From 7ae5983ba12080af998717a46a25f4059a70508c Mon Sep 17 00:00:00 2001
From: ALTracer <11005378+ALTracer@users.noreply.github.com>
Date: Tue, 28 Jan 2025 20:36:02 +0300
Subject: [PATCH 1/4] hosted/ftdi: Implement jtagtap_cycle like tms_seq

---
 src/platforms/hosted/ftdi_jtag.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/src/platforms/hosted/ftdi_jtag.c b/src/platforms/hosted/ftdi_jtag.c
index a88f0e4ee9c..2b22b2bc89f 100644
--- a/src/platforms/hosted/ftdi_jtag.c
+++ b/src/platforms/hosted/ftdi_jtag.c
@@ -149,6 +149,18 @@ static bool ftdi_jtag_next(const bool tms, const bool tdi)
 
 static void ftdi_jtag_cycle(const bool tms, const bool tdi, const size_t clock_cycles)
 {
-	for (size_t i = 0; i < clock_cycles; i++)
-		ftdi_jtag_next(tms, tdi);
+	DEBUG_PROBE("%s: %zu clock cycles with TMS %s and TDI %s\n", __func__, clock_cycles, tms ? "high" : "low",
+		tdi ? "high" : "low");
+	uint64_t tms_states = (UINT64_C(1) << clock_cycles) - 1ULL;
+	if (!tms)
+		tms_states = 0;
+	for (size_t cycle = 0U; cycle < clock_cycles; cycle += 7U) {
+		const uint8_t cmd[3U] = {
+			MPSSE_WRITE_TMS | MPSSE_LSB | MPSSE_BITMODE | MPSSE_WRITE_NEG,
+			MIN(7U, clock_cycles - cycle) - 1U,
+			(tdi ? 0x80U : 0U) | (tms_states & 0x7fU),
+		};
+		tms_states >>= 7U;
+		ftdi_buffer_write_arr(cmd);
+	}
 }

From a2ba626bdc4cde55db230429b8dc99dd7b557193 Mon Sep 17 00:00:00 2001
From: ALTracer <11005378+ALTracer@users.noreply.github.com>
Date: Tue, 28 Jan 2025 20:48:52 +0300
Subject: [PATCH 2/4] hosted/ftdi: Add buffering to jtagtap_cycle

---
 src/platforms/hosted/ftdi_jtag.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/src/platforms/hosted/ftdi_jtag.c b/src/platforms/hosted/ftdi_jtag.c
index 2b22b2bc89f..97dcf86dfa3 100644
--- a/src/platforms/hosted/ftdi_jtag.c
+++ b/src/platforms/hosted/ftdi_jtag.c
@@ -154,6 +154,8 @@ static void ftdi_jtag_cycle(const bool tms, const bool tdi, const size_t clock_c
 	uint64_t tms_states = (UINT64_C(1) << clock_cycles) - 1ULL;
 	if (!tms)
 		tms_states = 0;
+	uint8_t cmds[30U] = {0}; /* Up to ceil[64 // 7] = 10 commands, 3 byte each */
+	size_t cmd_count = 0;
 	for (size_t cycle = 0U; cycle < clock_cycles; cycle += 7U) {
 		const uint8_t cmd[3U] = {
 			MPSSE_WRITE_TMS | MPSSE_LSB | MPSSE_BITMODE | MPSSE_WRITE_NEG,
@@ -161,6 +163,7 @@ static void ftdi_jtag_cycle(const bool tms, const bool tdi, const size_t clock_c
 			(tdi ? 0x80U : 0U) | (tms_states & 0x7fU),
 		};
 		tms_states >>= 7U;
-		ftdi_buffer_write_arr(cmd);
+		memcpy(&cmds[sizeof(cmd) * cmd_count++], cmd, sizeof(cmd));
 	}
+	ftdi_buffer_write(cmds, 3U * cmd_count);
 }

From d095eaea62d09cf17a0cb650bbf84653a243c5b5 Mon Sep 17 00:00:00 2001
From: ALTracer <11005378+ALTracer@users.noreply.github.com>
Date: Tue, 28 Jan 2025 21:02:18 +0300
Subject: [PATCH 3/4] hosted/ftdi: Simplify jtagtap_cycle via
 CLK_BYTES/CLK_BITS opcode

