OpenMathLib/OpenBLAS
1
0
Fork 0
mirror of https://github.com/OpenMathLib/OpenBLAS synced 2026-06-15 07:51:43 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

Merge pull request #5022 from tingboliao/develop

Browse Source 
Replace the __riscv_vid_v_i32m2 and __riscv_vid_v_i64m2 with __riscv…_vid_v_u32m2 and __riscv_vid_v_u64m2 for riscv64-unknown-linux-gnu-gcc compiling.
This commit is contained in:
Martin Kroeker
2024-12-18 14:29:39 -08:00
committed by GitHub
parent 229d8a025e d00cc400b1
commit d3cc8c65ed
7 changed files with 40 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
kernel/riscv64/symm_lcopy_rvv_v1.c
View File

@@ -35,11 +35,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VSEV_FLOAT __riscv_vse32_v_f32m2
#define VLSEV_FLOAT __riscv_vlse32_v_f32m2
#define INT_V_T vint32m2_t
#define VID_V_INT __riscv_vid_v_i32m2
#define VID_V_INT __riscv_vid_v_u32m2
#define VADD_VX_INT __riscv_vadd_vx_i32m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i32m2_b16
#define VBOOL_T vbool16_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define VSETVL_MAX __riscv_vsetvlmax_e64m2()
@@ -48,11 +49,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VSEV_FLOAT __riscv_vse64_v_f64m2
#define VLSEV_FLOAT __riscv_vlse64_v_f64m2
#define INT_V_T vint64m2_t
#define VID_V_INT __riscv_vid_v_i64m2
#define VID_V_INT __riscv_vid_v_u64m2
#define VADD_VX_INT __riscv_vadd_vx_i64m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i64m2_b32
#define VBOOL_T vbool32_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f64m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u64m2_i64m2
#endif
// Optimizes the implementation in ../generic/symm_lcopy_4.c
@@ -70,7 +72,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
INT_V_T vindex_max, vindex;
size_t vl = VSETVL_MAX;
vindex_max = VID_V_INT(vl);
vindex_max = V_UM2_TO_IM2(VID_V_INT(vl));
for (js = n; js > 0; js -= vl, posX += vl) {
vl = VSETVL(js);
@@ -98,4 +100,3 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
return 0;
}

10
kernel/riscv64/symm_ucopy_rvv_v1.c
View File

@@ -35,11 +35,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VSEV_FLOAT __riscv_vse32_v_f32m2
#define VLSEV_FLOAT __riscv_vlse32_v_f32m2
#define INT_V_T vint32m2_t
#define VID_V_INT __riscv_vid_v_i32m2
#define VID_V_INT __riscv_vid_v_u32m2
#define VADD_VX_INT __riscv_vadd_vx_i32m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i32m2_b16
#define VBOOL_T vbool16_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define VSETVL_MAX __riscv_vsetvlmax_e64m2()
@@ -48,11 +49,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VSEV_FLOAT __riscv_vse64_v_f64m2
#define VLSEV_FLOAT __riscv_vlse64_v_f64m2
#define INT_V_T vint64m2_t
#define VID_V_INT __riscv_vid_v_i64m2
#define VID_V_INT __riscv_vid_v_u64m2
#define VADD_VX_INT __riscv_vadd_vx_i64m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i64m2_b32
#define VBOOL_T vbool32_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f64m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u64m2_i64m2
#endif
// Optimizes the implementation in ../generic/symm_ucopy_4.c
@@ -70,7 +72,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
INT_V_T vindex_max, vindex;
size_t vl = VSETVL_MAX;
vindex_max = VID_V_INT(vl);
vindex_max = V_UM2_TO_IM2(VID_V_INT(vl));
for (js = n; js > 0; js -= vl, posX += vl) {
vl = VSETVL(js);
@@ -97,4 +99,4 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
}
return 0;
}
}

9
kernel/riscv64/zhemm_ltcopy_rvv_v1.c
View File

@@ -41,7 +41,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e32_v_f32m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e32_v_f32m2x2
#define INT_V_T vint32m2_t
#define VID_V_INT __riscv_vid_v_i32m2
#define VID_V_INT __riscv_vid_v_u32m2
#define VADD_VX_INT __riscv_vadd_vx_i32m2
#define VFRSUB_VF_FLOAT __riscv_vfrsub_vf_f32m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i32m2_b16
@@ -50,6 +50,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VBOOL_T vbool16_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f32m2
#define VFMVVF_FLOAT __riscv_vfmv_v_f_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define VSETVL_MAX __riscv_vsetvlmax_e64m2()
@@ -64,7 +65,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e64_v_f64m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e64_v_f64m2x2
#define INT_V_T vint64m2_t
#define VID_V_INT __riscv_vid_v_i64m2
#define VID_V_INT __riscv_vid_v_u64m2
#define VADD_VX_INT __riscv_vadd_vx_i64m2
#define VFRSUB_VF_FLOAT __riscv_vfrsub_vf_f64m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i64m2_b32
@@ -73,6 +74,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VBOOL_T vbool32_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f64m2
#define VFMVVF_FLOAT __riscv_vfmv_v_f_f64m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u64m2_i64m2
#endif
@@ -92,7 +94,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
INT_V_T vindex_max, vindex;
size_t vl = VSETVL_MAX;
vindex_max = VID_V_INT(vl);
vindex_max = V_UM2_TO_IM2(VID_V_INT(vl));
vzero = VFMVVF_FLOAT(ZERO, vl);
for (js = n; js > 0; js -= vl, posX += vl) {
@@ -136,4 +138,3 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
return 0;
}

10
kernel/riscv64/zhemm_utcopy_rvv_v1.c
View File

@@ -41,7 +41,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e32_v_f32m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e32_v_f32m2x2
#define INT_V_T vint32m2_t
#define VID_V_INT __riscv_vid_v_i32m2
#define VID_V_INT __riscv_vid_v_u32m2
#define VADD_VX_INT __riscv_vadd_vx_i32m2
#define VFRSUB_VF_FLOAT __riscv_vfrsub_vf_f32m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i32m2_b16
@@ -50,6 +50,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VBOOL_T vbool16_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f32m2
#define VFMVVF_FLOAT __riscv_vfmv_v_f_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define VSETVL_MAX __riscv_vsetvlmax_e64m2()
@@ -64,7 +65,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e64_v_f64m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e64_v_f64m2x2
#define INT_V_T vint64m2_t
#define VID_V_INT __riscv_vid_v_i64m2
#define VID_V_INT __riscv_vid_v_u64m2
#define VADD_VX_INT __riscv_vadd_vx_i64m2
#define VFRSUB_VF_FLOAT __riscv_vfrsub_vf_f64m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i64m2_b32
@@ -73,6 +74,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VBOOL_T vbool32_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f64m2
#define VFMVVF_FLOAT __riscv_vfmv_v_f_f64m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u64m2_i64m2
#endif
@@ -90,7 +92,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
INT_V_T vindex_max, vindex;
size_t vl = VSETVL_MAX;
vindex_max = VID_V_INT(vl);
vindex_max = V_UM2_TO_IM2(VID_V_INT(vl));
vzero = VFMVVF_FLOAT(ZERO, vl);
for (js = n; js > 0; js -= vl, posX += vl) {
@@ -132,4 +134,4 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
}
return 0;
}
}

9
kernel/riscv64/zsymm_lcopy_rvv_v1.c
View File

@@ -41,11 +41,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e32_v_f32m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e32_v_f32m2x2
#define INT_V_T vint32m2_t
#define VID_V_INT __riscv_vid_v_i32m2
#define VID_V_INT __riscv_vid_v_u32m2
#define VADD_VX_INT __riscv_vadd_vx_i32m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i32m2_b16
#define VBOOL_T vbool16_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define VSETVL_MAX __riscv_vsetvlmax_e64m2()
@@ -60,11 +61,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e64_v_f64m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e64_v_f64m2x2
#define INT_V_T vint64m2_t
#define VID_V_INT __riscv_vid_v_i64m2
#define VID_V_INT __riscv_vid_v_u64m2
#define VADD_VX_INT __riscv_vadd_vx_i64m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i64m2_b32
#define VBOOL_T vbool32_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f64m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u64m2_i64m2
#endif
int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLONG posY, FLOAT *b)
@@ -81,7 +83,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
INT_V_T vindex_max, vindex;
size_t vl = VSETVL_MAX;
vindex_max = VID_V_INT(vl);
vindex_max = V_UM2_TO_IM2(VID_V_INT(vl));
for (js = n; js > 0; js -= vl, posX += vl) {
vl = VSETVL(js);
@@ -118,4 +120,3 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
return 0;
}

10
kernel/riscv64/zsymm_ucopy_rvv_v1.c
View File

@@ -41,11 +41,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e32_v_f32m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e32_v_f32m2x2
#define INT_V_T vint32m2_t
#define VID_V_INT __riscv_vid_v_i32m2
#define VID_V_INT __riscv_vid_v_u32m2
#define VADD_VX_INT __riscv_vadd_vx_i32m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i32m2_b16
#define VBOOL_T vbool16_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define VSETVL_MAX __riscv_vsetvlmax_e64m2()
@@ -60,11 +61,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VLSSEG2_FLOAT __riscv_vlsseg2e64_v_f64m2x2
#define VSSEG2_FLOAT __riscv_vsseg2e64_v_f64m2x2
#define INT_V_T vint64m2_t
#define VID_V_INT __riscv_vid_v_i64m2
#define VID_V_INT __riscv_vid_v_u64m2
#define VADD_VX_INT __riscv_vadd_vx_i64m2
#define VMSGT_VX_INT __riscv_vmsgt_vx_i64m2_b32
#define VBOOL_T vbool32_t
#define VMERGE_VVM_FLOAT __riscv_vmerge_vvm_f64m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u64m2_i64m2
#endif
@@ -83,7 +85,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
size_t vl = VSETVL_MAX;
vindex_max = VID_V_INT(vl);
vindex_max = V_UM2_TO_IM2(VID_V_INT(vl));
for (js = n; js > 0; js -= vl, posX += vl) {
vl = VSETVL(js);
@@ -118,4 +120,4 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
}
return 0;
}
}

12
kernel/riscv64/ztrmm_lncopy_rvv_v1.c
View File

@@ -42,10 +42,11 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VSSEG2_FLOAT __riscv_vsseg2e32_v_f32m2x2
#define VBOOL_T vbool16_t
#define UINT_V_T vint32m2_t
#define VID_V_UINT __riscv_vid_v_i32m2
#define VID_V_UINT __riscv_vid_v_u32m2
#define VMSGTU_VX_UINT __riscv_vmsgt_vx_i32m2_b16
#define VMSEQ_VX_UINT __riscv_vmseq_vx_i32m2_b16
#define VFMERGE_VFM_FLOAT __riscv_vfmerge_vfm_f32m2
#define V_UM2_TO_IM2 __riscv_vreinterpret_v_u32m2_i32m2
#else
#define VSETVL(n) __riscv_vsetvl_e64m2(n)
#define FLOAT_V_T vfloat64m2_t
@@ -63,6 +64,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define VMSGTU_VX_UINT __riscv_vmsgtu_vx_u64m2_b32
#define VMSEQ_VX_UINT __riscv_vmseq_vx_u64m2_b32
#define VFMERGE_VFM_FLOAT __riscv_vfmerge_vfm_f64m2
#define V_UM2_TO_IM2(values) values
#endif
int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLONG posY, FLOAT *b){
@@ -99,7 +101,7 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
}
i = 0;
do
do
{
if (X > posY)
{
@@ -119,9 +121,9 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
X ++;
i ++;
}
else
else
{
vindex = VID_V_UINT(vl);
vindex = V_UM2_TO_IM2(VID_V_UINT(vl));
for (unsigned int j = 0; j < vl; j++)
{
vax2 = VLSSEG2_FLOAT(ao, stride_lda, vl);
@@ -152,4 +154,4 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLON
}
return 0;
}
}