Merge pull request #5515 from iha-taisei/feature/ger_unroll

Improve single-thread performance of [SD]GER on A64FX and Neoverse V1

Author: martin-frbg

kernel/arm64/KERNEL.A64FX

@@ -10,3 +10,6 @@ SDOTKERNEL   = dot_sve_v8.c
 
 SAXPYKERNEL  = axpy_sve.c
 DAXPYKERNEL  = axpy_sve.c
+
+SGERKERNEL   = ger_sve_v1x3.c
+DGERKERNEL   = ger_sve_v1x3.c

kernel/arm64/KERNEL.NEOVERSEV1

@@ -70,3 +70,6 @@ SBGEMVNKERNEL = sbgemv_n_neon.c
 SBGEMVTKERNEL = sbgemv_t_bfdot.c
 
 endif
+
+SGERKERNEL   = ger_sve_v1x3.c
+DGERKERNEL   = ger_sve_v1x3.c

kernel/arm64/ger_sve_v1x3.c

@@ -0,0 +1,135 @@
+/***************************************************************************
+Copyright (c) 2025, The OpenBLAS Project
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+   1. Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+
+   2. Redistributions in binary form must reproduce the above copyright
+      notice, this list of conditions and the following disclaimer in
+      the documentation and/or other materials provided with the
+      distribution.
+   3. Neither the name of the OpenBLAS project nor the names of
+      its contributors may be used to endorse or promote products
+      derived from this software without specific prior written
+      permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <arm_sve.h>
+#include "common.h"
+
+#ifdef DOUBLE
+#define SV_COUNT svcntd
+#define SV_TYPE svfloat64_t
+#define SV_TRUE svptrue_b64
+#define SV_WHILE svwhilelt_b64_s64
+#define SV_DUP svdup_f64
+#else
+#define SV_COUNT svcntw
+#define SV_TYPE svfloat32_t
+#define SV_TRUE svptrue_b32
+#define SV_WHILE svwhilelt_b32_s64
+#define SV_DUP svdup_f32
+#endif
+
+int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha,
+	 FLOAT *x, BLASLONG incx,
+	 FLOAT *y, BLASLONG incy,
+	 FLOAT *a, BLASLONG lda, FLOAT *buffer){
+
+  FLOAT *X    = x;
+
+  if (incx != 1) {
+    X = buffer;
+    COPY_K(m, x, incx, X, 1);
+  }
+
+  BLASLONG width = (n + 3 - 1) / 3;
+  BLASLONG i, j;
+  BLASLONG sve_size = SV_COUNT();
+
+  FLOAT *y0_ptr = y + incy * width * 0;
+  FLOAT *y1_ptr = y + incy * width * 1;
+  FLOAT *y2_ptr = y + incy * width * 2;
+
+  for (j = 0; j < width; j++) {
+    svbool_t pg00 = (j + width * 0 < n) ? SV_TRUE() : svpfalse();
+    svbool_t pg01 = (j + width * 1 < n) ? SV_TRUE() : svpfalse();
+    svbool_t pg02 = (j + width * 2 < n) ? SV_TRUE() : svpfalse();
+
+    SV_TYPE temp0_vec = (j + width * 0 < n) ? SV_DUP(alpha * *y0_ptr) : SV_DUP(0.0);
+    SV_TYPE temp1_vec = (j + width * 1 < n) ? SV_DUP(alpha * *y1_ptr) : SV_DUP(0.0);
+    SV_TYPE temp2_vec = (j + width * 2 < n) ? SV_DUP(alpha * *y2_ptr) : SV_DUP(0.0);
+
+    FLOAT *x_ptr = X;
+    FLOAT *a0_ptr = a + lda * width * 0 + lda * j;
+    FLOAT *a1_ptr = a + lda * width * 1 + lda * j;
+    FLOAT *a2_ptr = a + lda * width * 2 + lda * j;
+
+    i = 0;
+    while (i + sve_size * 1 - 1 < m) {
+      SV_TYPE x0_vec = svld1_vnum(SV_TRUE(), x_ptr, 0);
+
+      SV_TYPE a00_vec = svld1_vnum(pg00, a0_ptr, 0);
+      SV_TYPE a01_vec = svld1_vnum(pg01, a1_ptr, 0);
+      SV_TYPE a02_vec = svld1_vnum(pg02, a2_ptr, 0);
+
+      a00_vec = svmla_x(pg00, a00_vec, temp0_vec, x0_vec);
+      a01_vec = svmla_x(pg01, a01_vec, temp1_vec, x0_vec);
+      a02_vec = svmla_x(pg02, a02_vec, temp2_vec, x0_vec);
+
+      svst1_vnum(pg00, a0_ptr, 0, a00_vec);
+      svst1_vnum(pg01, a1_ptr, 0, a01_vec);
+      svst1_vnum(pg02, a2_ptr, 0, a02_vec);
+
+      i += sve_size * 1;
+      x_ptr += sve_size * 1;
+      a0_ptr += sve_size * 1;
+      a1_ptr += sve_size * 1;
+      a2_ptr += sve_size * 1;
+    }
+
+    if (i < m) {
+      svbool_t pg0 = SV_WHILE(i + sve_size * 0, m);
+
+      pg00 = svand_z(SV_TRUE(), pg0, pg00);
+      pg01 = svand_z(SV_TRUE(), pg0, pg01);
+      pg02 = svand_z(SV_TRUE(), pg0, pg02);
+
+      SV_TYPE x0_vec = svld1_vnum(pg0, x_ptr, 0);
+
+      SV_TYPE a00_vec = svld1_vnum(pg00, a0_ptr, 0);
+      SV_TYPE a01_vec = svld1_vnum(pg01, a1_ptr, 0);
+      SV_TYPE a02_vec = svld1_vnum(pg02, a2_ptr, 0);
+
+      a00_vec = svmla_x(pg00, a00_vec, temp0_vec, x0_vec);
+      a01_vec = svmla_x(pg01, a01_vec, temp1_vec, x0_vec);
+      a02_vec = svmla_x(pg02, a02_vec, temp2_vec, x0_vec);
+
+      svst1_vnum(pg00, a0_ptr, 0, a00_vec);
+      svst1_vnum(pg01, a1_ptr, 0, a01_vec);
+      svst1_vnum(pg02, a2_ptr, 0, a02_vec);
+    }
+
+    y0_ptr += incy;
+    y1_ptr += incy;
+    y2_ptr += incy;
+  }
+  
+  return 0;
+}