---
 src/platforms/hosted/ftdi_jtag.c | 45 +++++++++++++++++++++++---------
 1 file changed, 32 insertions(+), 13 deletions(-)

diff --git a/src/platforms/hosted/ftdi_jtag.c b/src/platforms/hosted/ftdi_jtag.c
index 97dcf86dfa3..765859e499f 100644
--- a/src/platforms/hosted/ftdi_jtag.c
+++ b/src/platforms/hosted/ftdi_jtag.c
@@ -149,21 +149,40 @@ static bool ftdi_jtag_next(const bool tms, const bool tdi)
 
 static void ftdi_jtag_cycle(const bool tms, const bool tdi, const size_t clock_cycles)
 {
+	if (clock_cycles >= 0x80000U)
+		return;
 	DEBUG_PROBE("%s: %zu clock cycles with TMS %s and TDI %s\n", __func__, clock_cycles, tms ? "high" : "low",
 		tdi ? "high" : "low");
-	uint64_t tms_states = (UINT64_C(1) << clock_cycles) - 1ULL;
-	if (!tms)
-		tms_states = 0;
-	uint8_t cmds[30U] = {0}; /* Up to ceil[64 // 7] = 10 commands, 3 byte each */
-	size_t cmd_count = 0;
-	for (size_t cycle = 0U; cycle < clock_cycles; cycle += 7U) {
-		const uint8_t cmd[3U] = {
-			MPSSE_WRITE_TMS | MPSSE_LSB | MPSSE_BITMODE | MPSSE_WRITE_NEG,
-			MIN(7U, clock_cycles - cycle) - 1U,
-			(tdi ? 0x80U : 0U) | (tms_states & 0x7fU),
+
+	/* Update state of TMS & TDI using 0x4b opcode, then trigger 8a+b cycles using 0x8f, 0x8e. */
+	const uint8_t cmd[3] = {
+		MPSSE_WRITE_TMS | MPSSE_LSB | MPSSE_BITMODE | MPSSE_WRITE_NEG,
+		0,
+		(tdi ? 0x80U : 0U) | (tms ? 0x01U : 0U),
+	};
+	ftdi_buffer_write_arr(cmd);
+
+	/*
+	 * 0x8f CLK_BYTES: Length of 0..n will generate 1..1+n byte cycles,
+	 * i.e. emit 8..8*(n+1) clocks, so adjust clock_bytes down by 1.
+	 */
+	size_t clock_bytes = clock_cycles / 8U;
+	if (clock_bytes > 0) {
+		clock_bytes--;
+		const uint8_t cmd8[3U] = {
+			CLK_BYTES,
+			clock_bytes & 0xffU,
+			clock_bytes >> 8U,
 		};
-		tms_states >>= 7U;
-		memcpy(&cmds[sizeof(cmd) * cmd_count++], cmd, sizeof(cmd));
+		ftdi_buffer_write_arr(cmd8);
 	}
-	ftdi_buffer_write(cmds, 3U * cmd_count);
+	/* 0x8e CLK_BITS: Length of 0..7 will emit 1..8 clocks */
+	const uint8_t clock_bits = clock_cycles % 8U - 1U;
+	if (clock_bits >= 8U)
+		return;
+	const uint8_t cmd1[2U] = {
+		CLK_BITS,
+		clock_bits,
+	};
+	ftdi_buffer_write_arr(cmd1);
 }

From f9e0169a080844601eafb6b09255610765dd49d4 Mon Sep 17 00:00:00 2001
From: ALTracer <11005378+ALTracer@users.noreply.github.com>
Date: Fri, 30 Jan 2026 21:30:20 +0300
Subject: [PATCH 4/4] hosted/ftdi: jtagtap_cycle: Express divmod by 8 as shift
 and mask

---
 src/platforms/hosted/ftdi_jtag.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/platforms/hosted/ftdi_jtag.c b/src/platforms/hosted/ftdi_jtag.c
index 765859e499f..31a03c24f27 100644
--- a/src/platforms/hosted/ftdi_jtag.c
+++ b/src/platforms/hosted/ftdi_jtag.c
@@ -166,7 +166,7 @@ static void ftdi_jtag_cycle(const bool tms, const bool tdi, const size_t clock_c
 	 * 0x8f CLK_BYTES: Length of 0..n will generate 1..1+n byte cycles,
 	 * i.e. emit 8..8*(n+1) clocks, so adjust clock_bytes down by 1.
 	 */
-	size_t clock_bytes = clock_cycles / 8U;
+	size_t clock_bytes = clock_cycles >> 3U;
 	if (clock_bytes > 0) {
 		clock_bytes--;
 		const uint8_t cmd8[3U] = {
@@ -177,7 +177,7 @@ static void ftdi_jtag_cycle(const bool tms, const bool tdi, const size_t clock_c
 		ftdi_buffer_write_arr(cmd8);
 	}
 	/* 0x8e CLK_BITS: Length of 0..7 will emit 1..8 clocks */
-	const uint8_t clock_bits = clock_cycles % 8U - 1U;
+	const uint8_t clock_bits = (clock_cycles & 7U) - 1U;
 	if (clock_bits >= 8U)
 		return;
 	const uint8_t cmd1[2U] = {