Skip to content

S
simde
Commit a202d011 authored by Michael R. Crusoe's avatar Michael R. Crusoe Committed by Michael R. Crusoe
Browse files
Options

avx512/gather: mm512_{mask_,}i64gather_{epi32,epi64,ps,pd)

parent 95a6d081
Hide whitespace changes
Inline Side-by-side
Showing with 206 additions and 0 deletions
x86/avx512/gather.h
View file @ a202d011
......@@ -28,6 +28,7 @@
#define SIMDE_X86_AVX512_GATHER_H
#include "types.h"
#include "../avx2.h"
HEDLEY_DIAGNOSTIC_PUSH
SIMDE_DISABLE_UNWANTED_DIAGNOSTICS
......@@ -54,12 +55,217 @@ simde_mm512_i32gather_ps(simde__m512i vindex, const void* base_addr, const int32
}
#if defined(SIMDE_X86_AVX512F_NATIVE) && (!defined(__clang__) || SIMDE_DETECT_CLANG_VERSION_CHECK(10,0,0))
#define simde_mm512_i32gather_ps(vindex, base_addr, scale) _mm512_i32gather_ps((vindex), (base_addr), (scale))
#elif defined(SIMDE_X86_AVX2_NATIVE) && defined(SIMDE_STATEMENT_EXPR_)
#define simde_mm512_i32gather_ps(vindex, base_addr, scale) SIMDE_STATEMENT_EXPR_(({\
simde__m512_private simde_mm512_i32gather_ps_r_; \
simde__m512i_private simde_mm512_i32gather_ps_vindex_ = simde__m512i_to_private((vindex)); \
simde_mm512_i32gather_ps_r_.m256[0] = _mm256_i32gather_ps( \
HEDLEY_STATIC_CAST(float const*, (base_addr)), simde_mm512_i32gather_ps_vindex_.m256i[0], (scale)); \
simde_mm512_i32gather_ps_r_.m256[1] = _mm256_i32gather_ps( \
HEDLEY_STATIC_CAST(float const*, (base_addr)), simde_mm512_i32gather_ps_vindex_.m256i[1], (scale)); \
simde__m512_from_private(simde_mm512_i32gather_ps_r_); \
}))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_i32gather_ps
#define _mm512_i32gather_ps(vindex, base_addr, scale) simde_mm512_i32gather_ps((vindex), (base_addr), (scale))
#endif
SIMDE_FUNCTION_ATTRIBUTES
simde__m256i
simde_mm512_i64gather_epi32(simde__m512i vindex, const void* base_addr, const int32_t scale)
SIMDE_REQUIRE_CONSTANT(scale)
HEDLEY_REQUIRE_MSG((scale && scale <= 8 && !(scale & (scale - 1))), "`scale' must be a power of two less than or equal to 8") {
simde__m512i_private vindex_;
simde__m256i_private r_;
vindex_ = simde__m512i_to_private(vindex);
r_ = simde__m256i_to_private(simde_mm256_setzero_si256());
const uint8_t* addr = HEDLEY_REINTERPRET_CAST(const uint8_t*, base_addr);
SIMDE_VECTORIZE
for (size_t i = 0 ; i < (sizeof(vindex_.i64) / sizeof(vindex_.i64[0])) ; i++) {
const uint8_t* src = addr + (HEDLEY_STATIC_CAST(size_t , vindex_.i64[i]) * HEDLEY_STATIC_CAST(size_t , scale));
int32_t dst;
simde_memcpy(&dst, src, sizeof(dst));
r_.i32[i] = dst;
}
return simde__m256i_from_private(r_);
}
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_i64gather_epi32(vindex, base_addr, scale) _mm512_i64gather_epi32((vindex), (base_addr), (scale))
#elif defined(SIMDE_X86_AVX2_NATIVE) && defined(SIMDE_STATEMENT_EXPR_)
#define simde_mm512_i64gather_epi32(vindex, base_addr, scale) SIMDE_STATEMENT_EXPR_(({\
simde__m256i_private simde_mm512_i64gather_epi32_r_; \
simde__m512i_private simde_mm512_i64gather_epi32_vindex_ = simde__m512i_to_private((vindex)); \
simde_mm512_i64gather_epi32_r_.m128i[0] = _mm256_i64gather_epi32( \
HEDLEY_STATIC_CAST(int const*, (base_addr)), simde_mm512_i64gather_epi32_vindex_.m256i[0], (scale)); \
simde_mm512_i64gather_epi32_r_.m128i[1] = _mm256_i64gather_epi32( \
HEDLEY_STATIC_CAST(int const*, (base_addr)), simde_mm512_i64gather_epi32_vindex_.m256i[1], (scale)); \
simde__m256i_from_private(simde_mm512_i64gather_epi32_r_); \
}))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_i64gather_epi32
#define _mm512_i64gather_epi32(vindex, base_addr, scale) simde_mm512_i64gather_epi32((vindex), (base_addr), (scale))
#endif
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_mask_i64gather_epi32(src, k, vindex, base_addr, scale) _mm512_mask_i64gather_epi32((src), (k), (vindex), (base_addr), (scale))
#else
