28 gv100_gr_trap_sm(struct gf100_gr *gr, int gpc, int tpc, int sm)  in gv100_gr_trap_sm()  argument
32 	u32 werr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x730 + (sm * 0x80)));  in gv100_gr_trap_sm()
33 	u32 gerr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x734 + (sm * 0x80)));  in gv100_gr_trap_sm()
40 	nvkm_error(subdev, "GPC%i/TPC%i/SM%d trap: "  in gv100_gr_trap_sm()
42 		   gpc, tpc, sm, gerr, glob, werr, warp ? warp->name : "");  in gv100_gr_trap_sm()
44 	nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x730 + sm * 0x80), 0x00000000);  in gv100_gr_trap_sm()
45 	nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x734 + sm * 0x80), gerr);  in gv100_gr_trap_sm()
49 gv100_gr_trap_mp(struct gf100_gr *gr, int gpc, int tpc)  in gv100_gr_trap_mp()  argument
51 	gv100_gr_trap_sm(gr, gpc, tpc, 0);  in gv100_gr_trap_mp()
52 	gv100_gr_trap_sm(gr, gpc, tpc, 1);  in gv100_gr_trap_mp()
64 gv100_gr_init_shader_exceptions(struct gf100_gr *gr, int gpc, int tpc)  in gv100_gr_init_shader_exceptions()  argument
69 		nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x610), 0x00000001);  in gv100_gr_init_shader_exceptions()
70 		nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x72c + sm), 0x00000004);  in gv100_gr_init_shader_exceptions()
75 gv100_gr_init_504430(struct gf100_gr *gr, int gpc, int tpc)  in gv100_gr_init_504430()  argument
78 	nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x430), 0x403f0000);  in gv100_gr_init_504430()
89 gv100_gr_nonpes_aware_tpc(struct gf100_gr *gr, u32 gpc, u32 tpc)  in gv100_gr_nonpes_aware_tpc()  argument
93 	for (pes = 0; pes < gr->ppc_nr[gpc]; pes++) {  in gv100_gr_nonpes_aware_tpc()
94 		if (gr->ppc_tpc_mask[gpc][pes] & BIT(tpc))  in gv100_gr_nonpes_aware_tpc()
97 		tpc_new += gr->ppc_tpc_nr[gpc][pes];  in gv100_gr_nonpes_aware_tpc()
100 	temp = (BIT(tpc) - 1) & gr->ppc_tpc_mask[gpc][pes];  in gv100_gr_nonpes_aware_tpc()
115 	u32 average_tpcs = 0; /* Average of # of TPCs per GPC */  in gv100_gr_scg_estimate_perf()
121 	u32 gpc;  in gv100_gr_scg_estimate_perf()  local
134 	/* Calculate pix-perf-reduction-rate per GPC and find bottleneck TPC */  in gv100_gr_scg_estimate_perf()
135 	for (gpc = 0; gpc < gr->gpc_nr; gpc++) {  in gv100_gr_scg_estimate_perf()
136 		num_tpc_mask = gpc_tpc_mask[gpc];  in gv100_gr_scg_estimate_perf()
138 		if ((gpc == disable_gpc) && num_tpc_mask & BIT(disable_tpc)) {  in gv100_gr_scg_estimate_perf()
149 		num_tpc_gpc[gpc] = hweight32(num_tpc_mask);  in gv100_gr_scg_estimate_perf()
150 		average_tpcs += num_tpc_gpc[gpc];  in gv100_gr_scg_estimate_perf()
153 		max_tpc_gpc = num_tpc_gpc[gpc] > max_tpc_gpc ? num_tpc_gpc[gpc] : max_tpc_gpc;  in gv100_gr_scg_estimate_perf()
158 		 * ratio represents relative throughput of the GPC  in gv100_gr_scg_estimate_perf()
160 		scg_gpc_pix_perf = scale_factor * num_tpc_gpc[gpc] / gr->tpc_nr[gpc];  in gv100_gr_scg_estimate_perf()
165 		for (pes = 0; pes < gr->ppc_nr[gpc]; pes++) {  in gv100_gr_scg_estimate_perf()
167 			num_tpc_mask = gr->ppc_tpc_mask[gpc][pes] & gpc_tpc_mask[gpc];  in gv100_gr_scg_estimate_perf()
169 			if ((gpc == disable_gpc) && (num_tpc_mask & BIT(disable_tpc))) {  in gv100_gr_scg_estimate_perf()
194 	for (gpc = 0; gpc < gr->gpc_nr; gpc++) {  in gv100_gr_scg_estimate_perf()
195 		diff = average_tpcs - scale_factor * num_tpc_gpc[gpc];  in gv100_gr_scg_estimate_perf()
231 	u32 gpc, tpc, pes, gtpc;  in gv100_gr_oneinit_sm_id()  local
242 	for (gpc = 0; gpc < gr->gpc_nr; gpc++) {  in gv100_gr_oneinit_sm_id()
243 		for (pes = 0; pes < gr->ppc_nr[gpc]; pes++)  in gv100_gr_oneinit_sm_id()
244 			gpc_tpc_mask[gpc] |= gr->ppc_tpc_mask[gpc][pes];  in gv100_gr_oneinit_sm_id()
248 		for (maxperf = -1, gpc = 0; gpc < gr->gpc_nr; gpc++) {  in gv100_gr_oneinit_sm_id()
249 			for_each_set_bit(tpc, &gpc_tpc_mask[gpc], gr->tpc_nr[gpc]) {  in gv100_gr_oneinit_sm_id()
250 				ret = gv100_gr_scg_estimate_perf(gr, gpc_tpc_mask, gpc, tpc, &perf);  in gv100_gr_oneinit_sm_id()
257 					gpc_table[gtpc] = gpc;  in gv100_gr_oneinit_sm_id()
268 		gr->sm[gtpc].gpc = gpc_table[gtpc];  in gv100_gr_oneinit_sm_id()