMD_EMIT_EXTRACT_MASK },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_Max, OP_PMAXB_UN, SIMD_EMIT_BINARY },
- { SN_Min, OP_PMINB_UN, SIMD_EMIT_BINARY },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_SubtractWithSaturation, OP_PSUBB_SAT_UN, SIMD_EMIT_BINARY },
- { SN_SumOfAbsoluteDifferences, OP_PSUM_ABS_DIFF, SIMD_EMIT_BINARY },
- { SN_UnpackHigh, OP_UNPACK_HIGHB, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWB, SIMD_EMIT_BINARY },
- { SN_get_V0, 0, SIMD_EMIT_GETTER },
- { SN_get_V1, 1, SIMD_EMIT_GETTER },
- { SN_get_V10, 10, SIMD_EMIT_GETTER },
- { SN_get_V11, 11, SIMD_EMIT_GETTER },
- { SN_get_V12, 12, SIMD_EMIT_GETTER },
- { SN_get_V13, 13, SIMD_EMIT_GETTER },
- { SN_get_V14, 14, SIMD_EMIT_GETTER },
- { SN_get_V15, 15, SIMD_EMIT_GETTER },
- { SN_get_V2, 2, SIMD_EMIT_GETTER },
- { SN_get_V3, 3, SIMD_EMIT_GETTER },
- { SN_get_V4, 4, SIMD_EMIT_GETTER },
- { SN_get_V5, 5, SIMD_EMIT_GETTER },
- { SN_get_V6, 6, SIMD_EMIT_GETTER },
- { SN_get_V7, 7, SIMD_EMIT_GETTER },
- { SN_get_V8, 8, SIMD_EMIT_GETTER },
- { SN_get_V9, 9, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_PADDB, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_PCMPEQB, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_PCMPEQB, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_Subtraction, OP_PSUBB, SIMD_EMIT_BINARY },
- { SN_set_V0, 0, SIMD_EMIT_SETTER },
- { SN_set_V1, 1, SIMD_EMIT_SETTER },
- { SN_set_V10, 10, SIMD_EMIT_SETTER },
- { SN_set_V11, 11, SIMD_EMIT_SETTER },
- { SN_set_V12, 12, SIMD_EMIT_SETTER },
- { SN_set_V13, 13, SIMD_EMIT_SETTER },
- { SN_set_V14, 14, SIMD_EMIT_SETTER },
- { SN_set_V15, 15, SIMD_EMIT_SETTER },
- { SN_set_V2, 2, SIMD_EMIT_SETTER },
- { SN_set_V3, 3, SIMD_EMIT_SETTER },
- { SN_set_V4, 4, SIMD_EMIT_SETTER },
- { SN_set_V5, 5, SIMD_EMIT_SETTER },
- { SN_set_V6, 6, SIMD_EMIT_SETTER },
- { SN_set_V7, 7, SIMD_EMIT_SETTER },
- { SN_set_V8, 8, SIMD_EMIT_SETTER },
- { SN_set_V9, 9, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I1, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_AddWithSaturation, OP_PADDB_SAT_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_Average, OP_PAVGB_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareEqual, OP_PCMPEQB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_ExtractByteMask, 0, SIMD_VERSION_SSE1, SIMD_EMIT_EXTRACT_MASK },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_Max, OP_PMAXB_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_Min, OP_PMINB_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_SubtractWithSaturation, OP_PSUBB_SAT_UN, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_SumOfAbsoluteDifferences, OP_PSUM_ABS_DIFF, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackHigh, OP_UNPACK_HIGHB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V10, 10, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V11, 11, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V12, 12, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V13, 13, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V14, 14, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V15, 15, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V8, 8, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V9, 9, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_PADDB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_PCMPEQB, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_PCMPEQB, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_Subtraction, OP_PSUBB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V10, 10, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V11, 11, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V12, 12, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V13, 13, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V14, 14, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V15, 15, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V8, 8, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V9, 9, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
/*
setters
*/
static const SimdIntrinsc vector16sb_intrinsics[] = {
- { SN_ctor, OP_EXPAND_I1, SIMD_EMIT_CTOR },
- { SN_AddWithSaturation, OP_PADDB_SAT, SIMD_EMIT_BINARY },
- { SN_CompareEqual, OP_PCMPEQB, SIMD_EMIT_BINARY },
- { SN_CompareGreaterThan, OP_PCMPGTB, SIMD_EMIT_BINARY },
- { SN_ExtractByteMask, 0, SIMD_EMIT_EXTRACT_MASK },
- { SN_LoadAligned, 0, SIMD_EMIT_LOAD_ALIGNED },
- { SN_Max, OP_PMAXB, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_Min, OP_PMINB, SIMD_EMIT_BINARY, SIMD_VERSION_SSE41 },
- { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_0 },
- { SN_PrefetchTemporal1stLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_1 },
- { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_2 },
- { SN_PrefetchNonTemporal, 0, SIMD_EMIT_PREFETCH, SIMD_VERSION_SSE1, SIMD_PREFETCH_MODE_NTA },
- { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_EMIT_STORE },
- { SN_SubtractWithSaturation, OP_PSUBB_SAT, SIMD_EMIT_BINARY },
- { SN_UnpackHigh, OP_UNPACK_HIGHB, SIMD_EMIT_BINARY },
- { SN_UnpackLow, OP_UNPACK_LOWB, SIMD_EMIT_BINARY },
- { SN_get_V0, 0, SIMD_EMIT_GETTER },
- { SN_get_V1, 1, SIMD_EMIT_GETTER },
- { SN_get_V10, 10, SIMD_EMIT_GETTER },
- { SN_get_V11, 11, SIMD_EMIT_GETTER },
- { SN_get_V12, 12, SIMD_EMIT_GETTER },
- { SN_get_V13, 13, SIMD_EMIT_GETTER },
- { SN_get_V14, 14, SIMD_EMIT_GETTER },
- { SN_get_V15, 15, SIMD_EMIT_GETTER },
- { SN_get_V2, 2, SIMD_EMIT_GETTER },
- { SN_get_V3, 3, SIMD_EMIT_GETTER },
- { SN_get_V4, 4, SIMD_EMIT_GETTER },
- { SN_get_V5, 5, SIMD_EMIT_GETTER },
- { SN_get_V6, 6, SIMD_EMIT_GETTER },
- { SN_get_V7, 7, SIMD_EMIT_GETTER },
- { SN_get_V8, 8, SIMD_EMIT_GETTER },
- { SN_get_V9, 9, SIMD_EMIT_GETTER },
- { SN_op_Addition, OP_PADDB, SIMD_EMIT_BINARY },
- { SN_op_BitwiseAnd, OP_PAND, SIMD_EMIT_BINARY },
- { SN_op_BitwiseOr, OP_POR, SIMD_EMIT_BINARY },
- { SN_op_Equality, OP_PCMPEQB, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_EQ },
- { SN_op_ExclusiveOr, OP_PXOR, SIMD_EMIT_BINARY },
- { SN_op_Explicit, 0, SIMD_EMIT_CAST },
- { SN_op_Inequality, OP_PCMPEQB, SIMD_EMIT_EQUALITY, SIMD_VERSION_SSE1, SIMD_COMP_NEQ },
- { SN_op_Subtraction, OP_PSUBB, SIMD_EMIT_BINARY },
- { SN_set_V0, 0, SIMD_EMIT_SETTER },
- { SN_set_V1, 1, SIMD_EMIT_SETTER },
- { SN_set_V10, 10, SIMD_EMIT_SETTER },
- { SN_set_V11, 11, SIMD_EMIT_SETTER },
- { SN_set_V12, 12, SIMD_EMIT_SETTER },
- { SN_set_V13, 13, SIMD_EMIT_SETTER },
- { SN_set_V14, 14, SIMD_EMIT_SETTER },
- { SN_set_V15, 15, SIMD_EMIT_SETTER },
- { SN_set_V2, 2, SIMD_EMIT_SETTER },
- { SN_set_V3, 3, SIMD_EMIT_SETTER },
- { SN_set_V4, 4, SIMD_EMIT_SETTER },
- { SN_set_V5, 5, SIMD_EMIT_SETTER },
- { SN_set_V6, 6, SIMD_EMIT_SETTER },
- { SN_set_V7, 7, SIMD_EMIT_SETTER },
