turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack disassembly

turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack.s

	.section	.text.turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack,"ax",@progbits
	.p2align	4, 0x90
	.type	turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack,@function
turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack:
.Lfunc_begin1:
	.loc	1 1 0 is_stmt 1
	.cfi_startproc
	push	rbp
	.cfi_def_cfa_offset 16
	.cfi_offset rbp, -16
	mov	rbp, rsp
	.cfi_def_cfa_register rbp
.Ltmp2:
	push	r15
	push	r14
	push	r13
	push	r12
	push	rbx
	and	rsp, -64
	sub	rsp, 256
	.cfi_offset rbx, -56
	.cfi_offset r12, -48
	.cfi_offset r13, -40
	.cfi_offset r14, -32
	.cfi_offset r15, -24
	.loc	1 4 3 prologue_end
	mov	r9, qword ptr [rsi + 24]
	movzx	edi, word ptr [rdx + 8]
	mov	ebx, dword ptr [r9 + 12]
	mov	eax, dword ptr [r9 + 8]
	mov	r8d, edi
	shl	r8d, 6
	mov	qword ptr [rsp + 8], r8
	shl	rbx, 32
	lea	rcx, [rbx + rax]
	mov	qword ptr [rsp + 32], rcx
	cmp	r8, rcx
	jge	.LBB1_18
	.loc	1 0 3 is_stmt 0
	mov	r11d, 15361
	movzx	r10d, word ptr [rsi + 20]
	mov	r15d, 3538944000
	.loc	1 4 3
	or	rbx, rax
	mov	r8, qword ptr [rsi + 32]
	mov	ecx, dword ptr [rsi + 16]
	mov	r14d, dword ptr [rsi + 12]
	sub	rbx, qword ptr [rsp + 8]
	bextr	r9, qword ptr [r9], r11
	mov	r11d, dword ptr [rdx]
	mov	edx, dword ptr [rdx + 4]
	imul	rdi, r15
	imul	r15, r10
	mov	rsi, qword ptr [r8]
	shl	r10d, 6
	imul	rax, rdx, 409600
	mov	qword ptr [rsp + 64], rdx
	mov	qword ptr [rsp + 104], r11
	mov	qword ptr [rsp + 48], r10
	mov	qword ptr [rsp + 56], r15
	lea	r15, [4*rdx]
	mov	rdx, r11
	shl	rdx, 11
	shl	r11, 6
	add	rax, rdi
	mov	qword ptr [rsp + 80], r11
	add	rdx, rax
	mov	qword ptr [rsp + 40], r15
	lea	r15, [4*rcx]
	imul	rcx, rcx, 409600
	lea	rax, [rdx + 2*r9]
	add	rax, qword ptr [r8 + 8]
	mov	rdx, r14
	shl	rdx, 6
	shl	r14, 11
	mov	qword ptr [rsp + 96], r15
	mov	qword ptr [rsp + 112], rdx
	mov	qword ptr [rsp + 120], r14
	mov	qword ptr [rsp + 88], rcx
	mov	qword ptr [rsp + 16], rax
	jmp	.LBB1_2
	.p2align	4, 0x90
.LBB1_17:
	.loc	1 0 3
	mov	rdx, qword ptr [rsp + 16]
	mov	rax, qword ptr [rsp + 8]
	mov	rcx, qword ptr [rsp + 48]
	mov	rbx, qword ptr [rsp + 72]
	.loc	1 4 3
	add	rdx, qword ptr [rsp + 56]
	add	rax, rcx
	sub	rbx, rcx
	mov	qword ptr [rsp + 16], rdx
	mov	qword ptr [rsp + 8], rax
	cmp	rax, qword ptr [rsp + 32]
	jge	.LBB1_18
.LBB1_2:
	cmp	rbx, 64
	mov	edx, 64
	mov	eax, 1
	mov	qword ptr [rsp + 72], rbx
	cmovl	rdx, rbx
	cmp	rdx, 2
	cmovl	rdx, rax
	cmp	dword ptr [rsp + 64], 134
	ja	.LBB1_17
	.loc	1 0 3
	mov	rcx, qword ptr [rsp + 32]
	mov	rax, qword ptr [rsp + 16]
	mov	r11, qword ptr [rsp + 40]
	.loc	1 4 3
	sub	rcx, qword ptr [rsp + 8]
	mov	qword ptr [rsp + 24], rax
	jmp	.LBB1_4
	.p2align	4, 0x90
.LBB1_16:
	.loc	1 0 3
	mov	rax, qword ptr [rsp + 24]
	.loc	1 4 3
