r600g,radeonsi: add a driver query returning GPU load

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 141 insertions, 0 deletions

diff --git a/src/gallium/drivers/radeon/r600_gpu_load.c b/src/gallium/drivers/radeon/r600_gpu_load.c
new file mode 100644
index 0000000..a653834
--- /dev/null
+++ b/src/gallium/drivers/radeon/r600_gpu_load.c
@@ -0,0 +1,141 @@
+/*
+ * Copyright 2015 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors: Marek Olšák <maraeo@gmail.com>
+ *
+ */
+
+/* The GPU load is measured as follows.
+ *
+ * There is a thread which samples the GRBM_STATUS register at a certain
+ * frequency and the "busy" or "idle" counter is incremented based on
+ * whether the GUI_ACTIVE bit is set or not.
+ *
+ * Then, the user can sample the counters twice and calculate the average
+ * GPU load between the two samples.
+ */
+
+#include "r600_pipe_common.h"
+#include "os/os_time.h"
+
+/* For good accuracy at 1000 fps or lower. This will be inaccurate for higher
+ * fps (there are too few samples per frame). */
+#define SAMPLES_PER_SEC 10000
+
+#define GRBM_STATUS		0x8010
+#define GUI_ACTIVE(x)		(((x) >> 31) & 0x1)
+
+static bool r600_is_gpu_busy(struct r600_common_screen *rscreen)
+{
+	uint32_t value = 0;
+
+	rscreen->ws->read_registers(rscreen->ws, GRBM_STATUS, 1, &value);
+	return GUI_ACTIVE(value);
+}
+
+static PIPE_THREAD_ROUTINE(r600_gpu_load_thread, param)
+{
+	struct r600_common_screen *rscreen = (struct r600_common_screen*)param;
+	const int period_us = 1000000 / SAMPLES_PER_SEC;
+	int sleep_us = period_us;
+	int64_t cur_time, last_time = os_time_get();
+
+	while (!p_atomic_read(&rscreen->gpu_load_stop_thread)) {
+		if (sleep_us)
+			os_time_sleep(sleep_us);
+
+		/* Make sure we sleep the ideal amount of time to match
+		 * the expected frequency. */
+		cur_time = os_time_get();
+
+		if (os_time_timeout(last_time, last_time + period_us,
+				    cur_time))
+			sleep_us = MAX2(sleep_us - 1, 1);
+		else
+			sleep_us += 1;
+
+		/*printf("Hz: %.1f\n", 1000000.0 / (cur_time - last_time));*/
+		last_time = cur_time;
+
+		/* Update the counters. */
+		if (r600_is_gpu_busy(rscreen))
+			p_atomic_inc(&rscreen->gpu_load_counter_busy);
+		else
+			p_atomic_inc(&rscreen->gpu_load_counter_idle);
+	}
+	p_atomic_dec(&rscreen->gpu_load_stop_thread);
+	return 0;
+}
+
+void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen)
+{
+	if (!rscreen->gpu_load_thread)
+		return;
+
+	p_atomic_inc(&rscreen->gpu_load_stop_thread);
+	pipe_thread_wait(rscreen->gpu_load_thread);
+	rscreen->gpu_load_thread = 0;
+}
+
+static uint64_t r600_gpu_load_read_counter(struct r600_common_screen *rscreen)
+{
+	/* Start the thread if needed. */
+	if (!rscreen->gpu_load_thread) {
+		pipe_mutex_lock(rscreen->gpu_load_mutex);
+		/* Check again inside the mutex. */
+		if (!rscreen->gpu_load_thread)
+			rscreen->gpu_load_thread =
+				pipe_thread_create(r600_gpu_load_thread, rscreen);
+		pipe_mutex_unlock(rscreen->gpu_load_mutex);
+	}
+
+	/* The busy counter is in the lower 32 bits.
+	 * The idle counter is in the upper 32 bits. */
+	return p_atomic_read(&rscreen->gpu_load_counter_busy) |
+	       ((uint64_t)p_atomic_read(&rscreen->gpu_load_counter_idle) << 32);
+}
+
+/**
+ * Just return the counters.
+ */
+uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen)
+{
+	return r600_gpu_load_read_counter(rscreen);
+}
+
+unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin)
+{
+	uint64_t end = r600_gpu_load_read_counter(rscreen);
+	unsigned busy = (end & 0xffffffff) - (begin & 0xffffffff);
+	unsigned idle = (end >> 32) - (begin >> 32);
+
+	/* Calculate the GPU load.
+	 *
+	 * If no counters have been incremented, return the current load.
+	 * It's for the case when the load is queried faster than
+	 * the counters are updated.
+	 */
+	if (idle || busy)
+		return busy*100 / (busy + idle);
+	else
+		return r600_is_gpu_busy(rscreen) ? 100 : 0;
+}