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

LoongArch64: Handle NAN and INF

This commit is contained in:
gxw 2024-01-30 16:54:14 +08:00
parent d6a5174e9c
commit 83ce97a4ca
2 changed files with 4 additions and 275 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

149
kernel/loongarch64/cscal_lasx.S
View File

@ -99,7 +99,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L113 //alpha_r != 0.0 && alpha_i == 0.0 b .L113 //alpha_r != 0.0 && alpha_i == 0.0
.L14: .L14:
bceqz $fcc1, .L112 //alpha_r == 0.0 && alpha_i != 0.0 bceqz $fcc1, .L114 //alpha_r == 0.0 && alpha_i != 0.0
b .L111 //alpha_r == 0.0 && alpha_i == 0.0 b .L111 //alpha_r == 0.0 && alpha_i == 0.0
.align 3 .align 3
@ -117,38 +117,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L997 b .L997
.align 3 .align 3
.L112: //alpha_r == 0.0 && alpha_i != 0.0
xvld VX0, X, 0 * SIZE
#ifdef DOUBLE
xvld VX1, X, 4 * SIZE
xvpickev.d x1, VX1, VX0
xvpickod.d x2, VX1, VX0
xvfmul.d x3, VXAI, x2
xvfsub.d x3, VXZ, x3
xvfmul.d x4, VXAI, x1
xvilvl.d VX2, x4 ,x3
xvilvh.d VX3, x4, x3
xvst VX2, X, 0 * SIZE
xvst VX3, X, 4 * SIZE
addi.d X, X, 8 * SIZE
#else
xvld VX1, X, 8 * SIZE
xvpickev.w x1, VX1, VX0
xvpickod.w x2, VX1, VX0
xvfmul.s x3, VXAI, x2
xvfsub.s x3, VXZ, x3
xvfmul.s x4, VXAI, x1
xvilvl.w VX2, x4 ,x3
xvilvh.w VX3, x4, x3
xvst VX2, X, 0 * SIZE
xvst VX3, X, 8 * SIZE
addi.d X, X, 16 * SIZE
#endif
addi.d I, I, -1
blt $r0, I, .L112
b .L997
.align 3
.L113: //alpha_r != 0.0 && alpha_i == 0.0 .L113: //alpha_r != 0.0 && alpha_i == 0.0
xvld VX0, X, 0 * SIZE xvld VX0, X, 0 * SIZE
#ifdef DOUBLE #ifdef DOUBLE
@ -227,7 +195,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L223 //alpha_r != 0.0 && alpha_i == 0.0 b .L223 //alpha_r != 0.0 && alpha_i == 0.0
.L24: .L24:
bceqz $fcc1, .L222 //alpha_r == 0.0 && alpha_i != 0.0 bceqz $fcc1, .L224 //alpha_r == 0.0 && alpha_i != 0.0
b .L221 //alpha_r == 0.0 && alpha_i == 0.0 b .L221 //alpha_r == 0.0 && alpha_i == 0.0
.align 3 .align 3
@ -275,119 +243,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L997 b .L997
.align 3 .align 3
.L222: //alpha_r == 0.0 && alpha_i != 0.0
#ifdef DOUBLE
ld.d t1, X, 0 * SIZE
ld.d t2, X, 1 * SIZE
add.d X, X, INCX
ld.d t3, X, 0 * SIZE
ld.d t4, X, 1 * SIZE
add.d X, X, INCX
xvinsgr2vr.d x1, t1, 0
xvinsgr2vr.d x2, t2, 0
xvinsgr2vr.d x1, t3, 1
xvinsgr2vr.d x2, t4, 1
ld.d t1, X, 0 * SIZE
