mpn_zero_p

[Torbjörn Granlund]

bodrato at mail.dm.unipi.it writes:

  mpn/x86_64/fastsse/com.asm does support zero size with an initial
          test    n, n
          jz      L(don)
  while neither the generic C function for the library ( mpn/generic/com.c ),
  nor the inlined version in gmp-impl.h does.

I took a quick look at the code in that directory.  It can surely be
polished.  What about this initial patch:

diff -Nrc2 gmp-main.e5ccd9546c7d/mpn/x86_64/fastsse/com.asm gmp-main/mpn/x86_64/fastsse/com.asm
*** gmp-main.e5ccd9546c7d/mpn/x86_64/fastsse/com.asm	Wed Jun 24 11:59:18 2015
--- gmp-main/mpn/x86_64/fastsse/com.asm	Wed Jun 24 11:59:18 2015
***************
*** 79,85 ****
  	FUNC_ENTRY(3)
  
- 	test	n, n
- 	jz	L(don)
- 
  	pcmpeqb	%xmm7, %xmm7		C set to 111...111
  
--- 79,82 ----
diff -Nrc2 gmp-main.e5ccd9546c7d/mpn/x86_64/fastsse/copyd.asm gmp-main/mpn/x86_64/fastsse/copyd.asm
*** gmp-main.e5ccd9546c7d/mpn/x86_64/fastsse/copyd.asm	Wed Jun 24 11:59:18 2015
--- gmp-main/mpn/x86_64/fastsse/copyd.asm	Wed Jun 24 11:59:18 2015
***************
*** 81,87 ****
  	FUNC_ENTRY(3)
  
- 	test	n, n
- 	jz	L(don)
- 
  	lea	-16(rp,n,8), rp
  	lea	-16(up,n,8), up
--- 81,84 ----
diff -Nrc2 gmp-main.e5ccd9546c7d/mpn/x86_64/fastsse/copyi.asm gmp-main/mpn/x86_64/fastsse/copyi.asm
*** gmp-main.e5ccd9546c7d/mpn/x86_64/fastsse/copyi.asm	Wed Jun 24 11:59:18 2015
--- gmp-main/mpn/x86_64/fastsse/copyi.asm	Wed Jun 24 11:59:18 2015
***************
*** 81,85 ****
  	FUNC_ENTRY(3)
  
! 	cmp	$3, n
  	jc	L(bc)
  
--- 81,85 ----
  	FUNC_ENTRY(3)
  
! 	cmp	$3, n			C NB: bc code below assumes this limit
  	jc	L(bc)
  
***************
*** 144,150 ****
  	ALIGN(16)
  1:
! L(end):	test	$1, R8(n)
  	jz	1f
! 	mov	(up), %r8
  	mov	%r8, (rp)
  1:
--- 144,150 ----
  	ALIGN(16)
  1:
! 	test	$1, R8(n)
  	jz	1f
! L(1):	mov	(up), %r8
  	mov	%r8, (rp)
  1:
***************
*** 156,174 ****
  
  L(bc):	sub	$2, n
! 	jc	L(end)
  	ALIGN(16)
  1:	mov	(up), %rax
  	mov	8(up), %rcx
! 	lea	16(up), up
  	mov	%rax, (rp)
  	mov	%rcx, 8(rp)
! 	lea	16(rp), rp
! 	sub	$2, n
! 	jnc	1b
  
  	test	$1, R8(n)
  	jz	L(ret)
! 	mov	(up), %rax
! 	mov	%rax, (rp)
  L(ret):	FUNC_EXIT()
  	ret
--- 156,174 ----
  
  L(bc):	sub	$2, n
! 	jc	L(1)
  	ALIGN(16)
  1:	mov	(up), %rax
  	mov	8(up), %rcx
! dnl	lea	16(up), up
  	mov	%rax, (rp)
  	mov	%rcx, 8(rp)
! dnl	lea	16(rp), rp
! dnl	sub	$2, n
! dnl	jnc	1b
  
  	test	$1, R8(n)
  	jz	L(ret)
! 	mov	16(up), %rax
! 	mov	%rax, 16(rp)
  L(ret):	FUNC_EXIT()
  	ret



-- 
Torbjörn
Please encrypt, key id 0xC8601622