Chapter optimizari.

curs/curs-16-17-optimizari/.gitignore

@@ -0,0 +1,25 @@
+.vscode/
+
+# Build artifacts
+*.o
+*~
+
+# Example binaries
+gprof_demo/gprof_demo
+gprof_demo/profile_data.txt
+gprof_demo/gmon.out
+inline/no-inline
+inline/inline
+loop_unroll/normal_loop
+loop_unroll/unrolled_loop
+opt_demo/test
+opt_demo/test_o1
+opt_demo/test_o2
+opt_demo/test_o3
+opt_demo/test_unroll
+prefetch/prefetch
+prefetch/no-prefetch
+sse/test_sse
+time_demo/clk
+time_demo/rdtscp
+time_demo/papi

curs/curs-16-17-optimizari/inline/Makefile

@@ -1,4 +1,4 @@
-CFLAGS=-m32
+CFLAGS=-m64
 
 run: no-inline inline
 	./no-inline

curs/curs-16-17-optimizari/loop_unroll/Makefile

@@ -1,4 +1,4 @@
-CFLAGS=-m32
+CFLAGS=-m64
 
 run: normal_loop unrolled_loop
 	./normal_loop

curs/curs-16-17-optimizari/opt_demo/Makefile

@@ -1,19 +1,19 @@
 all: test test_o1 test_o2 test_o3 test_unroll
 
 test: test.c
-	gcc -Wall -m32 -O0 -o test test.c
+	gcc -Wall -m64 -O0 -o $@ $<
 
 test_o1: test.c
-	gcc -Wall -m32 -O1 -o test_o1 test.c
+	gcc -Wall -m64 -O1 -o $@ $<
 
 test_o2: test.c
-	gcc -Wall -m32 -O2 -o test_o2 test.c
+	gcc -Wall -m64 -O2 -o $@ $<
 
 test_o3: test.c
-	gcc -Wall -m32 -O3 -o test_o3 test.c
+	gcc -Wall -m64 -O3 -o $@ $<
 
 test_unroll: test.c
-	gcc -Wall -funroll-loops -m32 -O3 -o test_unroll test.c
+	gcc -Wall -funroll-loops -m64 -O3 -o $@ $<
 
 clean:
 	rm -rf test test_o* test_unroll

curs/curs-16-17-optimizari/sse/Makefile

@@ -1,13 +1,13 @@
-ASM_FLAGS = -f elf32 -g -F dwarf
-C_FLAGS = -g -m32 -O3
+ASM_FLAGS = -f elf64 -g -F dwarf
+C_FLAGS = -g -m64 -O3
 NASM=nasm
 CC=gcc
 CPP=g++
 
 all: test_sse
 
 test_sse: test_sse.o sse.o
-	gcc -g -O3 -m32 -o $@ $^
+	gcc -g -O3 -m64 -o $@ $^
 
 # test_sa.o: test_sa.asm
 # 	nasm -f elf32 -g -F dwarf test_sa.asm

curs/curs-16-17-optimizari/sse/sse.asm

@@ -1,74 +1,39 @@
-	BITS 32
+	BITS 64
+	DEFAULT REL
+
 	GLOBAL sum_array_sse
 sum_array_sse:
-	push ebp
-	mov ebp, esp
-	push esi
-	push edi
-
-	push ebx
-	mov ecx, [ebp + 20] ; ecx = n
-	mov esi, [ebp + 8] ; esi = a
-	mov edi, [ebp + 12] ; edi = b
-	mov ebx, [ebp + 16] ; ebx = c
-	; n = n / 16	shr ecx, 4		
+	; SysV x64 ABI: rdi=a, rsi=b, rdx=c, ecx=n
 	xor eax, eax
 
 	cmp eax, ecx
-	jge end
-begin:
-	movdqu xmm0, [esi]
-	movdqu xmm1, [edi]
+	jge .end
+.begin:
+	movdqu xmm0, [rdi + rax*4]
+	movdqu xmm1, [rsi + rax*4]
 	paddd xmm0, xmm1
-	movdqu [ebx], xmm0
-
-	add esi, 16
-	add edi, 16
-	add ebx, 16
-	add eax, 4	
+	movdqu [rdx + rax*4], xmm0
+
+	add eax, 4
 	cmp eax, ecx
-	jl begin
-end:
-	pop ebx
-	pop edi
-	pop esi
-	leave
+	jl .begin
+.end:
 	ret
 
-
-GLOBAL sum_array_plain 
+	GLOBAL sum_array_plain
 sum_array_plain:
-	push ebp
-	mov ebp, esp
-	push esi
-	push edi
-	push ebx
-	push edx
-
-	mov ecx, [ebp + 20] ; ecx = n
-	mov esi, [ebp + 8] ; esi = a
-	mov edi, [ebp + 12] ; edi = b
-	mov ebx, [ebp + 16] ; ebx = c
-
+	; SysV x64 ABI: rdi=a, rsi=b, rdx=c, ecx=n
 	xor eax, eax
 
 	cmp eax, ecx
-	jge pend
-pbegin:
-	mov edx, [esi+eax*4]
-	add edx, [edi+eax*4]
-	mov [ebx+eax*4], edx
-
-	;; add esi, 4		
-	;; add edi, 4
-	;; add ebx, 4
-	add eax, 1	
+	jge .pend
+.pbegin:
+	mov r8d, [rdi + rax*4]
+	add r8d, [rsi + rax*4]
+	mov [rdx + rax*4], r8d
+
+	add eax, 1
 	cmp eax, ecx
-	jl pbegin
-pend:
-	pop edx
-	pop ebx
-	pop edi
-	pop esi
-	leave
+	jl .pbegin
+.pend:
 	ret

curs/curs-16-17-optimizari/sse/test_sse.c

@@ -3,9 +3,10 @@
 #include <stdlib.h>
 
 uint64_t rdtsc(void) {
-    uint64_t result;
-    __asm__ __volatile__ ("rdtsc" : "=A" (result));
-    return result;
+	uint32_t lo, hi;
+
+	__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
+	return ((uint64_t)hi << 32) | lo;
 }
 
 void sum_array_sse(uint32_t *a, uint32_t *b, uint32_t *c, int n);

curs/curs-16-17-optimizari/time_demo/Makefile

@@ -1,10 +1,10 @@
-all: clk rdtscp papi
+all: clk rdtscp
 
 clk: clk.c
-	gcc -O3 -m32 -Wall clk.c -o clk
+	gcc -O3 -m64 -Wall clk.c -o clk
 
 rdtscp: rdtscp.c
-	gcc -O3 -m32 -Wall rdtscp.c -o rdtscp
+	gcc -O3 -m64 -Wall rdtscp.c -o rdtscp
 
 papi: papi.c
 	gcc -O3 -Wall papi.c -I$(PAPI_DIR) -L$(PAPI_DIR) -o papi -lpapi