	add	r11, qword ptr [rsp + 96]
	add	rax, qword ptr [rsp + 88]
	mov	qword ptr [rsp + 24], rax
	cmp	r11, 540
	jge	.LBB1_17
.LBB1_4:
	.loc	1 0 3
	cmp	dword ptr [rsp + 104], 49
	.loc	1 4 3
	ja	.LBB1_16
	.loc	1 0 3
	mov	r9, qword ptr [rsp + 24]
	mov	r15, qword ptr [rsp + 80]
	jmp	.LBB1_6
	.p2align	4, 0x90
.LBB1_15:
	.loc	1 4 3
	add	r15, qword ptr [rsp + 112]
	add	r9, qword ptr [rsp + 120]
	cmp	r15, 3200
	jge	.LBB1_16
.LBB1_6:
	.loc	1 0 3
	test	rcx, rcx
	.loc	1 4 3
	jle	.LBB1_15
	.loc	1 0 3
	mov	rdi, r9
	xor	eax, eax
	.p2align	4, 0x90
.LBB1_8:
	mov	r10, rdi
	xor	r14d, r14d
	.p2align	4, 0x90
.LBB1_9:
	lea	r8, [r14 + r11]
	mov	r12, r10
	xor	r13d, r13d
	imul	rbx, r8, 102400
	add	rbx, rsi
	.p2align	4, 0x90
.LBB1_10:
	.loc	1 4 3
	mov	r8, r13
	or	r8, r15
	vpinsrw	xmm0, xmm0, word ptr [rbx + 2*r8], 0
	vpinsrw	xmm1, xmm0, word ptr [rbx + 2*r8 + 6400], 0
	vpinsrw	xmm2, xmm0, word ptr [rbx + 2*r8 + 12800], 0
	vpinsrw	xmm3, xmm0, word ptr [rbx + 2*r8 + 19200], 0
	vpinsrw	xmm4, xmm0, word ptr [rbx + 2*r8 + 25600], 0
	vpinsrw	xmm5, xmm0, word ptr [rbx + 2*r8 + 32000], 0
	vpinsrw	xmm6, xmm0, word ptr [rbx + 2*r8 + 38400], 0
	vpinsrw	xmm7, xmm0, word ptr [rbx + 2*r8 + 44800], 0
	vpinsrw	xmm8, xmm0, word ptr [rbx + 2*r8 + 51200], 0
	vpinsrw	xmm9, xmm0, word ptr [rbx + 2*r8 + 57600], 0
	vpinsrw	xmm10, xmm0, word ptr [rbx + 2*r8 + 64000], 0
	vpinsrw	xmm11, xmm0, word ptr [rbx + 2*r8 + 70400], 0
	vpinsrw	xmm12, xmm0, word ptr [rbx + 2*r8 + 76800], 0
	vpinsrw	xmm13, xmm0, word ptr [rbx + 2*r8 + 83200], 0
	vpinsrw	xmm14, xmm0, word ptr [rbx + 2*r8 + 89600], 0
	vpinsrw	xmm15, xmm0, word ptr [rbx + 2*r8 + 96000], 0
	xor	r8d, r8d
	vpunpcklwd	xmm6, xmm6, xmm7
	vpunpcklwd	xmm14, xmm14, xmm15
	vpunpcklwd	xmm4, xmm4, xmm5
	vpunpcklwd	xmm7, xmm12, xmm13
	vpunpcklwd	xmm2, xmm2, xmm3
	vpunpcklwd	xmm5, xmm10, xmm11
	vpunpcklwd	xmm3, xmm8, xmm9
	vpunpcklwd	xmm0, xmm0, xmm1
	vinserti128	ymm6, ymm6, xmm14, 1
	vinserti128	ymm4, ymm4, xmm7, 1
	vinserti128	ymm2, ymm2, xmm5, 1
	vinserti128	ymm0, ymm0, xmm3, 1
	vpunpckldq	ymm4, ymm4, ymm6
	vpunpckldq	ymm0, ymm0, ymm2
	vpunpcklqdq	ymm0, ymm0, ymm4
	vmovdqa	ymmword ptr [rsp + 128], ymm0
	.p2align	4, 0x90
.LBB1_11:
	vpinsrw	xmm0, xmm0, word ptr [rsp + 2*r8 + 128], 0
	vpextrw	word ptr [r12 + 2*r8], xmm0, 0
	inc	r8
	cmp	r8, 16
	jne	.LBB1_11
	inc	r13
	add	r12, 32
	cmp	r13, 64
	jne	.LBB1_10
	inc	r14
	add	r10, 102400
	cmp	r14, 4
	jne	.LBB1_9
	inc	rax
	add	rdi, 55296000
	cmp	rax, rdx
	jne	.LBB1_8
	jmp	.LBB1_15
.LBB1_18:
	xor	eax, eax
	lea	rsp, [rbp - 40]
	.loc	1 4 3 epilogue_begin
	pop	rbx
	pop	r12
	pop	r13
	pop	r14
	pop	r15
	pop	rbp
	.cfi_def_cfa rsp, 8
	vzeroupper
	ret
.Ltmp3:
.Lfunc_end1:
	.size	turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack, .Lfunc_end1-turbine_llm_mmtfp_3d_8640_3200_f32f16_dispatch_1_generic_Dx8640x3200_f16_pack
	.cfi_endproc