generic-library/isa-l
1
0
Fork 0
mirror of https://github.com/intel/isa-l.git synced 2024-12-12 09:23:50 +01:00
Code Issues Projects Releases Wiki Activity

erasure_code: fix stack allocation

Browse Source 
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This commit is contained in:
Pablo de Lara 2023-11-17 08:42:28 +00:00 committed by Tomasz Kantecki
parent 801df41929
commit 07af4032ff
8 changed files with 117 additions and 144 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
erasure_code/gf_2vect_dot_prod_avx512.asm
View File

@ -73,7 +73,7 @@
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 9*16 + 5*8 ; must be an odd multiple of 8
%define stack_size 3*16 + 3*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
@ -82,16 +82,9 @@
vmovdqa [rsp + 0*16], xmm6
vmovdqa [rsp + 1*16], xmm7
vmovdqa [rsp + 2*16], xmm8
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
vmovdqa [rsp + 6*16], xmm12
vmovdqa [rsp + 7*16], xmm13
vmovdqa [rsp + 8*16], xmm14
save_reg r12, 9*16 + 0*8
save_reg r13, 9*16 + 1*8
save_reg r14, 9*16 + 2*8
save_reg r15, 9*16 + 3*8
save_reg r12, 3*16 + 0*8
save_reg r13, 3*16 + 1*8
save_reg r15, 3*16 + 2*8
end_prolog
mov arg4, arg(4)
%endmacro
@ -100,16 +93,9 @@
vmovdqa xmm6, [rsp + 0*16]
vmovdqa xmm7, [rsp + 1*16]
vmovdqa xmm8, [rsp + 2*16]
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
mov r12, [rsp + 9*16 + 0*8]
mov r13, [rsp + 9*16 + 1*8]
mov r14, [rsp + 9*16 + 2*8]
mov r15, [rsp + 9*16 + 3*8]
mov r12, [rsp + 3*16 + 0*8]
mov r13, [rsp + 3*16 + 1*8]
mov r15, [rsp + 3*16 + 2*8]
add rsp, stack_size
%endmacro
%endif

28
erasure_code/gf_3vect_dot_prod_avx512.asm
View File

@ -44,8 +44,6 @@
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
@ -73,15 +71,13 @@
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 9*16 + 5*8 ; must be an odd multiple of 8
%define stack_size 6*16 + 5*8 ; must be an odd multiple of 8
%define arg(x) [rsp + stack_size + PS + PS*x]
%define func(x) proc_frame x
@ -93,13 +89,10 @@
vmovdqa [rsp + 3*16], xmm9
vmovdqa [rsp + 4*16], xmm10
vmovdqa [rsp + 5*16], xmm11
vmovdqa [rsp + 6*16], xmm12
vmovdqa [rsp + 7*16], xmm13
vmovdqa [rsp + 8*16], xmm14
save_reg r12, 9*16 + 0*8
save_reg r13, 9*16 + 1*8
save_reg r14, 9*16 + 2*8
save_reg r15, 9*16 + 3*8
save_reg r12, 6*16 + 0*8
save_reg r13, 6*16 + 1*8
save_reg r14, 6*16 + 2*8
save_reg r15, 6*16 + 3*8
end_prolog
mov arg4, arg(4)
%endmacro
@ -111,13 +104,10 @@
vmovdqa xmm9, [rsp + 3*16]
vmovdqa xmm10, [rsp + 4*16]
vmovdqa xmm11, [rsp + 5*16]
vmovdqa xmm12, [rsp + 6*16]
vmovdqa xmm13, [rsp + 7*16]
vmovdqa xmm14, [rsp + 8*16]
mov r12, [rsp + 9*16 + 0*8]
mov r13, [rsp + 9*16 + 1*8]
mov r14, [rsp + 9*16 + 2*8]
mov r15, [rsp + 9*16 + 3*8]
mov r12, [rsp + 6*16 + 0*8]
mov r13, [rsp + 6*16 + 1*8]
mov r14, [rsp + 6*16 + 2*8]
mov r15, [rsp + 6*16 + 3*8]
add rsp, stack_size
%endmacro
%endif

23
erasure_code/gf_4vect_dot_prod_avx512.asm
View File

@ -44,8 +44,6 @@
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
@ -54,19 +52,22 @@
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 4*8
%define func(x) x: endbranch
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
sub rsp, stack_size
mov [rsp + 0*8], r12
mov [rsp + 1*8], r13
mov [rsp + 2*8], r14
mov [rsp + 3*8], r15
%endmacro
%macro FUNC_RESTORE 0
pop r15
pop r14
pop r13
pop r12
mov r12, [rsp + 0*8]
mov r13, [rsp + 1*8]
mov r14, [rsp + 2*8]
mov r15, [rsp + 3*8]
add rsp, stack_size
%endmacro
%endif
@ -79,8 +80,6 @@
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored

40
erasure_code/gf_4vect_mad_avx512.asm
View File

@ -64,16 +64,16 @@
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
vmovdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
@ -82,16 +82,16 @@
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
vmovdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size

