#736 Revert #733 patch to fix bus error on ARM.

This commit is contained in:
Zhang Xianyi 2016-01-12 22:19:58 +00:00
parent 708ad330ac
commit 3602a2cd1f
32 changed files with 241 additions and 282 deletions

View File

@ -368,11 +368,11 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.align 5
#if defined(DOUBLE)
vldr.f64 d0 , =0
vldr.f64 d1 , =0
vsub.f64 d0 , d0 , d0
vsub.f64 d1 , d1 , d1
#else
vldr.f32 s0 , =0
vldr.f32 s1 , =0
vsub.f32 s0 , s0 , s0
vsub.f32 s1 , s1 , s1
#endif
cmp N, #0

View File

@ -188,10 +188,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
mov Y, OLD_Y
ldr INC_Y, OLD_INC_Y
vldr.f32 s0 , =0
vldr.f32 s1 , =0
vldr.f32 s2 , =0
vldr.f32 s3 , =0
vsub.f32 s0 , s0 , s0
vsub.f32 s1 , s1 , s1
vsub.f32 s2 , s2 , s2
vsub.f32 s3 , s3 , s3
cmp N, #0
ble cdot_kernel_L999

View File

@ -138,7 +138,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -340,7 +340,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s12, s8
vmov.f32 s13, s8
@ -514,7 +514,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -681,7 +681,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
.endm

View File

@ -147,6 +147,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -367,6 +368,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s20, s16
vmov.f32 s21, s16
@ -548,6 +550,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -727,6 +730,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s24, s16
vmov.f32 s25, s16
@ -1004,12 +1008,9 @@ cgemm_kernel_L2_M2_32:
b cgemm_kernel_L2_M2_44
cgemm_f32zero:
.word 0x00000000
cgemm_kernel_L2_M2_40:
vldr.f32 s16 , cgemm_f32zero
INIT2x2
@ -1043,7 +1044,6 @@ cgemm_kernel_L2_M1_BEGIN:
cgemm_kernel_L2_M1_20:
vldr.f32 s16 , cgemm_f32zero
INIT1x2
mov BO, BC
@ -1219,7 +1219,6 @@ cgemm_kernel_L1_M2_32:
cgemm_kernel_L1_M2_40:
vldr.f32 s16 , =0
INIT2x1
@ -1253,7 +1252,6 @@ cgemm_kernel_L1_M1_BEGIN:
cgemm_kernel_L1_M1_20:
vldr.f32 s16 , =0
INIT1x1
mov BO, BC

View File

@ -117,7 +117,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F4
pld [ YO, #Y_PRE ]
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -220,7 +220,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
.endm
@ -267,7 +267,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S4
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -384,7 +384,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
.endm

View File

@ -116,10 +116,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vldr.f32 s12, =0
vldr.f32 s13, =0
vldr.f32 s14, =0
vldr.f32 s15, =0
vsub.f32 s12, s12, s12
vsub.f32 s13, s13, s13
vsub.f32 s14, s14, s14
vsub.f32 s15, s15, s15
.endm
@ -172,8 +172,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f32 s12, =0
vldr.f32 s13, =0
vsub.f32 s12, s12, s12
vsub.f32 s13, s13, s13
.endm
@ -215,10 +215,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vldr.f32 s12, =0
vldr.f32 s13, =0
vldr.f32 s14, =0
vldr.f32 s15, =0
vsub.f32 s12, s12, s12
vsub.f32 s13, s13, s13
vsub.f32 s14, s14, s14
vsub.f32 s15, s15, s15
.endm
@ -281,8 +281,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f32 s12, =0
vldr.f32 s13, =0
vsub.f32 s12, s12, s12
vsub.f32 s13, s13, s13
.endm

