Add VCN (Video Core Next) monitoring for RDNA

detect.c

@@ -361,6 +361,7 @@ void initbits(int fam) {
 	bits.gui = (1U << 31);
 	bits.uvd = 0;
 	bits.vce0 = 0;
+	bits.vcn = 0;
 
 	// R600 has a different texture bit, and only R600 has the TC, CR, SMX bits
 	if (fam < RV770) {
@@ -377,4 +378,17 @@ void initbits(int fam) {
 			bits.vce0 = (1U << 7);
 		}
 	}
+
+	// Video Core (VCN) on RDNA: bit 1 of SRBM_STATUS2
+	if (fam >= NAVI10) {
+		bits.vcn = (1U << 1);
+	}
+
+	// RDNA (GFX10+): VGT replaced by Geometry Engine at bit 21,
+	// Event Engine and Sequencer Instruction Cache bits undefined.
+	if (fam >= NAVI10) {
+		bits.ee = 0;
+		bits.vgt = (1U << 21);
+		bits.sh = 0;
+	}
 }

dump.c

@@ -98,6 +98,7 @@ void dumpdata(const unsigned int ticks, const char file[], const unsigned int li
 		float cb = 100 * results->cb * k;
 		float uvd = 100 * results->uvd * k;
 		float vce0 = 100 * results->vce0 * k;
+		float vcn = 100 * results->vcn * k;
 		float vram = 100.0f * results->vram / vramsize;
 		float vrammb = results->vram / 1024.0f / 1024.0f;
 		float gtt = 100.0f * results->gtt / gttsize;
@@ -108,15 +109,17 @@ void dumpdata(const unsigned int ticks, const char file[], const unsigned int li
 		float sclk_ghz = results->sclk * k / 1000.0f;
 
 		fprintf(f, "gpu %.2f%%, ", gui);
-		fprintf(f, "ee %.2f%%, ", ee);
-		fprintf(f, "vgt %.2f%%, ", vgt);
+		if (bits.ee)
+			fprintf(f, "ee %.2f%%, ", ee);
+		fprintf(f, "%s %.2f%%, ", bits.ee ? "vgt" : "ge", vgt);
 		fprintf(f, "ta %.2f%%, ", ta);
 
 		if (bits.tc)
 			fprintf(f, "tc %.2f%%, ", tc);
 
 		fprintf(f, "sx %.2f%%, ", sx);
-		fprintf(f, "sh %.2f%%, ", sh);
+		if (bits.sh)
+			fprintf(f, "sh %.2f%%, ", sh);
 		fprintf(f, "spi %.2f%%, ", spi);
 
 		if (bits.smx)
@@ -133,6 +136,8 @@ void dumpdata(const unsigned int ticks, const char file[], const unsigned int li
 			fprintf(f, ", uvd %.2f%%", uvd);
 		if (bits.vce0)
 			fprintf(f, ", vce0 %.2f%%", vce0);
+		if (bits.vcn)
+			fprintf(f, ", vc %.2f%%", vcn);
 
 		if (bits.vram)
 			fprintf(f, ", vram %.2f%% %.2fmb", vram, vrammb);

include/radeontop.h

@@ -160,6 +160,7 @@ struct bits_t {
 	unsigned int cr;
 	unsigned int uvd;
 	unsigned int vce0;
+	unsigned int vcn;
 	uint64_t vram;
 	uint64_t gtt;
 	unsigned int sclk;

ticks.c

@@ -43,8 +43,8 @@ static void *collector(void *arg) {
 		getgrbm(&stat);
 		unsigned int uvd;
 		if (bits.uvd) getsrbm(&uvd);
-		unsigned int vce;
-		if (bits.vce0) getsrbm2(&vce);
+		unsigned int srbm2;
+		if (bits.vce0 || bits.vcn) getsrbm2(&srbm2);
 
 		memset(&history[cur], 0, sizeof(struct bits_t));
 
@@ -63,7 +63,10 @@ static void *collector(void *arg) {
 		if (stat & bits.cr) history[cur].cr = 1;
 		if (stat & bits.cb) history[cur].cb = 1;
 		if (uvd & bits.uvd) history[cur].uvd = 1;
-		if (vce & bits.vce0) history[cur].vce0 = 1;
+		if (bits.vce0 || bits.vcn) {
+			if (srbm2 & bits.vce0) history[cur].vce0 = 1;
+			if (srbm2 & bits.vcn) history[cur].vcn = 1;
+		}
 		getsclk(&history[cur].sclk);
 		getmclk(&history[cur].mclk);
 
@@ -94,6 +97,7 @@ static void *collector(void *arg) {
 				res[curres].cr += history[i].cr;
 				res[curres].uvd += history[i].uvd;
 				res[curres].vce0 += history[i].vce0;
+				res[curres].vcn += history[i].vcn;
 				res[curres].mclk += history[i].mclk;
 				res[curres].sclk += history[i].sclk;
 			}

ui.c

@@ -153,6 +153,7 @@ void present(const unsigned int ticks, const char card[], unsigned int color,
 		float cb = 100 * results->cb * k;
 		float uvd = 100 * results->uvd * k;
 		float vce0 = 100 * results->vce0 * k;
+		float vcn = 100 * results->vcn * k;
 		float vram = 100.0f * results->vram / vramsize;
 		float vrammb = results->vram / 1024.0f / 1024.0f;
 		float vramsizemb = vramsize / 1024.0f / 1024.0f;
@@ -175,15 +176,20 @@ void present(const unsigned int ticks, const char card[], unsigned int color,
 
 		unsigned int start = 4;
 
-		percentage(start, w, ee);
-		printright(start++, hw, _("Event Engine %6.2f%%"), ee);
+		if (bits.ee) {
+			percentage(start, w, ee);
+			printright(start++, hw, _("Event Engine %6.2f%%"), ee);
 
-		// Enough height?
-		if (h > bigh) start++;
+			// Enough height?
+			if (h > bigh) start++;
+		}
 
 		if (color) attron(COLOR_PAIR(2));
 		percentage(start, w, vgt);
-		printright(start++, hw, _("Vertex Grouper + Tesselator %6.2f%%"), vgt);
+		if (bits.ee)
+			printright(start++, hw, _("Vertex Grouper + Tesselator %6.2f%%"), vgt);
+		else
+			printright(start++, hw, _("Geometry Engine %6.2f%%"), vgt);
 		if (color) attroff(COLOR_PAIR(2));
 
 		// Enough height?
@@ -207,8 +213,10 @@ void present(const unsigned int ticks, const char card[], unsigned int color,
 		percentage(start, w, sx);
 		printright(start++, hw, _("Shader Export %6.2f%%"), sx);
 
-		percentage(start, w, sh);
-		printright(start++, hw, _("Sequencer Instruction Cache %6.2f%%"), sh);
+		if (bits.sh) {
+			percentage(start, w, sh);
+			printright(start++, hw, _("Sequencer Instruction Cache %6.2f%%"), sh);
+		}
 
 		percentage(start, w, spi);
 		printright(start++, hw, _("Shader Interpolator %6.2f%%"), spi);
@@ -254,6 +262,10 @@ void present(const unsigned int ticks, const char card[], unsigned int color,
 			percentage(start, w, vce0);
 			printright(start++, hw, _("VCE %6.2f%%"), vce0);
 		}
+		if (bits.vcn) {
+			percentage(start, w, vcn);
+			printright(start++, hw, _("Video Core %6.2f%%"), vcn);
+		}
 
 		if (bits.vram || bits.gtt) {
 			// Enough height?