Add RISC-V vector add and triad benchmarks with corresponding Makefiles and assembly files

examples/add/Makefile

@@ -0,0 +1,24 @@
+# Makefile for RISC-V add example
+
+CC = gcc
+CFLAGS = -O3
+CFLAGS_VEC = -O3 -march=rv64gcv
+
+# Default target with -O3
+all: add_riscv add_riscv_vec
+
+# Build with -O3 optimization
+add_riscv: add_riscv.c
+	$(CC) $(CFLAGS) -o add_riscv add_riscv.c
+	$(CC) $(CFLAGS) -S -o add_riscv.s add_riscv.c
+
+# Build with vector extensions
+add_riscv_vec: add_riscv.c
+	$(CC) $(CFLAGS_VEC) -o add_riscv_vec add_riscv.c
+	$(CC) $(CFLAGS_VEC) -S -o add_riscv_vec.s add_riscv.c
+
+# Clean up
+clean:
+	rm -f add_riscv add_riscv_vec add_riscv.s add_riscv_vec.s
+
+.PHONY: all clean 

examples/add/add_riscv.c

@@ -0,0 +1,54 @@
+// Vector add benchmark for RISC-V testing
+// a[i] = b[i] + c[i]
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#define DTYPE double
+
+void kernel(DTYPE* a, DTYPE* b, DTYPE* c, const int size)
+{
+    // OSACA start marker will be added around this loop
+    for(int i=0; i<size; i++) {
+        a[i] = b[i] + c[i];
+    }
+    // OSACA end marker will be added
+}
+
+int main(int argc, char *argv[]) {
+    int size = 1000;
+    if(argc > 1) {
+        size = atoi(argv[1]);
+    }
+    
+    printf("RISC-V Vector add: a[i] = b[i] + c[i], size=%d\n", size);
+    
+    // Allocate memory
+    DTYPE* a = (DTYPE*)malloc(size * sizeof(DTYPE));
+    DTYPE* b = (DTYPE*)malloc(size * sizeof(DTYPE));
+    DTYPE* c = (DTYPE*)malloc(size * sizeof(DTYPE));
+    
+    // Initialize arrays
+    for(int i=0; i<size; i++) {
+        a[i] = 0.0;
+        b[i] = i;
+        c[i] = i * 2.0;
+    }
+    
+    // Run kernel
+    kernel(a, b, c, size);
+    
+    // Check result (to prevent optimization)
+    DTYPE checksum = 0.0;
+    for(int i=0; i<size; i++) {
+        checksum += a[i];
+    }
+    printf("Checksum: %f\n", checksum);
+    
+    // Cleanup
+    free(a);
+    free(b);
+    free(c);
+    
+    return 0;
+} 

examples/riscy_asm_files/add_riscv.s

@@ -0,0 +1,178 @@
+	.file	"add_riscv.c"
+	.option pic
+	.attribute arch, "rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0"
+	.attribute unaligned_access, 0
+	.attribute stack_align, 16
+	.text
+	.align	1
+	.globl	kernel
+	.type	kernel, @function
+kernel:
+.LFB22:
+	.cfi_startproc
+	ble	a3,zero,.L1
+	slli	a3,a3,3
+	add	a5,a1,a3
+.L3:
+	fld	fa5,0(a1)
+	fld	fa4,0(a2)
+	addi	a1,a1,8
+	addi	a2,a2,8
+	fadd.d	fa5,fa5,fa4
+	addi	a0,a0,8
+	fsd	fa5,-8(a0)
+	bne	a1,a5,.L3
+.L1:
+	ret
+	.cfi_endproc
+.LFE22:
+	.size	kernel, .-kernel
+	.section	.rodata.str1.8,"aMS",@progbits,1
+	.align	3
+.LC0:
+	.string	"RISC-V Vector add: a[i] = b[i] + c[i], size=%d\n"
+	.align	3
+.LC1:
+	.string	"Checksum: %f\n"
+	.section	.text.startup,"ax",@progbits
+	.align	1
+	.globl	main
+	.type	main, @function
+main:
+.LFB23:
+	.cfi_startproc
+	addi	sp,sp,-48
+	.cfi_def_cfa_offset 48
+	sd	ra,40(sp)
+	sd	s0,32(sp)
+	sd	s1,24(sp)
+	sd	s2,16(sp)
+	sd	s3,8(sp)
+	sd	s4,0(sp)
+	.cfi_offset 1, -8
+	.cfi_offset 8, -16
+	.cfi_offset 9, -24
+	.cfi_offset 18, -32
+	.cfi_offset 19, -40
+	.cfi_offset 20, -48
+	li	a5,1
+	bgt	a0,a5,.L21
+	li	a1,1000
+	lla	a0,.LC0
+	call	printf@plt
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s1,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s2,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	li	s0,8192
+	mv	s3,a0
+	addi	s0,s0,-192
+	li	s4,1000
+.L13:
+	slli	a5,s4,32
+	srli	a2,a5,29
+	li	a1,0
+	mv	a0,s1
+	call	memset@plt
+	mv	a4,s2
+	mv	a3,s2
+	li	a5,0
+.L9:
+	fcvt.d.w	fa5,a5
+	mv	a1,a5
+	addiw	a5,a5,1
+	fsd	fa5,0(a3)
+	addi	a3,a3,8
+	bne	a5,s4,.L9
+	mv	a2,s3
+	mv	a3,s3
+	li	a5,0
+.L10:
+	fcvt.d.w	fa5,a5
+	mv	a0,a5
+	addi	a3,a3,8
+	fadd.d	fa5,fa5,fa5
+	addiw	a5,a5,1
+	fsd	fa5,-8(a3)
+	bne	a0,a1,.L10
+	mv	a5,s1
+	add	a1,s2,s0
+	mv	a3,s1
+.L11:
+	fld	fa5,0(a4)
+	fld	fa4,0(a2)
+	addi	a4,a4,8
+	addi	a2,a2,8
+	fadd.d	fa5,fa5,fa4
+	addi	a3,a3,8
+	fsd	fa5,-8(a3)
+	bne	a4,a1,.L11
+	fmv.d.x	fa5,zero
+	add	s0,s1,s0
+.L12:
+	fld	fa4,0(a5)
+	addi	a5,a5,8
+	fadd.d	fa5,fa5,fa4
+	bne	a5,s0,.L12
+.L8:
+	fmv.x.d	a1,fa5
+	lla	a0,.LC1
+	call	printf@plt
+	mv	a0,s1
+	call	free@plt
+	mv	a0,s2
+	call	free@plt
+	mv	a0,s3
+	call	free@plt
+	ld	ra,40(sp)
+	.cfi_remember_state
+	.cfi_restore 1
+	ld	s0,32(sp)
+	.cfi_restore 8
+	ld	s1,24(sp)
+	.cfi_restore 9
+	ld	s2,16(sp)
+	.cfi_restore 18
+	ld	s3,8(sp)
+	.cfi_restore 19
+	ld	s4,0(sp)
+	.cfi_restore 20
+	li	a0,0
+	addi	sp,sp,48
+	.cfi_def_cfa_offset 0
+	jr	ra
+.L21:
+	.cfi_restore_state
+	ld	a0,8(a1)
+	li	a2,10
+	li	a1,0
+	call	strtol@plt
+	sext.w	s4,a0
+	mv	a1,s4
+	lla	a0,.LC0
+	call	printf@plt
+	slli	s0,s4,3
+	mv	a0,s0
+	call	malloc@plt
+	mv	s1,a0
+	mv	a0,s0
+	call	malloc@plt
+	mv	s2,a0
+	mv	a0,s0
+	call	malloc@plt
+	mv	s3,a0
+	bgt	s4,zero,.L13
+	fmv.d.x	fa5,zero
+	j	.L8
+	.cfi_endproc
+.LFE23:
+	.size	main, .-main
+	.ident	"GCC: (GNU) 14.2.1 20250207"
+	.section	.note.GNU-stack,"",@progbits

