Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expert-level Verilog and SystemVerilog knowledge following IEEE 1800 standards. Generates synthesizable RTL code with proper coding styles and constructs.
.claude/skills/a5c-ai-verilog-sv-language/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 263% | 0% |
| case-24 | ✗→✓ | ▲ Improved | 91% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 188% | 0% |
Expert skill for Verilog and SystemVerilog development following IEEE 1364 and IEEE 1800 standards. Provides deep expertise in synthesizable RTL code generation, proper construct usage, and modern coding practices.
The Verilog/SystemVerilog Language skill enables comprehensive HDL development for FPGA and ASIC designs, supporting:
Use SystemVerilog always block variants correctly:
systemverilog// Sequential logic - always_ff always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin counter <= '0; state <= IDLE; end else begin counter <= counter + 1'b1; state <= next_state; end end // Combinational logic - always_comb always_comb begin // Default assignments prevent latches next_state = state; output_valid = 1'b0; case (state) IDLE: begin if (start) next_state = RUN; end RUN: begin output_valid = 1'b1; if (done) next_state = IDLE; end default: next_state = IDLE; endcase end // Intentional latch - always_latch (rare) always_latch begin if (enable) latch_out = data_in; end
Create reusable parameterized modules:
systemverilogmodule sync_fifo #( parameter int DATA_WIDTH = 8, parameter int DEPTH = 16, parameter int ALMOST_FULL_THRESH = DEPTH - 2, parameter int ALMOST_EMPTY_THRESH = 2, // Derived parameters using localparam localparam int ADDR_WIDTH = $clog2(DEPTH), localparam int CNT_WIDTH = $clog2(DEPTH + 1) ) ( input logic clk, input logic rst_n, // Write interface input logic wr_en, input logic [DATA_WIDTH-1:0] wr_data, output logic full, output logic almost_full, // Read interface input logic rd_en, output logic [DATA_WIDTH-1:0] rd_data, output logic empty, output logic almost_empty, // Status output logic [CNT_WIDTH-1:0] fill_level ); // Memory array logic [DATA_WIDTH-1:0] mem [DEPTH]; // Pointers logic [ADDR_WIDTH-1:0] wr_ptr, rd_ptr; logic [CNT_WIDTH-1:0] count; // Write logic always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin wr_ptr <= '0; end else if (wr_en && !full) begin mem[wr_ptr] <= wr_data; wr_ptr <= wr_ptr + 1'b1; end end // Read logic always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin rd_ptr <= '0; end else if (rd_en && !empty) begin rd_ptr <= rd_ptr + 1'b1; end end // Read data (registered output) always_ff @(posedge clk) begin if (rd_en && !empty) begin rd_data <= mem[rd_ptr]; end end // Count logic always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin count <= '0; end else begin case ({wr_en && !full, rd_en && !empty}) 2'b10: count <= count + 1'b1; 2'b01: count <= count - 1'b1; default: count <= count; endcase end end // Status outputs assign full = (count == DEPTH); assign empty = (count == '0); assign almost_full = (count >= ALMOST_FULL_THRESH); assign almost_empty = (count <= ALMOST_EMPTY_THRESH); assign fill_level = count; endmodule
Define reusable interfaces with modports:
systemverilog// AXI-Stream interface definition interface axis_if #( parameter int DATA_WIDTH = 32, parameter int USER_WIDTH = 1, parameter int ID_WIDTH = 1 ) ( input logic aclk, input logic aresetn ); logic tvalid; logic tready; logic [DATA_WIDTH-1:0] tdata; logic [DATA_WIDTH/8-1:0] tstrb; logic [DATA_WIDTH/8-1:0] tkeep; logic tlast; logic [ID_WIDTH-1:0] tid; logic [ID_WIDTH-1:0] tdest; logic [USER_WIDTH-1:0] tuser; // Master modport modport master ( input aclk, aresetn, tready, output tvalid, tdata, tstrb, tkeep, tlast, tid, tdest, tuser ); // Slave modport modport slave ( input aclk, aresetn, tvalid, tdata, tstrb, tkeep, tlast, tid, tdest, tuser, output tready ); // Monitor modport for verification modport monitor ( input aclk, aresetn, tvalid, tready, tdata, tstrb, tkeep, tlast, tid, tdest, tuser ); // Helper tasks for verification task automatic wait_for_handshake(); @(posedge aclk); while (!(tvalid && tready)) @(posedge aclk); endtask endinterface // Using the interface in a module module axis_register #( parameter int DATA_WIDTH = 32 ) ( input logic clk, input logic rst_n, axis_if.slave s_axis, axis_if.master m_axis ); // Skid buffer implementation logic [DATA_WIDTH-1:0] data_reg; logic valid_reg; always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin valid_reg <= 1'b0; data_reg <= '0; end else if (s_axis.tready) begin valid_reg <= s_axis.tvalid; data_reg <= s_axis.tdata; end end assign m_axis.tvalid = valid_reg; assign m_axis.tdata = data_reg; assign s_axis.tready = m_axis.tready || !valid_reg; endmodule