View File

@ -136,7 +136,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -301,10 +301,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
flds s0, ALPHA_R
flds s1, ALPHA_I
vldr.f32 s4, =0
vldr.f32 s5, =0
vldr.f32 s6, =0
vldr.f32 s7, =0
vsub.f32 s4, s4, s4
vsub.f32 s5, s5, s5
vsub.f32 s6, s6, s6
vsub.f32 s7, s7, s7
FMAC_R1 s4 , s0 , s8
FMAC_I1 s5 , s0 , s9
@ -318,10 +318,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fstmias CO1, { s4 - s7 }
vldr.f32 s4, =0
vldr.f32 s5, =0
vldr.f32 s6, =0
vldr.f32 s7, =0
vsub.f32 s4, s4, s4
vsub.f32 s5, s5, s5
vsub.f32 s6, s6, s6
vsub.f32 s7, s7, s7
FMAC_R1 s4 , s0 , s12
FMAC_I1 s5 , s0 , s13
@ -343,7 +343,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s12, s8
vmov.f32 s13, s8
@ -490,8 +490,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
flds s0, ALPHA_R
flds s1, ALPHA_I
vldr.f32 s4, =0
vldr.f32 s5, =0
vsub.f32 s4, s4, s4
vsub.f32 s5, s5, s5
FMAC_R1 s4 , s0 , s8
FMAC_I1 s5 , s0 , s9
@ -500,8 +500,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fstmias CO1, { s4 - s5 }
vldr.f32 s4, =0
vldr.f32 s5, =0
vsub.f32 s4, s4, s4
vsub.f32 s5, s5, s5
FMAC_R1 s4 , s0 , s12
FMAC_I1 s5 , s0 , s13
@ -519,7 +519,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -663,10 +663,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
flds s0, ALPHA_R
flds s1, ALPHA_I
vldr.f32 s4, =0
vldr.f32 s5, =0
vldr.f32 s6, =0
vldr.f32 s7, =0
vsub.f32 s4, s4, s4
vsub.f32 s5, s5, s5
vsub.f32 s6, s6, s6
vsub.f32 s7, s7, s7
FMAC_R1 s4 , s0 , s8
FMAC_I1 s5 , s0 , s9
@ -689,7 +689,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
.endm
@ -795,8 +795,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
flds s0, ALPHA_R
flds s1, ALPHA_I
vldr.f32 s4, =0
vldr.f32 s5, =0
vsub.f32 s4, s4, s4
vsub.f32 s5, s5, s5
FMAC_R1 s4 , s0 , s8
FMAC_I1 s5 , s0 , s9

View File

@ -134,6 +134,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -350,6 +351,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s20, s16
vmov.f32 s21, s16
@ -527,6 +529,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -703,6 +706,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s24, s16
vmov.f32 s25, s16
@ -1016,12 +1020,9 @@ _L2_M2_32:
b _L2_M2_44
ctrmm_f32zero:
.word 0x00000000
_L2_M2_40:
vldr.f32 s16 , ctrmm_f32zero
INIT2x2
@ -1073,7 +1074,6 @@ _L2_M1_BEGIN:
_L2_M1_20:
vldr.f32 s16 , ctrmm_f32zero
INIT1x2
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1337,7 +1337,6 @@ _L1_M2_32:
_L1_M2_40:
vldr.f32 s16 , =0
INIT2x1
@ -1390,7 +1389,6 @@ _L1_M1_BEGIN:
_L1_M1_20:
vldr.f32 s16 , =0
INIT1x1
#if (defined(LEFT) && defined(TRANSA)) || \

View File

@ -152,8 +152,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
mov Y, OLD_Y
ldr INC_Y, OLD_INC_Y
vldr.f64 d0 , =0
vldr.f64 d1 , =0
vsub.f64 d0 , d0 , d0
vsub.f64 d1 , d1 , d1
cmp N, #0
ble ddot_kernel_L999

View File

@ -85,7 +85,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9, d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -173,7 +173,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9, d8
vmov.f64 d12, d8
vmov.f64 d13, d8
@ -233,7 +233,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d12, d8
.endm
@ -283,7 +283,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9, d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -338,7 +338,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
.endm
@ -380,7 +380,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
.endm