- { SN_set_V8, 8, SIMD_EMIT_SETTER },
- { SN_set_V9, 9, SIMD_EMIT_SETTER },
+ { SN_ctor, OP_EXPAND_I1, SIMD_VERSION_SSE1, SIMD_EMIT_CTOR },
+ { SN_AddWithSaturation, OP_PADDB_SAT, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareEqual, OP_PCMPEQB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_CompareGreaterThan, OP_PCMPGTB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_ExtractByteMask, 0, SIMD_VERSION_SSE1, SIMD_EMIT_EXTRACT_MASK },
+ { SN_LoadAligned, 0, SIMD_VERSION_SSE1, SIMD_EMIT_LOAD_ALIGNED },
+ { SN_Max, OP_PMAXB, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_Min, OP_PMINB, SIMD_VERSION_SSE41, SIMD_EMIT_BINARY },
+ { SN_PrefetchTemporalAllCacheLevels, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_0 },
+ { SN_PrefetchTemporal1stLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_1 },
+ { SN_PrefetchTemporal2ndLevelCache, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_2 },
+ { SN_PrefetchNonTemporal, 0, SIMD_VERSION_SSE1, SIMD_EMIT_PREFETCH, SIMD_PREFETCH_MODE_NTA },
+ { SN_StoreAligned, OP_STOREX_ALIGNED_MEMBASE_REG, SIMD_VERSION_SSE1, SIMD_EMIT_STORE },
+ { SN_SubtractWithSaturation, OP_PSUBB_SAT, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackHigh, OP_UNPACK_HIGHB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_UnpackLow, OP_UNPACK_LOWB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_get_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V10, 10, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V11, 11, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V12, 12, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V13, 13, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V14, 14, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V15, 15, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V8, 8, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_get_V9, 9, SIMD_VERSION_SSE1, SIMD_EMIT_GETTER },
+ { SN_op_Addition, OP_PADDB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseAnd, OP_PAND, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_BitwiseOr, OP_POR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Equality, OP_PCMPEQB, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_EQ },
+ { SN_op_ExclusiveOr, OP_PXOR, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_op_Explicit, 0, SIMD_VERSION_SSE1, SIMD_EMIT_CAST },
+ { SN_op_Inequality, OP_PCMPEQB, SIMD_VERSION_SSE1, SIMD_EMIT_EQUALITY, SIMD_COMP_NEQ },
+ { SN_op_Subtraction, OP_PSUBB, SIMD_VERSION_SSE1, SIMD_EMIT_BINARY },
+ { SN_set_V0, 0, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V1, 1, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V10, 10, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V11, 11, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V12, 12, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V13, 13, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V14, 14, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V15, 15, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V2, 2, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V3, 3, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V4, 4, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V5, 5, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V6, 6, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V7, 7, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V8, 8, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