examples/riscy_asm_files/add_riscv_vec.s

@@ -0,0 +1,225 @@
+	.file	"add_riscv.c"
+	.option pic
+	.attribute arch, "rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_v1p0_zicsr2p0_zifencei2p0_zve32f1p0_zve32x1p0_zve64d1p0_zve64f1p0_zve64x1p0_zvl128b1p0_zvl32b1p0_zvl64b1p0"
+	.attribute unaligned_access, 0
+	.attribute stack_align, 16
+	.text
+	.align	1
+	.globl	kernel
+	.type	kernel, @function
+kernel:
+.LFB22:
+	.cfi_startproc
+	ble	a3,zero,.L10
+	addiw	a5,a3,-1
+	li	a4,2
+	bleu	a5,a4,.L3
+	addi	a5,a2,8
+	addi	a4,a1,8
+	sub	a5,a0,a5
+	sub	a4,a0,a4
+	bgtu	a5,a4,.L14
+.L4:
+	csrr	a4,vlenb
+	addi	a4,a4,-16
+	bleu	a5,a4,.L3
+.L5:
+	vsetvli	a5,a3,e64,m1,ta,ma
+	vle64.v	v1,0(a1)
+	vle64.v	v2,0(a2)
+	slli	a4,a5,3
+	sub	a3,a3,a5
+	add	a1,a1,a4
+	add	a2,a2,a4
+	vfadd.vv	v1,v1,v2
+	vse64.v	v1,0(a0)
+	add	a0,a0,a4
+	bne	a3,zero,.L5
+	ret
+.L3:
+	slli	a3,a3,3
+	add	a3,a1,a3
+.L7:
+	fld	fa5,0(a1)
+	fld	fa4,0(a2)
+	addi	a1,a1,8
+	addi	a2,a2,8
+	fadd.d	fa5,fa5,fa4
+	addi	a0,a0,8
+	fsd	fa5,-8(a0)
+	bne	a1,a3,.L7
+.L10:
+	ret
+.L14:
+	mv	a5,a4
+	j	.L4
+	.cfi_endproc
+.LFE22:
+	.size	kernel, .-kernel
+	.section	.rodata.str1.8,"aMS",@progbits,1
+	.align	3
+.LC0:
+	.string	"RISC-V Vector add: a[i] = b[i] + c[i], size=%d\n"
+	.align	3
+.LC2:
+	.string	"Checksum: %f\n"
+	.section	.text.startup,"ax",@progbits
+	.align	1
+	.globl	main
+	.type	main, @function
+main:
+.LFB23:
+	.cfi_startproc
+	addi	sp,sp,-48
+	.cfi_def_cfa_offset 48
+	sd	ra,40(sp)
+	sd	s0,32(sp)
+	sd	s1,24(sp)
+	sd	s2,16(sp)
+	sd	s3,8(sp)
+	sd	s4,0(sp)
+	.cfi_offset 1, -8
+	.cfi_offset 8, -16
+	.cfi_offset 9, -24
+	.cfi_offset 18, -32
+	.cfi_offset 19, -40
+	.cfi_offset 20, -48
+	li	a5,1
+	bgt	a0,a5,.L35
+	li	a1,1000
+	lla	a0,.LC0
+	call	printf@plt
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s0,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s2,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	li	s3,8192
+	mv	s1,a0
+	addi	s3,s3,-192
+	li	s4,1000
+.L22:
+	slli	a5,s4,32
+	srli	a2,a5,29
+	li	a1,0
+	mv	a0,s0
+	call	memset@plt
+	mv	a4,s2
+	li	a5,0
+.L18:
+	fcvt.d.w	fa5,a5
+	addiw	a5,a5,1
+	addi	a4,a4,8
+	fsd	fa5,-8(a4)
+	bne	a5,s4,.L18
+	fld	fa5,.LC1,a4
+	vsetvli	a3,zero,e64,m1,ta,ma
+	mv	a2,a5
+	vfmv.v.f	v4,fa5
+	vsetvli	zero,zero,e32,mf2,ta,ma
+	vid.v	v2
+	mv	a3,s1
+.L19:
+	vsetvli	a4,a2,e32,mf2,ta,ma
+	vfwcvt.f.x.v	v1,v2
+	vsetvli	a0,zero,e32,mf2,ta,ma
+	vmv.v.x	v3,a4
+	vsetvli	zero,a4,e64,m1,ta,ma
+	vfmul.vv	v1,v1,v4
+	vsetvli	a0,zero,e32,mf2,ta,ma
+	vadd.vv	v2,v2,v3
+	vsetvli	zero,a4,e64,m1,ta,ma
+	slli	a1,a4,3
+	sub	a2,a2,a4
+	vse64.v	v1,0(a3)
+	add	a3,a3,a1
+	bne	a2,zero,.L19
+	mv	a3,s0
+	mv	a0,s1
+	mv	a1,s2
+.L20:
+	vsetvli	a4,a5,e64,m1,ta,ma
+	vle64.v	v2,0(a1)
+	vle64.v	v1,0(a0)
+	slli	a2,a4,3
+	sub	a5,a5,a4
+	add	a1,a1,a2
+	add	a0,a0,a2
+	vfadd.vv	v1,v1,v2
+	vse64.v	v1,0(a3)
+	add	a3,a3,a2
+	bne	a5,zero,.L20
+	fmv.d.x	fa5,zero
+	add	s3,s0,s3
+	mv	a5,s0
+.L21:
+	fld	fa4,0(a5)
+	addi	a5,a5,8
+	fadd.d	fa5,fa5,fa4
+	bne	s3,a5,.L21
+.L17:
+	fmv.x.d	a1,fa5
+	lla	a0,.LC2
+	call	printf@plt
+	mv	a0,s0
+	call	free@plt
+	mv	a0,s2
+	call	free@plt
+	mv	a0,s1
+	call	free@plt
+	ld	ra,40(sp)
+	.cfi_remember_state
+	.cfi_restore 1
+	ld	s0,32(sp)
+	.cfi_restore 8
+	ld	s1,24(sp)
+	.cfi_restore 9
+	ld	s2,16(sp)
+	.cfi_restore 18
+	ld	s3,8(sp)
+	.cfi_restore 19
+	ld	s4,0(sp)
+	.cfi_restore 20
+	li	a0,0
+	addi	sp,sp,48
+	.cfi_def_cfa_offset 0
+	jr	ra
+.L35:
+	.cfi_restore_state
+	ld	a0,8(a1)
+	li	a2,10
+	li	a1,0
+	call	strtol@plt
+	sext.w	s4,a0
+	mv	a1,s4
+	lla	a0,.LC0
+	call	printf@plt
+	slli	s3,s4,3
+	mv	a0,s3
+	call	malloc@plt
+	mv	s0,a0
+	mv	a0,s3
+	call	malloc@plt
+	mv	s2,a0
+	mv	a0,s3
+	call	malloc@plt
+	mv	s1,a0
+	bgt	s4,zero,.L22
+	fmv.d.x	fa5,zero
+	j	.L17
+	.cfi_endproc
+.LFE23:
+	.size	main, .-main
+	.section	.rodata.cst8,"aM",@progbits,8
+	.align	3
+.LC1:
+	.word	0
+	.word	1073741824
+	.ident	"GCC: (GNU) 14.2.1 20250207"
+	.section	.note.GNU-stack,"",@progbits