#define simde_mm512_mask_i64gather_epi32(src, k, vindex, base_addr, scale) simde_mm256_mask_mov_epi32(src, k, simde_mm512_i64gather_epi32((vindex), (base_addr), (scale)))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_mask_i64gather_epi32
#define _mm512_mask_i64gather_epi32(src, k, vindex, base_addr, scale) simde_mm512_mask_i64gather_epi32((src), (k), (vindex), (base_addr), (scale))
#endif
SIMDE_FUNCTION_ATTRIBUTES
simde__m512i
simde_mm512_i64gather_epi64(simde__m512i vindex, const void* base_addr, const int32_t scale)
SIMDE_REQUIRE_CONSTANT(scale)
HEDLEY_REQUIRE_MSG((scale && scale <= 8 && !(scale & (scale - 1))), "`scale' must be a power of two less than or equal to 8") {
simde__m512i_private
vindex_ = simde__m512i_to_private(vindex),
r_ = simde__m512i_to_private(simde_mm512_setzero_si512());
const uint8_t* addr = HEDLEY_REINTERPRET_CAST(const uint8_t*, base_addr);
SIMDE_VECTORIZE
for (size_t i = 0 ; i < (sizeof(vindex_.i64) / sizeof(vindex_.i64[0])) ; i++) {
const uint8_t* src = addr + (HEDLEY_STATIC_CAST(size_t , vindex_.i64[i]) * HEDLEY_STATIC_CAST(size_t , scale));
int64_t dst;
simde_memcpy(&dst, src, sizeof(dst));
r_.i64[i] = dst;
}
return simde__m512i_from_private(r_);
}
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_i64gather_epi64(vindex, base_addr, scale) _mm512_i64gather_epi64((vindex), (base_addr), (scale))
#elif defined(SIMDE_X86_AVX2_NATIVE) && defined(SIMDE_STATEMENT_EXPR_)
#define simde_mm512_i64gather_epi64(vindex, base_addr, scale) SIMDE_STATEMENT_EXPR_(({\
simde__m512i_private simde_mm512_i64gather_epi64_r_, \
simde_mm512_i64gather_epi64_vindex_ = simde__m512i_to_private((vindex)); \
simde_mm512_i64gather_epi64_r_.m256i[0] = _mm256_i64gather_epi64( \
HEDLEY_STATIC_CAST(int64_t const*, (base_addr)), simde_mm512_i64gather_epi64_vindex_.m256i[0], (scale)); \
simde_mm512_i64gather_epi64_r_.m256i[1] = _mm256_i64gather_epi64( \
HEDLEY_STATIC_CAST(int64_t const*, (base_addr)), simde_mm512_i64gather_epi64_vindex_.m256i[1], (scale)); \
simde__m512i_from_private(simde_mm512_i64gather_epi64_r_); \
}))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_i64gather_epi64
#define _mm512_i64gather_epi64(vindex, base_addr, scale) simde_mm512_i64gather_epi64(vindex, (base_addr), (scale))
#endif
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_mask_i64gather_epi64(src, k, vindex, base_addr, scale) _mm512_mask_i64gather_epi64((src), (k), (vindex), (base_addr), (scale))
#else
#define simde_mm512_mask_i64gather_epi64(src, k, vindex, base_addr, scale) simde_mm512_mask_mov_epi64((src), (k), simde_mm512_i64gather_epi64((vindex), (base_addr), (scale)))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_mask_i64gather_epi64
#define _mm512_mask_i64gather_epi64(src, k, vindex, base_addr, scale) simde_mm512_mask_i64gather_epi64((src), (k), (vindex), (base_addr), (scale))
#endif
SIMDE_FUNCTION_ATTRIBUTES
simde__m512d
simde_mm512_i64gather_pd(simde__m512i vindex, const void* base_addr, const int32_t scale)
SIMDE_REQUIRE_CONSTANT(scale)
HEDLEY_REQUIRE_MSG((scale && scale <= 8 && !(scale & (scale - 1))), "`scale' must be a power of two less than or equal to 8") {
simde__m512i_private vindex_;
simde__m512d_private r_;
vindex_ = simde__m512i_to_private(vindex);
r_ = simde__m512d_to_private(simde_mm512_setzero_pd());
const uint8_t* addr = HEDLEY_REINTERPRET_CAST(const uint8_t*, base_addr);
SIMDE_VECTORIZE
for (size_t i = 0 ; i < (sizeof(vindex_.i64) / sizeof(vindex_.i64[0])) ; i++) {
const uint8_t* src = addr + (HEDLEY_STATIC_CAST(size_t , vindex_.i64[i]) * HEDLEY_STATIC_CAST(size_t , scale));
simde_float64 dst;
simde_memcpy(&dst, src, sizeof(dst));
r_.f64[i] = dst;
}
return simde__m512d_from_private(r_);
}
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_i64gather_pd(vindex, base_addr, scale) _mm512_i64gather_pd((vindex), (base_addr), (scale))
#elif defined(SIMDE_X86_AVX2_NATIVE) && defined(SIMDE_STATEMENT_EXPR_)
#define simde_mm512_i64gather_pd(vindex, base_addr, scale) SIMDE_STATEMENT_EXPR_(({\