+ { SN_set_V9, 9, SIMD_VERSION_SSE1, SIMD_EMIT_SETTER },
};
static guint32 simd_supported_versions;
/*Scan the first basic block looking xzeros not used*/
for (ins = first_bb->code; ins; ins = ins->next) {
+ int num_sregs;
+ int sregs [MONO_MAX_SRC_REGS];
+
if (ins->opcode == OP_XZERO) {
if (!(vreg_flags [ins->dreg] & VREG_HAS_OTHER_OP_BB0)) {
DEBUG (printf ("[simd-simplify] R%d has vzero: ", ins->dreg); mono_print_ins(ins));
}
if (ins->opcode == OP_LDADDR && apply_vreg_first_block_interference (cfg, ins, ((MonoInst*)ins->inst_p0)->dreg, max_vreg, vreg_flags))
continue;
-
if (apply_vreg_first_block_interference (cfg, ins, ins->dreg, max_vreg, vreg_flags))
continue;
- if (apply_vreg_first_block_interference (cfg, ins, ins->sreg1, max_vreg, vreg_flags))
- continue;
- if (apply_vreg_first_block_interference (cfg, ins, ins->sreg2, max_vreg, vreg_flags))
- continue;
+ num_sregs = mono_inst_get_src_registers (ins, sregs);
+ for (i = 0; i < num_sregs; ++i) {
+ if (apply_vreg_first_block_interference (cfg, ins, sregs [i], max_vreg, vreg_flags))
+ break;
+ }
}
if (IS_DEBUG_ON (cfg)) {
for (bb = first_bb->next_bb; bb; bb = bb->next_bb) {
for (ins = bb->code; ins; ins = ins->next) {
-
+ int num_sregs;
+ int sregs [MONO_MAX_SRC_REGS];
+
if (ins->opcode == OP_LDADDR && apply_vreg_following_block_interference (cfg, ins, ((MonoInst*)ins->inst_p0)->dreg, bb, max_vreg, vreg_flags, target_bb))
continue;
if (apply_vreg_following_block_interference (cfg, ins, ins->dreg, bb, max_vreg, vreg_flags, target_bb))
continue;
- if (apply_vreg_following_block_interference (cfg, ins, ins->sreg1, bb, max_vreg, vreg_flags, target_bb))
- continue;
- if (apply_vreg_following_block_interference (cfg, ins, ins->sreg2, bb, max_vreg, vreg_flags, target_bb))
- continue;
+ num_sregs = mono_inst_get_src_registers (ins, sregs);
+ for (i = 0; i < num_sregs; ++i) {
+ if (apply_vreg_following_block_interference (cfg, ins, sregs [i], bb,
+ max_vreg, vreg_flags, target_bb))
+ continue;
+ }
}
}
if (!(vreg_flags [var->dreg] & VREG_SINGLE_BB_USE))
continue;
for (ins = target_bb [var->dreg]->code; ins; ins = ins->next) {
+ int num_sregs, j;
+ int sregs [MONO_MAX_SRC_REGS];
+ gboolean found = FALSE;
+
+ num_sregs = mono_inst_get_src_registers (ins, sregs);
+ for (j = 0; j < num_sregs; ++j) {
+ if (sregs [j] == var->dreg)
+ found = TRUE;
+ }
/*We can avoid inserting the XZERO if the first use doesn't depend on the zero'ed value.*/
- if (ins->dreg == var->dreg && ins->sreg1 != var->dreg && ins->sreg2 != var->dreg) {
+ if (ins->dreg == var->dreg && !found) {
+ DEBUG (printf ("[simd-simplify] INGORING R%d on BB %d because first op is a def", i, target_bb [var->dreg]->block_num););
break;
- } else if (ins->sreg1 == var->dreg || ins->sreg2 == var->dreg) {
+ } else if (found) {
+ DEBUG (printf ("[simd-simplify] Adding XZERO for R%d on BB %d: ", i, target_bb [var->dreg]->block_num); );
MonoInst *tmp;
MONO_INST_NEW (cfg, tmp, OP_XZERO);
tmp->dreg = var->dreg;
}
for (ins = first_bb->code; ins; ins = ins->next) {
- if (ins->opcode == OP_XZERO && (vreg_flags [ins->dreg] & VREG_SINGLE_BB_USE))
+ if (ins->opcode == OP_XZERO && (vreg_flags [ins->dreg] & VREG_SINGLE_BB_USE)) {
+ DEBUG (printf ("[simd-simplify] Nullify %d on first BB: ", ins->dreg); mono_print_ins(ins));
NULLIFY_INS (ins);
+ }
}
g_free (vreg_flags);
* This function will load the value if needed.
