TM4C1294NCPDT: How to globally disable/enable interrupts?

I just noticed that you want direct register manipulation. I doubt if that would have any performance benefit over calling IntMasterDisable() ... IntMasterEnable() and furthermore I prefer easy-to-understand C APIs for maintainability, but if you insist, the answer is in the implementation of IntMasterEnable() and IntMasterDisable():

//*****************************************************************************
//
//! Enables the processor interrupt.
//!
//! This function allows the processor to respond to interrupts.  This function
//! does not affect the set of interrupts enabled in the interrupt controller;
//! it just gates the single interrupt from the controller to the processor.
//!
//! \b Example: Enable interrupts to the processor.
//!
//! \verbatim
//! //
//! // Enable interrupts to the processor.
//! //
//! IntMasterEnable();
//!
//! \endverbatim
//!
//! \return Returns \b true if interrupts were disabled when the function was
//! called or \b false if they were initially enabled.
//
//*****************************************************************************
bool
IntMasterEnable(void)
{
    //
    // Enable processor interrupts.
    //
    return(CPUcpsie());
}

//*****************************************************************************
//
//! Disables the processor interrupt.
//!
//! This function prevents the processor from receiving interrupts.  This
//! function does not affect the set of interrupts enabled in the interrupt
//! controller; it just gates the single interrupt from the controller to the
//! processor.
//!
//! \note Previously, this function had no return value.  As such, it was
//! possible to include <tt>interrupt.h</tt> and call this function without
//! having included <tt>hw_types.h</tt>.  Now that the return is a
//! <tt>bool</tt>, a compiler error occurs in this case.  The solution
//! is to include <tt>hw_types.h</tt> before including <tt>interrupt.h</tt>.
//!
//! \b Example: Disable interrupts to the processor.
//!
//! \verbatim
//! //
//! // Disable interrupts to the processor.
//! //
//! IntMasterDisable();
//!
//! \endverbatim
//!
//! \return Returns \b true if interrupts were already disabled when the
//! function was called or \b false if they were initially enabled.
//
//*****************************************************************************
bool
IntMasterDisable(void)
{
    //
    // Disable processor interrupts.
    //
    return(CPUcpsid());
}

And looking at CPUcpsie() and CPUcpsid() in C:\ti\TivaWare_C_Series-2.1.4.178\driverlib\cpu.c if you installed TivaWare to the default location, these are implemented in assembly. You can call these directly but I suggest just sticking with the TivaWare APIs for self-documented code and maintainability.

First, thank you for starting a new thread. When a thread has not been active for over a month it is closed. If additional questions come up, the proper thing to do is exactly what you did, create a new thread with a link to the old thread.

I am not sure why IntMasterDisable() and IntMasterEnable() are not doing what you want. (You may need to clarify.) If you have a critical section of code that cannot be interrupted, a call to IntMasterDisable() disables all programmable interrupts and returns TRUE if interrupts were already disabled, FALSE if they were previously enabled. After the section of critical code, you call IntMasterEnable() if you need to re-enable the interrupts. IntMasterEnable() also returns TRUE if interrupts were disabled when this was called.

If you want to see under the hood, page 96 of the TM4C1294NCPDT datasheet describes the PRIMASK register. These functions simply read and write to this Cortex-M4 register. If you need to read this register without writing it, there is a "function" uint32_t CPUprimask(void), in the driver library file cpu.c that returns the value of the PRIMASK register.

There is additional information about the exception architecture of the TM4C1294 in section 2.5 of the datasheet.