icache: Log 36 bits of instruction rather than 32

This expands the field in the log buffer that stores the instruction
fetched from the icache to 36 bits, so that we get the insn_code and
illegal instruction indication.  To do this, we reclaim 3 unused bits
from execute1's portion and one other unused bit (previously just set
to 0 in core.vhdl).

This also alters the trigger behaviour to stop after one quarter of
the log buffer has been filled with samples after the trigger, or 256
entries, whichever is less.  This is to ensure that the trigger event
doesn't get overwritten when the log buffer is small.

This updates fmt_log to the new log format.  Valid instructions are
printed as a decimal insn_code value followed by the bottom 26 bits of
the instruction.  Illegal instructions are printed as "ill" followed
by the full 32 bits of the instruction.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>

core.vhdl

@@ -267,7 +267,7 @@ begin
             wishbone_in => wishbone_insn_in,
             wb_snoop_in => wb_snoop_in,
             events => icache_events,
-            log_out => log_data(96 downto 43)
+            log_out => log_data(100 downto 43)
             );
 
     icache_stall_in <= decode1_busy;
@@ -288,7 +288,7 @@ begin
             d_out => decode1_to_decode2,
             f_out => decode1_to_fetch1,
             r_out => decode1_to_register_file,
-            log_out => log_data(109 downto 97)
+            log_out => log_data(113 downto 101)
             );
 
     decode1_stall_in <= decode2_stall_out;
@@ -320,7 +320,7 @@ begin
             writeback_bypass => writeback_bypass,
             dbg_spr_req => dbg_spr_req,
             dbg_spr_addr => dbg_spr_addr,
-            log_out => log_data(119 downto 110)
+            log_out => log_data(123 downto 114)
             );
     decode2_busy_in <= ex1_busy_out;
 
@@ -399,7 +399,7 @@ begin
             dbg_spr_data => dbg_spr_data,
             sim_dump => sim_ex_dump,
             sim_dump_done => sim_cr_dump,
-            log_out => log_data(134 downto 120),
+            log_out => log_data(135 downto 124),
             log_rd_addr => log_rd_addr,
             log_rd_data => log_rd_data,
             log_wr_addr => log_wr_addr
@@ -501,7 +501,7 @@ begin
             );
 
     log_data(150) <= '0';
-    log_data(139 downto 135) <= "00000";
+    log_data(139 downto 136) <= "0000";
 
     debug_0: entity work.core_debug
         generic map (

core_debug.vhdl

@@ -175,7 +175,8 @@ begin
                 gspr_index <= (others => '0');
             else
                 if do_log_trigger = '1' or log_trigger_delay /= 0 then
-                    if log_trigger_delay = 255 then
+                    if log_trigger_delay = 255 or
+                        (LOG_LENGTH < 1024 and log_trigger_delay = LOG_LENGTH / 4) then
                         log_dmi_trigger(1) <= '1';
                         log_trigger_delay <= 0;
                     else

execute1.vhdl

@@ -65,7 +65,7 @@ entity execute1 is
         sim_dump      : in std_ulogic;
         sim_dump_done : out std_ulogic;
 
-        log_out : out std_ulogic_vector(14 downto 0);
+        log_out : out std_ulogic_vector(11 downto 0);
         log_rd_addr : out std_ulogic_vector(31 downto 0);
         log_rd_data : in std_ulogic_vector(63 downto 0);
         log_wr_addr : in std_ulogic_vector(31 downto 0)
@@ -1802,7 +1802,7 @@ begin
     end generate;
 
     e1_log: if LOG_LENGTH > 0 generate
-        signal log_data : std_ulogic_vector(14 downto 0);
+        signal log_data : std_ulogic_vector(11 downto 0);
     begin
         ex1_log : process(clk)
         begin
@@ -1812,7 +1812,6 @@ begin
                             exception_log &
                             irq_valid_log &
                             interrupt_in.intr &
-                            "000" &
                             ex2.e.write_enable &
                             ex2.e.valid &
                             (ex2.e.redirect or ex2.e.interrupt) &

icache.vhdl

@@ -71,7 +71,7 @@ entity icache is
         wb_snoop_in  : in wishbone_master_out := wishbone_master_out_init;
 
         events       : out IcacheEventType;
-        log_out      : out std_ulogic_vector(53 downto 0)
+        log_out      : out std_ulogic_vector(57 downto 0)
         );
 end entity icache;
 
@@ -244,6 +244,8 @@ architecture rtl of icache is
     signal snoop_index : index_t;
     signal snoop_hits  : cache_way_valids_t;
 