*/
static int
-load_simd_vreg (MonoCompile *cfg, MonoMethod *cmethod, MonoInst *src)
+load_simd_vreg (MonoCompile *cfg, MonoMethod *cmethod, MonoInst *src, gboolean *indirect)
{
+ if (indirect)
+ *indirect = FALSE;
if (src->opcode == OP_XMOVE) {
return src->sreg1;
} else if (src->opcode == OP_LDADDR) {
return src->dreg;
} else if (src->type == STACK_PTR || src->type == STACK_MP) {
MonoInst *ins;
+ if (indirect)
+ *indirect = TRUE;
MONO_INST_NEW (cfg, ins, OP_LOADX_MEMBASE);
ins->klass = cmethod->klass;
g_assert_not_reached ();
}
-static MonoInst*
-get_int_to_float_spill_area (MonoCompile *cfg)
-{
- if (!cfg->iconv_raw_var) {
- cfg->iconv_raw_var = mono_compile_create_var (cfg, &mono_defaults.int32_class->byval_arg, OP_LOCAL);
- cfg->iconv_raw_var->flags |= MONO_INST_VOLATILE; /*FIXME, use the don't regalloc flag*/
- }
- return cfg->iconv_raw_var;
-}
-
/*We share the var with fconv_to_r8_x to save some stack space.*/
static MonoInst*
get_double_spill_area (MonoCompile *cfg)
return cfg->simd_ctor_var;
}
+static int
+mono_type_to_expand_op (MonoType *type)
+{
+ switch (type->type) {
+ case MONO_TYPE_I1:
+ case MONO_TYPE_U1:
+ return OP_EXPAND_I1;
+ case MONO_TYPE_I2:
+ case MONO_TYPE_U2:
+ return OP_EXPAND_I2;
+ case MONO_TYPE_I4:
+ case MONO_TYPE_U4:
+ return OP_EXPAND_I4;
+ case MONO_TYPE_I8:
+ case MONO_TYPE_U8:
+ return OP_EXPAND_I8;
+ case MONO_TYPE_R4:
+ return OP_EXPAND_R4;
+ case MONO_TYPE_R8:
+ return OP_EXPAND_R8;
+ default:
+ g_assert_not_reached ();
+ }
+}
+
+static int
+get_simd_vreg_or_expanded_scalar (MonoCompile *cfg, MonoMethod *cmethod, MonoInst *src, int position)
+{
+ MonoInst *ins;
+ MonoMethodSignature *sig = mono_method_signature (cmethod);
+ int expand_op;
+
+ g_assert (sig->param_count == 2);
+ g_assert (position == 0 || position == 1);
+
+ if (mono_class_from_mono_type (sig->params [position])->simd_type)
+ return get_simd_vreg (cfg, cmethod, src);
+
+ expand_op = mono_type_to_expand_op (sig->params [position]);
+ MONO_INST_NEW (cfg, ins, expand_op);
+ ins->klass = cmethod->klass;
+ ins->sreg1 = src->dreg;
+ ins->type = STACK_VTYPE;
+ ins->dreg = alloc_ireg (cfg);
+ MONO_ADD_INS (cfg->cbb, ins);
+
+ if (expand_op == OP_EXPAND_R4)
+ ins->backend.spill_var = mini_get_int_to_float_spill_area (cfg);
+ else if (expand_op == OP_EXPAND_R8)
+ ins->backend.spill_var = get_double_spill_area (cfg);
+
+ return ins->dreg;
+}
+
static MonoInst*
simd_intrinsic_emit_binary (const SimdIntrinsc *intrinsic, MonoCompile *cfg, MonoMethod *cmethod, MonoInst **args)
{
MonoInst* ins;
int left_vreg, right_vreg;
- left_vreg = get_simd_vreg (cfg, cmethod, args [0]);
- right_vreg = get_simd_vreg (cfg, cmethod, args [1]);
-
+ left_vreg = get_simd_vreg_or_expanded_scalar (cfg, cmethod, args [0], 0);
+ right_vreg = get_simd_vreg_or_expanded_scalar (cfg, cmethod, args [1], 1);
+
MONO_INST_NEW (cfg, ins, intrinsic->opcode);
ins->klass = cmethod->klass;
ins->sreg1 = left_vreg;
ins->sreg2 = right_vreg;
ins->type = STACK_VTYPE;
- ins->klass = cmethod->klass;
ins->dreg = alloc_ireg (cfg);
ins->inst_c0 = intrinsic->flags;
MONO_ADD_INS (cfg->cbb, ins);
case MONO_TYPE_U4:
case MONO_TYPE_R4:
return OP_EXTRACT_I4;
+ default:
+ g_assert_not_reached ();
}
- g_assert_not_reached ();
}
/*Returns the amount to shift the element index to get the dword it belongs to*/
case MONO_TYPE_U4:
case MONO_TYPE_R4:
return 0;
+ default:
+ g_assert_not_reached ();
+ }
+}
+
+static G_GNUC_UNUSED int
+mono_type_to_insert_op (MonoType *type)
+{
+ switch (type->type) {
+ case MONO_TYPE_I1:
+ case MONO_TYPE_U1:
+ return OP_INSERT_I1;
+ case MONO_TYPE_I2:
+ case MONO_TYPE_U2:
+ return OP_INSERT_I2;
+ case MONO_TYPE_I4:
+ case MONO_TYPE_U4:
+ return OP_INSERT_I4;
+ case MONO_TYPE_I8:
+ case MONO_TYPE_U8:
+ return OP_INSERT_I8;
+ case MONO_TYPE_R4:
+ return OP_INSERT_R4;
+ case MONO_TYPE_R8:
+ return OP_INSERT_R8;
+ default:
+ g_assert_not_reached ();
}
- g_assert_not_reached ();
}
static int
return OP_INSERTX_R4_SLOW;
case MONO_TYPE_R8:
return OP_INSERTX_R8_SLOW;
+ default:
+ g_assert_not_reached ();
}
- g_assert_not_reached ();
}
static MonoInst*
MonoInst *ins;
MonoMethodSignature *sig = mono_method_signature (cmethod);
int size, align;
+ gboolean indirect;
+ int dreg;
+
size = mono_type_size (sig->params [0], &align);
- if (size == 2 || size == 4 || size == 8) {
+ if (COMPILE_LLVM (cfg)) {
+ MONO_INST_NEW (cfg, ins, mono_type_to_insert_op (sig->params [0]));
+ ins->klass = cmethod->klass;
+ ins->dreg = ins->sreg1 = dreg = load_simd_vreg (cfg, cmethod, args [0], &indirect);
+ ins->sreg2 = args [1]->dreg;
+ ins->inst_c0 = intrinsic->opcode;
+ MONO_ADD_INS (cfg->cbb, ins);
+ } else if (size == 2 || size == 4 || size == 8) {
MONO_INST_NEW (cfg, ins, mono_type_to_slow_insert_op (sig->params [0]));
ins->klass = cmethod->klass;
/*This is a partial load so we encode the dependency on the previous value by setting dreg and sreg1 to the same value.*/
- ins->dreg = ins->sreg1 = load_simd_vreg (cfg, cmethod, args [0]);
+ ins->dreg = ins->sreg1 = dreg = load_simd_vreg (cfg, cmethod, args [0], &indirect);
ins->sreg2 = args [1]->dreg;
ins->inst_c0 = intrinsic->opcode;
if (sig->params [0]->type == MONO_TYPE_R4)
- ins->backend.spill_var = get_int_to_float_spill_area (cfg);
+ ins->backend.spill_var = mini_get_int_to_float_spill_area (cfg);
else if (sig->params [0]->type == MONO_TYPE_R8)
ins->backend.spill_var = get_double_spill_area (cfg);
MONO_ADD_INS (cfg->cbb, ins);
MONO_INST_NEW (cfg, ins, OP_EXTRACTX_U2);
ins->klass = cmethod->klass;
- ins->sreg1 = sreg = load_simd_vreg (cfg, cmethod, args [0]);
+ ins->sreg1 = sreg = dreg = load_simd_vreg (cfg, cmethod, args [0], &indirect);
ins->type = STACK_I4;
ins->dreg = vreg = alloc_ireg (cfg);
ins->inst_c0 = intrinsic->opcode / 2;
ins->dreg = sreg;
ins->inst_c0 = intrinsic->opcode;
MONO_ADD_INS (cfg->cbb, ins);
+ }
+ if (indirect) {
+ MONO_INST_NEW (cfg, ins, OP_STOREX_MEMBASE);
+ ins->klass = cmethod->klass;
+ ins->dreg = args [0]->dreg;
+ ins->sreg1 = dreg;
+ MONO_ADD_INS (cfg->cbb, ins);
}
return ins;
}
MonoMethodSignature *sig = mono_method_signature (cmethod);
int vreg, shift_bits = mono_type_elements_shift_bits (sig->ret);
- vreg = load_simd_vreg (cfg, cmethod, args [0]);
+ vreg = load_simd_vreg (cfg, cmethod, args [0], NULL);
- if (intrinsic->opcode >> shift_bits) {