View File

@ -102,7 +102,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x4
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -376,7 +376,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x4
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d20, d16
vmov.f64 d21, d16
@ -470,7 +470,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x4
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d20, d16
vmov.f64 d24, d16
vmov.f64 d28, d16
@ -533,7 +533,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -617,7 +617,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d20, d16
vmov.f64 d21, d16
@ -678,7 +678,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d20, d16
.endm
@ -723,7 +723,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -782,7 +782,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
.endm
@ -826,7 +826,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
.endm

View File

@ -90,7 +90,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9, d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -165,7 +165,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9, d8
vmov.f64 d12, d8
vmov.f64 d13, d8
@ -220,7 +220,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d12, d8
.endm
@ -268,7 +268,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9, d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -318,7 +318,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
.endm
@ -357,7 +357,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
.endm

View File

@ -89,7 +89,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x4
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -386,7 +386,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x4
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d20, d16
vmov.f64 d21, d16
@ -468,7 +468,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x4
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d20, d16
vmov.f64 d24, d16
vmov.f64 d28, d16
@ -527,7 +527,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -601,7 +601,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d20, d16
vmov.f64 d21, d16
@ -656,7 +656,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d20, d16
.endm
@ -699,7 +699,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -753,7 +753,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
.endm
@ -794,7 +794,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
.endm

View File

@ -79,7 +79,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
pld [ YO , #Y_PRE ]
pld [ YO , #Y_PRE+32 ]
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10 , d8
vmov.f64 d11 , d8
@ -158,7 +158,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f64 d12 , =0
vsub.f64 d12 , d12 , d12
.endm
@ -185,7 +185,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S4
vldr.f64 d12 , =0
vsub.f64 d12 , d12 , d12
vmov.f64 d13 , d12
vmov.f64 d14 , d12
vmov.f64 d15 , d12
@ -245,7 +245,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f64 d12 , =0
vsub.f64 d12 , d12 , d12
.endm
@ -279,7 +279,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
pld [ YO , #Y_PRE ]
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
vmov.f32 s10 , s8
vmov.f32 s11 , s8
@ -357,7 +357,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f32 s12 , =0
vsub.f32 s12 , s12 , s12
.endm
@ -384,7 +384,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S4
vldr.f32 s12 , =0
vsub.f32 s12 , s12 , s12
vmov.f32 s13 , s12
vmov.f32 s14 , s12
vmov.f32 s15 , s12
@ -445,7 +445,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f32 s12 , =0
vsub.f32 s12 , s12 , s12
.endm

View File

@ -79,7 +79,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
pld [ YO , #Y_PRE ]
pld [ YO , #Y_PRE+32 ]
vldr.f64 d24 , =0
vsub.f64 d24 , d24 , d24
vmov.f64 d25 , d24
vmov.f64 d26 , d24
vmov.f64 d27 , d24
@ -147,7 +147,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f64 d24 , =0
vsub.f64 d24 , d24 , d24
.endm
@ -175,7 +175,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S8
vldr.f64 d24 , =0
vsub.f64 d24 , d24 , d24
vmov.f64 d25 , d24
vmov.f64 d26 , d24
vmov.f64 d27 , d24
@ -269,7 +269,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f64 d24 , =0
vsub.f64 d24 , d24 , d24
.endm
@ -302,7 +302,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
pld [ YO , #Y_PRE ]
vldr.f32 s24 , =0
vsub.f32 s24 , s24 , s24
vmov.f32 s25 , s24
vmov.f32 s26 , s24
vmov.f32 s27 , s24
@ -368,7 +368,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f32 s24 , =0
vsub.f32 s24 , s24 , s24
.endm
@ -396,7 +396,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S8
vldr.f32 s24 , =0
vsub.f32 s24 , s24 , s24
vmov.f32 s25 , s24
vmov.f32 s26 , s24
vmov.f32 s27 , s24
@ -489,7 +489,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f32 s24 , =0
vsub.f32 s24 , s24 , s24
.endm