examples/riscy_asm_files/triad.s

@@ -0,0 +1,184 @@
+	.file	"triad.c"
+	.option pic
+	.attribute arch, "rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0"
+	.attribute unaligned_access, 0
+	.attribute stack_align, 16
+	.text
+	.align	1
+	.globl	kernel
+	.type	kernel, @function
+kernel:
+.LFB22:
+	.cfi_startproc
+	ble	a3,zero,.L1
+	slli	a3,a3,3
+	add	a5,a1,a3
+.L3:
+	fld	fa5,0(a2)
+	fld	fa4,0(a1)
+	addi	a1,a1,8
+	addi	a2,a2,8
+	fmadd.d	fa5,fa5,fa0,fa4
+	addi	a0,a0,8
+	fsd	fa5,-8(a0)
+	bne	a1,a5,.L3
+.L1:
+	ret
+	.cfi_endproc
+.LFE22:
+	.size	kernel, .-kernel
+	.section	.rodata.str1.8,"aMS",@progbits,1
+	.align	3
+.LC0:
+	.string	"RISC-V STREAM triad: a[i] = b[i] + s * c[i], size=%d\n"
+	.align	3
+.LC2:
+	.string	"Checksum: %f\n"
+	.section	.text.startup,"ax",@progbits
+	.align	1
+	.globl	main
+	.type	main, @function
+main:
+.LFB23:
+	.cfi_startproc
+	addi	sp,sp,-48
+	.cfi_def_cfa_offset 48
+	sd	ra,40(sp)
+	sd	s0,32(sp)
+	sd	s1,24(sp)
+	sd	s2,16(sp)
+	sd	s3,8(sp)
+	sd	s4,0(sp)
+	.cfi_offset 1, -8
+	.cfi_offset 8, -16
+	.cfi_offset 9, -24
+	.cfi_offset 18, -32
+	.cfi_offset 19, -40
+	.cfi_offset 20, -48
+	li	a5,1
+	bgt	a0,a5,.L21
+	li	a1,1000
+	lla	a0,.LC0
+	call	printf@plt
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s1,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s2,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	li	s0,8192
+	mv	s3,a0
+	addi	s0,s0,-192
+	li	s4,1000
+.L13:
+	slli	a5,s4,32
+	srli	a2,a5,29
+	li	a1,0
+	mv	a0,s1
+	call	memset@plt
+	mv	a4,s2
+	mv	a3,s2
+	li	a5,0
+.L9:
+	fcvt.d.w	fa5,a5
+	mv	a1,a5
+	addiw	a5,a5,1
+	fsd	fa5,0(a3)
+	addi	a3,a3,8
+	bne	a5,s4,.L9
+	mv	a2,s3
+	mv	a3,s3
+	li	a5,0
+.L10:
+	fcvt.d.w	fa5,a5
+	mv	a0,a5
+	addi	a3,a3,8
+	fadd.d	fa5,fa5,fa5
+	addiw	a5,a5,1
+	fsd	fa5,-8(a3)
+	bne	a0,a1,.L10
+	fld	fa3,.LC1,a5
+	add	a1,s2,s0
+	mv	a5,s1
+	mv	a3,s1
+.L11:
+	fld	fa5,0(a2)
+	fld	fa4,0(a4)
+	addi	a4,a4,8
+	addi	a2,a2,8
+	fmadd.d	fa5,fa5,fa3,fa4
+	addi	a3,a3,8
+	fsd	fa5,-8(a3)
+	bne	a4,a1,.L11
+	fmv.d.x	fa5,zero
+	add	s0,s1,s0
+.L12:
+	fld	fa4,0(a5)
+	addi	a5,a5,8
+	fadd.d	fa5,fa5,fa4
+	bne	a5,s0,.L12
+.L8:
+	fmv.x.d	a1,fa5
+	lla	a0,.LC2
+	call	printf@plt
+	mv	a0,s1
+	call	free@plt
+	mv	a0,s2
+	call	free@plt
+	mv	a0,s3
+	call	free@plt
+	ld	ra,40(sp)
+	.cfi_remember_state
+	.cfi_restore 1
+	ld	s0,32(sp)
+	.cfi_restore 8
+	ld	s1,24(sp)
+	.cfi_restore 9
+	ld	s2,16(sp)
+	.cfi_restore 18
+	ld	s3,8(sp)
+	.cfi_restore 19
+	ld	s4,0(sp)
+	.cfi_restore 20
+	li	a0,0
+	addi	sp,sp,48
+	.cfi_def_cfa_offset 0
+	jr	ra
+.L21:
+	.cfi_restore_state
+	ld	a0,8(a1)
+	li	a2,10
+	li	a1,0
+	call	strtol@plt
+	sext.w	s4,a0
+	mv	a1,s4
+	lla	a0,.LC0
+	call	printf@plt
+	slli	s0,s4,3
+	mv	a0,s0
+	call	malloc@plt
+	mv	s1,a0
+	mv	a0,s0
+	call	malloc@plt
+	mv	s2,a0
+	mv	a0,s0
+	call	malloc@plt
+	mv	s3,a0
+	bgt	s4,zero,.L13
+	fmv.d.x	fa5,zero
+	j	.L8
+	.cfi_endproc
+.LFE23:
+	.size	main, .-main
+	.section	.rodata.cst8,"aM",@progbits,8
+	.align	3
+.LC1:
+	.word	1374389535
+	.word	1074339512
+	.ident	"GCC: (GNU) 14.2.1 20250207"
+	.section	.note.GNU-stack,"",@progbits

