From: Øyvind Harboe <oyvind.harboe@zylin.com>
Date: Tue, 9 Feb 2010 13:26:57 +0000 (+0100)
Subject: arm11: improve performance using minidriver hook
X-Git-Tag: v0.5.0-rc1~915
X-Git-Url: https://git.sur5r.net/?a=commitdiff_plain;h=faef631a4d1429f48c34da13d446dcd64d1523bf;p=openocd

arm11: improve performance using minidriver hook

zy1000 performance for GDB load went from 100kBytes/s
to 300kBytes/s @ 8 MHz by implementing the inner loop
of unack arm11 memory writes directly on top of the hw
fifo.

Profiling info:

 78.57      0.77     0.77                             arm11_run_instr_data_to_core_noack_inner
  5.10      0.82     0.05                             memcpy
  4.08      0.86     0.04                             jtag_tap_next_enabled
  3.06      0.89     0.03                             gdb_input

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
---

diff --git a/src/jtag/zy1000/zy1000.c b/src/jtag/zy1000/zy1000.c
index da5aa366..d920c30b 100644
--- a/src/jtag/zy1000/zy1000.c
+++ b/src/jtag/zy1000/zy1000.c
@@ -845,11 +845,73 @@ void embeddedice_write_dcc(struct jtag_tap *tap, int reg_addr, uint8_t *buffer,
 }
 
 
-int arm11_run_instr_data_to_core_noack_inner_default(struct arm11_common * arm11, uint32_t opcode, uint32_t * data, size_t count);
 
-int arm11_run_instr_data_to_core_noack_inner(struct arm11_common * arm11, uint32_t opcode, uint32_t * data, size_t count)
+int arm11_run_instr_data_to_core_noack_inner(struct jtag_tap * tap, uint32_t opcode, uint32_t * data, size_t count)
 {
-	return arm11_run_instr_data_to_core_noack_inner_default(arm11, opcode, data, count);
+#if 0
+	int arm11_run_instr_data_to_core_noack_inner_default(struct jtag_tap * tap, uint32_t opcode, uint32_t * data, size_t count);
+	return arm11_run_instr_data_to_core_noack_inner_default(tap, opcode, data, count);
+#else
+	static const int bits[] = {32, 2};
+	uint32_t values[] = {0, 0};
+
+	/* FIX!!!!!! the target_write_memory() API started this nasty problem
+	 * with unaligned uint32_t * pointers... */
+	const uint8_t *t = (const uint8_t *)data;
+
+	while (count--)
+	{
+		values[0] = *t++;
+		values[0] |= (*t++<<8);
+		values[0] |= (*t++<<16);
+		values[0] |= (*t++<<24);
+
+		if (count > 0)
+		{
+			jtag_add_dr_out(tap,
+				2,
+				bits,
+				values,
+				TAP_DRPAUSE);
+
+#if 1
+			/* copy & paste from arm11_dbgtap.c */
+			//TAP_DREXIT2, TAP_DRUPDATE, TAP_IDLE, TAP_IDLE, TAP_IDLE, TAP_DRSELECT, TAP_DRCAPTURE, TAP_DRSHIFT
+
+			waitIdle();
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 1);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 1);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 1);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x28, 0);
+			waitIdle();
+			ZY1000_POKE(ZY1000_JTAG_BASE + 0x20, TAP_DRSHIFT);
+#else
+			static const tap_state_t arm11_MOVE_DRPAUSE_IDLE_DRPAUSE_with_delay[] =
+			{
+				TAP_DREXIT2, TAP_DRUPDATE, TAP_IDLE, TAP_IDLE, TAP_IDLE, TAP_DRSELECT, TAP_DRCAPTURE, TAP_DRSHIFT
+			};
+
+			jtag_add_pathmove(ARRAY_SIZE(arm11_MOVE_DRPAUSE_IDLE_DRPAUSE_with_delay),
+				arm11_MOVE_DRPAUSE_IDLE_DRPAUSE_with_delay);
+#endif
+		} else
+		{
+			/* This will happen on the last iteration updating the current tap state
+			 * so we don't have to track it during the common code path */
+			jtag_add_dr_out(tap,
+				2,
+				bits,
+				values,
+				TAP_IDLE);
+		}
+	}
+
+	return jtag_execute_queue();
+#endif
 }