Rev |
Log message |
Author |
Age |
Path |
298 |
Fixed a long-standing bug in the SBC instruction where the 1 wasn't being added to complete the 2's complement of Rn. This was causing off-by-one errors in subtraction and negating carry only subtractions. |
jshamlet |
848d 07h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
290 |
Added an additional generic "Rotation_Ignores_Carry" that removes the carry logic from the ROL/ROR instructions such that they now rotate 'normally',
Added an alias for PSR_GP4 named PSR_S, as it is now used to switch the function of the RSP instruction. The internal opcode hasn't changed, but it allows assembly code to use PSR_S or BRS/BNS when performing RSP related operations. |
jshamlet |
1283d 05h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
270 |
Moved CPU internal constants to o8_cpu.vhd and replace the generic that set the RSP direction flag with a constant instead. This removes the need to expose internal architectural flags externally.
Also added a hard-coded version register that takes a major and minor value as bytes using generics. This is a read-only register to the CPU. |
jshamlet |
1475d 12h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
269 |
Modified the write data path to use separate enumerated states rather than reuse the .reg field to improve performance. |
jshamlet |
1478d 02h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
264 |
Updated comments |
jshamlet |
1580d 08h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
263 |
Fixed a very old bug in the CPU core where autoincrements weren't affecting the upper register in the pair, causing it to loop around the lower 256 bytes. This only affected LDX/LDO, as the proper ALU signals were being generated in STO/STX and UPP. Wow, that bug has been in there for AGES.
Also separated the SDLC TX and RX interrupts so that they could be handled separately. |
jshamlet |
1580d 09h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
260 |
Added missing comments for Sequential_Interrupts generic, as well as comments explaining portions of the CPU operations. |
jshamlet |
1609d 11h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
256 |
Removed unused generic from the status_led.vhd and cleaned up comments on the CPU |
jshamlet |
1610d 12h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
255 |
Modified code to make ModelSim happy (It didn't like the generate blocks for some reason). Also added a block describing the new generic. |
jshamlet |
1610d 17h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
254 |
Simplified the ISR address logic so that the upper 12 bits are constant (set by generic) and only the lower 4 bits are registered/computed. |
jshamlet |
1611d 07h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
253 |
Fixed spelling error in comment |
jshamlet |
1611d 08h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
252 |
(This time the CPU model was included...)
Added the ability to use unsigned offsets to the LDO/STO instructions. The original behavior of signed offsets is preserved if the Unsigned_Index_Offsets is left unset or set to FALSE. While inserting this code, pipeline registers were also inserted into the address generation logic for indexed instructions. This simplifies the final multiplexor and improves FMax at the slight expense of LDO/SDO now taking one additional clock cycle to execute. |
jshamlet |
1611d 08h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
248 |
Removed Default_Int_Flag generic from CPU, as it is duplicated by Supervisor_Mode. |
jshamlet |
1627d 16h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
245 |
Modified the CPU's Supervisor_Mode to also protect SMSK and RSP instructions,
Added an external interrupt manager, o8_int_mgr.vhd. |
jshamlet |
1630d 13h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
244 |
Added two new generics to the CPU model. The first is a supervisory mode that disables the STP PSR_I instruction. This prevents errant code execution from setting the I bit, and disabling any subsequent memory protection logic. The second allows the default state of the I bit to be set at startup. If set true, initialization code will run with the I bit set, allowing it to bypass memory protection.
Also modified the RAM models to include write mask logic, where the mask register is write-protected by the I bit in the CPU. When enabled, the models will prevent code from writing to memory regions which do not have their mask bits set. The upshot is that code can effectively "write protect" the RAM - which is useful for multitasking applications.
Also, most modules have been updated with write qualification inputs, allowing a similar scheme to be used for I/O, though not as elegantly. I use a register module, whose own write qual line is attached to the external copy of the I bit as an I/O write protect register.