simde__m512d_private simde_mm512_i64gather_pd_r_; \
simde__m512i_private simde_mm512_i64gather_pd_vindex_ = simde__m512i_to_private((vindex)); \
simde_mm512_i64gather_pd_r_.m256d[0] = _mm256_i64gather_pd( \
HEDLEY_STATIC_CAST(double const*, (base_addr)), simde_mm512_i64gather_pd_vindex_.m256i[0], (scale)); \
simde_mm512_i64gather_pd_r_.m256d[1] = _mm256_i64gather_pd( \
HEDLEY_STATIC_CAST(double const*, (base_addr)), simde_mm512_i64gather_pd_vindex_.m256i[1], (scale)); \
simde__m512d_from_private(simde_mm512_i64gather_pd_r_); \
}))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_i64gather_pd
#define _mm512_i64gather_pd(vindex, base_addr, scale) simde_mm512_i64gather_pd((vindex), (base_addr), (scale))
#endif
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_mask_i64gather_pd(src, k, vindex, base_addr, scale) _mm512_mask_i64gather_pd((src), (k), (vindex), (base_addr), (scale))
#else
#define simde_mm512_mask_i64gather_pd(src, k, vindex, base_addr, scale) simde_mm512_mask_mov_pd((src), (k), simde_mm512_i64gather_pd((vindex), (base_addr), (scale)))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_mask_i64gather_pd
#define _mm512_mask_i64gather_pd(src, k, vindex, base_addr, scale) simde_mm512_mask_i64gather_pd((src), (k), (vindex), (base_addr), (scale))
#endif
SIMDE_FUNCTION_ATTRIBUTES
simde__m256
simde_mm512_i64gather_ps(simde__m512i vindex, const void* base_addr, const int32_t scale)
SIMDE_REQUIRE_CONSTANT(scale)
HEDLEY_REQUIRE_MSG((scale && scale <= 8 && !(scale & (scale - 1))), "`scale' must be a power of two less than or equal to 8") {
simde__m512i_private vindex_;
simde__m256_private r_;
vindex_ = simde__m512i_to_private(vindex);
r_ = simde__m256_to_private(simde_mm256_setzero_ps());
const uint8_t* addr = HEDLEY_REINTERPRET_CAST(const uint8_t*, base_addr);
SIMDE_VECTORIZE
for (size_t i = 0 ; i < (sizeof(vindex_.i64) / sizeof(vindex_.i64[0])) ; i++) {
const uint8_t* src = addr + (HEDLEY_STATIC_CAST(size_t , vindex_.i64[i]) * HEDLEY_STATIC_CAST(size_t , scale));
simde_float32 dst;
simde_memcpy(&dst, src, sizeof(dst));
r_.f32[i] = dst;
}
return simde__m256_from_private(r_);
}
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_i64gather_ps(vindex, base_addr, scale) _mm512_i64gather_ps((vindex), (base_addr), (scale))
#elif defined(SIMDE_X86_AVX2_NATIVE) && defined(SIMDE_STATEMENT_EXPR_)
#define simde_mm512_i64gather_ps(vindex, base_addr, scale) SIMDE_STATEMENT_EXPR_(({\
simde__m256_private simde_mm512_i64gather_ps_r_; \
simde__m512i_private simde_mm512_i64gather_ps_vindex_ = simde__m512i_to_private((vindex)); \
simde_mm512_i64gather_ps_r_.m128[0] = _mm256_i64gather_ps( \
HEDLEY_STATIC_CAST(float const*, (base_addr)), simde_mm512_i64gather_ps_vindex_.m256i[0], (scale)); \
simde_mm512_i64gather_ps_r_.m128[1] = _mm256_i64gather_ps( \
HEDLEY_STATIC_CAST(float const*, (base_addr)), simde_mm512_i64gather_ps_vindex_.m256i[1], (scale)); \
simde__m256_from_private(simde_mm512_i64gather_ps_r_); \
}))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_i64gather_ps
#define _mm512_i64gather_ps(vindex, base_addr, scale) simde_mm512_i64gather_ps((vindex), (base_addr), (scale))
#endif
#if defined(SIMDE_X86_AVX512F_NATIVE)
#define simde_mm512_mask_i64gather_ps(src, k, vindex, base_addr, scale) _mm512_mask_i64gather_ps((src), (k), (vindex), (base_addr), (scale))
#else
#define simde_mm512_mask_i64gather_ps(src, k, vindex, base_addr, scale) simde_mm256_mask_mov_ps((src), (k), simde_mm512_i64gather_ps((vindex), (base_addr), (scale)))
#endif
#if defined(SIMDE_X86_AVX512F_ENABLE_NATIVE_ALIASES)
#undef _mm512_mask_i64gather_ps
#define _mm512_mask_i64gather_ps(src, k, vindex, base_addr, scale) simde_mm512_mask_i64gather_ps((src), (k), (vindex), (base_addr), (scale))
#endif
SIMDE_END_DECLS_
HEDLEY_DIAGNOSTIC_POP
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment