Merge pull request #5265 from guoyuanplct/develop

kernel/riscv64:Added support for omatcopy on RISCV64_ZVL256B

.github/workflows/c910v.yml

@@ -83,9 +83,39 @@ jobs:
 
       - name: test
         run: |
-          export PATH=$GITHUB_WORKSPACE/qemu-install/bin/:$PATH
-          qemu-riscv64 ./utest/openblas_utest
-          qemu-riscv64 ./utest/openblas_utest_ext
+          run_with_retry() {
+              local cmd="$1"
+              local time_out=10
+              local retries=10
+              local attempt=0
+
+              for ((i=1; i<=retries; i++)); do
+                  attempt=$((i))
+                  if timeout -s 12  --preserve-status $time_out $cmd; then
+                      echo "Command succeeded on attempt $i."
+                      return 0
+                  else
+                      local exit_code=$?
+                      if [ $exit_code -eq 140 ]; then
+                          echo "Attempt $i timed out (retrying...)"
+                          time_out=$((time_out + 5))
+                      else
+                          echo "Attempt $i failed with exit code $exit_code. Aborting workflow."
+                          exit $exit_code
+                      fi
+                  fi
+              done
+              echo "All $retries attempts failed, giving up."
+              echo "Final failure was due to timeout."
+              echo "Aborting workflow."
+              exit $exit_code
+          }
+          export PATH=$GITHUB_WORKSPACE/qemu-install/bin:$PATH
+          which qemu-riscv64
+          export QEMU_BIN=$(which qemu-riscv64)
+          run_with_retry "$QEMU_BIN ./utest/openblas_utest"
+          run_with_retry "$QEMU_BIN ./utest/openblas_utest_ext"
+
           OPENBLAS_NUM_THREADS=2 qemu-riscv64 ./ctest/xscblat1
           OPENBLAS_NUM_THREADS=2 qemu-riscv64 ./ctest/xdcblat1
           OPENBLAS_NUM_THREADS=2 qemu-riscv64 ./ctest/xccblat1

kernel/riscv64/KERNEL.RISCV64_ZVL256B

@@ -201,3 +201,9 @@ endif
 ifndef ZGEMM_BETA
 ZGEMM_BETA = ../generic/zgemm_beta.c
 endif
+
+ZOMATCOPY_CN = zomatcopy_cn_vector.c
+COMATCOPY_CN = zomatcopy_cn_vector.c
+
+DOMATCOPY_CN = omatcopy_cn_vector.c
+SOMATCOPY_CN = omatcopy_cn_vector.c

kernel/riscv64/omatcopy_cn_vector.c

@@ -0,0 +1,123 @@
+/***************************************************************************
+Copyright (c) 2013, 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 "common.h"
+#if !defined(DOUBLE)
+#define VSETVL_MAX				RISCV_RVV(vsetvlmax_e32m4)()
+#define VSETVL(n)               RISCV_RVV(vsetvl_e32m4)(n)
+#define FLOAT_V_T               vfloat32m4_t
+#define VLEV_FLOAT              RISCV_RVV(vle32_v_f32m4)
+#define VSEV_FLOAT              RISCV_RVV(vse32_v_f32m4)
+#define VFMULVF_FLOAT           RISCV_RVV(vfmul_vf_f32m4)
+#define VFMVVF_FLOAT            RISCV_RVV(vfmv_v_f_f32m4)
+#else
+#define VSETVL_MAX				RISCV_RVV(vsetvlmax_e64m4)()
+#define VSETVL(n)               RISCV_RVV(vsetvl_e64m4)(n)
+#define FLOAT_V_T               vfloat64m4_t
+#define VLEV_FLOAT              RISCV_RVV(vle64_v_f64m4)
+#define VSEV_FLOAT              RISCV_RVV(vse64_v_f64m4)
+#define VFMULVF_FLOAT           RISCV_RVV(vfmul_vf_f64m4)
+#define VFMVVF_FLOAT            RISCV_RVV(vfmv_v_f_f64m4)
+#endif
+
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG ldb)
+{
+	BLASLONG i,j;
+	FLOAT *aptr,*bptr;
+	size_t vl;
+
+	FLOAT_V_T va, vb,va1,vb1;
+	if ( rows <= 0 )  return(0);
+	if ( cols <= 0 )  return(0);
+
+	aptr = a;
+	bptr = b;
+
+	if ( alpha == 0.0 )
+	{
+		vl = VSETVL_MAX;
+		va = VFMVVF_FLOAT(0, vl);
+		for ( i=0; i<cols ; i++ )
+		{
+			for(j=0; j<rows; j+=vl)
+			{
+				vl = VSETVL(rows - j);
+				VSEV_FLOAT(bptr + j, va, vl);
+			}
+			bptr += ldb;
+		}
+		return(0);
+	}
+
+	if ( alpha == 1.0 )
+	{
+		for ( i=0; i<cols ; i++ )
+		{
+			for(j=0; j<rows; j+=vl)
+			{
+				vl = VSETVL(rows - j);
+				va = VLEV_FLOAT(aptr + j, vl);
+				VSEV_FLOAT(bptr + j, va, vl);
+			}
+			aptr += lda;
+			bptr += ldb;
+		}
+		return(0);
+	}
+	i = 0;
+	if( cols % 2  ){
+		
+		for(j=0; j<rows; j+=vl)
+		{
+			vl = VSETVL(rows - j);
+			va = VLEV_FLOAT(aptr + j, vl);
+			va = VFMULVF_FLOAT(va, alpha, vl);
+			VSEV_FLOAT(bptr + j, va, vl);
+		}
+		aptr +=  lda;
+		bptr +=  ldb;
+		i = 1;
+	}
+	for ( ; i<cols ; i+=2 )
+	{
+		for(j=0; j<rows; j+=vl)
+		{
+			vl = VSETVL(rows - j);
+			va = VLEV_FLOAT(aptr + j, vl);
+			va1= VLEV_FLOAT(aptr + lda + j, vl);
+			va = VFMULVF_FLOAT(va, alpha, vl);
+			va1= VFMULVF_FLOAT(va1, alpha, vl);
+			VSEV_FLOAT(bptr + j, va, vl);
+			VSEV_FLOAT(bptr + ldb + j, va1, vl);
+		}
+		aptr += 2 * lda;
+		bptr += 2 * ldb;
+	}
+
+	return(0);
+}

