Position independent loads fixed #21

src/asm/program_prologue_linux.inc

@@ -12,8 +12,28 @@
 	mov rcx, rdi
 	mov rbp, qword ptr [rsi]    ;# "mx", "ma"
 	mov rdi, qword ptr [rsi+8]  ;# uint8_t* dataset
-	mov rsi, rdx                ;# convertible_t* scratchpad
+	mov rsi, rdx                ;# uint8_t* scratchpad
 
-	#include "program_prologue_load.inc"
+	mov rax, rbp
+
+	;# zero integer registers
+	xor r8, r8
+	xor r9, r9
+	xor r10, r10
+	xor r11, r11
+	xor r12, r12
+	xor r13, r13
+	xor r14, r14
+	xor r15, r15
+
+	;# load constant registers
+	lea rcx, [rcx+120]
+	movapd xmm8, xmmword ptr [rcx+72]
+	movapd xmm9, xmmword ptr [rcx+88]
+	movapd xmm10, xmmword ptr [rcx+104]
+	movapd xmm11, xmmword ptr [rcx+120]
+	movapd xmm13, xmmword ptr minDbl[rip]
+	movapd xmm14, xmmword ptr absMask[rip]
+	movapd xmm15, xmmword ptr signMask[rip]
 
 	jmp DECL(randomx_program_loop_begin)

src/asm/program_prologue_win64.inc

@@ -20,13 +20,33 @@
 	movdqu xmmword ptr [rsp+16], xmm14
 	movdqu xmmword ptr [rsp+0], xmm15
 
-	; function arguments
-	push rcx                    ; RegisterFile& registerFile
-	mov rbp, qword ptr [rdx]    ; "mx", "ma"
-	mov rdi, qword ptr [rdx+8]  ; uint8_t* dataset
-	mov rsi, r8                 ; convertible_t* scratchpad
-	mov rbx, r9                 ; loop counter
+	;# function arguments
+	push rcx                    ;# RegisterFile& registerFile
+	mov rbp, qword ptr [rdx]    ;# "mx", "ma"
+	mov rdi, qword ptr [rdx+8]  ;# uint8_t* dataset
+	mov rsi, r8                 ;# uint8_t* scratchpad
+	mov rbx, r9                 ;# loop counter
 
-	include program_prologue_load.inc
+	mov rax, rbp
+
+	;# zero integer registers
+	xor r8, r8
+	xor r9, r9
+	xor r10, r10
+	xor r11, r11
+	xor r12, r12
+	xor r13, r13
+	xor r14, r14
+	xor r15, r15
+
+	;# load constant registers
+	lea rcx, [rcx+120]
+	movapd xmm8, xmmword ptr [rcx+72]
+	movapd xmm9, xmmword ptr [rcx+88]
+	movapd xmm10, xmmword ptr [rcx+104]
+	movapd xmm11, xmmword ptr [rcx+120]
+	movapd xmm13, xmmword ptr [minDbl]
+	movapd xmm14, xmmword ptr [absMask]
+	movapd xmm15, xmmword ptr [signMask]
 
 	jmp randomx_program_loop_begin