mat4_mul_mat4.lasm

local ffi = require "ffi"
local dasm = require "dasm"

|.arch x64
|.actionlist actions

local Dst = dasm.new( actions )

|  sub     rsp, 104
|  movups  xmm3, [rdx]
|  lea     rax, qword [rcx+8]
|  movups  xmm4, [rdx+16]
|  mov     ecx, 4
|  movups  xmm5, [rdx+32]
|  mov     r9d, ecx
|  movaps  [rsp+80], xmm6
|  movups  xmm6, [rdx+48]
|  lea     rdx, qword [rsp+0]
|  movaps  [rsp+64], xmm7
|  xorps   xmm7, xmm7
|->L1:
|  movss   xmm2, dword [rax-8]
|  movss   xmm0, dword [rax-4]
|  movss   xmm1, dword [rax]
|  shufps  xmm2, xmm2, 0
|  mulps   xmm2, xmm3
|  shufps  xmm0, xmm0, 0
|  mulps   xmm0, xmm4
|  addps   xmm2, xmm7
|  shufps  xmm1, xmm1, 0
|  mulps   xmm1, xmm5
|  addps   xmm2, xmm0
|  movss   xmm0, dword [rax+4]
|  shufps  xmm0, xmm0, 0
|  add     rax, 16
|  mulps   xmm0, xmm6
|  addps   xmm2, xmm1
|  addps   xmm2, xmm0
|  movups  [rdx], xmm2
|  add     rdx, 16
|  sub     r9, 1
|  jne ->L1
|  lea     rax, qword [rsp+0]
|->L2:
|  movups  xmm0, [rax]
|  lea     rax, qword [rax+16]
|  movups  [r8], xmm0
|  lea     r8, qword [r8+16]
|  sub     rcx, 1
|  jne ->L2
|  movaps  xmm6, [rsp+80]
|  movaps  xmm7, [rsp+64]
|  add     rsp, 104
|  ret     0


local code = Dst:build()
local fptr = ffi.cast( "void __cdecl (*) (__m128* x, __m128* y, __m128* result)", code )

debug.getregistry().__DASM_CODEGEN = debug.getregistry().__DASM_CODEGEN or {}
table.insert( debug.getregistry().__DASM_CODEGEN, code )

return fptr