Streaming SIMD Extensions Intrinsics Implementation

Regular Streaming SIMD Extensions intrinsics work on 4 32-bit single precision values. On ItaniumŪ-based systems basic operations like add or compare will require two SIMD instructions. Both can be executed in the same cycle so the throughput is one basic Streaming SIMD Extensions operation per cycle or 4 32-bit single precision operations per cycle.

Key to the table entries

Intrinsic
Name		 Alternate
Name		 Across
All IA		 MMX(TM
Technology		 Streaming
SIMD
Extensions

Streaming
SIMD
Extensions 2

 ItaniumŪ
Architecture
_mm_add_ss   N/A N/A B B
_mm_add_ps   N/A N/A A A
_mm_sub_ss   N/A N/A B B
_mm_sub_ps   N/A N/A A A
_mm_mul_ss   N/A N/A B B
_mm_mul_ps   N/A N/A A A
_mm_div_ss   N/A N/A B B
_mm_div_ps   N/A N/A A A
_mm_sqrt_ss   N/A N/A B B
_mm_sqrt_ps   N/A N/A A A
_mm_rcp_ss   N/A N/A B B
_mm_rcp_ps   N/A N/A A A
_mm_rsqrt_ss   N/A N/A B B
_mm_rsqrt_ps   N/A N/A A A
_mm_min_ss   N/A N/A B B
_mm_min_ps   N/A N/A A A
_mm_max_ss   N/A N/A B B
_mm_max_ps   N/A N/A A A
_mm_and_ps   N/A N/A A A
_mm_andnot_ps   N/A N/A A A
_mm_or_ps   N/A N/A A A
_mm_xor_ps   N/A N/A A A
_mm_cmpeq_ss   N/A N/A B B
_mm_cmpeq_ps   N/A N/A A A
_mm_cmplt_ss   N/A N/A B B
_mm_cmplt_ps   N/A N/A A A
_mm_cmple_ss   N/A N/A B B
_mm_cmple_ps   N/A N/A A A
_mm_cmpgt_ss   N/A N/A B B
_mm_cmpgt_ps   N/A N/A A A
_mm_cmpge_ss   N/A N/A B B
_mm_cmpge_ps   N/A N/A A A
_mm_cmpneq_ss   N/A N/A B B
_mm_cmpneq_ps   N/A N/A A A
_mm_cmpnlt_ss   N/A N/A B B
_mm_cmpnlt_ps   N/A N/A A A
_mm_cmpnle_ss   N/A N/A B B
_mm_cmpnle_ps   N/A N/A A A
_mm_cmpngt_ss   N/A N/A B B
_mm_cmpngt_ps   N/A N/A A A
_mm_cmpnge_ss   N/A N/A B B
_mm_cmpnge_ps   N/A N/A A A
_mm_cmpord_ss   N/A N/A B B
_mm_cmpord_ps   N/A N/A A A
_mm_cmpunord_ss   N/A N/A B B
_mm_cmpunord_ps   N/A N/A A A
_mm_comieq_ss   N/A N/A B B
_mm_comilt_ss   N/A N/A B B
_mm_comile_ss   N/A N/A B B
_mm_comigt_ss   N/A N/A B B
_mm_comige_ss   N/A N/A B B
_mm_comineq_ss   N/A N/A B B
_mm_ucomieq_ss   N/A N/A B B
_mm_ucomilt_ss   N/A N/A B B
_mm_ucomile_ss   N/A N/A B B
_mm_ucomigt_ss   N/A N/A B B
_mm_ucomige_ss   N/A N/A B B
_mm_ucomineq_ss   N/A N/A B B
_mm_cvt_ss2si _mm_cvtss_si32 N/A N/A A B
_mm_cvt_ps2pi _mm_cvtps_pi32 N/A N/A A A
_mm_cvtt_ss2si _mm_cvttss_si32 N/A N/A A B
_mm_cvtt_ps2pi _mm_cvttps_pi32 N/A N/A A A
_mm_cvt_si2ss _mm_cvtsi32_ss N/A N/A A B
_mm_cvt_pi2ps _mm_cvtpi32_ps N/A N/A A C
_mm_cvtpi16_ps   N/A N/A A C
_mm_cvtpu16_ps   N/A N/A A C
_mm_cvtpi8_ps   N/A N/A A C
_mm_cvtpu8_ps   N/A N/A A C
_mm_cvtpi32x2_ps   N/A N/A A C
_mm_cvtps_pi16   N/A N/A A C
_mm_cvtps_pi8   N/A N/A A C
_mm_move_ss   N/A N/A A A
_mm_shuffle_ps   N/A N/A A A
_mm_unpackhi_ps   N/A N/A A A
_mm_unpacklo_ps   N/A N/A A A
_mm_movehl_ps   N/A N/A A A
_mm_movelh_ps   N/A N/A A A
_mm_movemask_ps   N/A N/A A C
_mm_getcsr   N/A N/A A A
_mm_setcsr   N/A N/A A A
_mm_loadh_pi   N/A N/A A A
_mm_loadl_pi   N/A N/A A A
_mm_load_ss   N/A N/A A B
_mm_load_ps1 _mm_load1_ps N/A N/A A A
_mm_load_ps   N/A N/A A A
_mm_loadu_ps   N/A N/A A A
_mm_loadr_ps   N/A N/A A A
_mm_storeh_pi   N/A N/A A A
_mm_storel_pi   N/A N/A A A
_mm_store_ss   N/A N/A A A
_mm_store_ps   N/A N/A A A
_mm_store_ps1 _mm_store1_ps N/A N/A A A
_mm_storeu_ps   N/A N/A A A
_mm_storer_ps   N/A N/A A A
_mm_set_ss   N/A N/A A A
_mm_set_ps1 _mm_set1_ps N/A N/A A A
_mm_set_ps   N/A N/A A A
_mm_setr_ps   N/A N/A A A
_mm_setzero_ps   N/A N/A A A
_mm_prefetch   N/A N/A A A
_mm_stream_pi   N/A N/A A A
_mm_stream_ps   N/A N/A A A
_mm_sfence   N/A N/A A A
_m_pextrw _mm_extract_pi16 N/A N/A A A
_m_pinsrw _mm_insert_pi16 N/A N/A A A
_m_pmaxsw _mm_max_pi16 N/A N/A A A
_m_pmaxub _mm_max_pu8 N/A N/A A A
_m_pminsw _mm_min_pi16 N/A N/A A A
_m_pminub _mm_min_pu8 N/A N/A A A
_m_pmovmskb _mm_movemask_pi8 N/A N/A A C
_m_pmulhuw _mm_mulhi_pu16 N/A N/A A A
_m_pshufw _mm_shuffle_pi16 N/A N/A A A
_m_maskmovq _mm_maskmove_si64 N/A N/A A C
_m_pavgb _mm_avg_pu8 N/A N/A A A
_m_pavgw _mm_avg_pu16 N/A N/A A A
_m_psadbw _mm_sad_pu8 N/A N/A A A