examples/riscy_asm_files/triad_vec.s

@@ -0,0 +1,234 @@
+	.file	"triad.c"
+	.option pic
+	.attribute arch, "rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_v1p0_zicsr2p0_zifencei2p0_zve32f1p0_zve32x1p0_zve64d1p0_zve64f1p0_zve64x1p0_zvl128b1p0_zvl32b1p0_zvl64b1p0"
+	.attribute unaligned_access, 0
+	.attribute stack_align, 16
+	.text
+	.align	1
+	.globl	kernel
+	.type	kernel, @function
+kernel:
+.LFB22:
+	.cfi_startproc
+	ble	a3,zero,.L10
+	addiw	a5,a3,-1
+	li	a4,4
+	bleu	a5,a4,.L3
+	addi	a5,a2,8
+	addi	a4,a1,8
+	sub	a5,a0,a5
+	sub	a4,a0,a4
+	bgtu	a5,a4,.L14
+.L4:
+	csrr	a4,vlenb
+	addi	a4,a4,-16
+	bleu	a5,a4,.L3
+	vsetvli	a5,zero,e64,m1,ta,ma
+	vfmv.v.f	v3,fa0
+.L5:
+	vsetvli	a5,a3,e64,m1,ta,ma
+	vle64.v	v2,0(a1)
+	vle64.v	v1,0(a2)
+	slli	a4,a5,3
+	sub	a3,a3,a5
+	add	a1,a1,a4
+	add	a2,a2,a4
+	vfmadd.vv	v1,v3,v2
+	vse64.v	v1,0(a0)
+	add	a0,a0,a4
+	bne	a3,zero,.L5
+	ret
+.L3:
+	slli	a3,a3,3
+	add	a3,a1,a3
+.L7:
+	fld	fa5,0(a2)
+	fld	fa4,0(a1)
+	addi	a1,a1,8
+	addi	a2,a2,8
+	fmadd.d	fa5,fa0,fa5,fa4
+	addi	a0,a0,8
+	fsd	fa5,-8(a0)
+	bne	a3,a1,.L7
+.L10:
+	ret
+.L14:
+	mv	a5,a4
+	j	.L4
+	.cfi_endproc
+.LFE22:
+	.size	kernel, .-kernel
+	.section	.rodata.str1.8,"aMS",@progbits,1
+	.align	3
+.LC0:
+	.string	"RISC-V STREAM triad: a[i] = b[i] + s * c[i], size=%d\n"
+	.align	3
+.LC3:
+	.string	"Checksum: %f\n"
+	.section	.text.startup,"ax",@progbits
+	.align	1
+	.globl	main
+	.type	main, @function
+main:
+.LFB23:
+	.cfi_startproc
+	addi	sp,sp,-48
+	.cfi_def_cfa_offset 48
+	sd	ra,40(sp)
+	sd	s0,32(sp)
+	sd	s1,24(sp)
+	sd	s2,16(sp)
+	sd	s3,8(sp)
+	sd	s4,0(sp)
+	.cfi_offset 1, -8
+	.cfi_offset 8, -16
+	.cfi_offset 9, -24
+	.cfi_offset 18, -32
+	.cfi_offset 19, -40
+	.cfi_offset 20, -48
+	li	a5,1
+	bgt	a0,a5,.L35
+	li	a1,1000
+	lla	a0,.LC0
+	call	printf@plt
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s0,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	mv	s2,a0
+	li	a0,8192
+	addi	a0,a0,-192
+	call	malloc@plt
+	li	s3,8192
+	mv	s1,a0
+	addi	s3,s3,-192
+	li	s4,1000
+.L22:
+	slli	a5,s4,32
+	srli	a2,a5,29
+	li	a1,0
+	mv	a0,s0
+	call	memset@plt
+	mv	a4,s2
+	li	a5,0
+.L18:
+	fcvt.d.w	fa5,a5
+	addiw	a5,a5,1
+	addi	a4,a4,8
+	fsd	fa5,-8(a4)
+	bne	a5,s4,.L18
+	fld	fa5,.LC1,a4
+	vsetvli	a3,zero,e64,m1,ta,ma
+	mv	a2,a5
+	vfmv.v.f	v4,fa5
+	vsetvli	zero,zero,e32,mf2,ta,ma
+	vid.v	v2
+	mv	a3,s1
+.L19:
+	vsetvli	a4,a2,e32,mf2,ta,ma
+	vfwcvt.f.x.v	v1,v2
+	vsetvli	a0,zero,e32,mf2,ta,ma
+	vmv.v.x	v3,a4
+	vsetvli	zero,a4,e64,m1,ta,ma
+	vfmul.vv	v1,v1,v4
+	vsetvli	a0,zero,e32,mf2,ta,ma
+	vadd.vv	v2,v2,v3
+	vsetvli	zero,a4,e64,m1,ta,ma
+	slli	a1,a4,3
+	sub	a2,a2,a4
+	vse64.v	v1,0(a3)
+	add	a3,a3,a1
+	bne	a2,zero,.L19
+	fld	fa5,.LC2,a4
+	vsetvli	a0,zero,e64,m1,ta,ma
+	mv	a3,s0
+	vfmv.v.f	v3,fa5
+	mv	a0,s1
+	mv	a1,s2
+.L20:
+	vsetvli	a4,a5,e64,m1,ta,ma
+	vle64.v	v2,0(a1)
+	vle64.v	v1,0(a0)
+	slli	a2,a4,3
+	sub	a5,a5,a4
+	add	a1,a1,a2
+	add	a0,a0,a2
+	vfmadd.vv	v1,v3,v2
+	vse64.v	v1,0(a3)
+	add	a3,a3,a2
+	bne	a5,zero,.L20
+	fmv.d.x	fa5,zero
+	add	s3,s0,s3
+	mv	a5,s0
+.L21:
+	fld	fa4,0(a5)
+	addi	a5,a5,8
+	fadd.d	fa5,fa5,fa4
+	bne	s3,a5,.L21
+.L17:
+	fmv.x.d	a1,fa5
+	lla	a0,.LC3
+	call	printf@plt
+	mv	a0,s0
+	call	free@plt
+	mv	a0,s2
+	call	free@plt
+	mv	a0,s1
+	call	free@plt
+	ld	ra,40(sp)
+	.cfi_remember_state
+	.cfi_restore 1
+	ld	s0,32(sp)
+	.cfi_restore 8
+	ld	s1,24(sp)
+	.cfi_restore 9
+	ld	s2,16(sp)
+	.cfi_restore 18
+	ld	s3,8(sp)
+	.cfi_restore 19
+	ld	s4,0(sp)
+	.cfi_restore 20
+	li	a0,0
+	addi	sp,sp,48
+	.cfi_def_cfa_offset 0
+	jr	ra
+.L35:
+	.cfi_restore_state
+	ld	a0,8(a1)
+	li	a2,10
+	li	a1,0
+	call	strtol@plt
+	sext.w	s4,a0
+	mv	a1,s4
+	lla	a0,.LC0
+	call	printf@plt
+	slli	s3,s4,3
+	mv	a0,s3
+	call	malloc@plt
+	mv	s0,a0
+	mv	a0,s3
+	call	malloc@plt
+	mv	s2,a0
+	mv	a0,s3
+	call	malloc@plt
+	mv	s1,a0
+	bgt	s4,zero,.L22
+	fmv.d.x	fa5,zero
+	j	.L17
+	.cfi_endproc
+.LFE23:
+	.size	main, .-main
+	.section	.rodata.cst8,"aM",@progbits,8
+	.align	3
+.LC1:
+	.word	0
+	.word	1073741824
+	.align	3
+.LC2:
+	.word	1374389535
+	.word	1074339512
+	.ident	"GCC: (GNU) 14.2.1 20250207"
+	.section	.note.GNU-stack,"",@progbits

