floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix declaration of assembly arguments in SSYMV and DSYMV microkernels 

Arguments 0 and 1 are both input and output
This commit is contained in:
Martin Kroeker 2019-02-12 16:00:18 +01:00 committed by GitHub
parent 91481a3e4e
commit b824fa70eb
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
8 changed files with 24 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
kernel/x86_64/dsymv_U_microk_bulldozer-2.c
View File

@ -106,9 +106,9 @@ static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"vmovsd %%xmm3 ,24(%9) \n\t" // save temp2 "vmovsd %%xmm3 ,24(%9) \n\t" // save temp2
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/dsymv_U_microk_haswell-2.c
View File

@ -107,9 +107,9 @@ static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"vzeroupper \n\t" "vzeroupper \n\t"
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/dsymv_U_microk_nehalem-2.c
View File

@ -101,9 +101,9 @@ static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"movsd %%xmm3 , 24(%9) \n\t" // save temp2 "movsd %%xmm3 , 24(%9) \n\t" // save temp2
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/dsymv_U_microk_sandy-2.c
View File

@ -116,9 +116,9 @@ static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"vzeroupper \n\t" "vzeroupper \n\t"
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/ssymv_U_microk_bulldozer-2.c
View File

@ -90,9 +90,9 @@ static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"vmovss %%xmm3 ,12(%9) \n\t" // save temp2 "vmovss %%xmm3 ,12(%9) \n\t" // save temp2
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/ssymv_U_microk_haswell-2.c
View File

@ -112,9 +112,9 @@ static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"vzeroupper \n\t" "vzeroupper \n\t"
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/ssymv_U_microk_nehalem-2.c
View File

@ -106,9 +106,9 @@ static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"movss %%xmm3 , 12(%9) \n\t" // save temp2 "movss %%xmm3 , 12(%9) \n\t" // save temp2
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4

4
kernel/x86_64/ssymv_U_microk_sandy-2.c
View File

@ -120,9 +120,9 @@ static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT
"vzeroupper \n\t" "vzeroupper \n\t"
: :
"+r" (i), // 0
"+r" (n) // 1
: :
"r" (i), // 0
"r" (n), // 1
"r" (x), // 2 "r" (x), // 2
"r" (y), // 3 "r" (y), // 3
"r" (a0), // 4 "r" (a0), // 4