From 77e13f522a0555721d312f610b743adae96cc2e8 Mon Sep 17 00:00:00 2001
From: Qi Xuesong <1411979083@qq.com>
Date: Wed, 16 Jul 2025 20:44:48 +0800
Subject: [PATCH 1/2] [bitmanip] Remove redundant logic

---
 rtl/ibex_alu.sv | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/rtl/ibex_alu.sv b/rtl/ibex_alu.sv
index 563f571cf4..b0e907a01f 100644
--- a/rtl/ibex_alu.sv
+++ b/rtl/ibex_alu.sv
@@ -458,7 +458,7 @@ module ibex_alu #(
     always_comb begin
       unique case (1'b1)
         zbe_op:      bitcnt_bits = operand_b_i;
-        bitcnt_cz:   bitcnt_bits = bitcnt_bit_mask & ~bitcnt_mask_op; // clz / ctz
+        bitcnt_cz:   bitcnt_bits = bitcnt_bit_mask; // clz / ctz
         default:     bitcnt_bits = operand_a_i; // cpop
       endcase
     end

From a754dbca8fe0683991146053703948dee0bc3a67 Mon Sep 17 00:00:00 2001
From: Qi Xuesong <1411979083@qq.com>
Date: Mon, 21 Jul 2025 22:43:08 +0800
Subject: [PATCH 2/2] [bitmanip] Remove redundant logic in converting clz/ctz
 to cpop

---
 rtl/ibex_alu.sv | 25 ++++++++++---------------
 1 file changed, 10 insertions(+), 15 deletions(-)

diff --git a/rtl/ibex_alu.sv b/rtl/ibex_alu.sv
index b0e907a01f..afbbcfbc94 100644
--- a/rtl/ibex_alu.sv
+++ b/rtl/ibex_alu.sv
@@ -437,28 +437,23 @@ module ibex_alu #(
     assign bitcnt_cz     = bitcnt_ctz | bitcnt_clz;
     assign bitcnt_result = bitcnt_partial[31];
 
-    // Bit-mask generation for clz and ctz:
-    // The bit mask is generated by spreading the lowest-order set bit in the operand to all
-    // higher order bits. The resulting mask is inverted to cover the lowest order zeros. In order
-    // to create the bit mask for leading zeros, the input operand needs to be reversed.
-    assign bitcnt_mask_op = bitcnt_clz ? operand_a_rev : operand_a_i;
+    // Bitcnt_bits generation for clz and ctz:
+    // The operand of ctz can be divided into 3 parts: the don't-care bits in the high-order bits,
+    // the lowest-order set bit, and the trailing zeros. Bitcnt_bit_mask is obtained by subtracting 1
+    // from bitcnt_mask_op, which clears the lowest-order set bit and sets all the trailing zeros to 1.
+    // Then, bitcnt_bit_mask & ~bitcnt_mask_op clears the don't-care bits.
+    // At this point, counting the number of 1s gives the number of trailing zeros.
+    // In order to create the bit mask for leading zeros, the input operand needs to be reversed.
 
-    always_comb begin
-      bitcnt_bit_mask = bitcnt_mask_op;
-      bitcnt_bit_mask |= bitcnt_bit_mask << 1;
-      bitcnt_bit_mask |= bitcnt_bit_mask << 2;
-      bitcnt_bit_mask |= bitcnt_bit_mask << 4;
-      bitcnt_bit_mask |= bitcnt_bit_mask << 8;
-      bitcnt_bit_mask |= bitcnt_bit_mask << 16;
-      bitcnt_bit_mask = ~bitcnt_bit_mask;
-    end
+    assign bitcnt_mask_op = bitcnt_clz ? operand_a_rev : operand_a_i;
+    assign bitcnt_bit_mask = bitcnt_mask_op - 32'h1;
 
     assign zbe_op = (operator_i == ALU_BCOMPRESS) | (operator_i == ALU_BDECOMPRESS);
 
     always_comb begin
       unique case (1'b1)
         zbe_op:      bitcnt_bits = operand_b_i;
-        bitcnt_cz:   bitcnt_bits = bitcnt_bit_mask; // clz / ctz
+        bitcnt_cz:   bitcnt_bits = bitcnt_bit_mask & ~bitcnt_mask_op; // clz / ctz
         default:     bitcnt_bits = operand_a_i; // cpop
       endcase
     end