foxtran

      !>
      !> @brief Perform B <- alpha * A^T + B
      !>
      !> @note OpenMP parallelization is used for faster matrix transposing.
      !>       The speed-up is about 3-10 times in comparison
      !>         with original implementation via BLAS-1 routines.
      !>
      !> @param[in]    n       size of square matrix
      !> @param[in]    alpha   scaling factor
      !> @param[in]    mat     n-by-n matrix A

foxtran

module helper_mod {
  public i32 sudctoi(c8 c) {
    consteval i32 C0 = i32('0'), C9 = i32('9'), CA = i32('A'), CZ = i32('Z')
    if (irepr >= C0 .and. irepr <= C9) {
      return i32(c) - C0
    } else if (irepr >= CA .and. irepr <= CZ) {
      return i32(c) - CA + 10
    } else {
      return 0
    }

foxtran

#!/bin/bash

## This script does the following:
## 1. It checks out and builds trunk LLVM.
## 2. It checks out and builds the create_llvm_prof tool.
## 3. It builds multiple clang binaries towards building a
##    propeller optimized clang binary.
## 4. It runs performance comparisons of a baseline clang
##    binary and the Propeller optimized clang binary.

foxtran

#!/bin/bash

## This script does the following:
## 1. It checks out and builds trunk LLVM.
## 2. It checks out and builds the create_llvm_prof tool.
## 3. It builds multiple clang binaries towards building a
##    propeller optimized clang binary.
## 4. It runs performance comparisons of a baseline clang
##    binary and the Propeller optimized clang binary.

foxtran

Benchmark was compiled using the following compiler:

GCC version 12.2.0

Benchmark was compiled with the following options:

-mabi=lp64d -mcpu=sifive-u74 -misa-spec=20191213 -march=rv64imafdc_zicsr -O3 -ffree-line-length-none -fpre-include=/usr/include/finclude/riscv64-linux-gnu/math-vector-fortran.h

foxtran

Benchmark was compiled using the following compiler:

GCC version 12.2.0

Benchmark was compiled with the following options:

-mabi=lp64d -mcpu=sifive-u74 -misa-spec=20191213 -march=rv64imafdc_zicsr -O3 -ffree-line-length-none -fpre-include=/usr/include/finclude/riscv64-linux-gnu/math-vector-fortran.h

Number of repeats is: 100000

foxtran

Benchmark was compiled using the following compiler:

GCC version 12.2.0

Benchmark was compiled with the following options:

-mabi=lp64d -mcpu=sifive-s76 -misa-spec=20191213 -march=rv64ifd_zicsr -O3 -ffree-line-length-none -fpre-include=/usr/include/finclude/riscv64-linux-gnu/math-vector-fortran.h

foxtran

Benchmark was compiled using the following compiler:

GCC version 11.3.0

Benchmark was compiled with the following options:

-mabi=lp64d -mcpu=sifive-u74 -misa-spec=2.2 -march=rv64imafdc -O3 -Wall -Wextra -ffree-line-length-none -fpre-include=/usr/include/finclude/riscv64-linux-gnu/math-vector-fortran.h

foxtran

24
symmetry c1
C       -0.162106268      0.986301416      2.475612750
C       -0.162101268     -0.406019584      2.475612750
C        1.043582732     -1.102098584      2.475260750
C        2.249277732     -0.406018584      2.475612750
C        2.249273732      0.986302416      2.475612750
C        1.043588732      1.682382416      2.475260750
H       -1.092522268      1.524132416      2.470409750
H       -1.092523268     -0.943850584      2.470409750

foxtran

PROGRAM test_div
  IMPLICIT NONE
  INTEGER :: count1, count_rate1, count_max1
  INTEGER :: count2, count_rate2, count_max2
  INTEGER :: N
  REAL(8), allocatable :: a(:), b(:), c(:)
  read(*,*) N
  allocate(a(N), b(N), c(N))
  CALL RANDOM_NUMBER(a)
  CALL RANDOM_NUMBER(b)