README.md

   1 # IEEE754 Floating-Point ALU, in nmigen
   2
   3 This project implements a pipelined IEEE754 floating-point ALU that
   4 supports FP16, FP32 and FP64.  It is a general-purpose unit that
   5 may be used in any project (not limited to one specific processor).
   6
   7 # Requirements
   8
   9 * nmigen
  10 * yosys (latest git repository, required by nmigen)
  11 * sfpy (running unit tests).  provides python bindings to berkeley softfloat-3
  12
  13 # Building sfpy
  14
  15 The standard sfpy will not work without being modified to the type of
  16 IEEE754 FP emulation being tested.  This FPU is emulating RISC-V, and
  17 there is some weirdness in x86 IEEE754 implementations when it comes
  18 to FP16 non-canonical NaNs.
  19
  20 The following modifications are required to the sfpy berkeley-softfloat-3
  21 Makefile:
  22
  23     diff --git a/build/Linux-x86_64-GCC/Makefile b/build/Linux-x86_64-GCC/Makefile
  24     index 2ee5dad..566d225 100644
  25     --- a/build/Linux-x86_64-GCC/Makefile
  26     +++ b/build/Linux-x86_64-GCC/Makefile
  27     @@ -35,7 +35,7 @@
  28      #=============================================================================
  29
  30      SOURCE_DIR ?= ../../source
  31     -SPECIALIZE_TYPE ?= 8086-SSE
  32     +SPECIALIZE_TYPE ?= RISCV
  33
  34      SOFTFLOAT_OPTS ?= \
  35        -DSOFTFLOAT_ROUND_ODD -DINLINE_LEVEL=5 -DSOFTFLOAT_FAST_DIV32TO16 \
  36     @@ -45,7 +45,7 @@ DELETE = rm -f
  37      C_INCLUDES = -I. -I$(SOURCE_DIR)/$(SPECIALIZE_TYPE) -I$(SOURCE_DIR)/include
  38      COMPILE_C = \
  39        gcc -c -Werror-implicit-function-declaration -DSOFTFLOAT_FAST_INT64 \
  40     -    $(SOFTFLOAT_OPTS) $(C_INCLUDES) -O2 -o $@
  41     +    $(SOFTFLOAT_OPTS) $(C_INCLUDES) -O2 -fPIC -o $@
  42      MAKELIB = ar crs $@
  43
  44      OBJ = .o
  45
  46
  47 The following modifications are required to the sfpy SoftPosit Makefile:
  48
  49     diff --git a/build/Linux-x86_64-GCC/Makefile b/build/Linux-x86_64-GCC/Makefile
  50     index 7affd4b..25dd39e 100644
  51     --- a/build/Linux-x86_64-GCC/Makefile
  52     +++ b/build/Linux-x86_64-GCC/Makefile
  53     @@ -69,7 +69,7 @@ endif
  54      C_INCLUDES = -I. -I$(SOURCE_DIR)/$(SPECIALIZE_TYPE) -I$(SOURCE_DIR)/include
  55      OPTIMISATION  = -O2 #-march=core-avx2
  56      COMPILE_C = \
  57     -  $(COMPILER) -c -Werror-implicit-function-declaration -DSOFTPOSIT_FAST_INT64 \
  58     +  $(COMPILER) -fPIC -c -Werror-implicit-function-declaration -DSOFTPOSIT_FAST_INT64 \
  59          $(SOFTPOSIT_OPTS) $(C_INCLUDES) $(OPTIMISATION) \
  60          -o $@
  61      MAKELIB = ar crs $@
  62
  63 # Useful resources
  64
  65 * https://en.wikipedia.org/wiki/IEEE_754-1985
  66 * http://weitz.de/ieee/
  67 * https://steve.hollasch.net/cgindex/coding/ieeefloat.html
  68