added C-extension support

2026-06-18 19:10:32 +03:00
parent caf64b689b
commit 3391969951
6 changed files with 378 additions and 15 deletions
@@ -2,21 +2,21 @@ CC = riscv64-linux-gnu-gcc
 OBJCOPY = riscv64-linux-gnu-objcopy
 EMULATOR = ./amach

-ASFLAGS = -march=rv32ima -mabi=ilp32 -nostdlib -s -static -fno-pic -fno-pie -mno-relax
+ASFLAGS = -march=rv32imac -mabi=ilp32 -nostdlib -s -static -fno-pic -fno-pie -mno-relax
 LDFLAGS = -Wl,-Ttext=0x100b0 -Wl,--build-id=none

-all: rv32ima_tests.bin
+all: rv32imac_tests.bin

-rv32ima_tests.elf: rv32ima_tests.s
-	$(CC) $(ASFLAGS) $(LDFLAGS) rv32ima_tests.s -o rv32ima_tests.elf
+rv32imac_tests.elf: rv32imac_tests.s
+	$(CC) $(ASFLAGS) $(LDFLAGS) rv32imac_tests.s -o rv32imac_tests.elf

-rv32ima_tests.bin: rv32ima_tests.elf
-	$(OBJCOPY) -O binary rv32ima_tests.elf rv32ima_tests.bin
+rv32imac_tests.bin: rv32imac_tests.elf
+	$(OBJCOPY) -O binary rv32imac_tests.elf rv32imac_tests.bin

-run: rv32ima_tests.bin
-	$(EMULATOR) rv32ima_tests.bin
+run: rv32imac_tests.bin
+	$(EMULATOR) rv32imac_tests.bin

 clean:
-	rm -f rv32ima_tests.elf rv32ima_tests.bin
+	rm -f rv32imac_tests.elf rv32imac_tests.bin

 .PHONY: all run clean
@@ -1,13 +1,13 @@
-# A-machine: a minimal RISC-V (RV32IMA) emulator in POSIX AWK
+# A-machine: a minimal RISC-V (RV32IMAC) emulator in POSIX AWK

 ## About

-A-machine is a simple and straightforward implementation of the RV32I specification with M (multiplication/division) and A (atomic operations) extenstions. It consists of two files:
+A-machine is a simple and straightforward implementation of the RV32I specification with M (multiplication/division), A (atomic operations), and C (compressed instructions) extensions. It consists of two files:

 - [amach.awk](amach.awk), the emulator core itself that supports .dec files,
 - and [amach](amach) shell script which is a two-liner that uses POSIX `od` to convert .bin files into .dec files on the fly and then passes them to the emulator core.

-Additionally, A-machine is shipped with a reference test suite that assumes that you have a full RISC-V toolchain installed under the `riscv64-linux-gnu-` prefix (adjustable in the `Makefile`). Just run `make run` to build and run all tests. In case you don't have any toolchain available, a file called `rv32ima_tests.bin` is also included into the repo.
+Additionally, A-machine is shipped with a reference test suite that assumes that you have a full RISC-V toolchain installed under the `riscv64-linux-gnu-` prefix (adjustable in the `Makefile`). Just run `make run` to build and run all tests. In case you don't have any toolchain available, a file called `rv32imac_tests.bin` is also included into the repo.

 The project is to be considered highly experimental and not production-ready in any way.

@@ -32,11 +32,10 @@ POSIXLY_CORRECT=1 awk -f amach.awk [-v LVA=...] -- path/to/program.dec

 ## Support

-Currently, A-machine only supports the **unprivileged** RV32IMA instruction set with a few additional `ecall` syscalls: 63 (sys_read, for fd 0 only), 64 (sys_write, for fd 1 and 2 only) and 93 (sys_exit). For completeness, `fence` instruction is also supported but yields a no-op.
+Currently, A-machine only supports the **unprivileged** RV32IMAC instruction set with a few additional `ecall` syscalls: 63 (sys_read, for fd 0 only), 64 (sys_write, for fd 1 and 2 only) and 93 (sys_exit). For completeness, `fence` instruction is also supported but yields a no-op.

