updated gemv kernel for armv6

This commit is contained in:
Werner Saar 2016-01-24 16:31:19 +01:00
parent eaa63165df
commit ceaa931e48
3 changed files with 40 additions and 21 deletions

View File

@ -59,6 +59,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define I r12
#define FP_ZERO [fp, #-228]
#define FP_ZERO_0 [fp, #-228]
#define FP_ZERO_1 [fp, #-224]
#define M [fp, #-252 ]
#define A [fp, #-256 ]
@ -79,7 +83,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
pld [ YO , #Y_PRE ]
pld [ YO , #Y_PRE+32 ]
vsub.f64 d8 , d8 , d8
fldd d8 , FP_ZERO
vmov.f64 d9 , d8
vmov.f64 d10 , d8
vmov.f64 d11 , d8
@ -158,7 +162,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vsub.f64 d12 , d12 , d12
fldd d12 , FP_ZERO
.endm
@ -185,7 +189,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S4
vsub.f64 d12 , d12 , d12
fldd d12 , FP_ZERO
vmov.f64 d13 , d12
vmov.f64 d14 , d12
vmov.f64 d15 , d12
@ -245,7 +249,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vsub.f64 d12 , d12 , d12
fldd d12 , FP_ZERO
.endm
@ -279,7 +283,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
pld [ YO , #Y_PRE ]
vsub.f32 s8 , s8 , s8
flds s8 , FP_ZERO
vmov.f32 s9 , s8
vmov.f32 s10 , s8
vmov.f32 s11 , s8
@ -357,7 +361,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vsub.f32 s12 , s12 , s12
flds s12 , FP_ZERO
.endm
@ -384,7 +388,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S4
vsub.f32 s12 , s12 , s12
flds s12 , FP_ZERO
vmov.f32 s13 , s12
vmov.f32 s14 , s12
vmov.f32 s15 , s12
@ -445,7 +449,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vsub.f32 s12 , s12 , s12
flds s12 , FP_ZERO
.endm
@ -494,6 +498,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
vstm r12, { s8 - s15 } // store floating point registers
#endif
movs r12, #0
str r12, FP_ZERO
str r12, FP_ZERO_1
cmp OLD_M, #0
ble gemvn_kernel_L999

View File

@ -59,6 +59,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define I r12
#define FP_ZERO [fp, #-228]
#define FP_ZERO_0 [fp, #-228]
#define FP_ZERO_1 [fp, #-224]
#define N [fp, #-252 ]
#define A [fp, #-256 ]
@ -75,8 +79,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vsub.f64 d2 , d2 , d2
vsub.f64 d3 , d3 , d3
fldd d2, FP_ZERO
vmov.f64 d3 , d2
.endm
@ -123,7 +127,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vsub.f64 d2 , d2 , d2
fldd d2, FP_ZERO
vmov.f64 d3 , d2
.endm
@ -160,8 +165,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vsub.f64 d2 , d2 , d2
vsub.f64 d3 , d3 , d3
fldd d2, FP_ZERO
vmov.f64 d3 , d2
.endm
@ -224,7 +229,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vsub.f64 d2 , d2 , d2
fldd d2, FP_ZERO
vmov.f64 d3 , d2
.endm
@ -276,8 +282,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vsub.f32 s2 , s2 , s2
vsub.f32 s3 , s3 , s3
flds s2 , FP_ZERO
vmov.f32 s3 , s2
.endm
@ -321,7 +328,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vsub.f32 s2 , s2 , s2
flds s2 , FP_ZERO
.endm
@ -356,8 +363,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vsub.f32 s2 , s2 , s2
vsub.f32 s3 , s3 , s3
flds s2 , FP_ZERO
vmov.f32 s3 , s2
.endm
@ -418,7 +425,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vsub.f32 s2 , s2 , s2
flds s2 , FP_ZERO
.endm
@ -488,6 +495,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
vstm r12, { s8 - s15 } // store floating point registers
#endif
movs r12, #0
str r12, FP_ZERO
str r12, FP_ZERO_1
cmp M, #0
ble gemvt_kernel_L999

View File

@ -10,7 +10,7 @@ NEP: Data file for testing Nonsymmetric Eigenvalue Problem routines
0 5 7 3 200 Values of INIBL (nibble crossover point)
1 2 4 2 1 Values of ISHFTS (number of simultaneous shifts)
0 1 2 0 1 Values of IACC22 (select structured matrix multiply: 0, 1 or 2)
20.0 Threshold value
30.0 Threshold value
T Put T to test the error exits
1 Code to interpret the seed
NEP 21