+ if ((intrinsic->opcode >> shift_bits) && !cfg->compile_llvm) {
MONO_INST_NEW (cfg, ins, OP_PSHUFLED);
ins->klass = cmethod->klass;
ins->sreg1 = vreg;
ins->sreg1 = vreg;
ins->type = STACK_I4;
ins->dreg = vreg = alloc_ireg (cfg);
- ins->inst_c0 = intrinsic->opcode & ((1 << shift_bits) - 1);
+ if (cfg->compile_llvm)
+ ins->inst_c0 = intrinsic->opcode;
+ else
+ ins->inst_c0 = intrinsic->opcode & ((1 << shift_bits) - 1);
MONO_ADD_INS (cfg->cbb, ins);
if (sig->ret->type == MONO_TYPE_R4) {
- MONO_INST_NEW (cfg, ins, OP_ICONV_TO_R8_RAW);
+ MONO_INST_NEW (cfg, ins, cfg->r4fp ? OP_ICONV_TO_R4_RAW : OP_MOVE_I4_TO_F);
ins->klass = mono_defaults.single_class;
ins->sreg1 = vreg;
- ins->type = STACK_R8;
+ ins->type = cfg->r4_stack_type;
ins->dreg = alloc_freg (cfg);
- ins->backend.spill_var = get_int_to_float_spill_area (cfg);
- MONO_ADD_INS (cfg->cbb, ins);
+ ins->backend.spill_var = mini_get_int_to_float_spill_area (cfg);
+ MONO_ADD_INS (cfg->cbb, ins);
}
return ins;
}
int vreg;
gboolean is_r8 = mono_method_signature (cmethod)->ret->type == MONO_TYPE_R8;
- vreg = load_simd_vreg (cfg, cmethod, args [0]);
+ vreg = load_simd_vreg (cfg, cmethod, args [0], NULL);
MONO_INST_NEW (cfg, ins, is_r8 ? OP_EXTRACT_R8 : OP_EXTRACT_I8);
ins->klass = cmethod->klass;
MONO_ADD_INS (cfg->cbb, ins);
if (sig->params [0]->type == MONO_TYPE_R4)
- ins->backend.spill_var = get_int_to_float_spill_area (cfg);
+ ins->backend.spill_var = mini_get_int_to_float_spill_area (cfg);
else if (sig->params [0]->type == MONO_TYPE_R8)
ins->backend.spill_var = get_double_spill_area (cfg);
simd_intrinsic_emit_shuffle (const SimdIntrinsc *intrinsic, MonoCompile *cfg, MonoMethod *cmethod, MonoInst **args)
{
MonoInst *ins;
- int vreg;
+ int vreg, vreg2 = -1;
+ int param_count = mono_method_signature (cmethod)->param_count;
- /*TODO Exposing shuffle is not a good thing as it's non obvious. We should come up with better abstractions*/
-
- if (args [1]->opcode != OP_ICONST) {
- g_warning ("Shuffle with non literals is not yet supported");
- g_assert_not_reached ();
+ if (args [param_count - 1]->opcode != OP_ICONST) {
+ /*TODO Shuffle with non literals is not yet supported */
+ return NULL;
}
+
vreg = get_simd_vreg (cfg, cmethod, args [0]);
- NULLIFY_INS (args [1]);
+ if (param_count == 3)
+ vreg2 = get_simd_vreg (cfg, cmethod, args [1]);
+
+ NULLIFY_INS (args [param_count - 1]);
+
MONO_INST_NEW (cfg, ins, intrinsic->opcode);
ins->klass = cmethod->klass;
ins->sreg1 = vreg;
- ins->inst_c0 = args [1]->inst_c0;
+ ins->sreg2 = vreg2;
+ ins->inst_c0 = args [param_count - 1]->inst_c0;
ins->type = STACK_VTYPE;
ins->dreg = alloc_ireg (cfg);
MONO_ADD_INS (cfg->cbb, ins);
+
+ if (param_count == 3 && ins->opcode == OP_PSHUFLED)
+ ins->opcode = OP_SHUFPS;
return ins;
}
static MonoInst*
emit_intrinsics (MonoCompile *cfg, MonoMethod *cmethod, MonoMethodSignature *fsig, MonoInst **args, const SimdIntrinsc *intrinsics, guint32 size)
{
- const SimdIntrinsc * result = bsearch (cmethod->name, intrinsics, size, sizeof (SimdIntrinsc), &simd_intrinsic_compare_by_name);
+ const SimdIntrinsc * result = mono_binary_search (cmethod->name, intrinsics, size, sizeof (SimdIntrinsc), &simd_intrinsic_compare_by_name);