examples/triad/Makefile

@@ -0,0 +1,24 @@
+# Makefile for RISC-V triad example
+
+CC = gcc
+CFLAGS = -O3
+CFLAGS_VEC = -O3 -march=rv64gcv
+
+# Default target with -O3
+all: triad triad_vec
+
+# Build with -O3 optimization
+triad: triad.c
+	$(CC) $(CFLAGS) -o triad triad.c
+	$(CC) $(CFLAGS) -S -o triad.s triad.c
+
+# Build with vector extensions
+triad_vec: triad.c
+	$(CC) $(CFLAGS_VEC) -o triad_vec triad.c
+	$(CC) $(CFLAGS_VEC) -S -o triad_vec.s triad.c
+
+# Clean up
+clean:
+	rm -f triad triad_vec triad.s triad_vec.s
+
+.PHONY: all clean 

examples/triad/triad.c

@@ -0,0 +1,55 @@
+// STREAM triad benchmark for RISC-V testing
+// a[i] = b[i] + s * c[i]
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#define DTYPE double
+
+void kernel(DTYPE* a, DTYPE* b, DTYPE* c, const DTYPE s, const int size)
+{
+    // OSACA start marker will be added around this loop
+    for(int i=0; i<size; i++) {
+        a[i] = b[i] + s * c[i];
+    }
+    // OSACA end marker will be added
+}
+
+int main(int argc, char *argv[]) {
+    int size = 1000;
+    if(argc > 1) {
+        size = atoi(argv[1]);
+    }
+    
+    printf("RISC-V STREAM triad: a[i] = b[i] + s * c[i], size=%d\n", size);
+    
+    // Allocate memory
+    DTYPE* a = (DTYPE*)malloc(size * sizeof(DTYPE));
+    DTYPE* b = (DTYPE*)malloc(size * sizeof(DTYPE));
+    DTYPE* c = (DTYPE*)malloc(size * sizeof(DTYPE));
+    
+    // Initialize arrays
+    for(int i=0; i<size; i++) {
+        a[i] = 0.0;
+        b[i] = i;
+        c[i] = i * 2.0;
+    }
+    
+    // Run kernel
+    DTYPE scalar = 3.14;
+    kernel(a, b, c, scalar, size);
+    
+    // Check result (to prevent optimization)
+    DTYPE checksum = 0.0;
+    for(int i=0; i<size; i++) {
+        checksum += a[i];
+    }
+    printf("Checksum: %f\n", checksum);
+    
+    // Cleanup
+    free(a);
+    free(b);
+    free(c);
+    
+    return 0;
+} 

osaca/data/isa/riscv.yml

@@ -1,5 +1,5 @@
 ---
-osaca_version: 0.6.1
+osaca_version: 0.7.0
 isa: riscv
 # Contains all operand-irregular instruction forms OSACA supports for RISC-V.
 # Operand-regular for a RISC-V instruction form with N operands in the shape of
@@ -8,259 +8,6 @@ isa: riscv
 # For vector instructions with suffixes (.v, .vv, .vf), the operand behavior follows
 # the base instruction pattern.
 instruction_forms:
-- name: addi
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: immediate
-    imd: int
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: add
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: sub
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: mul
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: div
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: and
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: or
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: xor
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: sll
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: srl
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
-- name: sra
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: false
-    destination: true
-  - class: flag
-    name: N
-    source: false
-    destination: true
 - name: lw
   operands:
   - class: register
@@ -364,11 +111,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: true
-    destination: false
 - name: bne
   operands:
   - class: register
@@ -382,11 +124,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: Z
-    source: true
-    destination: false
 - name: blt
   operands:
   - class: register
@@ -400,11 +137,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: N
-    source: true
-    destination: false
 - name: bge
   operands:
   - class: register
@@ -418,11 +150,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: N
-    source: true
-    destination: false
 - name: jal
   operands:
   - class: register
@@ -432,67 +159,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-- name: jalr
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-  - class: immediate
-    imd: int
-    source: true
-    destination: false
-- name: lui
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: immediate
-    imd: int
-    source: true
-    destination: false
-- name: auipc
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: immediate
-    imd: int
-    source: true
-    destination: false
-- name: li
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: immediate
-    imd: int
-    source: true
-    destination: false
-- name: mv
-  operands:
-  - class: register
-    prefix: x
-    source: false