View File

@ -75,8 +75,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vldr.f64 d2 , =0
vldr.f64 d3 , =0
vsub.f64 d2 , d2 , d2
vsub.f64 d3 , d3 , d3
.endm
@ -123,7 +123,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f64 d2 , =0
vsub.f64 d2 , d2 , d2
.endm
@ -160,8 +160,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vldr.f64 d2 , =0
vldr.f64 d3 , =0
vsub.f64 d2 , d2 , d2
vsub.f64 d3 , d3 , d3
.endm
@ -224,7 +224,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f64 d2 , =0
vsub.f64 d2 , d2 , d2
.endm
@ -276,8 +276,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vldr.f32 s2 , =0
vldr.f32 s3 , =0
vsub.f32 s2 , s2 , s2
vsub.f32 s3 , s3 , s3
.endm
@ -321,7 +321,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f32 s2 , =0
vsub.f32 s2 , s2 , s2
.endm
@ -356,8 +356,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vldr.f32 s2 , =0
vldr.f32 s3 , =0
vsub.f32 s2 , s2 , s2
vsub.f32 s3 , s3 , s3
.endm
@ -418,7 +418,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f32 s2 , =0
vsub.f32 s2 , s2 , s2
.endm

View File

@ -75,8 +75,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vldr.f64 d4 , =0
vldr.f64 d5 , =0
vsub.f64 d4 , d4 , d4
vsub.f64 d5 , d5 , d5
.endm
@ -123,8 +123,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vldr.f64 d4 , =0
vldr.f64 d5 , =0
vsub.f64 d4 , d4 , d4
vsub.f64 d5 , d5 , d5
.endm
@ -183,7 +183,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f64 d4 , =0
vsub.f64 d4 , d4 , d4
.endm
@ -220,7 +220,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f64 d4 , =0
vsub.f64 d4 , d4 , d4
.endm
@ -268,8 +268,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vldr.f32 s4 , =0
vldr.f32 s5 , =0
vsub.f32 s4 , s4 , s4
vsub.f32 s5 , s5 , s5
.endm
@ -313,8 +313,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vldr.f32 s4 , =0
vldr.f32 s5 , =0
vsub.f32 s4 , s4 , s4
vsub.f32 s5 , s5 , s5
.endm
@ -371,7 +371,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f32 s4 , =0
vsub.f32 s4 , s4 , s4
.endm
@ -406,7 +406,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f32 s4 , =0
vsub.f32 s4 , s4 , s4
.endm

View File

@ -342,9 +342,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
push {r4}
#if defined(DOUBLE)
vldr.f64 d0 , =0
vsub.f64 d0 , d0 , d0
#else
vldr.f32 s0 , =0
vsub.f32 s0 , s0 , s0
#endif
mov INDEX, #0

View File

@ -446,12 +446,12 @@ nrm2_begin:
#if defined(COMPLEX)
#if defined(DOUBLE)
vldr.f64 d0 , =0 // scale=0.0
vsub.f64 d0 , d0 , d0 // scale=0.0
vldr.64 d1 , znrm2_one // ssq=1.0
vmov.f64 d7 , d1 // value 1.0
vmov.f64 d6 , d0 // value 0.0
#else
vldr.f32 s0 , =0 // scale=0.0
vsub.f32 s0 , s0 , s0 // scale=0.0
vldr.32 s1 , cnrm2_one // ssq=1.0
vmov.f32 s7 , s1 // value 1.0
vmov.f32 s6 , s0 // value 0.0
@ -460,12 +460,12 @@ nrm2_begin:
#else
#if defined(DOUBLE)
vldr.f64 d0 , =0 // scale=0.0
vsub.f64 d0 , d0 , d0 // scale=0.0
vldr.64 d1 , dnrm2_one // ssq=1.0
vmov.f64 d7 , d1 // value 1.0
vmov.f64 d6 , d0 // value 0.0
#else
vldr.f32 s0 , =0 // scale=0.0
vsub.f32 s0 , s0 , s0 // scale=0.0
vldr.32 s1 , snrm2_one // ssq=1.0
vmov.f32 s7 , s1 // value 1.0
vmov.f32 s6 , s0 // value 0.0