+    signal log_insn : std_ulogic_vector(35 downto 0);
+
     -- Return the cache line index (tag index) for an address
     function get_index(addr: std_ulogic_vector) return index_t is
     begin
@@ -623,6 +625,7 @@ begin
 	end if;
         i_out.insn <= insn(31 downto 0);
         i_out.icode <= icode;
+        log_insn <= cache_wr_data(35 downto 0);
 	i_out.valid <= r.hit_valid;
 	i_out.nia <= r.hit_nia;
 	i_out.stop_mark <= r.hit_smark;
@@ -872,7 +875,7 @@ begin
 
     icache_log: if LOG_LENGTH > 0 generate
         -- Output data to logger
-        signal log_data    : std_ulogic_vector(53 downto 0);
+        signal log_data    : std_ulogic_vector(57 downto 0);
     begin
         data_log: process(clk)
             variable lway: way_t;
@@ -885,7 +888,7 @@ begin
                     wstate := '1';
                 end if;
                 log_data <= i_out.valid &
-                            i_out.insn &
+                            log_insn &
                             wishbone_in.ack &
                             r.wb.adr(2 downto 0) &
                             r.wb.stb & r.wb.cyc &

scripts/fmt_log/fmt_log.c

@@ -22,7 +22,7 @@ struct log_entry {
 	u64	ic_wb_adr: 3;
 	u64	ic_wb_ack: 1;
 
-	u64	ic_insn: 32;
+	u64	ic_insn: 36;
 	u64	ic_valid: 1;
 	u64	d1_valid: 1;
 	u64	d1_unit: 2;
@@ -39,9 +39,8 @@ struct log_entry {
 	u64	e1_stall_out: 1;
 	u64	e1_redirect: 1;
 	u64	e1_valid: 1;
-	u64	e1_write_enable: 1;
-	u64	e1_unused: 3;
 
+	u64	e1_write_enable: 1;
 	u64	e1_irq_state: 1;
 	u64	e1_irq: 1;
 	u64	e1_exception: 1;
@@ -49,7 +48,7 @@ struct log_entry {
 	u64	e1_msr_ir: 1;
 	u64	e1_msr_pr: 1;
 	u64	e1_msr_ee: 1;
-	u64	pad1: 5;
+	u64	pad1: 4;
 	u64	ls_state: 3;
 	u64	ls_dw_done: 1;
 	u64	ls_min_done: 1;
@@ -134,9 +133,9 @@ int main(int ac, char **av)
 		full_nia[log.nia_lo & 0xf] = (log.nia_hi? 0xc000000000000000: 0) |
 			(log.nia_lo << 2);
 		if (lineno % 20 == 1) {
-			printf("        fetch1 NIA      icache                         decode1       decode2   execute1         loadstore  dcache       CR   GSPR\n");
+			printf("        fetch1 NIA      icache                             decode1       decode2   execute1         loadstore  dcache       CR   GSPR\n");
-			printf("     ----------------   TAHW S -WB-- pN --insn--    pN un op         pN byp    FR IIE MSR  WC   SD MM CE   SRTO DE -WB-- c ms reg val\n");
+			printf("     ----------------   TAHW S -WB-- pN  ic --insn--    pN un op         pN byp    FR IIE MSR  WC   SD MM CE   SRTO DE -WB-- c ms reg val\n");
-			printf("                        LdMy t csnSa IA             IA it            IA abc    le srx EPID em   tw rd mx   tAwp vr csnSa 0 k\n");
+			printf("                        LdMy t csnSa IA                 IA it            IA abc    le srx EPID em   tw rd mx   tAwp vr csnSa 0 k\n");
 		}
 		printf("%4ld %c0000%.11llx %c ", lineno,
 		       (log.nia_hi? 'c': '0'),
@@ -154,12 +153,16 @@ int main(int ac, char **av)
 		       FLAG(ic_wb_stall, 'S'),
 		       FLAG(ic_wb_ack, 'a'),
 		       PNIA(ic_part_nia));
-		if (log.ic_valid)
+		if (log.ic_valid) {
-			printf("%.8x", log.ic_insn);
+			if (log.ic_insn & (1ul << 35))
-		else if (log.ic_fetch_failed)
+				printf("ill %.8lx", log.ic_insn & 0xfffffffful);
-			printf("!!!!!!!!");
+			else
+				printf("%3lu x%.7lx", (long)(log.ic_insn >> 26),
+				       (unsigned long)(log.ic_insn & 0x3ffffff));
+		} else if (log.ic_fetch_failed)
+			printf("    !!!!!!!!");
 		else
-			printf("--------");
+			printf("--- --------");
 		printf(" %c%c %.2llx ",
 		       FLAG(ic_valid, '>'),
 		       FLAG(d2_stall_out, '|'),