if (!result) {
DEBUG (printf ("function doesn't have a simd intrinsic %s::%s/%d\n", cmethod->klass->name, cmethod->name, fsig->param_count));
return NULL;
mono_print_ins (args [i]);
}
}
- if (result->simd_version && !((1 << result->simd_version) & simd_supported_versions)) {
- if (IS_DEBUG_ON (cfg))
- printf ("function %s::%s/%d requires unsuported SIMD instruction set %s \n", cmethod->klass->name, cmethod->name, fsig->param_count, simd_version_name (result->simd_version));
+ if (result->simd_version_flags && !(result->simd_version_flags & simd_supported_versions)) {
+ if (IS_DEBUG_ON (cfg)) {
+ int x;
+ printf ("function %s::%s/%d requires one of unsuported SIMD instruction set(s): ", cmethod->klass->name, cmethod->name, fsig->param_count);
+ for (x = 1; x <= SIMD_VERSION_INDEX_END; x++)
+ if (result->simd_version_flags & (1 << x))
+ printf ("%s ", simd_version_name (1 << x));
+
+ printf ("\n");
+ }
return NULL;
}
MONO_EMIT_NEW_BIALU_IMM (cfg, OP_MUL_IMM, mult_reg, index2_reg, size);
MONO_EMIT_NEW_BIALU (cfg, OP_PADD, add_reg, array_reg, mult_reg);
- NEW_BIALU_IMM (cfg, ins, OP_PADD_IMM, add_reg, add_reg, G_STRUCT_OFFSET (MonoArray, vector));
+ NEW_BIALU_IMM (cfg, ins, OP_PADD_IMM, add_reg, add_reg, MONO_STRUCT_OFFSET (MonoArray, vector));
ins->type = STACK_PTR;
MONO_ADD_INS (cfg->cbb, ins);
static MonoInst*
emit_array_extension_intrinsics (MonoCompile *cfg, MonoMethod *cmethod, MonoMethodSignature *fsig, MonoInst **args)
{
- if (!strcmp ("GetVector", cmethod->name) || !strcmp ("GetVectorAligned", cmethod->name)) {
+ if ((!strcmp ("GetVector", cmethod->name) || !strcmp ("GetVectorAligned", cmethod->name)) && fsig->param_count == 2) {
MonoInst *load;
int addr = mono_emit_vector_ldelema (cfg, fsig->params [0], args [0], args [1], TRUE);
return load;
}
- if (!strcmp ("SetVector", cmethod->name) || !strcmp ("SetVectorAligned", cmethod->name)) {
+ if ((!strcmp ("SetVector", cmethod->name) || !strcmp ("SetVectorAligned", cmethod->name)) && fsig->param_count == 3) {
MonoInst *store;
int vreg = get_simd_vreg (cfg, cmethod, args [1]);
int addr = mono_emit_vector_ldelema (cfg, fsig->params [0], args [0], args [2], TRUE);
return store;
}
- if (!strcmp ("IsAligned", cmethod->name)) {
+ if (!strcmp ("IsAligned", cmethod->name) && fsig->param_count == 2) {
MonoInst *ins;
int addr = mono_emit_vector_ldelema (cfg, fsig->params [0], args [0], args [1], FALSE);
static MonoInst*
emit_simd_runtime_intrinsics (MonoCompile *cfg, MonoMethod *cmethod, MonoMethodSignature *fsig, MonoInst **args)
{
- if (!strcmp ("get_AccelMode", cmethod->name)) {
+ if (!strcmp ("get_AccelMode", cmethod->name) && fsig->param_count == 0) {
MonoInst *ins;
EMIT_NEW_ICONST (cfg, ins, simd_supported_versions);
return ins;
{
const char *class_name;
- if (strcmp ("Mono.Simd", cmethod->klass->name_space))
+ if (strcmp ("Mono.Simd", cmethod->klass->image->assembly->aname.name) ||
+ strcmp ("Mono.Simd", cmethod->klass->name_space))
return NULL;
class_name = cmethod->klass->name;
return NULL;
}
-#endif
+#endif /* DISABLE_JIT */
+#endif /* MONO_ARCH_SIMD_INTRINSICS */
projects / mono.git / blobdiff