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

Power8 inline assembly fixes

Browse Source 
Quoting patch author amodra from #1078
Lots of issues here.
- The vsx regs weren't listed as clobbered.
- Poor choice of vsx regs, which along with the lack of clobbers led to
  trashing v0..v21 and fr14..fr23.  Ideally you'd let gcc choose all
  temp vsx regs, but asms currently have a limit of 30 i/o parms.
- Other regs were clobbered unnecessarily, seemingly in an attempt to
  clobber inputs, with gcc-7 complaining about the clobber of r2.
  (Changed inputs should be also listed as outputs or as an i/o.)
- "r" constraint used instead of "b" for gprs used in insns where the
  r0 encoding means zero rather than r0.
- There were unused asm inputs too.
- All memory was clobbered rather than hooking up memory outputs with
  proper memory constraints, and that and the lack of proper memory
  input constraints meant the asms needed to be volatile and their
  containing function noinline.
- Some parameters were being passed unnecessarily via memory.
- When a copy of a
This commit is contained in:
Martin Kroeker
2017-02-13 23:38:50 +01:00
committed by GitHub
parent e2489c9a92
commit 9e2f316ede
38 changed files with 2923 additions and 3249 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
kernel/power/drot.c
View File

@@ -46,7 +46,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#ifndef HAVE_KERNEL_16
static void drot_kernel_16(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *c, FLOAT *s)
static void drot_kernel_16(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT c, FLOAT s)
{
BLASLONG i=0;
@@ -56,8 +56,6 @@ static void drot_kernel_16(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *c, FLOAT *s)
FLOAT y00, y01, y02, y03;
FLOAT *x1=x;
FLOAT *y1=y;
FLOAT c1=*c;
FLOAT s1=*s;
while ( i<n )
{
@@ -71,14 +69,14 @@ static void drot_kernel_16(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *c, FLOAT *s)
x03 = x1[3];
y03 = y1[3];
f0 = c1*x00 + s1*y00;
g0 = c1*y00 - s1*x00;
f1 = c1*x01 + s1*y01;
g1 = c1*y01 - s1*x01;
f2 = c1*x02 + s1*y02;
g2 = c1*y02 - s1*x02;
f3 = c1*x03 + s1*y03;
g3 = c1*y03 - s1*x03;
f0 = c*x00 + s*y00;
g0 = c*y00 - s*x00;
f1 = c*x01 + s*y01;
g1 = c*y01 - s*x01;
f2 = c*x02 + s*y02;
g2 = c*y02 - s*x02;
f3 = c*x03 + s*y03;
g3 = c*y03 - s*x03;
x1[0] = f0;
y1[0] = g0;
@@ -106,8 +104,6 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
{
BLASLONG i=0;
BLASLONG ix=0,iy=0;
FLOAT c1[4] __attribute__ ((aligned (16)));;
FLOAT s1[4] __attribute__ ((aligned (16)));;
FLOAT *x1=x;
FLOAT *y1=y;
FLOAT temp;
@@ -120,15 +116,7 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
BLASLONG n1 = n & -16;
if ( n1 > 0 )
{
c1[0]=c;
c1[1]=c;
c1[2]=c;
c1[3]=c;
s1[0]=s;
s1[1]=s;
s1[2]=s;
s1[3]=s;
drot_kernel_16(n1, x1, y1, c1, s1);
drot_kernel_16(n1, x1, y1, c, s);
i=n1;
}