synth_intel_alm: alternative synthesis for Intel FPGAs

By operating at a layer of abstraction over the rather clumsy Intel primitives,
we can avoid special hacks like `dffinit -highlow` in favour of simple techmapping.

This also makes the primitives much easier to manipulate, and more descriptive
(no more cyclonev_lcell_comb to mean anything from a LUT2 to a LUT6).

1 files changed, 64 insertions, 0 deletions

diff --git a/techlibs/intel_alm/common/arith_alm_map.v b/techlibs/intel_alm/common/arith_alm_map.v
new file mode 100644
index 000000000..ddf81d9d0
--- /dev/null
+++ b/techlibs/intel_alm/common/arith_alm_map.v
@@ -0,0 +1,64 @@
+`default_nettype none
+
+module \$alu (A, B, CI, BI, X, Y, CO);
+
+parameter A_SIGNED = 0;
+parameter B_SIGNED = 0;
+parameter A_WIDTH = 1;
+parameter B_WIDTH = 1;
+parameter Y_WIDTH = 1;
+
+parameter _TECHMAP_CONSTMSK_CI_ = 0;
+parameter _TECHMAP_CONSTVAL_CI_ = 0;
+
+input [A_WIDTH-1:0] A;
+input [B_WIDTH-1:0] B;
+input CI, BI;
+output [Y_WIDTH-1:0] X, Y, CO;
+
+wire [Y_WIDTH-1:0] A_buf, B_buf;
+\$pos #(.A_SIGNED(A_SIGNED), .A_WIDTH(A_WIDTH), .Y_WIDTH(Y_WIDTH)) A_conv (.A(A), .Y(A_buf));
+\$pos #(.A_SIGNED(B_SIGNED), .A_WIDTH(B_WIDTH), .Y_WIDTH(Y_WIDTH)) B_conv (.A(B), .Y(B_buf));
+
+wire [Y_WIDTH-1:0] AA = A_buf;
+wire [Y_WIDTH-1:0] BB = BI ? ~B_buf : B_buf;
+wire [Y_WIDTH-1:0] BX = B_buf;
+wire [Y_WIDTH:0] ALM_CARRY;
+
+// Start of carry chain
+generate
+    if (_TECHMAP_CONSTMSK_CI_ == 1) begin
+        assign ALM_CARRY[0] = _TECHMAP_CONSTVAL_CI_;
+    end else begin
+        MISTRAL_ALUT_ARITH #(
+            .LUT0(16'b1010_1010_1010_1010), // Q = A
+            .LUT1(16'b0000_0000_0000_0000), // Q = 0 (LUT1's input to the adder is inverted)
+        ) alm_start (
+            .A(CI), .B(1'b1), .C(1'b1), .D0(1'b1), .D1(1'b1),
+            .CI(1'b0),
+            .CO(ALM_CARRY[0])
+        );
+    end
+endgenerate
+
+// Carry chain
+genvar i;
+generate for (i = 0; i < Y_WIDTH; i = i + 1) begin:slice
+    // TODO: mwk suggests that a pass could merge pre-adder logic into this.
+    MISTRAL_ALUT_ARITH #(
+        .LUT0(16'b1010_1010_1010_1010), // Q = A
+        .LUT1(16'b1100_0011_1100_0011), // Q = C ? B : ~B (LUT1's input to the adder is inverted)
+    ) alm_i (
+        .A(AA[i]), .B(BX[i]), .C(BI), .D0(1'b1), .D1(1'b1),
+        .CI(ALM_CARRY[i]),
+        .SO(Y[i]),
+        .CO(ALM_CARRY[i+1])
+    );
+
+    // ALM carry chain is not directly accessible, so calculate the carry through soft logic if really needed.
+    assign CO[i] = (AA[i] && BB[i]) || ((Y[i] ^ AA[i] ^ BB[i]) && (AA[i] || BB[i]));
+end endgenerate
+
+assign X = AA ^ BB;
+
+endmodule