Lastly, added a new externally triggered timer, which can generate pulses with programmable delays and widths, and which can interrupt on either the input trigger, the output rising edge, or output falling edge. The time base can be either the internal microsecond tick signal, or an external clock. |
jshamlet |
1631d 09h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
227 |
Added a demonstration Open8_cfg.vhd file, which is used to configure the system constants. It also provides a function that makes it easy to merge read buses. |
jshamlet |
1665d 11h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
225 |
Added Halt_Ack to go with Halt_Req. |
jshamlet |
1665d 14h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
224 |
Finished new Open8 bus record, which now includes the clock, reset and a microsecond tick. The CPU now accepts a clock and pll_locked signal, which it uses to generate the system reset in the bus record. It also contains a simple microsecond counter to feed the usec_tick in the record. This logic was removed from the real time clock and system timer entities, which now use the global version. Bus connections should be dramatically simplified, as only the read logic and interrupts are still run as separate signals. |
jshamlet |
1665d 16h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
223 |
Added an OPEN8_BUS_TYPE record to simplify connection to Open8 modules. The CPU now passes and Open8_Bus out, which supplies the bus address, write enable, write data, and read enable. Read data and interrupts are still handled as separate signals, since they are muxed/connected at the next level up. |
jshamlet |
1666d 09h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
210 |
Modified the timers to reset on new interval write. This avoids an issue in the original design where the timer had to reach zero before updating, potentially causing unwanted interrupts.
Also added a flag to the CPU to allow interrupts to be processed sequentially based on the state of the I bit. This one is set to false by default, as it is a significant change in interrupt behavior. Long, and reentrant, ISRs can clear the I bit prematurely to allow themselves to be interrupted.
Lastly, added the I bit to the exported flags for possible use in memory protection schemes. |
jshamlet |
1672d 17h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
209 |
Fixed an issue in the PIT timer that caused an immediate interrupt on interval write,
Fixed an issue in the epoch timer that resulted in a spurious interrupt due to extra LSB's being set by default in the set point register,
While cleaning elsewhere, founding a spacing issue in the CPU HDL,
Added a 4k ROM and MW core. |
jshamlet |
1673d 06h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
194 |
Cleaned up licensing sections |
jshamlet |
1681d 12h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
191 |
Cleaned up comments, added back the OPEN8_NULLBUS constant, and added some new modules for ADCs and LCD displays.
Also made the button input module more configurable by moving the debounce code to a separate entity and using generics to instantiate it. |
jshamlet |
1681d 13h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
190 |
Fixed a bug in CPU where RTI/RTS wasn't idling the instruction cache, causing intermittent failures where RTI would execute as RTS, corrupting the stack;
Fixed a bug in the real-time clock where the uSec tick generator would stop if the PIT timer value was left/set to 0x00. |
jshamlet |
1693d 10h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
188 |
Added a generic to alter the behavior of RTI so that it can optionally skip restoring the general purpose flags GP4 to GP7, allowing ISR's to make persistent changes to them. Also exported these flags to the top level for use outside the CPU. |
jshamlet |
1694d 14h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
187 |
Added the CPU_Halt input, only now as an input to the instruction decoder. The CPU_Halt line will assert the registered CPU_Halt_Req, which will cause the instruction decoder to abort the current instruction, reset the PC, then enter a hold state until the line is deasserted. Additionally, a very minor bug that could cause the SMSK instruction to effectively execute twice if interrupted was fixed. Lastly, cleaned up the comments even more. |
jshamlet |
1696d 10h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
186 |
Merged the interrupt override logic into the case structure, simplifying how interrupts are processed. |
jshamlet |
1699d 10h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
185 |
1) Fixed an apparently long-standing bug where the interrupt bit wasn't being cleared after an RTI
2) Modified the program counter logic to be simpler. It now always increments, and states control the increment using the offset field. A new set of constants was added to replace the old states.
3) Modified the ALU to always use Operand1 instead of ALU_Ctrl.Data (and removed the field in the record). A new ALU command, ALU_GMSK, was added, as it was the only instruction to set the .Data field to something other than Operand1 (Int_Mask)
4) Modified the package file so that flag names match what the assembler calls them. FL_Z is now PSR_Z, FL_GP1 is now PSR_GP4, etc.
5) Cleaned up the comments and code formatting |
jshamlet |
1699d 12h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
183 |
Renamed core to o8_cpu to match new naming scheme |
jshamlet |
1701d 13h |
/open8_urisc/trunk/VHDL/o8_cpu.vhd |
182 |
Simplified the address generation logic at the expense of making LDX take one additional clock cycle. This allowed the address logic to be split out of the main state machine and simplified (greatly). During this process, a bug in SDO was found and fixed that caused it to return through the wrong pipe fill state wnen auto increment was disabled. |
jshamlet |
1701d 13h |
/open8_urisc/trunk/VHDL/Open8.vhd |