31
erasure_code/gf_5vect_dot_prod_avx512.asm
View File

@ -44,8 +44,6 @@
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
@ -56,23 +54,26 @@
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 6*8
%define func(x) x: endbranch
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
push rbp
push rbx
sub rsp, stack_size
mov [rsp + 0*8], r12
mov [rsp + 1*8], r13
mov [rsp + 2*8], r14
mov [rsp + 3*8], r15
mov [rsp + 4*8], rbp
mov [rsp + 5*8], rbx
%endmacro
%macro FUNC_RESTORE 0
pop rbx
pop rbp
pop r15
pop r14
pop r13
pop r12
mov r12, [rsp + 0*8]
mov r13, [rsp + 1*8]
mov r14, [rsp + 2*8]
mov r15, [rsp + 3*8]
mov rbp, [rsp + 4*8]
mov rbx, [rsp + 5*8]
add rsp, stack_size
%endmacro
%endif
@ -85,8 +86,6 @@
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored

40
erasure_code/gf_5vect_mad_avx512.asm
View File

@ -66,16 +66,16 @@
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
vmovdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
end_prolog
@ -84,16 +84,16 @@
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
vmovdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
add rsp, stack_size

31
erasure_code/gf_6vect_dot_prod_avx512.asm
View File

@ -44,8 +44,6 @@
%define arg5 r9
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r12 ; must be saved and restored
@ -56,23 +54,26 @@
%define return rax
%define PS 8
%define LOG_PS 3
%define stack_size 6*8
%define func(x) x: endbranch
%macro FUNC_SAVE 0
push r12
push r13
push r14
push r15
push rbp
push rbx
sub rsp, stack_size
mov [rsp + 0*8], r12
mov [rsp + 1*8], r13
mov [rsp + 2*8], r14
mov [rsp + 3*8], r15
mov [rsp + 4*8], rbp
mov [rsp + 5*8], rbx
%endmacro
%macro FUNC_RESTORE 0
pop rbx
pop rbp
pop r15
pop r14
pop r13
pop r12
mov r12, [rsp + 0*8]
mov r13, [rsp + 1*8]
mov r14, [rsp + 2*8]
mov r15, [rsp + 3*8]
mov rbp, [rsp + 4*8]
mov rbx, [rsp + 5*8]
add rsp, stack_size
%endmacro
%endif
@ -85,8 +86,6 @@
%define arg4 r12 ; must be saved, loaded and restored
%define arg5 r15 ; must be saved and restored
%define tmp r11
%define tmp.w r11d
%define tmp.b r11b
%define tmp2 r10
%define tmp3 r13 ; must be saved and restored
%define tmp4 r14 ; must be saved and restored

40
erasure_code/gf_6vect_mad_avx512.asm
View File

@ -72,16 +72,16 @@
%macro FUNC_SAVE 0
sub rsp, stack_size
movdqa [rsp+16*0],xmm6
movdqa [rsp+16*1],xmm7
movdqa [rsp+16*2],xmm8
movdqa [rsp+16*3],xmm9
movdqa [rsp+16*4],xmm10
movdqa [rsp+16*5],xmm11
movdqa [rsp+16*6],xmm12
movdqa [rsp+16*7],xmm13
movdqa [rsp+16*8],xmm14
movdqa [rsp+16*9],xmm15
vmovdqa [rsp+16*0],xmm6
vmovdqa [rsp+16*1],xmm7
vmovdqa [rsp+16*2],xmm8
vmovdqa [rsp+16*3],xmm9
vmovdqa [rsp+16*4],xmm10
vmovdqa [rsp+16*5],xmm11
vmovdqa [rsp+16*6],xmm12
vmovdqa [rsp+16*7],xmm13
vmovdqa [rsp+16*8],xmm14
vmovdqa [rsp+16*9],xmm15
save_reg r12, 10*16 + 0*8
save_reg r15, 10*16 + 1*8
save_reg r13, 10*16 + 2*8
@ -91,16 +91,16 @@
%endmacro
%macro FUNC_RESTORE 0
movdqa xmm6, [rsp+16*0]
movdqa xmm7, [rsp+16*1]
movdqa xmm8, [rsp+16*2]
movdqa xmm9, [rsp+16*3]
movdqa xmm10, [rsp+16*4]
movdqa xmm11, [rsp+16*5]
movdqa xmm12, [rsp+16*6]
movdqa xmm13, [rsp+16*7]
movdqa xmm14, [rsp+16*8]
movdqa xmm15, [rsp+16*9]
vmovdqa xmm6, [rsp+16*0]
vmovdqa xmm7, [rsp+16*1]
vmovdqa xmm8, [rsp+16*2]
vmovdqa xmm9, [rsp+16*3]
vmovdqa xmm10, [rsp+16*4]
vmovdqa xmm11, [rsp+16*5]
vmovdqa xmm12, [rsp+16*6]
vmovdqa xmm13, [rsp+16*7]
vmovdqa xmm14, [rsp+16*8]
vmovdqa xmm15, [rsp+16*9]
mov r12, [rsp + 10*16 + 0*8]
mov r15, [rsp + 10*16 + 1*8]
mov r13, [rsp + 10*16 + 2*8]