-    destination: true
-  - class: register
-    prefix: x
-    source: true
-    destination: false
-- name: ret
-  operands:
-  - class: register
-    prefix: x
-    name: '1'      # x1 is the return address register (ra)
-    source: true
-    destination: false
 - name: j
   operands:
   - class: identifier
@@ -534,82 +200,6 @@ instruction_forms:
     post_indexed: '*'
     source: false
     destination: true
-- name: fadd
-  operands:
-  - class: register
-    prefix: f
-    source: false
-    destination: true
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: FSR
-    source: true
-    destination: true
-- name: fsub
-  operands:
-  - class: register
-    prefix: f
-    source: false
-    destination: true
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: FSR
-    source: true
-    destination: true
-- name: fmul
-  operands:
-  - class: register
-    prefix: f
-    source: false
-    destination: true
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: FSR
-    source: true
-    destination: true
-- name: fdiv
-  operands:
-  - class: register
-    prefix: f
-    source: false
-    destination: true
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  - class: register
-    prefix: f
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: FSR
-    source: true
-    destination: true
 - name: fmv.x.w
   operands:
   - class: register
@@ -652,11 +242,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: false
-    destination: true
 - name: vsetivli
   operands:
   - class: register
@@ -673,11 +258,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: false
-    destination: true
 - name: vle32.v
   operands:
   - class: register
@@ -693,11 +273,6 @@ instruction_forms:
     post_indexed: '*'
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: true
-    destination: true
 - name: vse32.v
   operands:
   - class: register
@@ -713,30 +288,6 @@ instruction_forms:
     post_indexed: '*'
     source: false
     destination: true
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: true
-    destination: true
-- name: vadd.vv
-  operands:
-  - class: register
-    prefix: v
-    source: false
-    destination: true
-  - class: register
-    prefix: v
-    source: true
-    destination: false
-  - class: register
-    prefix: v
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: true
-    destination: true
 - name: vfmv.v.f
   operands:
   - class: register
@@ -747,30 +298,6 @@ instruction_forms:
     prefix: f
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: true
-    destination: true
-- name: vfmadd.vv
-  operands:
-  - class: register
-    prefix: v
-    source: false
-    destination: true
-  - class: register
-    prefix: v
-    source: true
-    destination: false
-  - class: register
-    prefix: v
-    source: true
-    destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: true
-    destination: true
 - name: vfmacc.vf
   operands:
   - class: register
@@ -785,12 +312,7 @@ instruction_forms:
     prefix: v
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: VSR
-    source: true
-    destination: true
-  # CSR instructions
+# CSR instructions
 - name: csrr
   operands:
   - class: register
@@ -800,11 +322,6 @@ instruction_forms:
   - class: identifier
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: CSR
-    source: true
-    destination: false
 - name: csrw
   operands:
   - class: identifier
@@ -814,11 +331,6 @@ instruction_forms:
     prefix: x
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: CSR
-    source: false
-    destination: true
 - name: csrs
   operands:
   - class: identifier
@@ -828,11 +340,6 @@ instruction_forms:
     prefix: x
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: CSR
-    source: true
-    destination: true
 - name: csrc
   operands:
   - class: identifier
@@ -842,12 +349,7 @@ instruction_forms:
     prefix: x
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: CSR
-    source: true
-    destination: true
-  # Atomic instructions
+# Atomic instructions
 - name: lr.w
   operands:
   - class: register
@@ -863,11 +365,6 @@ instruction_forms:
     post_indexed: '*'
     source: true
     destination: false
-  hidden_operands:
-  - class: flag
-    name: CSR
-    source: true
-    destination: true
 - name: sc.w
   operands:
   - class: register
@@ -886,9 +383,4 @@ instruction_forms:
     pre_indexed: '*'
     post_indexed: '*'
     source: true
-    destination: true
-  hidden_operands:
-  - class: flag
-    name: CSR
-    source: true
-    destination: true
+    destination: true

