code wiki / (root) / nx_simd_i8x32_test.nx

nx_simd_i8x32_test.nx

buildroot/runtime/nx_simd_i8x32_test.nx

3851 B106 linesdepth 6pulls 8 transitivereach 0 importersview sourcekind gate/prooftopic simd

docs dependencies structs consts functions

about

nx_simd_i8x32_test.nx -- exercise the i8x32 SIMD intrinsics. 32 lanes of 8-bit ints in 256-bit vectors. Widest practical SIMD. Useful for crypto (AES rounds), string ops, INT8 quantized ML inference. No VMUL because AVX2 lacks vpmullb -- callers that need byte-mul split to i16x16 first.

dependencies 2 imports · 0 importers

imports: nx_kernel_v2.nxnx_log.nx

imported by: nobody (leaf or entry point)

call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown

structs

none

consts

none

functions

12	func pack8(b0: i64, b1: i64, b2: i64, b3: i64, called by 4: t1_load_store t2_vadd t3_vbroadcast t4_vreduce_sum
24	func t1_load_store() -> i64 called by 1: main calls 1: pack8
42	func t2_vadd() -> i64 called by 1: main calls 1: pack8
66	func t3_vbroadcast() -> i64 called by 1: main calls 1: pack8
79	func t4_vreduce_sum() -> i64 called by 1: main calls 1: pack8
93	func main() -> nx_exit calls 4: t1_load_store t2_vadd t3_vbroadcast t4_vreduce_sum