def conv_3x64_7x32_acc_i (i) :
global V
print " // accumulate r2-r4 to", [read_V(acc_r(i,j)) for j in range(3)]
print_ldr("r10", acc_r(i,0), "limb 0")
print_ldr("r12", acc_r(i,1), "limb 1")
print_ldr("r14", acc_r(i,2), "limb 2")
print " add r2, r2, r10"
print " add r3, r3, r12"
print " add r4, r4, r14"
print " asr r8, r11, #6"
def conv_3x64_7x32_store_end (i) : # store 4 accumulators at end of thread
global V
print " // store r6-r8 to", [read_V(acc_r(i,j)) for j in range(4,7)]
for j in range(4,7) :
print_str("r"+str(j+2), acc_r(i,j), "limb %d" % (j))
print " // compress and store r2-r5"
print_ldr("r6","hh","reload cursor")
print_ldr("r7","q","load q")
print_ldr("r8","q32","load round(-2^32/q)")
print " br_32x2 r2, r3, r7, r8, r9"
print " br_32x2 r4, r5, r7, r8, r9"
print " str r2, [r6], #4"
print " str r4, [r6], #4"
print_str("r6","hh","store cursor")
def conv_3x64_7x32_acc (i) :
global V
print " // accumulate to", [read_V(acc_r(i,j)) for j in range(7)]
print_ldr("r10", acc_r(i,0), "limb 0")
print_ldr("r12", acc_r(i,1), "limb 1")
print_ldr("r14", acc_r(i,2), "limb 2")
print_ldr("r9", acc_r(i,3), "limb 3")
print " add r2, r2, r10"
print " add r3, r3, r12"
print " add r4, r4, r14"
print " add r5, r5, r9"
print_ldr("r10", acc_r(i,4), "limb 4")
print_ldr("r12", acc_r(i,5), "limb 5")
print_ldr("r8", acc_r(i,6), "limb 6")
print " add r6, r6, r10"
print " add r7, r7, r12"
print " add r8, r8, r11, asr #6"
def v (s) :
return read_V(s)
def conv_3x64_7x32_store (i) :
global V
print " // store r2-r8 to", [read_V(acc_r(i,j)) for j in range(7)]
for j in range(7) :
print_str("r"+str(j+2), acc_r(i,j), "limb %d" % (j))