floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update sgemm kernel 1x4 for C910.

This commit is contained in:
Xianyi Zhang 2020-12-18 11:53:23 +08:00
parent 7834c10e2f
commit a3cac9cca0
1 changed files with 94 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

109
kernel/riscv64/sgemm_kernel_16x4_c910v.c
View File

@ -382,7 +382,7 @@ int CNAME(BLASLONG bm,BLASLONG bn,BLASLONG bk,FLOAT alpha,FLOAT* ba,FLOAT* bb,FL
{ {
BLASLONG i,j,k; BLASLONG i,j,k;
FLOAT *C0,*C1,*C2,*C3; FLOAT *C0,*C1,*C2,*C3;
FLOAT *ptrba,*ptrbb; FLOAT *ptrba,*ptrbb, *tmpc;
FLOAT loadb0,loadb1,loadb2,loadb3; FLOAT loadb0,loadb1,loadb2,loadb3;
FLOAT load0,load1,load2,load3,load4,load5,load6,load7; FLOAT load0,load1,load2,load3,load4,load5,load6,load7;
@ -392,6 +392,7 @@ int CNAME(BLASLONG bm,BLASLONG bn,BLASLONG bk,FLOAT alpha,FLOAT* ba,FLOAT* bb,FL
FLOAT res8,res9,res10,res11; FLOAT res8,res9,res10,res11;
FLOAT res12,res13,res14,res15; FLOAT res12,res13,res14,res15;
for (j=0; j<bn/4; j+=1){ for (j=0; j<bn/4; j+=1){
C0 = C; C0 = C;
C1 = C0+ldc; C1 = C0+ldc;
@ -942,53 +943,109 @@ int CNAME(BLASLONG bm,BLASLONG bn,BLASLONG bk,FLOAT alpha,FLOAT* ba,FLOAT* bb,FL
} }
if(bm&1){ if(bm&1){
ptrbb = bb; ptrbb = bb;
//t0 for k
//ft0-ft3,ft4-ft7,v8-v15 for B, t1-t3 for PB1-3
//v0-v3,v4-v7 for A, t4-t6 for PA1-3
//v16-v31 for temp C
res0 = 0; FLOAT tmp[4];
tmpc=tmp;
//t1-t3 for PB
//v0-v4 for A, v8-v11 for B
//v16-v19 for C
asm volatile(
"vsetvli zero, zero, e32,m1 \n\t"
"fmv.w.x ft11, zero \n\t"
res4 = 0; "vfmv.v.f v16, ft11 \n\t"
"vfmv.v.f v17, ft11 \n\t"
"vfmv.v.f v18, ft11 \n\t"
"vfmv.v.f v19, ft11 \n\t"
//unloop 4
res8 = 0; "srli t0, %[BK], 2 \n\t"
"blez t0, M1x4_TAIL \n\t"
res12 = 0; "addi t1, %[PB], 4*4 \n\t"
"addi t2, %[PB], 8*4 \n\t"
"addi t3, %[PB], 12*4 \n\t"
for(k=0; k<bk; k+=1){ ".align 4 \n\t"
loadb0 = ptrbb[0]; "M1x4_MAINLOOP: \n\t"
loadb1 = ptrbb[1];
load0 = ptrba[0]; "vle.v v4, (%[PA]) \n\t"
"addi %[PA], %[PA], 4*4 \n\t"
"vrgather.vi v0, v4, 0 \n\t"
res0 = res0 + load0 * loadb0; "vle.v v8, (%[PB]) \n\t"
"addi %[PB], %[PB], 16*4 \n\t"
"vrgather.vi v1, v4, 1 \n\t"
res4 = res4 + load0 * loadb1; "vle.v v9, (t1) \n\t"
"addi t1, t1, 16*4 \n\t"
"vrgather.vi v2, v4, 2 \n\t"
loadb2 = ptrbb[2]; "vle.v v10, (t2) \n\t"
loadb3 = ptrbb[3]; "addi t2, t2, 16*4 \n\t"
"vrgather.vi v3, v4, 3 \n\t"
res8 = res8 + load0 * loadb2; "vle.v v11, (t3) \n\t"
"addi t3, t3, 16*4 \n\t"
res12 = res12 + load0 * loadb3; "vfmacc.vv v16, v8, v0 \n\t"
"vfmacc.vv v17, v9, v1 \n\t"
"vfmacc.vv v18, v10, v2 \n\t"
"vfmacc.vv v19, v11, v3 \n\t"
ptrba += 1; "addi t0, t0, -1 \n\t"
ptrbb += 4; "bgtz t0, M1x4_MAINLOOP \n\t"
}
res0 = res0 * alpha; "M1x4_TAIL: \n\t"
"andi t0, %[BK], 3 \n\t"
"blez t0, M1x4_SAVERESULT \n\t"
res4 = res4 * alpha; "M1x4_TAILLOOP: \n\t"
"flw ft0, (%[PA]) \n\t"
"addi %[PA], %[PA], 1*4 \n\t"
"vle.v v8, (%[PB]) \n\t"
"addi %[PB], %[PB], 4*4 \n\t"
"vfmv.v.f v0, ft0 \n\t"
"vfmacc.vv v16, v8, v0 \n\t"
res8 = res8 * alpha; "addi t0, t0, -1 \n\t"
"bgtz t0, M1x4_TAILLOOP \n\t"
res12 = res12 * alpha; "M1x4_SAVERESULT: \n\t"
//merge v16-v19
"vfadd.vv v16, v16, v17 \n\t"
"vfadd.vv v18, v18, v19 \n\t"
"vfadd.vv v16, v16, v18 \n\t"
C0[0] += res0; "vfmv.v.f v8, %[ALPHA] \n\t"
C1[0] += res4; "vfmul.vv v16, v8, v16 \n\t"
C2[0] += res8; "vse.v v16, (%[TMP_C]) \n\t"
C3[0] += res12; "M1x4_END: \n\t"
:[TMP_C]"+r"(tmpc),
[PA]"+r"(ptrba), [PB]"+r"(ptrbb)
:[ALPHA]"f"(alpha), [BK]"r"(bk)
:"cc", "t0", "t3","t1","t2",
"ft0", "ft11",
"v0", "v1", "v2", "v3","v4",
"v8", "v9", "v10", "v11",
"v16", "v17","v18", "v19"
);
C0[0] += tmp[0];
C1[0] += tmp[1];
C2[0] += tmp[2];
C3[0] += tmp[3];
/* don't need move c point
C0 += 1; C0 += 1;
C1 += 1; C1 += 1;
C2 += 1; C2 += 1;
C3 += 1; C3 += 1;
*/
} }
k = bk<<2; k = bk<<2;