View File

@ -400,22 +400,17 @@ KERNEL_S1_END_\@:
* End of macro definitions
**************************************************************************************/
nrm2_zeros:
.align 5
.word 0x00000000
.word 0x00000000
PROLOGUE
.align 5
#if defined(DOUBLE)
vldr.f64 d0 , nrm2_zeros // scale=0.0
vsub.f64 d0 , d0 , d0 // scale=0.0
vmov.f64 d1 , #1.0 // ssq=1.0
vmov.f64 d7 , d1 // value 1.0
vmov.f64 d6 , d0 // value 0.0
#else
vldr.f32 s0 , nrm2_zeros // scale=0.0
vsub.f32 s0 , s0 , s0 // scale=0.0
vmov.f32 s1 , #1.0 // ssq=1.0
vmov.f32 s7 , s1 // value 1.0
vmov.f32 s6 , s0 // value 0.0

View File

@ -242,13 +242,13 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#if defined(DSDOT)
vldr.f64 d0 , =0
vldr.f64 d1 , =0
vsub.f64 d0 , d0 , d0
vsub.f64 d1 , d1 , d1
#else
vldr.f32 s0 , =0
vldr.f32 s1 , =0
vsub.f32 s0 , s0 , s0
vsub.f32 s1 , s1 , s1
#endif

View File

@ -85,7 +85,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9, s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -161,7 +161,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9, s8
vmov.f32 s12, s8
vmov.f32 s13, s8
@ -221,7 +221,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s12, s8
.endm
@ -271,7 +271,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9, s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -326,7 +326,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
.endm
@ -368,7 +368,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
.endm