 ## Plans (from higher to lower priority)

- Implement the C-extension (compressed instruction set).
 - Implement a simple RV32IMAC disassembler in POSIX AWK.
 - Introduce more popular Linux-compatible syscalls (that make sense to implement in AWK but don't increase the overall project complexity).
 - Test for RV32EMC compatibility (that the code built for the "embedded" variant runs here).
@@ -1,5 +1,5 @@
 #!/usr/bin/env awk -f
-# A-Machine: an experimental RISC-V (RV32IMA) emulator in POSIX AWK
+# A-Machine: an experimental RISC-V (RV32IMAC) emulator in POSIX AWK
 # with a very small subset of supported ECALLs
 # Accepts a headerless binary previously converted to .dec format
 # e.g. with POSIX od: od -An -v -tu1 program.bin > program.dec
@@ -275,8 +275,148 @@ function amach_atomic(f3, f5, aq, rl, rd, rs1, rs2,  r1, r2, val, uval, ur2) {
  } else trapout(sprintf("Illegal atomic instruction at 0x%X", pc-4))
 }

+# C-extension (compressed instructions) handler
+function amach_comp(instr,  op, f3, b12, r11_7, r9_7, r6_2, r4_2, shamt, imm, base, val) {
+  op = instr % 4
+  f3 = int(instr / 8192) % 8
+  b12 = int(instr / 4096) % 2
+  r11_7 = int(instr / 128) % 32
+  r9_7 = (int(instr / 128) % 8) + 8
+  r6_2 = int(instr / 4) % 32
+  r4_2 = (int(instr / 4) % 8) + 8
+  shamt = r6_2 + b12 * 32
+
+  if (op == 0) {
+    if (f3 == 0) { # C.ADDI4SPN
+      imm = (int(instr / 128) % 16) * 64 + (int(instr / 2048) % 4) * 16 + (int(instr / 32) % 2) * 8 + (int(instr / 64) % 2) * 4
+      if (imm == 0) trapout(sprintf("Illegal instruction C.ADDI4SPN at 0x%X", pc-2))
+      setreg(r4_2, getreg(2) + imm)
+    } else if (f3 == 2) { # C.LW
+      imm = (int(instr / 32) % 2) * 64 + (int(instr / 1024) % 8) * 8 + (int(instr / 64) % 2) * 4
+      base = (getreg(r9_7) + imm) % 4294967296
+      if (base < 0) base += 4294967296
+      val = MEM[base] + MEM[(base+1)%4294967296]*256 + MEM[(base+2)%4294967296]*65536 + MEM[(base+3)%4294967296]*16777216
+      setreg(r4_2, val)
+    } else if (f3 == 6) { # C.SW
+      imm = (int(instr / 32) % 2) * 64 + (int(instr / 1024) % 8) * 8 + (int(instr / 64) % 2) * 4
+      base = (getreg(r9_7) + imm) % 4294967296
+      if (base < 0) base += 4294967296
+      val = getreg(r4_2)
+      if (val < 0) val += 4294967296
+      MEM[base] = val % 256
+      MEM[(base+1)%4294967296] = int(val/256) % 256
+      MEM[(base+2)%4294967296] = int(val/65536) % 256
+      MEM[(base+3)%4294967296] = int(val/16777216) % 256
+    } else trapout(sprintf("Illegal instruction at 0x%X", pc-2))
+  } else if (op == 1) {
+    if (f3 == 0) { # C.NOP / C.ADDI
+      imm = r6_2 + b12 * 32
+      if (imm >= 32) imm -= 64
+      setreg(r11_7, getreg(r11_7) + imm)
+    } else if (f3 == 1 || f3 == 5) { # C.JAL / C.J
+      imm = b12 * 2048 + (int(instr / 256) % 2) * 1024 + (int(instr / 512) % 4) * 256 + (int(instr / 64) % 2) * 128 + (int(instr / 128) % 2) * 64 + (int(instr / 4) % 2) * 32 + (int(instr / 2048) % 2) * 16 + (int(instr / 8) % 8) * 2
+      if (imm >= 2048) imm -= 4096
+      if (f3 == 1) setreg(1, pc)
+      pc = (pc - 2) + imm
+    } else if (f3 == 2) { # C.LI
+      imm = r6_2 + b12 * 32
+      if (imm >= 32) imm -= 64
+      setreg(r11_7, imm)
+    } else if (f3 == 3) { # C.ADDI16SP / C.LUI
+      if (r11_7 == 2) { # C.ADDI16SP
+        imm = b12 * 512 + (int(instr / 8) % 4) * 128 + (int(instr / 32) % 2) * 64 + (int(instr / 4) % 2) * 32 + (int(instr / 64) % 2) * 16
+        if (imm >= 512) imm -= 1024
+        if (imm == 0) trapout(sprintf("Illegal instruction C.ADDI16SP at 0x%X", pc-2))
+        setreg(2, getreg(2) + imm)
+      } else if (r11_7 != 0) { # C.LUI
+        imm = r6_2 + b12 * 32
+        if (imm >= 32) imm -= 64
+        imm = imm * 4096
+        if (imm == 0) trapout(sprintf("Illegal instruction C.LUI at 0x%X", pc-2))
