lib/raid6/loongarch_simd.c

26 #define NSIZE 16
52 	for (d = 0; d < bytes; d += NSIZE*4) {
53 		/* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
54 		asm volatile("vld $vr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
55 		asm volatile("vld $vr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
56 		asm volatile("vld $vr2, %0" : : "m"(dptr[z0][d+2*NSIZE]));
57 		asm volatile("vld $vr3, %0" : : "m"(dptr[z0][d+3*NSIZE]));
63 			/* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
64 			asm volatile("vld $vr8, %0" : : "m"(dptr[z][d+0*NSIZE]));
65 			asm volatile("vld $vr9, %0" : : "m"(dptr[z][d+1*NSIZE]));
66 			asm volatile("vld $vr10, %0" : : "m"(dptr[z][d+2*NSIZE]));
67 			asm volatile("vld $vr11, %0" : : "m"(dptr[z][d+3*NSIZE]));
99 		/* *(unative_t *)&p[d+NSIZE*$$] = wp$$; */
100 		asm volatile("vst $vr0, %0" : "=m"(p[d+NSIZE*0]));
101 		asm volatile("vst $vr1, %0" : "=m"(p[d+NSIZE*1]));
102 		asm volatile("vst $vr2, %0" : "=m"(p[d+NSIZE*2]));
103 		asm volatile("vst $vr3, %0" : "=m"(p[d+NSIZE*3]));
104 		/* *(unative_t *)&q[d+NSIZE*$$] = wq$$; */
105 		asm volatile("vst $vr4, %0" : "=m"(q[d+NSIZE*0]));
106 		asm volatile("vst $vr5, %0" : "=m"(q[d+NSIZE*1]));
107 		asm volatile("vst $vr6, %0" : "=m"(q[d+NSIZE*2]));
108 		asm volatile("vst $vr7, %0" : "=m"(q[d+NSIZE*3]));
134 	for (d = 0; d < bytes; d += NSIZE*4) {
136 		/* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
137 		asm volatile("vld $vr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
138 		asm volatile("vld $vr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
139 		asm volatile("vld $vr2, %0" : : "m"(dptr[z0][d+2*NSIZE]));
140 		asm volatile("vld $vr3, %0" : : "m"(dptr[z0][d+3*NSIZE]));
146 			/* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
147 			asm volatile("vld $vr8, %0" : : "m"(dptr[z][d+0*NSIZE]));
148 			asm volatile("vld $vr9, %0" : : "m"(dptr[z][d+1*NSIZE]));
149 			asm volatile("vld $vr10, %0" : : "m"(dptr[z][d+2*NSIZE]));
150 			asm volatile("vld $vr11, %0" : : "m"(dptr[z][d+3*NSIZE]));
207 		 * *(unative_t *)&p[d+NSIZE*$$] ^= wp$$;
208 		 * *(unative_t *)&q[d+NSIZE*$$] ^= wq$$;
235 			: "+m"(p[d+NSIZE*0]), "+m"(p[d+NSIZE*1]),
236 			  "+m"(p[d+NSIZE*2]), "+m"(p[d+NSIZE*3]),
237 			  "+m"(q[d+NSIZE*0]), "+m"(q[d+NSIZE*1]),
238 			  "+m"(q[d+NSIZE*2]), "+m"(q[d+NSIZE*3])
253 #undef NSIZE
257 #define NSIZE 32
283 	for (d = 0; d < bytes; d += NSIZE*2) {
284 		/* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
285 		asm volatile("xvld $xr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
286 		asm volatile("xvld $xr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
290 			/* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
291 			asm volatile("xvld $xr4, %0" : : "m"(dptr[z][d+0*NSIZE]));
292 			asm volatile("xvld $xr5, %0" : : "m"(dptr[z][d+1*NSIZE]));
312 		/* *(unative_t *)&p[d+NSIZE*$$] = wp$$; */
313 		asm volatile("xvst $xr0, %0" : "=m"(p[d+NSIZE*0]));
314 		asm volatile("xvst $xr1, %0" : "=m"(p[d+NSIZE*1]));
315 		/* *(unative_t *)&q[d+NSIZE*$$] = wq$$; */
316 		asm volatile("xvst $xr2, %0" : "=m"(q[d+NSIZE*0]));
317 		asm volatile("xvst $xr3, %0" : "=m"(q[d+NSIZE*1]));
343 	for (d = 0; d < bytes; d += NSIZE*2) {
345 		/* wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE]; */
346 		asm volatile("xvld $xr0, %0" : : "m"(dptr[z0][d+0*NSIZE]));
347 		asm volatile("xvld $xr1, %0" : : "m"(dptr[z0][d+1*NSIZE]));
351 			/* wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE]; */
352 			asm volatile("xvld $xr4, %0" : : "m"(dptr[z][d+0*NSIZE]));
353 			asm volatile("xvld $xr5, %0" : : "m"(dptr[z][d+1*NSIZE]));
390 		 * *(unative_t *)&p[d+NSIZE*$$] ^= wp$$;
391 		 * *(unative_t *)&q[d+NSIZE*$$] ^= wq$$;
406 			: "+m"(p[d+NSIZE*0]), "+m"(p[d+NSIZE*1]),
407 			  "+m"(q[d+NSIZE*0]), "+m"(q[d+NSIZE*1])
421 #undef NSIZE