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

Tag arguments 0 and 1 as both input and output

This commit is contained in:
Martin Kroeker 2019-01-16 23:30:03 +01:00 committed by GitHub
parent 63cdd8f4a0
commit b6136be686
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 15 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
kernel/x86_64/cscal_microk_haswell-2.c
View File

@ -116,11 +116,11 @@ static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x)
"vzeroupper \n\t" "vzeroupper \n\t"
: :
: "+r" (n), // 0
"r" (n), // 0 "+r" (x), // 1
"r" (x), // 1 :
"r" (alpha) // 2 "r" (alpha) // 2
: "cc", //"0", "1", : "cc",
"%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7", "%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11", "%xmm8", "%xmm9", "%xmm10", "%xmm11",
@ -208,9 +208,9 @@ static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
"vzeroupper \n\t" "vzeroupper \n\t"
: :
: "+r" (n), // 0
"r" (n), // 0 "+r" (x), // 1
"r" (x), // 1 :
"r" (alpha) // 2 "r" (alpha) // 2
: "cc", // "0", "1", : "cc", // "0", "1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm0", "%xmm1", "%xmm2", "%xmm3",
@ -285,9 +285,9 @@ static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
"vzeroupper \n\t" "vzeroupper \n\t"
: :
: "+r" (n), // 0
"r" (n), // 0 "+r" (x), // 1
"r" (x), // 1 :
"r" (alpha) // 2 "r" (alpha) // 2
: "cc", //"%0", "%1", : "cc", //"%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm0", "%xmm1", "%xmm2", "%xmm3",
@ -329,12 +329,12 @@ static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
"vzeroupper \n\t" "vzeroupper \n\t"
: :
: "+r" (n), // 0
"r" (n), // 0 "+r" (x), // 1
"r" (x), // 1 :
"r" (alpha) // 2 "r" (alpha) // 2
: "cc", //"0", "1", : "cc",
"%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7", "%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11", "%xmm8", "%xmm9", "%xmm10", "%xmm11",