+        setreg(r11_7, imm)
+      }
+    } else if (f3 == 4) {
+      val = int(instr / 1024) % 4
+      if (val == 0) { # C.SRLI
+        if (shamt >= 32) trapout(sprintf("Illegal shift amount %d at 0x%X", shamt, pc-2))
+        imm = getreg(r9_7)
+        if (imm < 0) imm += 4294967296
+        setreg(r9_7, int(imm / (2^shamt)))
+      } else if (val == 1) { # C.SRAI
+        if (shamt >= 32) trapout(sprintf("Illegal shift amount %d at 0x%X", shamt, pc-2))
+        setreg(r9_7, floor(getreg(r9_7) / (2^shamt)))
+      } else if (val == 2) { # C.ANDI
+        imm = r6_2 + b12 * 32
+        if (imm >= 32) imm -= 64
+        setreg(r9_7, bw_and(getreg(r9_7), imm))
+      } else if (val == 3) { # C.SUB, C.XOR, C.OR, C.AND
+        imm = int(instr / 32) % 4
+        if (b12 != 0) trapout(sprintf("Illegal register-register instruction at 0x%X", pc-2))
+        if (imm == 0) setreg(r9_7, getreg(r9_7) - getreg(r4_2))
+        else if (imm == 1) setreg(r9_7, bw_xor(getreg(r9_7), getreg(r4_2)))
+        else if (imm == 2) setreg(r9_7, bw_or(getreg(r9_7), getreg(r4_2)))
+        else if (imm == 3) setreg(r9_7, bw_and(getreg(r9_7), getreg(r4_2)))
+      }
+    } else if (f3 == 6 || f3 == 7) { # C.BEQZ / C.BNEZ
+      imm = b12 * 256 + (int(instr / 32) % 4) * 64 + (int(instr / 4) % 2) * 32 + (int(instr / 1024) % 4) * 8 + (int(instr / 8) % 4) * 2
+      if (imm >= 256) imm -= 512
+      val = (f3 == 6) ? (getreg(r9_7) == 0) : (getreg(r9_7) != 0)
+      pc = val ? (pc - 2) + imm : pc
+    }
+  } else if (op == 2) {
+    if (f3 == 0) { # C.SLLI
+      if (shamt >= 32) trapout(sprintf("Illegal shift amount %d at 0x%X", shamt, pc-2))
+      if (r11_7 != 0) {
+        setreg(r11_7, (getreg(r11_7) * (2^shamt)) % 4294967296)
+      }
+    } else if (f3 == 2) { # C.LWSP
+      if (r11_7 == 0) trapout(sprintf("Illegal instruction C.LWSP with rd=0 at 0x%X", pc-2))
+      imm = (int(instr / 4) % 4) * 64 + b12 * 32 + (int(instr / 16) % 8) * 4
+      base = (getreg(2) + imm) % 4294967296
+      if (base < 0) base += 4294967296
+      val = MEM[base] + MEM[(base+1)%4294967296]*256 + MEM[(base+2)%4294967296]*65536 + MEM[(base+3)%4294967296]*16777216
+      setreg(r11_7, val)
+    } else if (f3 == 4) { # C.JR, C.MV, C.JALR, C.ADD
+      if (b12 == 0) {
+        if (r6_2 == 0) { # C.JR
+          if (r11_7 == 0) trapout(sprintf("Illegal instruction C.JR with rs1=0 at 0x%X", pc-2))
+          pc = getreg(r11_7)
+        } else { # C.MV
+          setreg(r11_7, getreg(r6_2))
+        }
+      } else {
+        if (r6_2 == 0) {
+          if (r11_7 == 0) { # C.EBREAK
+            trapout(sprintf("EBREAK at 0x%X", pc-2))
+          } else { # C.JALR
+            imm = getreg(r11_7)
+            setreg(1, pc)
+            pc = imm
+          }
+        } else { # C.ADD
+          setreg(r11_7, getreg(r11_7) + getreg(r6_2))
+        }
+      }
+    } else if (f3 == 6) { # C.SWSP
+      imm = (int(instr / 128) % 4) * 64 + (int(instr / 512) % 16) * 4
+      base = (getreg(2) + imm) % 4294967296
+      if (base < 0) base += 4294967296
+      val = getreg(r6_2)
+      if (val < 0) val += 4294967296
+      MEM[base] = val % 256
+      MEM[(base+1)%4294967296] = int(val/256) % 256
+      MEM[(base+2)%4294967296] = int(val/65536) % 256
+      MEM[(base+3)%4294967296] = int(val/16777216) % 256
+    } else trapout(sprintf("Illegal instruction at 0x%X", pc-2))
+  } else trapout(sprintf("Illegal instruction at 0x%X", pc-2))
+}
+
 # main instruction decoding and execution routine
 function amach_exec(instr,  opcode, rd, rs1, rs2, imm, funct3, funct7) {
+  if ((instr % 4) != 3) {
+    amach_comp(instr)
+    return
+  }
  opcode = instr % 128
  instr = int(instr / 128)
  rd = instr % 32
@@ -1,4 +1,5 @@
 .section .text
+.option norvc
 .global _start