ld.d t2, X, 1 * SIZE
add.d X, X, INCX
ld.d t3, X, 0 * SIZE
ld.d t4, X, 1 * SIZE
xvinsgr2vr.d x1, t1, 2
xvinsgr2vr.d x2, t2, 2
xvinsgr2vr.d x1, t3, 3
xvinsgr2vr.d x2, t4, 3
add.d X, X, INCX
xvfmul.d x3, VXAI, x2
xvfsub.d x3, VXZ, x3
xvfmul.d x4, VXAI, x1
addi.d I, I, -1
xvstelm.d x3, XX, 0 * SIZE, 0
xvstelm.d x4, XX, 1 * SIZE, 0
add.d XX, XX, INCX
xvstelm.d x3, XX, 0 * SIZE, 1
xvstelm.d x4, XX, 1 * SIZE, 1
add.d XX, XX, INCX
xvstelm.d x3, XX, 0 * SIZE, 2
xvstelm.d x4, XX, 1 * SIZE, 2
add.d XX, XX, INCX
xvstelm.d x3, XX, 0 * SIZE, 3
xvstelm.d x4, XX, 1 * SIZE, 3
#else
ld.w t1, X, 0 * SIZE
ld.w t2, X, 1 * SIZE
add.d X, X, INCX
ld.w t3, X, 0 * SIZE
ld.w t4, X, 1 * SIZE
add.d X, X, INCX
xvinsgr2vr.w x1, t1, 0
xvinsgr2vr.w x2, t2, 0
xvinsgr2vr.w x1, t3, 1
xvinsgr2vr.w x2, t4, 1
ld.w t1, X, 0 * SIZE
ld.w t2, X, 1 * SIZE
add.d X, X, INCX
ld.w t3, X, 0 * SIZE
ld.w t4, X, 1 * SIZE
xvinsgr2vr.w x1, t1, 2
xvinsgr2vr.w x2, t2, 2
xvinsgr2vr.w x1, t3, 3
xvinsgr2vr.w x2, t4, 3
add.d X, X, INCX
ld.w t1, X, 0 * SIZE
ld.w t2, X, 1 * SIZE
add.d X, X, INCX
ld.w t3, X, 0 * SIZE
ld.w t4, X, 1 * SIZE
add.d X, X, INCX
xvinsgr2vr.w x1, t1, 4
xvinsgr2vr.w x2, t2, 4
xvinsgr2vr.w x1, t3, 5
xvinsgr2vr.w x2, t4, 5
ld.w t1, X, 0 * SIZE
ld.w t2, X, 1 * SIZE
add.d X, X, INCX
ld.w t3, X, 0 * SIZE
ld.w t4, X, 1 * SIZE
xvinsgr2vr.w x1, t1, 6
xvinsgr2vr.w x2, t2, 6
xvinsgr2vr.w x1, t3, 7
xvinsgr2vr.w x2, t4, 7
add.d X, X, INCX
xvfmul.s x3, VXAI, x2
xvfsub.s x3, VXZ, x3
xvfmul.s x4, VXAI, x1
addi.d I, I, -1
xvstelm.w x3, XX, 0 * SIZE, 0
xvstelm.w x4, XX, 1 * SIZE, 0
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 1
xvstelm.w x4, XX, 1 * SIZE, 1
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 2
xvstelm.w x4, XX, 1 * SIZE, 2
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 3
xvstelm.w x4, XX, 1 * SIZE, 3
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 4
xvstelm.w x4, XX, 1 * SIZE, 4
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 5
xvstelm.w x4, XX, 1 * SIZE, 5
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 6
xvstelm.w x4, XX, 1 * SIZE, 6
add.d XX, XX, INCX
xvstelm.w x3, XX, 0 * SIZE, 7
xvstelm.w x4, XX, 1 * SIZE, 7
#endif
add.d XX, XX, INCX
blt $r0, I, .L222
b .L997
.align 3
.L223: //alpha_r != 0.0 && alpha_i == 0.0 .L223: //alpha_r != 0.0 && alpha_i == 0.0
#ifdef DOUBLE #ifdef DOUBLE
ld.d t1, X, 0 * SIZE ld.d t1, X, 0 * SIZE