osaca/data/rv64.yml

@@ -1,5 +1,5 @@
 ---
-osaca_version: 0.6.1
+osaca_version: 0.7.0
 micro_architecture: rv64
 arch_code: rv64
 isa: riscv
@@ -653,4 +653,4 @@ instruction_forms:
     operands: []
     latency: 1
     throughput: 1
-    port_pressure: [[1, ["ALU"]]]
+    port_pressure: [[1, ["ALU"]]]

osaca/osaca.py

@@ -107,8 +107,8 @@ def create_parser(parser=None):
         "--arch",
         type=str,
         help="Define architecture (SNB, IVB, HSW, BDW, SKX, CSX, ICL, ICX, SPR, ZEN1, ZEN2, ZEN3, "
-        "ZEN4, TX2, N1, A64FX, TSV110, A72, M1, V2). If no architecture is given, OSACA assumes a "
-        "default uarch for x86/AArch64.",
+        "ZEN4, TX2, N1, A64FX, TSV110, A72, M1, V2, RV64). If no architecture is given, OSACA assumes a "
+        "default uarch for the detected ISA.",
     )
     parser.add_argument(
         "--fixed",
@@ -325,15 +325,28 @@ def inspect(args, output_file=sys.stdout):
     except Exception as e:
         # probably the wrong parser based on heuristic
         if args.arch is None:
-            # change ISA and try again
-            arch = (
-                DEFAULT_ARCHS["x86"]
-                if BaseParser.detect_ISA(code) == "aarch64"
-                else DEFAULT_ARCHS["aarch64"]
-            )
-            isa = MachineModel.get_isa_for_arch(arch)
-            parser = get_asm_parser(arch)
-            parsed_code = parser.parse_file(code)
+            # Try all supported ISAs in order if auto-detection may have failed
+            detected_isa = BaseParser.detect_ISA(code)
+            fallback_isas = ["x86", "aarch64", "riscv"]
+            # Remove already tried ISA from fallback options
+            if detected_isa in fallback_isas:
+                fallback_isas.remove(detected_isa)
+            
+            # Try each remaining ISA until one works
+            for fallback_isa in fallback_isas:
+                try:
+                    arch = DEFAULT_ARCHS[fallback_isa]
+                    isa = MachineModel.get_isa_for_arch(arch)
+                    parser = get_asm_parser(arch)
+                    parsed_code = parser.parse_file(code)
+                    # If parsing succeeds, break out of the loop
+                    break
+                except Exception:
+                    # Continue trying with next ISA
+                    continue
+            else:
+                # If none of the parsers work, raise the original exception
+                raise e
         else:
             raise e
 

osaca/parser/base_parser.py

@@ -34,12 +34,24 @@ class BaseParser(object):
         heuristics_x86ATT = [r"%[xyz]mm[0-9]", r"%[er][abcd]x[0-9]"]
         # 2) check for v and z vector registers and x/w general-purpose registers
         heuristics_aarch64 = [r"[vz][0-9][0-9]?\.[0-9][0-9]?[bhsd]", r"[wx][0-9]"]
-        matches = {"x86": 0, "aarch64": 0}
+        # 3) check for RISC-V registers (x0-x31, a0-a7, t0-t6, s0-s11) and instructions
+        heuristics_riscv = [
+            r"\bx[0-9]|x[1-2][0-9]|x3[0-1]\b",  # x0-x31 registers
+            r"\ba[0-7]\b",                       # a0-a7 registers
+            r"\bt[0-6]\b",                       # t0-t6 registers 
+            r"\bs[0-9]|s1[0-1]\b",               # s0-s11 registers
+            r"\bzero\b|\bra\b|\bsp\b|\bgp\b",    # zero, ra, sp, gp registers
+            r"\bvsetvli\b|\bvle\b|\bvse\b",      # RV Vector instructions
+            r"\baddi\b|\bsd\b|\bld\b|\bjal\b"    # Common RISC-V instructions
+        ]
+        matches = {"x86": 0, "aarch64": 0, "riscv": 0}
 