 .macro INC_TESTS
@@ -530,6 +531,229 @@ _start:
  # ==========================================
  fence

+  # ==========================================
+  # RV32C Compressed Instruction Tests
+  # ==========================================
+  .option rvc
+
+  # 1. c.addi4spn
+  INC_TESTS
+  mv s0, sp
+  c.addi4spn a0, sp, 40
+  sub a0, a0, s0
+  ASSERT_EQ a0, 40, c_addi4spn_val
+
+  # 2. c.lw and c.sw
+  INC_TESTS
+  la a1, data_buffer
+  li a0, 0x76543210
+  sw a0, 4(a1)
+  c.lw a2, 4(a1)
+  ASSERT_EQ a2, 0x76543210, c_lw_val
+
+  INC_TESTS
+  li a0, 0x11223344
+  c.sw a0, 8(a1)
+  lw a2, 8(a1)
+  ASSERT_EQ a2, 0x11223344, c_sw_val
+
+  # 3. c.nop
+  INC_TESTS
+  c.nop
+  li a0, 1
+  ASSERT_EQ a0, 1, c_nop_val
+
+  # 4. c.addi
+  INC_TESTS
+  li a0, 10
+  c.addi a0, -3
+  ASSERT_EQ a0, 7, c_addi_val
+
+  # 5. c.jal
+  INC_TESTS
+3:
+  c.jal 1f
+  li a0, 0
+  j 2f
+1:
+  li a0, 1
+2:
+  ASSERT_EQ a0, 1, c_jal_val
+  
+  INC_TESTS
+  la a1, 3b + 2
+  ASSERT_EQ_REG ra, a1, c_jal_link
+
+  # 6. c.li
+  INC_TESTS
+  c.li a0, -15
+  ASSERT_EQ a0, -15, c_li_val
+
+  # 7. c.addi16sp
+  INC_TESTS
+  mv s0, sp
+  c.addi16sp sp, -32
+  sub a0, s0, sp
+  ASSERT_EQ a0, 32, c_addi16sp_val
+  mv sp, s0
+
+  # 8. c.lui
+  INC_TESTS
+  c.lui a0, 15
+  ASSERT_EQ a0, 0xf000, c_lui_val
+
+  # 9. c.srli, c.srai, c.andi
+  INC_TESTS
+  li a0, 0x80
+  c.srli a0, 3
+  ASSERT_EQ a0, 0x10, c_srli_val
+
+  INC_TESTS
+  li a0, -16
+  c.srai a0, 2
+  ASSERT_EQ a0, -4, c_srai_val
+
+  INC_TESTS
+  li a0, 15
+  c.andi a0, 9
+  ASSERT_EQ a0, 9, c_andi_val
+
+  # 10. c.sub, c.xor, c.or, c.and
+  INC_TESTS
+  li a0, 15
+  li a1, 5
+  c.sub a0, a1
+  ASSERT_EQ a0, 10, c_sub_val
+
+  INC_TESTS
+  li a0, 5
+  li a1, 3
+  c.xor a0, a1
+  ASSERT_EQ a0, 6, c_xor_val
+
+  INC_TESTS