View File

@ -102,6 +102,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x4
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -348,6 +349,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x4
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s20, s16
vmov.f32 s21, s16
@ -441,6 +443,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x4
vsub.f32 s16 , s16 , s16
vmov.f32 s20, s16
vmov.f32 s24, s16
vmov.f32 s28, s16
@ -503,6 +506,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -586,6 +590,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s20, s16
vmov.f32 s21, s16
@ -646,6 +651,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vsub.f32 s16 , s16 , s16
vmov.f32 s20, s16
.endm
@ -690,6 +696,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -748,6 +755,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
.endm
@ -789,6 +797,13 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/******************************************************************************/
.macro INIT1x1
vsub.f32 s16 , s16 , s16
.endm
.macro KERNEL1x1_SUB
@ -915,24 +930,9 @@ sgemm_kernel_L4_M4_32:
b sgemm_kernel_L4_M4_44
/* Note on loading zero into a fp register
* vsub doesn't work since it cannot handle NaN and infinity
* vmov.Fn doesn't work with 0
* vmov.In and veor are neon
* vldr , =0 doesn't work since the function is larger than 2KB
* and the assembler is not able to insert constant pool inside
* the function body.
*
* Therefore, the best way I've found is to manually create multiple
* copies of the zero constant and `vldr` from different ones depending
* on where the instruction is.
*/
sgemm_f32zero:
.word 0x00000000
sgemm_kernel_L4_M4_40:
vldr.f32 s16 , sgemm_f32zero
INIT4x4
@ -969,7 +969,6 @@ sgemm_kernel_L4_M2_BEGIN:
sgemm_kernel_L4_M2_20:
vldr.f32 s16 , sgemm_f32zero
INIT2x4
mov BO, BC
@ -1015,14 +1014,9 @@ sgemm_kernel_L4_M1_BEGIN:
tst I, #1 // I = I % 2
ble sgemm_kernel_L4_END
b sgemm_kernel_L4_M1_20
sgemm_f32zero4:
.word 0x00000000
sgemm_kernel_L4_M1_20:
vldr.f32 s16 , sgemm_f32zero4
INIT1x4
mov BO, BC
@ -1106,7 +1100,6 @@ sgemm_kernel_L2_M4_BEGIN:
sgemm_kernel_L2_M4_20:
vldr.f32 s16 , sgemm_f32zero3
INIT4x2
mov BO, BC
@ -1128,6 +1121,7 @@ sgemm_kernel_L2_M4_22:
subs L, L, #1
bgt sgemm_kernel_L2_M4_22
sgemm_kernel_L2_M4_40:
ands L , K1, #7 // L = L % 8
@ -1148,10 +1142,7 @@ sgemm_kernel_L2_M4_END:
subs I, I, #1
bgt sgemm_kernel_L2_M4_20
b sgemm_kernel_L2_M2_BEGIN
sgemm_f32zero3:
.word 0x00000000
sgemm_kernel_L2_M2_BEGIN:
@ -1164,7 +1155,6 @@ sgemm_kernel_L2_M2_BEGIN:
sgemm_kernel_L2_M2_20:
vldr.f32 s16 , sgemm_f32zero3
INIT2x2
mov BO, BC
@ -1213,7 +1203,6 @@ sgemm_kernel_L2_M1_BEGIN:
sgemm_kernel_L2_M1_20:
vldr.f32 s16 , sgemm_f32zero3
INIT1x2
mov BO, BC
@ -1289,7 +1278,6 @@ sgemm_kernel_L1_M4_BEGIN:
sgemm_kernel_L1_M4_20:
vldr.f32 s16 , sgemm_f32zero3
INIT4x1
mov BO, BC
@ -1345,7 +1333,6 @@ sgemm_kernel_L1_M2_BEGIN:
sgemm_kernel_L1_M2_20:
vldr.f32 s16 , sgemm_f32zero2
INIT2x1
mov BO, BC
@ -1394,7 +1381,7 @@ sgemm_kernel_L1_M1_BEGIN:
sgemm_kernel_L1_M1_20:
vldr.f32 s16 , sgemm_f32zero2
INIT1x1
mov BO, BC
asrs L , K1, #3 // L = L / 8
@ -1447,5 +1434,3 @@ sgemm_kernel_L999:
EPILOGUE
sgemm_f32zero2:
.word 0x00000000

View File

@ -90,7 +90,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9, s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -156,7 +156,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9, s8
vmov.f32 s12, s8
vmov.f32 s13, s8
@ -211,7 +211,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s12, s8
.endm
@ -259,7 +259,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9, s8
vmov.f32 s10, s8
vmov.f32 s11, s8
@ -309,7 +309,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
vmov.f32 s9 , s8
.endm
@ -348,7 +348,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f32 s8 , =0
vsub.f32 s8 , s8 , s8
.endm