         for h in heuristics_x86ATT:
             matches["x86"] += len(re.findall(h, file_content))
         for h in heuristics_aarch64:
             matches["aarch64"] += len(re.findall(h, file_content))
+        for h in heuristics_riscv:
+            matches["riscv"] += len(re.findall(h, file_content))
 
         return max(matches.items(), key=operator.itemgetter(1))[0]
 

osaca/parser/parser_RISCV.py

@@ -1,12 +1,9 @@
 #!/usr/bin/env python3
 import re
 import os
-import logging
 from copy import deepcopy
 import pyparsing as pp
 
-logger = logging.getLogger(__name__)
-
 from osaca.parser import BaseParser
 from osaca.parser.instruction_form import InstructionForm
 from osaca.parser.operand import Operand
@@ -30,7 +27,32 @@ class ParserRISCV(BaseParser):
 
     def __init__(self):
         super().__init__()
-        self.isa = "riscv"
+        # Initialize parser, but don't set 'isa' directly as an attribute
+        self._isa_str = "riscv"
+
+    def isa(self):
+        """Return the ISA string."""
+        return self._isa_str
+
+    def start_marker(self):
+        """Return the OSACA start marker for RISC-V assembly."""
+        # Parse the RISC-V start marker (li a1, 111 followed by NOP)
+        # This matches how start marker is defined in marker_utils.py for RISC-V
+        marker_str = (
+            "li        a1, 111    # OSACA START MARKER\n"
+            ".byte     19,0,0,0   # OSACA START MARKER\n"
+        )
+        return self.parse_file(marker_str)
+
+    def end_marker(self):
+        """Return the OSACA end marker for RISC-V assembly."""
+        # Parse the RISC-V end marker (li a1, 222 followed by NOP)
+        # This matches how end marker is defined in marker_utils.py for RISC-V
+        marker_str = (
+            "li        a1, 222    # OSACA END MARKER\n"
+            ".byte     19,0,0,0   # OSACA END MARKER\n"
+        )
+        return self.parse_file(marker_str)
 
     def construct_parser(self):
         """Create parser for RISC-V ISA."""
@@ -52,10 +74,15 @@ class ParserRISCV(BaseParser):
         vector_identifier = pp.Word(pp.alphas, pp.alphanums)
         special_identifier = pp.Word(pp.alphas + "%")
         
+        # First character of an identifier 
         first = pp.Word(pp.alphas + "_.", exact=1)
+        # Rest of the identifier
         rest = pp.Word(pp.alphanums + "_.")
+        # PLT suffix (@plt) for calls to shared libraries
+        plt_suffix = pp.Optional(pp.Literal("@") + pp.Word(pp.alphas))
+        
         identifier = pp.Group(
-            pp.Combine(first + pp.Optional(rest)).setResultsName("name")
+            (pp.Combine(first + pp.Optional(rest) + plt_suffix)).setResultsName("name")
             + pp.Optional(
                 pp.Suppress(pp.Literal("+"))
                 + (hex_number | decimal_number).setResultsName("offset")
@@ -346,7 +373,6 @@ class ParserRISCV(BaseParser):
             return return_dict
             
         except Exception as e:
-            logger.debug(f"Error parsing instruction: {instruction} - {str(e)}")
             # For special vector instructions or ones with % in them
             if "%" in instruction or instruction.startswith("v"):
                 parts = instruction.split("#")[0].strip().split(None, 1)
@@ -640,4 +666,53 @@ class ParserRISCV(BaseParser):
         elif name.startswith("csr"):
             return "csr"  # Control and Status Register
             
-        return "unknown" 
+        return "unknown"
+
+    def normalize_instruction_form(self, instruction_form, isa_model, arch_model):
+        """
+        Normalize instruction form for RISC-V instructions.
+        
+        :param instruction_form: instruction form to normalize
+        :param isa_model: ISA model to use for normalization
+        :param arch_model: architecture model to use for normalization
+        """
+        if instruction_form.normalized:
+            return
+            
+        if instruction_form.mnemonic is None:
+            instruction_form.normalized = True
+            return
+            
+        # Normalize the mnemonic if needed
+        if instruction_form.mnemonic:
+            # Handle any RISC-V specific mnemonic normalization
+            # For example, convert aliases or pseudo-instructions to their base form
+            pass
+            
+        # Normalize the operands if needed
+        for i, operand in enumerate(instruction_form.operands):
+            if isinstance(operand, ImmediateOperand):
+                # Normalize immediate operands
+                instruction_form.operands[i] = self.normalize_imd(operand)
+            elif isinstance(operand, RegisterOperand):
+                # Convert register names to canonical form if needed
+                pass
+                
+        instruction_form.normalized = True 
+
+    def get_regular_source_operands(self, instruction_form):
+        """Get source operand of given instruction form assuming regular src/dst behavior."""
+        # For RISC-V, the first operand is typically the destination,
+        # and the rest are sources
+        if len(instruction_form.operands) == 1:
+            return [instruction_form.operands[0]]
+        else:
+            return [op for op in instruction_form.operands[1:]]
+    
+    def get_regular_destination_operands(self, instruction_form):
+        """Get destination operand of given instruction form assuming regular src/dst behavior."""
+        # For RISC-V, the first operand is typically the destination
+        if len(instruction_form.operands) == 1:
+            return []
+        else:
+            return instruction_form.operands[:1]