130
kernel/loongarch64/cscal_lsx.S
View File

@ -97,7 +97,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L113 //alpha_r != 0.0 && alpha_i == 0.0 b .L113 //alpha_r != 0.0 && alpha_i == 0.0
.L14: .L14:
bceqz $fcc1, .L112 //alpha_r == 0.0 && alpha_i != 0.0 bceqz $fcc1, .L114 //alpha_r == 0.0 && alpha_i != 0.0
b .L111 //alpha_r == 0.0 && alpha_i == 0.0 b .L111 //alpha_r == 0.0 && alpha_i == 0.0
.align 3 .align 3
@ -116,48 +116,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L997 b .L997
.align 3 .align 3
.L112: //alpha_r == 0.0 && alpha_i != 0.0
vld VX0, X, 0 * SIZE
#ifdef DOUBLE
vld VX1, X, 2 * SIZE
vpickev.d x1, VX1, VX0
vpickod.d x2, VX1, VX0
vfmul.d x3, VXAI, x2
vfsub.d x3, VXZ, x3
vfmul.d x4, VXAI, x1
vilvl.d VX2, x4 ,x3
vilvh.d VX3, x4, x3
vst VX2, X, 0 * SIZE
vst VX3, X, 2 * SIZE
vld VX0, X, 4 * SIZE
vld VX1, X, 6 * SIZE
vpickev.d x1, VX1, VX0
vpickod.d x2, VX1, VX0
vfmul.d x3, VXAI, x2
vfsub.d x3, VXZ, x3
vfmul.d x4, VXAI, x1
vilvl.d VX2, x4 ,x3
vilvh.d VX3, x4, x3
vst VX2, X, 4 * SIZE
vst VX3, X, 6 * SIZE
#else
vld VX1, X, 4 * SIZE
vpickev.w x1, VX1, VX0
vpickod.w x2, VX1, VX0
vfmul.s x3, VXAI, x2
vfsub.s x3, VXZ, x3
vfmul.s x4, VXAI, x1
vilvl.w VX2, x4 ,x3
vilvh.w VX3, x4, x3
vst VX2, X, 0 * SIZE
vst VX3, X, 4 * SIZE
#endif
addi.d X, X, 8 * SIZE
addi.d I, I, -1
blt $r0, I, .L112
b .L997
.align 3
.L113: //alpha_r != 0.0 && alpha_i == 0.0 .L113: //alpha_r != 0.0 && alpha_i == 0.0
vld VX0, X, 0 * SIZE vld VX0, X, 0 * SIZE
#ifdef DOUBLE #ifdef DOUBLE
@ -256,7 +214,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L223 //alpha_r != 0.0 && alpha_i == 0.0 b .L223 //alpha_r != 0.0 && alpha_i == 0.0
.L24: .L24:
bceqz $fcc1, .L222 //alpha_r == 0.0 && alpha_i != 0.0 bceqz $fcc1, .L224 //alpha_r == 0.0 && alpha_i != 0.0
b .L221 //alpha_r == 0.0 && alpha_i == 0.0 b .L221 //alpha_r == 0.0 && alpha_i == 0.0
.align 3 .align 3
@ -292,90 +250,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
b .L997 b .L997
.align 3 .align 3
.L222: //alpha_r == 0.0 && alpha_i != 0.0
#ifdef DOUBLE
ld.d t1, X, 0 * SIZE
ld.d t2, X, 1 * SIZE
add.d X, X, INCX
ld.d t3, X, 0 * SIZE
ld.d t4, X, 1 * SIZE
add.d X, X, INCX
vinsgr2vr.d x1, t1, 0
vinsgr2vr.d x2, t2, 0
vinsgr2vr.d x1, t3, 1
vinsgr2vr.d x2, t4, 1
vfmul.d x3, VXAI, x2
vfsub.d x3, VXZ, x3
vfmul.d x4, VXAI, x1
vstelm.d x3, XX, 0 * SIZE, 0
vstelm.d x4, XX, 1 * SIZE, 0
add.d XX, XX, INCX
vstelm.d x3, XX, 0 * SIZE, 1
vstelm.d x4, XX, 1 * SIZE, 1
add.d XX, XX, INCX
ld.d t1, X, 0 * SIZE
ld.d t2, X, 1 * SIZE
add.d X, X, INCX
ld.d t3, X, 0 * SIZE
ld.d t4, X, 1 * SIZE
vinsgr2vr.d x1, t1, 0
vinsgr2vr.d x2, t2, 0
vinsgr2vr.d x1, t3, 1
vinsgr2vr.d x2, t4, 1
add.d X, X, INCX
vfmul.d x3, VXAI, x2
vfsub.d x3, VXZ, x3
vfmul.d x4, VXAI, x1
addi.d I, I, -1
vstelm.d x3, XX, 0 * SIZE, 0
vstelm.d x4, XX, 1 * SIZE, 0
add.d XX, XX, INCX
vstelm.d x3, XX, 0 * SIZE, 1
vstelm.d x4, XX, 1 * SIZE, 1
#else
ld.w t1, X, 0 * SIZE
ld.w t2, X, 1 * SIZE
add.d X, X, INCX
ld.w t3, X, 0 * SIZE
ld.w t4, X, 1 * SIZE
add.d X, X, INCX
vinsgr2vr.w x1, t1, 0
vinsgr2vr.w x2, t2, 0
vinsgr2vr.w x1, t3, 1
vinsgr2vr.w x2, t4, 1
ld.w t1, X, 0 * SIZE
ld.w t2, X, 1 * SIZE
add.d X, X, INCX
ld.w t3, X, 0 * SIZE
ld.w t4, X, 1 * SIZE
vinsgr2vr.w x1, t1, 2
vinsgr2vr.w x2, t2, 2
vinsgr2vr.w x1, t3, 3
vinsgr2vr.w x2, t4, 3
add.d X, X, INCX
vfmul.s x3, VXAI, x2
vfsub.s x3, VXZ, x3
vfmul.s x4, VXAI, x1
addi.d I, I, -1
vstelm.w x3, XX, 0 * SIZE, 0
vstelm.w x4, XX, 1 * SIZE, 0
add.d XX, XX, INCX
vstelm.w x3, XX, 0 * SIZE, 1
vstelm.w x4, XX, 1 * SIZE, 1
add.d XX, XX, INCX
vstelm.w x3, XX, 0 * SIZE, 2
vstelm.w x4, XX, 1 * SIZE, 2
add.d XX, XX, INCX
vstelm.w x3, XX, 0 * SIZE, 3
vstelm.w x4, XX, 1 * SIZE, 3
#endif
add.d XX, XX, INCX
blt $r0, I, .L222
b .L997
.align 3
.L223: //alpha_r != 0.0 && alpha_i == 0.0 .L223: //alpha_r != 0.0 && alpha_i == 0.0
#ifdef DOUBLE #ifdef DOUBLE
ld.d t1, X, 0 * SIZE ld.d t1, X, 0 * SIZE