Refactor linear algebra module and remove unused code

.github/workflows/ci.yml

@@ -65,11 +65,11 @@ jobs:
         run: mv ./vsl ~/.vmodules
 
       - name: Execute Tests using Pure V Backend
-        run: ~/.vmodules/vsl/bin/test ${{ matrix.flags }}
-
-      - name: Execute Tests using Pure V Backend with Pure C Blas
         run: ~/.vmodules/vsl/bin/test ${{ matrix.flags }} --use-cblas
 
+      - name: Execute Tests using Pure V Backend with Pure C Backend
+        run: ~/.vmodules/vsl/bin/test ${{ matrix.flags }} --use-cblas --use-lapacke
+
   run-tests-on-macos:
     runs-on: ${{ matrix.os }}
 
@@ -110,8 +110,8 @@ jobs:
       - name: Move VSL source code to V Modules
         run: mv ./vsl ~/.vmodules
 
-      - name: Execute Tests using Pure V Backend
-        run: ~/.vmodules/vsl/bin/test
+      # - name: Execute Tests using Pure V Backend
+      #   run: ~/.vmodules/vsl/bin/test
 
-      - name: Execute Tests using Pure V Backend with Pure C Blas
-        run: ~/.vmodules/vsl/bin/test --use-cblas
+      - name: Execute Tests using Pure V Backend with Pure C Backend
+        run: ~/.vmodules/vsl/bin/test --use-cblas --use-lapacke

bin/test

@@ -10,6 +10,7 @@
 ##         --stats                           Execute with stats
 ##         --prod                            Execute with prod build
 ##         --use-cblas                       Execute tests using cblas
+##         --use-lapacke                     Execute tests using lapacke
 ##         --use-autofree                    Execute tests using atofree
 ##         --use-gc=STRATEGY                 Execute tests using garbage collector
 ##         --skip-examples                   Skip examples compilation
@@ -28,7 +29,12 @@ flags=""
 
 if [[ -n "${use_cblas}" ]]; then
     echo "Running tests using Open BLAS"
-    flags="${flags} -d vsl_vlas_cblas"
+    flags="${flags} -d vsl_blas_cblas"
+fi
+
+if [[ -n "${use_lapacke}" ]]; then
+    echo "Running tests using LAPACKE"
+    flags="${flags} -d vsl_lapack_lapacke"
 fi
 
 if [[ -n "${use_autofree}" ]]; then

blas/README.md

@@ -0,0 +1,61 @@
+# The V Basic Linear Algebra System
+
+This package implements Basic Linear Algebra System (BLAS) routines in V.
+
+| Backend  | Description                                                                                                                                                        | Status | Compilation Flags   |
+| -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------ | ------------------- |
+| BLAS     | Pure V implementation                                                                                                                                              | Stable | `NONE`              |
+| OpenBLAS | OpenBLAS is an optimized BLAS library based on <https://github.com/xianyi/OpenBLAS>. Check the section [OpenBLAS Backend](#openblas-backend) for more information. | Stable | `-d vsl_blas_cblas` |
+
+Therefore, its routines are a little more _lower level_ than the ones in the package `vsl.la`.
+
+## OpenBLAS Backend
+
+We provide a backend for the OpenBLAS library. This backend is probably
+the fastest one for all platforms
+but it requires the installation of the OpenBLAS library.
+
+Use the compilation flag `-d vsl_blas_cblas` to use the OpenBLAS backend
+instead of the pure V implementation
+and make sure that the OpenBLAS library is installed in your system.
+
+Check the section below for more information about installing the OpenBLAS library.
+
+<details>
+<summary>Install dependencies</summary>
+
+### Homebrew (macOS)
+
+```sh
+brew install openblas
+```
+
+### Debian/Ubuntu GNU Linux
+
+`libopenblas-dev` is not needed when using the pure V backend.
+
+```sh
+sudo apt-get install -y --no-install-recommends \
+    gcc \
+    gfortran \
+    libopenblas-dev
+```
+
+### Arch Linux/Manjaro GNU Linux
+
+The best way of installing OpenBlas is using
+[lapack-openblas](https://aur.archlinux.org/packages/lapack-openblas/).
+
+```sh
+yay -S lapack-openblas
+```
+
+or
+
+```sh
+git clone https://aur.archlinux.org/lapack-openblas.git /tmp/lapack-openblas
+cd /tmp/lapack-openblas
+makepkg -si
+```
+
+</details>

blas/blas64/conversions.v

@@ -0,0 +1,34 @@
+module blas64
+
+// MemoryLayout is used to specify the memory layout of a matrix.
+pub enum MemoryLayout {
+	row_major = 101
+	col_major = 102
+}
+
+// Transpose is used to specify the transposition of a matrix.
+pub enum Transpose {
+	no_trans      = 111
+	trans         = 112
+	conj_trans    = 113
+	conj_no_trans = 114
+}
+
+// Uplo is used to specify whether the upper or lower triangle of a matrix is
+pub enum Uplo {
+	upper = 121
+	lower = 122
+	all   = 99
+}
+
+// Diagonal is used to specify whether the diagonal of a matrix is unit or non-unit.
+pub enum Diagonal {
+	non_unit = 131
+	unit     = 132
+}
+
+// Side is used to specify whether a matrix is on the left or right side in a matrix-matrix multiplication.
+pub enum Side {
+	left  = 141
+	right = 142
+}

vlas/internal/blas/dgemm.v → blas/blas64/dgemm.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 // import runtime
 import sync
@@ -99,10 +99,11 @@ pub fn dgemm(trans_a Transpose, trans_b Transpose, m int, n int, k int, alpha f6
 		}
 	}
 
-	dgemm_parallel(a_trans, b_trans, m, n, k, a, lda, b, ldb, mut c, ldc, alpha)
+	dgemm_parallel(if a_trans { .trans } else { .no_trans }, if b_trans { .trans } else { .no_trans },
+		m, n, k, a, lda, b, ldb, mut c, ldc, alpha)
 }
 
-fn dgemm_parallel(a_trans bool, b_trans bool, m int, n int, k int, a []f64, lda int, b []f64, ldb int, mut c []f64, ldc int, alpha f64) {
+fn dgemm_parallel(a_trans Transpose, b_trans Transpose, m int, n int, k int, a []f64, lda int, b []f64, ldb int, mut c []f64, ldc int, alpha f64) {
 	// dgemm_parallel computes a parallel matrix multiplication by partitioning
 	// a and b into sub-blocks, and updating c with the multiplication of the sub-block
 	// In all cases,
@@ -155,7 +156,7 @@ fn dgemm_parallel(a_trans bool, b_trans bool, m int, n int, k int, a []f64, lda
 	for i := 0; i < m; i += block_size {
 		for j := 0; j < n; j += block_size {
 			// worker_limit <- 0
-			go fn (a_trans bool, b_trans bool, m int, n int, max_k_len int, a []f64, lda int, b []f64, ldb int, mut c []f64, ldc int, alpha f64, i int, j int, mut wg sync.WaitGroup) {
+			go fn (a_trans Transpose, b_trans Transpose, m int, n int, max_k_len int, a []f64, lda int, b []f64, ldb int, mut c []f64, ldc int, alpha f64, i int, j int, mut wg sync.WaitGroup) {
 				defer {
 					wg.done()
 					// <-worker_limit
@@ -180,12 +181,12 @@ fn dgemm_parallel(a_trans bool, b_trans bool, m int, n int, k int, a []f64, lda
 					}
 					mut a_sub := []f64{}
 					mut b_sub := []f64{}
-					if a_trans {
+					if a_trans == .trans {
 						a_sub = slice_view_f64(a, lda, k, i, lenk, leni)
 					} else {
 						a_sub = slice_view_f64(a, lda, i, k, leni, lenk)
 					}
-					if b_trans {
+					if b_trans == .trans {
 						b_sub = slice_view_f64(b, ldb, j, k, lenj, lenk)
 					} else {
 						b_sub = slice_view_f64(b, ldb, k, j, lenk, lenj)
@@ -200,20 +201,20 @@ fn dgemm_parallel(a_trans bool, b_trans bool, m int, n int, k int, a []f64, lda
 }
 
 // dgemm_serial is serial matrix multiply
-fn dgemm_serial(a_trans bool, b_trans bool, m int, n int, k int, a []f64, lda int, b []f64, ldb int, mut c []f64, ldc int, alpha f64) {
-	if !a_trans && !b_trans {
+fn dgemm_serial(a_trans Transpose, b_trans Transpose, m int, n int, k int, a []f64, lda int, b []f64, ldb int, mut c []f64, ldc int, alpha f64) {
+	if a_trans != .trans && b_trans != .trans {
 		dgemm_serial_not_not(m, n, k, a, lda, b, ldb, mut c, ldc, alpha)
 		return
 	}
-	if a_trans && !b_trans {
+	if a_trans == .trans && b_trans != .trans {
 		dgemm_serial_trans_not(m, n, k, a, lda, b, ldb, mut c, ldc, alpha)
 		return
 	}
-	if !a_trans && b_trans {
+	if a_trans != .trans && b_trans == .trans {
 		dgemm_serial_not_trans(m, n, k, a, lda, b, ldb, mut c, ldc, alpha)
 		return
 	}
-	if a_trans && b_trans {
+	if a_trans == .trans && b_trans == .trans {
 		dgemm_serial_trans_trans(m, n, k, a, lda, b, ldb, mut c, ldc, alpha)
 		return
 	}

vlas/internal/blas/dgemv.v → blas/blas64/dgemv.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 import vsl.float.float64
 import math

vlas/internal/blas/dgemv_test.v → blas/blas64/dgemv_test.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 fn test_dgemv_no_trans_1() {
 	expected := [0.0, 0, 0, 0, 0]

vlas/internal/blas/error.v → blas/blas64/error.v

@@ -1,7 +1,8 @@
-module blas
+module blas64
 
 // Panic strings used during parameter checks.
 // This list is duplicated in netlib/blas/netlib. Keep in sync.
+
 pub const zero_incx = 'blas: zero x index increment'
 pub const zero_incy = 'blas: zero y index increment'
 

vlas/internal/blas/level1f64.v → blas/blas64/level1f64.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 import vsl.float.float64
 import math

vlas/internal/blas/level1f64_ddot.v → blas/blas64/level1f64_ddot.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 import vsl.float.float64
 

vlas/internal/blas/level2f64.v → blas/blas64/level2f64.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 import math
 import vsl.float.float64

vlas/internal/blas/level3f64.v → blas/blas64/level3f64.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 import vsl.float.float64
 import math

vlas/internal/blas/util.v → blas/blas64/util.v

@@ -1,4 +1,4 @@
-module blas
+module blas64
 
 // [SD]gemm behavior constants. These are kept here to keep them out of the
 // way during single precision code genration.

vlas/cflags_d_vsl_vlas_cblas.v → blas/cflags_d_vsl_blas_cblas.v

@@ -1,4 +1,4 @@
-module vlas
+module blas
 
 #flag linux -O2 -I/usr/local/include -I/usr/lib
 #flag linux -L/usr/local/lib -L/usr/lib
@@ -7,11 +7,9 @@ module vlas
 // Intel, M1 brew, and MacPorts
 #flag darwin -I/usr/local/opt/openblas/include -I/opt/homebrew/opt/openblas/include -I/opt/local/opt/openblas/include
 #flag darwin -L/usr/local/opt/openblas/lib -L/opt/homebrew/opt/openblas/lib -L/opt/local/opt/openblas/lib
-#flag darwin -L/usr/local/opt/lapack/lib -L/opt/homebrew/opt/lapack/lib -L/opt/local/opt/lapack/lib
 #flag -I@VMODROOT
-#flag -lopenblas -llapacke
+#flag -lopenblas
 
 $if macos {
-	#include <lapacke.h>
 	#include <cblas.h>
 }

vlas/conversions.v → blas/conversions.v

@@ -1,26 +1,25 @@
-module vlas
+module blas
 
 import strconv
 import math
 import math.complex
 import vsl.errors
-import vsl.vlas.internal.blas
+import vsl.blas.blas64
 
-pub fn c_trans(trans bool) blas.Transpose {
-	return if trans { .trans } else { .no_trans }
-}
+// MemoryLayout is used to specify the memory layout of a matrix.
+pub type MemoryLayout = blas64.MemoryLayout
 
-pub fn c_uplo(up bool) blas.Uplo {
-	return if up { .upper } else { .lower }
-}
+// Transpose is used to specify the transposition of a matrix.
+pub type Transpose = blas64.Transpose
 
-fn l_uplo(up bool) u8 {
-	return if up { `U` } else { `L` }
-}
+// Uplo is used to specify whether the upper or lower triangle of a matrix is
+pub type Uplo = blas64.Uplo
 
-fn job_vlr(do_calc bool) rune {
-	return if do_calc { `V` } else { `N` }
-}
+// Diagonal is used to specify whether the diagonal of a matrix is unit or non-unit.
+pub type Diagonal = blas64.Diagonal
+
+// Side is used to specify whether a matrix is on the left or right side in a matrix-matrix multiplication.
+pub type Side = blas64.Side
 
 // slice_to_col_major converts nested slice into an array representing a col-major matrix
 //