+  li a0, 5
+  li a1, 3
+  c.or a0, a1
+  ASSERT_EQ a0, 7, c_or_val
+
+  INC_TESTS
+  li a0, 5
+  li a1, 3
+  c.and a0, a1
+  ASSERT_EQ a0, 1, c_and_val
+
+  # 11. c.j
+  INC_TESTS
+  c.j 1f
+  li a0, 0
+  j 2f
+1:
+  li a0, 1
+2:
+  ASSERT_EQ a0, 1, c_j_val
+
+  # 12. c.beqz, c.bnez
+  INC_TESTS
+  li a0, 0
+  c.beqz a0, 1f
+  li a1, 0
+  j 2f
+1:
+  li a1, 1
+2:
+  ASSERT_EQ a1, 1, c_beqz_taken
+
+  INC_TESTS
+  li a0, 1
+  c.beqz a0, 1f
+  li a1, 0
+  j 2f
+1:
+  li a1, 1
+2:
+  ASSERT_EQ a1, 0, c_beqz_not_taken
+
+  INC_TESTS
+  li a0, 1
+  c.bnez a0, 1f
+  li a1, 0
+  j 2f
+1:
+  li a1, 1
+2:
+  ASSERT_EQ a1, 1, c_bnez_taken
+
+  INC_TESTS
+  li a0, 0
+  c.bnez a0, 1f
+  li a1, 0
+  j 2f
+1:
+  li a1, 1
+2:
+  ASSERT_EQ a1, 0, c_bnez_not_taken
+
+  # 13. c.slli
+  INC_TESTS
+  li a0, 5
+  c.slli a0, 2
+  ASSERT_EQ a0, 20, c_slli_val
+
+  # 14. c.lwsp, c.swsp
+  INC_TESTS
+  mv s0, sp
+  c.addi16sp sp, -32
+  li a0, 0x11223344
+  sw a0, 16(sp)
+  c.lwsp a1, 16(sp)
+  ASSERT_EQ a1, 0x11223344, c_lwsp_val
+  
+  INC_TESTS
+  li a0, 0x55667788
+  c.swsp a0, 20(sp)
+  lw a1, 20(sp)
+  ASSERT_EQ a1, 0x55667788, c_swsp_val
+  mv sp, s0
+
+  # 15. c.jr, c.mv, c.jalr, c.add
+  INC_TESTS
+  la a0, 1f
+  c.jr a0
+  li a1, 0
+  j 2f
+1:
+  li a1, 1
+2:
+  ASSERT_EQ a1, 1, c_jr_val
+
+  INC_TESTS
+  li a1, 42
+  c.mv a0, a1
+  ASSERT_EQ a0, 42, c_mv_val
+
+  INC_TESTS
+  la a0, 1f
+3:
+  c.jalr a0
+  li a1, 0
+  j 2f
+1:
+  li a1, 1
+2:
+  ASSERT_EQ a1, 1, c_jalr_val
+  INC_TESTS
+  la a2, 3b + 2
+  ASSERT_EQ_REG ra, a2, c_jalr_link
+
+  INC_TESTS
+  li a0, 10
+  li a1, 20
+  c.add a0, a1
+  ASSERT_EQ a0, 30, c_add_val
+
+  .option norvc
+
  # ==========================================
  # Report Results
  # ==========================================