View File

@ -88,6 +88,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x4
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -321,6 +322,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x4
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s20, s16
vmov.f32 s21, s16
@ -403,6 +405,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x4
vsub.f32 s16 , s16 , s16
vmov.f32 s20, s16
vmov.f32 s24, s16
vmov.f32 s28, s16
@ -461,6 +464,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -534,6 +538,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s20, s16
vmov.f32 s21, s16
@ -588,6 +593,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vsub.f32 s16 , s16 , s16
vmov.f32 s20, s16
.endm
@ -630,6 +636,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT4x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
vmov.f32 s18, s16
vmov.f32 s19, s16
@ -683,6 +690,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vsub.f32 s16 , s16 , s16
vmov.f32 s17, s16
.endm
@ -721,6 +729,13 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/******************************************************************************/
.macro INIT1x1
vsub.f32 s16 , s16 , s16
.endm
.macro KERNEL1x1_SUB
@ -940,24 +955,9 @@ _L4_M4_32:
b _L4_M4_44
/* Note on loading zero into a fp register
* vsub doesn't work since it cannot handle NaN and infinity
* vmov.Fn doesn't work with 0
* vmov.In and veor are neon
* vldr , =0 doesn't work since the function is larger than 2KB
* and the assembler is not able to insert constant pool inside
* the function body.
*
* Therefore, the best way I've found is to manually create multiple
* copies of the zero constant and `vldr` from different ones depending
* on where the instruction is.
*/
strmm_f32zero:
.word 0x00000000
_L4_M4_40:
vldr.f32 s16 , strmm_f32zero
INIT4x4
@ -1014,7 +1014,6 @@ _L4_M2_BEGIN:
_L4_M2_20:
vldr.f32 s16 , strmm_f32zero
INIT2x4
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1113,7 +1112,6 @@ _L4_M1_BEGIN:
_L4_M1_20:
vldr.f32 s16 , strmm_f32zero3
INIT1x4
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1254,14 +1252,9 @@ _L2_M4_BEGIN:
ldr I, M
asrs I, I, #2 // I = I / 4
ble _L2_M2_BEGIN
b _L2_M4_20
strmm_f32zero3:
.word 0x00000000
_L2_M4_20:
vldr.f32 s16 , strmm_f32zero3
INIT4x2
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1367,7 +1360,6 @@ _L2_M2_BEGIN:
_L2_M2_20:
vldr.f32 s16 , strmm_f32zero3
INIT2x2
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1466,7 +1458,6 @@ _L2_M1_BEGIN:
_L2_M1_20:
vldr.f32 s16 , strmm_f32zero4
INIT1x2
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1600,14 +1591,9 @@ _L1_M4_BEGIN:
ldr I, M
asrs I, I, #2 // I = I / 4
ble _L1_M2_BEGIN
b _L1_M4_20
strmm_f32zero4:
.word 0x00000000
_L1_M4_20:
vldr.f32 s16 , strmm_f32zero4
INIT4x1
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1714,7 +1700,6 @@ _L1_M2_BEGIN:
_L1_M2_20:
vldr.f32 s16 , strmm_f32zero2
INIT2x1
#if (defined(LEFT) && defined(TRANSA)) || \
@ -1813,7 +1798,7 @@ _L1_M1_BEGIN:
_L1_M1_20:
vldr.f32 s16 , strmm_f32zero2
INIT1x1
#if (defined(LEFT) && defined(TRANSA)) || \
(!defined(LEFT) && !defined(TRANSA))
@ -1897,5 +1882,3 @@ _L999:
EPILOGUE
strmm_f32zero2:
.word 0x00000000

View File

@ -190,10 +190,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
mov Y, OLD_Y
ldr INC_Y, OLD_INC_Y
vldr.f64 d0 , =0
vldr.f64 d1 , =0
vldr.f64 d2 , =0
vldr.f64 d3 , =0
vsub.f64 d0 , d0 , d0
vsub.f64 d1 , d1 , d1
vsub.f64 d2 , d2 , d2
vsub.f64 d3 , d3 , d3
cmp N, #0
ble zdot_kernel_L999

View File

@ -131,7 +131,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -383,7 +383,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d12, d8
vmov.f64 d13, d8
@ -557,7 +557,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -724,7 +724,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
.endm

View File

@ -147,7 +147,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -404,7 +404,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d20, d16
vmov.f64 d21, d16
@ -586,7 +586,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -766,7 +766,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d24, d16
vmov.f64 d25, d16

View File

@ -117,7 +117,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F4
pld [ YO, #Y_PRE ]
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -222,7 +222,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
.endm
@ -269,7 +269,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S4
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -386,7 +386,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
.endm

