generic-library/isa-l
1
0
Fork 0
mirror of https://github.com/intel/isa-l.git synced 2025-01-06 15:10:02 +01:00
Code Issues Projects Releases Wiki Activity

erasure_code: check for size on gf_vect_mul_sse/avx

Browse Source 
gf_vect_mul requires length to be multiple of 32 bytes,
so this check is added in the SSE/AVX implementations.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This commit is contained in:
Pablo de Lara 2024-01-12 15:16:53 +00:00 committed by Tomasz Kantecki
parent 275977156d
commit 91e7906f3f
2 changed files with 19 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
erasure_code/gf_vect_mul_avx.asm
View File

@ -38,9 +38,6 @@
%define arg1 rsi %define arg1 rsi
%define arg2 rdx %define arg2 rdx
%define arg3 rcx %define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax %define return rax
%define func(x) x: endbranch %define func(x) x: endbranch
%define FUNC_SAVE %define FUNC_SAVE
@ -81,6 +78,7 @@
%define src arg2 %define src arg2
%define dest arg3 %define dest arg3
%define pos return %define pos return
%define tmp r11
;;; Use Non-temporal load/stor ;;; Use Non-temporal load/stor
@ -113,7 +111,14 @@ section .text
align 16 align 16
mk_global gf_vect_mul_avx, function mk_global gf_vect_mul_avx, function
func(gf_vect_mul_avx) func(gf_vect_mul_avx)
; Check if length is multiple of 32 bytes
mov tmp, len
and tmp, 0x1f
jnz return_fail
FUNC_SAVE FUNC_SAVE
mov pos, 0 mov pos, 0
vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte vmovdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
vmovdqu xgft_lo, [mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ... vmovdqu xgft_lo, [mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
@ -140,14 +145,13 @@ loop32:
XSTR [dest+pos-16], xtmp2b ;Store +16B result XSTR [dest+pos-16], xtmp2b ;Store +16B result
jl loop32 jl loop32
FUNC_RESTORE
return_pass: return_pass:
FUNC_RESTORE xor return, return
sub pos, len
ret ret
return_fail: return_fail:
FUNC_RESTORE
mov return, 1 mov return, 1
ret ret

15
erasure_code/gf_vect_mul_sse.asm
View File

@ -38,9 +38,6 @@
%define arg1 rsi %define arg1 rsi
%define arg2 rdx %define arg2 rdx
%define arg3 rcx %define arg3 rcx
%define arg4 r8
%define arg5 r9
%define tmp r11
%define return rax %define return rax
%define func(x) x: endbranch %define func(x) x: endbranch
%define FUNC_SAVE %define FUNC_SAVE
@ -81,6 +78,7 @@
%define src arg2 %define src arg2
%define dest arg3 %define dest arg3
%define pos return %define pos return
%define tmp r11
;;; Use Non-temporal load/stor ;;; Use Non-temporal load/stor
@ -114,7 +112,13 @@ section .text
align 16 align 16
mk_global gf_vect_mul_sse, function mk_global gf_vect_mul_sse, function
func(gf_vect_mul_sse) func(gf_vect_mul_sse)
; Check if length is multiple of 32 bytes
mov tmp, len
and tmp, 0x1f
jnz return_fail
FUNC_SAVE FUNC_SAVE
mov pos, 0 mov pos, 0
movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte movdqa xmask0f, [mask0f] ;Load mask of lower nibble in each byte
movdqu xgft_lo, [mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ... movdqu xgft_lo, [mul_array] ;Load array Cx{00}, Cx{01}, Cx{02}, ...
@ -147,15 +151,14 @@ loop32:
cmp pos, len cmp pos, len
jl loop32 jl loop32
FUNC_RESTORE
return_pass: return_pass:
sub pos, len xor return, return
FUNC_RESTORE
ret ret
return_fail: return_fail:
mov return, 1 mov return, 1
FUNC_RESTORE
ret ret
endproc_frame endproc_frame