xref: /freebsd/contrib/llvm-project/compiler-rt/lib/builtins/i386/ashldi3.S (revision 2e3507c25e42292b45a5482e116d278f5515d04d)
1// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
2// See https://llvm.org/LICENSE.txt for license information.
3// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
4
5#include "../assembly.h"
6
7// di_int __ashldi3(di_int input, int count);
8
9// This routine has some extra memory traffic, loading the 64-bit input via two
10// 32-bit loads, then immediately storing it back to the stack via a single 64-bit
11// store.  This is to avoid a write-small, read-large stall.
12// However, if callers of this routine can be safely assumed to store the argument
13// via a 64-bt store, this is unnecessary memory traffic, and should be avoided.
14// It can be turned off by defining the TRUST_CALLERS_USE_64_BIT_STORES macro.
15
16#ifdef __i386__
17#ifdef __SSE2__
18
19.text
20.balign 4
21DEFINE_COMPILERRT_FUNCTION(__ashldi3)
22	movd	  12(%esp),		%xmm2	// Load count
23#ifndef TRUST_CALLERS_USE_64_BIT_STORES
24	movd	   4(%esp),		%xmm0
25	movd	   8(%esp),		%xmm1
26	punpckldq	%xmm1,		%xmm0	// Load input
27#else
28	movq	   4(%esp),		%xmm0	// Load input
29#endif
30	psllq		%xmm2,		%xmm0	// shift input by count
31	movd		%xmm0,		%eax
32	psrlq		$32,		%xmm0
33	movd		%xmm0,		%edx
34	ret
35END_COMPILERRT_FUNCTION(__ashldi3)
36
37#else // Use GPRs instead of SSE2 instructions, if they aren't available.
38
39.text
40.balign 4
41DEFINE_COMPILERRT_FUNCTION(__ashldi3)
42	movl	  12(%esp),		%ecx	// Load count
43	movl	   8(%esp),		%edx	// Load high
44	movl	   4(%esp),		%eax	// Load low
45
46	testl		$0x20,		%ecx	// If count >= 32
47	jnz		1f			//    goto 1
48	shldl		%cl, %eax,	%edx	// left shift high by count
49	shll		%cl,		%eax	// left shift low by count
50	ret
51
521:	movl		%eax,		%edx	// Move low to high
53	xorl		%eax,		%eax	// clear low
54	shll		%cl,		%edx	// shift high by count - 32
55	ret
56END_COMPILERRT_FUNCTION(__ashldi3)
57
58#endif // __SSE2__
59#endif // __i386__
60
61NO_EXEC_STACK_DIRECTIVE
62
63