Commit 9b5a1df8 by Andreas Marek

Single precision SSE BLOCK1 complex kernel

parent cb3da78c
 ... @@ -930,6 +930,8 @@ function solve_evp_complex_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, & ... @@ -930,6 +930,8 @@ function solve_evp_complex_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, & if ( (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC) .or. & if ( (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX_BLOCK1) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX_BLOCK1) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX_BLOCK2) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_SSE_BLOCK1) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_SSE) ) then (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_SSE) ) then else else print *,"At the moment single precision only works with the generic kernels" print *,"At the moment single precision only works with the generic kernels" ... @@ -1255,6 +1257,8 @@ function solve_evp_complex_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, & ... @@ -1255,6 +1257,8 @@ function solve_evp_complex_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, & if ( (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC) .or. & if ( (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_GENERIC_SIMPLE) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX_BLOCK1) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX_BLOCK1) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_AVX_BLOCK2) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_SSE_BLOCK1) .or. & (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_SSE) ) then (THIS_COMPLEX_ELPA_KERNEL .eq. COMPLEX_ELPA_KERNEL_SSE) ) then else else print *,"At the moment single precision only works with the generic kernels" print *,"At the moment single precision only works with the generic kernels" ... ...
This diff is collapsed.
 ... @@ -312,7 +312,7 @@ module compute_hh_trafo_complex ... @@ -312,7 +312,7 @@ module compute_hh_trafo_complex #if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL) #if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL) endif endif #endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */ #endif /* WITH_NO_SPECIFIC_COMPLEX_KERNEL */ #endif /* WITH_COMPLEX_SSE_BLOCK1_KERNE */ #endif /* WITH_COMPLEX_SSE_BLOCK1_KERNEL */ #if defined(WITH_COMPLEX_AVX_BLOCK1_KERNEL) || defined(WITH_COMPLEX_AVX2_BLOCK1_KERNEL) #if defined(WITH_COMPLEX_AVX_BLOCK1_KERNEL) || defined(WITH_COMPLEX_AVX2_BLOCK1_KERNEL) #if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL) #if defined(WITH_NO_SPECIFIC_COMPLEX_KERNEL) ... @@ -437,6 +437,33 @@ module compute_hh_trafo_complex ... @@ -437,6 +437,33 @@ module compute_hh_trafo_complex #else #else nl = merge(stripe_width, last_stripe_width, istripe
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment