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

added optimized zgemv_t kernel for haswell

This commit is contained in:
wernsaar 2014-08-11 16:57:52 +02:00
parent 09fcd3a341
commit 58b075daef
3 changed files with 30 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
kernel/x86_64/KERNEL.HASWELL
View File

@ -6,6 +6,7 @@ DGEMVTKERNEL = dgemv_t.c
ifndef OS_WINDOWS
ZGEMVNKERNEL = zgemv_n.c
endif
ZGEMVTKERNEL = zgemv_t.c
SGEMMKERNEL = sgemm_kernel_16x4_haswell.S
SGEMMINCOPY = ../generic/gemm_ncopy_16.c

2
kernel/x86_64/zgemv_t.c
View File

@ -31,6 +31,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#if defined(BULLDOZER)
#include "zgemv_t_microk_bulldozer-2.c"
#elif defined(HASWELL)
#include "zgemv_t_microk_haswell-2.c"
#endif

29
kernel/x86_64/zgemv_t_microk_haswell-2.c
View File

@ -49,6 +49,7 @@ static void zgemv_kernel_16x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y)
".align 16 \n\t"
".L01LOOP%=: \n\t"
"prefetcht0 384(%2,%0,8) \n\t"
"vmovddup (%2,%0,8), %%xmm0 \n\t" // real value from x0
"vmovddup 8(%2,%0,8), %%xmm1 \n\t" // imag value from x0
"vmovddup 16(%2,%0,8), %%xmm2 \n\t" // real value from x1
@ -56,9 +57,13 @@ static void zgemv_kernel_16x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y)
"vinsertf128 $1, %%xmm2, %%ymm0 , %%ymm0 \n\t" // real values from x0 and x1
"vinsertf128 $1, %%xmm3, %%ymm1 , %%ymm1 \n\t" // imag values from x0 and x1
"prefetcht0 384(%4,%0,8) \n\t"
"vmovups (%4,%0,8), %%ymm4 \n\t" // 2 complex values from a0
"prefetcht0 384(%5,%0,8) \n\t"
"vmovups (%5,%0,8), %%ymm5 \n\t" // 2 complex values from a1
"prefetcht0 384(%6,%0,8) \n\t"
"vmovups (%6,%0,8), %%ymm6 \n\t" // 2 complex values from a2
"prefetcht0 384(%7,%0,8) \n\t"
"vmovups (%7,%0,8), %%ymm7 \n\t" // 2 complex values from a3
"vfmadd231pd %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
@ -70,9 +75,29 @@ static void zgemv_kernel_16x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y)
"vfmadd231pd %%ymm7 , %%ymm0, %%ymm14 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231pd %%ymm7 , %%ymm1, %%ymm15 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vmovddup 32(%2,%0,8), %%xmm0 \n\t" // real value from x0
"vmovddup 40(%2,%0,8), %%xmm1 \n\t" // imag value from x0
"vmovddup 48(%2,%0,8), %%xmm2 \n\t" // real value from x1
"vmovddup 56(%2,%0,8), %%xmm3 \n\t" // imag value from x1
"vinsertf128 $1, %%xmm2, %%ymm0 , %%ymm0 \n\t" // real values from x0 and x1
"vinsertf128 $1, %%xmm3, %%ymm1 , %%ymm1 \n\t" // imag values from x0 and x1
"addq $4 , %0 \n\t"
"subq $2 , %1 \n\t"
"vmovups 32(%4,%0,8), %%ymm4 \n\t" // 2 complex values from a0
"vmovups 32(%5,%0,8), %%ymm5 \n\t" // 2 complex values from a1
"vmovups 32(%6,%0,8), %%ymm6 \n\t" // 2 complex values from a2
"vmovups 32(%7,%0,8), %%ymm7 \n\t" // 2 complex values from a3
"vfmadd231pd %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231pd %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231pd %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231pd %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231pd %%ymm6 , %%ymm0, %%ymm12 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231pd %%ymm6 , %%ymm1, %%ymm13 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231pd %%ymm7 , %%ymm0, %%ymm14 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231pd %%ymm7 , %%ymm1, %%ymm15 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $8 , %0 \n\t"
"subq $4 , %1 \n\t"
"jnz .L01LOOP%= \n\t"
#if ( !defined(CONJ) && !defined(XCONJ) ) || ( defined(CONJ) && defined(XCONJ) )