kernel/riscv64/zomatcopy_cn_vector.c

@@ -0,0 +1,106 @@
+/***************************************************************************
+Copyright (c) 2013, 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 "common.h"
+
+
+#if !defined(DOUBLE)
+#define VSETVL(n) RISCV_RVV(vsetvl_e32m4)(n)
+#define FLOAT_V_T vfloat32m4_t
+#define VLEV_FLOAT RISCV_RVV(vle32_v_f32m4)
+#define VLSEV_FLOAT RISCV_RVV(vlse32_v_f32m4)
+#define VSEV_FLOAT RISCV_RVV(vse32_v_f32m4)
+#define VSSEV_FLOAT RISCV_RVV(vsse32_v_f32m4)
+#define VFMACCVF_FLOAT RISCV_RVV(vfmacc_vf_f32m4)
+#define VFNMSACVF_FLOAT RISCV_RVV(vfnmsac_vf_f32m4)
+#define VFMUL_VF_FLOAT RISCV_RVV(vfmul_vf_f32m4)
+#define VSEV_FLOAT RISCV_RVV(vse32_v_f32m4)
+#define VLSEG2_FLOAT RISCV_RVV(vlseg2e32_v_f32m4x2)
+#define VSSEG2_FLOAT RISCV_RVV(vsseg2e32_v_f32m4x2)
+#define FLOAT_VX2_T vfloat32m4x2_t
+#define VGET_VX2 RISCV_RVV(vget_v_f32m4x2_f32m4)
+#define VSET_VX2 RISCV_RVV(vset_v_f32m4_f32m4x2)
+#else
+#define VSETVL(n) RISCV_RVV(vsetvl_e64m4)(n)
+#define FLOAT_V_T vfloat64m4_t
+#define VLEV_FLOAT RISCV_RVV(vle64_v_f64m4)
+#define VLSEV_FLOAT RISCV_RVV(vlse64_v_f64m4)
+#define VSEV_FLOAT RISCV_RVV(vse64_v_f64m4)
+#define VSSEV_FLOAT RISCV_RVV(vsse64_v_f64m4)
+#define VFMACCVF_FLOAT RISCV_RVV(vfmacc_vf_f64m4)
+#define VFNMSACVF_FLOAT RISCV_RVV(vfnmsac_vf_f64m4)
+#define VFMUL_VF_FLOAT RISCV_RVV(vfmul_vf_f64m4)
+#define VSEV_FLOAT RISCV_RVV(vse64_v_f64m4)
+#define VLSEG2_FLOAT RISCV_RVV(vlseg2e64_v_f64m4x2)
+#define VSSEG2_FLOAT RISCV_RVV(vsseg2e64_v_f64m4x2)
+#define FLOAT_VX2_T vfloat64m4x2_t
+#define VGET_VX2 RISCV_RVV(vget_v_f64m4x2_f64m4)
+#define VSET_VX2 RISCV_RVV(vset_v_f64m4_f64m4x2)
+#endif
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG ldb)
+{
+	BLASLONG i,j,ia;
+	FLOAT *aptr,*bptr;
+	FLOAT_V_T bptr_v0 , bptr_v1 , aptr_v0 ,aptr_v1;
+	FLOAT_VX2_T va, vb;
+	unsigned int gvl = 0;
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+	bptr = b;
+
+	lda *= 2;
+	ldb *= 2;
+	for ( i=0; i<cols ; i++ )
+	{
+		ia = 0;
+		for(j=0; j<rows ; j+=gvl)
+		{
+			gvl = VSETVL(rows - j);
+			va = VLSEG2_FLOAT(aptr + ia, gvl);
+			aptr_v0 = VGET_VX2(va, 0);
+			aptr_v1 = VGET_VX2(va, 1);
+			bptr_v1 = VFMUL_VF_FLOAT( aptr_v1, alpha_r,gvl);
+			bptr_v1 = VFMACCVF_FLOAT(bptr_v1, alpha_i, aptr_v0, gvl);
+			bptr_v0 = VFMUL_VF_FLOAT(  aptr_v0,alpha_r, gvl);
+			bptr_v0 = VFNMSACVF_FLOAT(bptr_v0, alpha_i, aptr_v1, gvl);
+			vb = VSET_VX2(vb, 0, bptr_v0);
+			vb = VSET_VX2(vb, 1, bptr_v1);
+			VSSEG2_FLOAT(&bptr[ia], vb, gvl);
+			ia += gvl * 2 ;
+
+		}
+		aptr += lda;
+		bptr += ldb;
+	}
+
+	return(0);
+
+}