// Tile size: 4x3
	// Accumulators: 0-11
	// Col regs: ymm12
	// Row regs: ymm13-15

	// Load col of A
	vmovaps			ymm12,	[rax]

	// Fill 3 cols of B
	vbroadcastss	ymm13,	dword ptr [rcx + 0]
	vbroadcastss	ymm14,	dword ptr [rcx + 4]
	vbroadcastss	ymm15,	dword ptr [rcx + 8]

	// N.B. Stepping cols in inner loop
	vfmadd231ps		ymm0,	ymm12, ymm13
	vfmadd231ps		ymm4,	ymm12, ymm14
	vfmadd231ps		ymm8,	ymm12, ymm15

	vmovaps			ymm12,	[rax+32]

	vfmadd231ps		ymm1,	ymm12, ymm13
	vfmadd231ps		ymm5,	ymm12, ymm14
	vfmadd231ps		ymm9,	ymm12, ymm15

	vmovaps			ymm12,	[rax+64]

	vfmadd231ps		ymm2,	ymm12, ymm13
	vfmadd231ps		ymm6,	ymm12, ymm14
	vfmadd231ps		ymm10,	 ymm12, ymm15

	vmovaps			ymm12,	[rax+96]

	vfmadd231ps		ymm3,	ymm12, ymm13
	vfmadd231ps		ymm7,	ymm12, ymm14
	vfmadd231ps		ymm11,	ymm12, ymm15

	// Load col of A, switching col!
	vmovaps			ymm13,	[rax + 128]

	// Fill 3 cols of B
	vbroadcastss	ymm14,	dword ptr [rcx + 12]
	vbroadcastss	ymm15,	dword ptr [rcx + 16]
	vbroadcastss	ymm12,	dword ptr [rcx + 20]

	// N.B. Stepping cols in inner loop
	vfmadd231ps		ymm0,	ymm13, ymm14
	vfmadd231ps		ymm4,	ymm13, ymm15
	vfmadd231ps		ymm8,	ymm13, ymm12

	vmovaps			ymm13,	[rax + 160]

	vfmadd231ps		ymm1,	ymm13, ymm14
	vfmadd231ps		ymm5,	ymm13, ymm15
	vfmadd231ps		ymm9,	ymm13, ymm12

	vmovaps			ymm13,	[rax + 192]

	vfmadd231ps		ymm2,	ymm13, ymm14
	vfmadd231ps		ymm6,	ymm13, ymm15
	vfmadd231ps		ymm10,	 ymm13, ymm12

	vmovaps			ymm13,	[rax + 224]

	vfmadd231ps		ymm3,	ymm13, ymm14
	vfmadd231ps		ymm7,	ymm13, ymm15
	vfmadd231ps		ymm11,	ymm13, ymm12

    add             rcx,    24
    add             rax,    256