Create and use SystemVerilog packages:
systemverilog// Package definition package fpga_pkg; // Type definitions typedef enum logic [2:0] { IDLE = 3'b000, INIT = 3'b001, RUN = 3'b010, PAUSE = 3'b011, DONE = 3'b100, ERROR = 3'b101 } state_t; // Struct definitions typedef struct packed { logic valid; logic [31:0] data; logic [3:0] strb; logic last; } axi_data_t; // Constants localparam int CLK_FREQ_HZ = 100_000_000; localparam int TIMEOUT_CYCLES = CLK_FREQ_HZ / 1000; // 1ms // Functions function automatic int clog2(int value); int result = 0; value = value - 1; while (value > 0) begin result++; value = value >> 1; end return result; endfunction function automatic logic [31:0] reverse_bits(logic [31:0] data); for (int i = 0; i < 32; i++) begin reverse_bits[i] = data[31-i]; end endfunction endpackage // Using the package module my_design import fpga_pkg::*; ( input logic clk, input logic rst_n, input logic start, output state_t current_state, output axi_data_t output_data ); state_t state, next_state; always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) state <= IDLE; else state <= next_state; end always_comb begin next_state = state; case (state) IDLE: if (start) next_state = RUN; RUN: next_state = DONE; DONE: next_state = IDLE; default: next_state = IDLE; endcase end assign current_state = state; endmodule
Use arrays correctly for synthesis:
systemverilogmodule array_examples #( parameter int WIDTH = 8, parameter int DEPTH = 4 ) ( input logic clk, input logic rst_n, input logic [WIDTH-1:0] data_in, output logic [WIDTH-1:0] data_out ); // Unpacked array - multiple memory locations logic [WIDTH-1:0] memory_array [DEPTH]; // Memory inference // Packed array - single contiguous bit vector logic [DEPTH-1:0][WIDTH-1:0] shift_reg; // Shift register // Multi-dimensional packed array logic [3:0][7:0] packed_data; // 32-bit value as 4 bytes // Multi-dimensional unpacked array logic [7:0] mem_2d [4][8]; // 4x8 array of bytes // Shift register using packed array always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin shift_reg <= '0; end else begin shift_reg <= {shift_reg[DEPTH-2:0], data_in}; end end // Memory write logic [$clog2(DEPTH)-1:0] wr_addr; always_ff @(posedge clk) begin memory_array[wr_addr] <= data_in; end // Byte access to packed data assign packed_data[0] = 8'hAB; // Least significant byte assign packed_data[3] = 8'hCD; // Most significant byte assign data_out = shift_reg[DEPTH-1]; endmodule
Apply assignments correctly:
systemverilog// CORRECT: Non-blocking for sequential logic always_ff @(posedge clk) begin reg_a <= data_in; // Non-blocking reg_b <= reg_a; // Non-blocking - creates pipeline reg_c <= reg_b; // Non-blocking end // CORRECT: Blocking for combinational logic always_comb begin temp = a & b; // Blocking result = temp | c; // Blocking - uses updated temp end // INCORRECT: Don't mix in sequential blocks always_ff @(posedge clk) begin temp = data_in; // WRONG: blocking in sequential reg_a <= temp; end // CORRECT: Use intermediate signals logic temp_comb; always_comb begin temp_comb = data_in & enable; end always_ff @(posedge clk) begin reg_a <= temp_comb; // Non-blocking end
Apply vendor-specific attributes:
systemverilogmodule synthesis_attributes ( input logic clk, input logic rst_n, input logic async_in, output logic sync_out ); // Xilinx: ASYNC_REG for synchronizers (* ASYNC_REG = "TRUE" *) logic [1:0] sync_reg; // Xilinx: Keep signal for debugging (* KEEP = "TRUE" *) logic debug_signal; // Xilinx: RAM style control (* RAM_STYLE = "block" *) logic [7:0] block_mem [1024]; (* RAM_STYLE = "distributed" *) logic [7:0] dist_mem [16]; // Xilinx: FSM encoding (* FSM_ENCODING = "one_hot" *) enum logic [2:0] { IDLE, RUN, DONE } state; // Intel: RAM inference (* ramstyle = "M20K" *) logic [7:0] intel_mem [1024]; // Synchronizer implementation always_ff @(posedge clk or negedge rst_n) begin if (!rst_n) begin sync_reg <= '0; end else begin sync_reg <= {sync_reg[0], async_in}; end end assign sync_out = sync_reg[1]; endmodule
This skill integrates with the following processes:
| Process | Integration Point | |---------|-------------------| | verilog-systemverilog-design.js | Primary Verilog/SV development | | testbench-development.js | SV testbench generation | | rtl-module-architecture.js | Module design | | uvm-testbench.js | UVM environment development |
json{ "module": { "name": "sync_fifo", "language": "SystemVerilog", "standard": "IEEE 1800-2017", "parameters": [ { "name": "DATA_WIDTH", "type": "int", "default": 8 }, { "name": "DEPTH", "type": "int", "default": 16 } ], "ports": [ { "name": "clk", "direction": "input", "type": "logic" }, { "name": "rst_n", "direction": "input", "type": "logic" } ], "interfaces": [] }, "package": { "name": "fpga_pkg", "types": ["state_t", "axi_data_t"], "functions": ["clog2", "reverse_bits"], "constants": ["CLK_FREQ_HZ", "TIMEOUT_CYCLES"] }, "compliance": { "synthesizable": true, "alwaysBlocksCorrect": true, "noInferredLatches": true, "noBlockingInSequential": true }, "artifacts": [ "src/sync_fifo.sv", "src/fpga_pkg.sv", "src/axis_if.sv", "tb/sync_fifo_tb.sv" ] }
always_ff for flip-flops (sequential logic)always_comb for combinational logicalways_latch only for intentional latches (rare)always @* in SystemVerilog<=) in always_ff=) in always_combparameter for configurable valueslocalparam for derived/internal constants$clog2() for address width calculationlogic instead of wire/reg'0 and '1 for all-zeros/all-onesverilog-systemverilog-design.js - Verilog/SV development processuvm-testbench.js - UVM testbench development| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 38,621 | 36,098 | -7% | 1 | 1 | 0% | 8,280 | 10,628 | +28% | 0 | 0 | — |
case-02 | pass→pass | 39,354 | 34,032 | -14% | 1 | 1 | 0% | 7,273 | 11,212 | +54% | 0 | 0 | — |
case-03 | fail→pass | 34,157 | 16,287 | -52% | 1 | 1 | 0% | 5,957 | 7,869 | +32% | 0 | 0 | — |
case-04 | pass→pass | 30,624 | 33,253 | +9% | 1 | 1 | 0% | 5,057 | 9,967 | +97% | 0 | 0 | — |
case-05 | pass→fail | 16,083 | 26,853 | +67% | 1 | 1 | 0% | 3,607 | 10,382 | +188% | 0 | 0 | — |
case-06 | pass→pass | 9,050 | 12,922 | +43% | 1 | 1 | 0% | 1,691 | 6,653 | +293% | 0 | 0 | — |
case-07 | pass→pass | 8,824 | 9,292 | +5% | 1 | 1 | 0% | 1,325 | 6,058 | +357% | 0 | 0 | — |
case-08 | pass→pass | 12,664 | 11,387 | -10% | 1 | 1 | 0% | 1,886 | 6,428 | +241% | 0 | 0 | — |
case-09 | pass→pass | 11,720 | 16,941 | +45% | 1 | 1 | 0% | 2,357 | 7,578 | +222% | 0 | 0 | — |
case-10 | pass→pass | 19,544 | 17,156 | -12% | 1 | 1 | 0% | 3,117 | 7,815 | +151% | 0 | 0 | — |
case-11 | pass→pass | 15,918 | 13,342 | -16% | 1 | 1 | 0% | 2,392 | 6,351 | +166% | 0 | 0 | — |
case-12 | pass→pass | 16,751 | 17,029 | +2% | 1 | 1 | 0% | 3,043 | 7,412 | +144% | 0 | 0 | — |
case-13 | pass→pass | 9,586 | 16,683 | +74% | 1 | 1 | 0% | 1,867 | 6,782 | +263% | 0 | 0 | — |
case-14 | pass→pass | 14,994 | 22,358 | +49% | 1 | 1 | 0% | 2,716 | 8,392 | +209% | 0 | 0 | — |
case-15 | pass→pass | 13,040 | 11,761 | -10% | 1 | 1 | 0% | 2,068 | 6,644 | +221% | 0 | 0 | — |
case-16 | pass→pass | 9,465 | 10,000 | +6% | 1 | 1 | 0% | 1,473 | 5,830 | +296% | 0 | 0 | — |
case-17 | pass→pass | 15,963 | 29,435 | +84% | 1 | 1 | 0% | 2,289 | 8,730 | +281% | 0 | 0 | — |
case-18 | pass→pass | 18,130 | 16,301 | -10% | 1 | 1 | 0% | 2,781 | 7,656 | +175% | 0 | 0 | — |
case-19 | pass→pass | 6,269 | 13,508 | +115% | 1 | 1 | 0% | 1,028 | 6,209 | +504% | 0 | 0 | — |
case-20 | pass→pass | 18,666 | 19,874 | +6% | 1 | 1 | 0% | 2,959 | 7,546 | +155% | 0 | 0 | — |
case-21 | pass→pass | 19,167 | 20,976 | +9% | 1 | 1 | 0% | 3,852 | 8,503 | +121% | 0 | 0 | — |
case-22 | fail→pass | 12,454 | 15,925 | +28% | 1 | 1 | 0% | 1,861 | 6,761 | +263% | 0 | 0 | — |
case-23 | pass→pass | 7,175 | 10,057 | +40% | 1 | 1 | 0% | 1,111 | 5,692 | +412% | 0 | 0 | — |
case-24 | fail→pass | 31,022 | 25,108 | -19% | 1 | 1 | 0% | 5,166 | 9,887 | +91% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted. The headline lift of +13 percentage points is the difference between those two pass rates over the 24 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.