View File

@ -117,10 +117,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F2
vldr.f64 d12, =0
vldr.f64 d13, =0
vldr.f64 d14, =0
vldr.f64 d15, =0
vsub.f64 d12, d12, d12
vsub.f64 d13, d13, d13
vsub.f64 d14, d14, d14
vsub.f64 d15, d15, d15
.endm
@ -173,8 +173,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_F1
vldr.f64 d12, =0
vldr.f64 d13, =0
vsub.f64 d12, d12, d12
vsub.f64 d13, d13, d13
.endm
@ -216,10 +216,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S2
vldr.f64 d12, =0
vldr.f64 d13, =0
vldr.f64 d14, =0
vldr.f64 d15, =0
vsub.f64 d12, d12, d12
vsub.f64 d13, d13, d13
vsub.f64 d14, d14, d14
vsub.f64 d15, d15, d15
.endm
@ -282,8 +282,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT_S1
vldr.f64 d12, =0
vldr.f64 d13, =0
vsub.f64 d12, d12, d12
vsub.f64 d13, d13, d13
.endm

View File

@ -140,7 +140,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -356,10 +356,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fldd d0, ALPHA_R
fldd d1, ALPHA_I
vldr.f64 d4, =0
vldr.f64 d5, =0
vldr.f64 d6, =0
vldr.f64 d7, =0
vsub.f64 d4, d4 , d4
vsub.f64 d5, d5 , d5
vsub.f64 d6, d6 , d6
vsub.f64 d7, d7 , d7
FMAC_R1 d4 , d0 , d8
FMAC_I1 d5 , d0 , d9
@ -373,10 +373,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fstmiad CO1, { d4 - d7 }
vldr.f64 d4, =0
vldr.f64 d5, =0
vldr.f64 d6, =0
vldr.f64 d7, =0
vsub.f64 d4, d4 , d4
vsub.f64 d5, d5 , d5
vsub.f64 d6, d6 , d6
vsub.f64 d7, d7 , d7
FMAC_R1 d4 , d0 , d12
FMAC_I1 d5 , d0 , d13
@ -398,7 +398,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d12, d8
vmov.f64 d13, d8
@ -545,8 +545,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fldd d0, ALPHA_R
fldd d1, ALPHA_I
vldr.f64 d4, =0
vldr.f64 d5, =0
vsub.f64 d4, d4 , d4
vsub.f64 d5, d5 , d5
FMAC_R1 d4 , d0 , d8
FMAC_I1 d5 , d0 , d9
@ -555,8 +555,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fstmiad CO1, { d4 - d5 }
vldr.f64 d4, =0
vldr.f64 d5, =0
vsub.f64 d4, d4 , d4
vsub.f64 d5, d5 , d5
FMAC_R1 d4 , d0 , d12
FMAC_I1 d5 , d0 , d13
@ -574,7 +574,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
vmov.f64 d10, d8
vmov.f64 d11, d8
@ -718,10 +718,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fldd d0, ALPHA_R
fldd d1, ALPHA_I
vldr.f64 d4, =0
vldr.f64 d5, =0
vldr.f64 d6, =0
vldr.f64 d7, =0
vsub.f64 d4, d4 , d4
vsub.f64 d5, d5 , d5
vsub.f64 d6, d6 , d6
vsub.f64 d7, d7 , d7
FMAC_R1 d4 , d0 , d8
FMAC_I1 d5 , d0 , d9
@ -744,7 +744,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d8 , =0
vsub.f64 d8 , d8 , d8
vmov.f64 d9 , d8
.endm
@ -850,8 +850,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
fldd d0, ALPHA_R
fldd d1, ALPHA_I
vldr.f64 d4, =0
vldr.f64 d5, =0
vsub.f64 d4, d4 , d4
vsub.f64 d5, d5 , d5
FMAC_R1 d4 , d0 , d8
FMAC_I1 d5 , d0 , d9

View File

@ -134,7 +134,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -388,7 +388,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x2
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d20, d16
vmov.f64 d21, d16
@ -566,7 +566,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT2x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d18, d16
vmov.f64 d19, d16
@ -743,7 +743,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.macro INIT1x1
vldr.f64 d16 , =0
vsub.f64 d16 , d16 , d16
vmov.f64 d17, d16
vmov.f64 d24, d16
vmov.f64 d25, d16